Upload
trinhhanh
View
215
Download
0
Embed Size (px)
Citation preview
1
Statewide Models
Statewide Model Topicsbull Why do we need Statewide Models
bull Statewide model theory and typical steps
bull Similarities and differences with urban models
bull Data needs and considerations
bull Validation and reasonableness checking transferability
bull Interfacing statewide and urban models
bull Shortcomings and areas for improvement
Topic ndash Why do we need Statewide Models
bull Rationale
bull Unique features
bull Current implementation
bull Contextunique issues
bull Typical uses
bull Examples of statewide models
bull Potential for micro-simulation
bull Documenting current practice
Source TRB presentation on Wisconsin Statewide Model WISDOT Cambridge Systematics Inc
Rationale ndash Why Statewide Modelsbull Consistency in statewide planning
ndash policy issues
bull Often motivated by specific project requirements
bull Issues or projects affecting entire state or beyond single MPO
bull Travel analyses for rural areas outside MPOs
bull Forecasting MPO external trips
bull Examining freight issues
Source University of Connecticut Connecticut Dept of Transportation
Unique Features of Statewide Models
bull Geographic scale and resolution In-state traffic analysis zones ldquoHalordquo or ldquoborderrdquo set of detailed
zones in adjacent state(s) where significant activity lies across the border
ldquoExternalrdquo zones for passenger travel at state line or halo border
Other states or groups of states for freight modeling (see next example map)
Source Florida Department of Transportation 2005 Statewide Model
Unique Features of Statewide Models (Contrsquod)
bull More aggregate than MPO models (networkszones)
bull Simplified approaches used at times (not necessarily recommended) Three step structure Synthetic or borrowed model structures and
parameters
bull Explicit treatment of visitors and long-distance travel
bull Freight a larger emphasis
bull Potential for inconsistencies with urban models
bull Some linkages with land use and economic models
Source Florida Department of Transportation 2005 Statewide Model
Current Status of Statewide Models
Operational (28)
Dormant (3)
No model (11)
Developing (8)
Source A Horowitz Statewide Travel Forecasting Models NCHRP Synthesis 358 (2006) with recent updates
Statewide ModelsContext for statewide modeling
MacroscopicMegascopic Mesoscopic Microscopic
Urban travel models (tour-
based or sequential)
Regional model of economic
trade and land use trends
Dynamic traffic assignment to
planning network
Detailed traffic analysis of key facilities and
corridors
Transport costs disutilities and
travel times
Multimodal person and freight tours
Time-sliced demand flows by mode
Congestion indices link and
node delay
Economic triggers for freight flows
Accessibilities and transport costs
Source Federal Highway Administration Pilot Course on Statewide Travel Forecasting 2007
Issues unique to statewide models
bull Explicit linkages to economic models
bull Scale issues Modeling states at an urban scale Expensive to build and maintain
bull Need for interfaces with urban models Consistency Data exchange
bull Lack of standard practicebull Unique components
Freight (commodities rarr vehicles) Long distance travel Economic impacts
Data and knowledge gaps
Source Michigan Department of Transportation
Typical uses of Statewide ModelsLess common usesbull Air quality conformity analysesbull Freight and intermodal
planningbull Traffic impact studiesbull Economic development studiesbull Long-term investment studiesbull Emergency evacuation studiesbull Multi-state corridors (possibly
with multi-state models)
Most commonly citedbull Corridor planning (including
community bypasses)bull Forecasting freighttruck
flowsbull External trip forecasts for
MPO amp regional models bull Expansion of MPO amp regional
models
Source Pennsylvania Department of Transportation TRB presentation on 2006-2030 Statewide Model
Louisianarsquos and Virginiarsquos statewide modelsMarket segments 2000 source amp estimation procedure Forecast source amp estimation
procedure
Macrolevel
1 Interstate and interparishtruck trip table
2000 macro zone level Transearch data some matrix estimation
2030 macro zone forecast of 2000 Transearch data interpolation and extrapolation to forecast other years
2 Long distance interstate business tourist and other auto trip tables
1995 American Travel Survey data disaggregated from state and MSA to county level and Fratar balanced to 2000 using household and employment growth
Fratar balance year 2000 county-level trip tables to forecast year using household and employment growth
3 Long distance intrastate business tourist and other auto trip tables
4 Short distance Interstate Louisiana HBW trip table
1990 Census county level JTW data Fratar balanced to 2000 using household and employment growth
Microlevel
5 Short distance intraparish truck trip table
Synthesized from 2000 intraparish Transearch data
Fratar balance 2000 intraparish trip table to forecast year using household and employment growth6 Short distance intrastate
Louisiana HBW trip table1990 tract-level JTW data Fratar balanced to 2000 using household and employment growth
7 Short distance intrastate HBO and NHB trip tables
Urban model style household-based trip generation and gravity model distribution
Urban style household-based trip generation and distribution models log-linear regression used to estimate external trip ends
Source Virginia Department of Transportation
MicrosimulationActivity-based Approach
bull Models dynamic evolution and actions of discrete agents across time and space
bull Observe emergent behavior Interaction Cooperation Coordination
bull Primary emphasis on understanding the system
bull Maybe forecast it
Between autonomous perceptive and deliberative agents
Source Ohio Department of Transportation
Key Differences ndash Traditional vs MicrosimulationldquoTraditional modelsrdquobull Closed form equilibriumbull Well understoodbull Typically deterministicbull Presumes central tendencies
in underlying databull Well-behavedbull ldquoManageablerdquo variationrdquobull Computationally efficient
Microsimulationbull Disequilibriumbull Large systems NWUbull Stochasticbull Diverse behaviorbull Handles wide variancesbull Accommodates weak to strong
understandingbull Computationally heavy
Zonal Land Use Data
Worker Traveler Generation
Vehicle Assignment
Starting Time Assignment
Post-Processor
Next Stop Purpose Choice
Model
Next Stop Location
Choice Model
Dynamic Activity Pattern Generation
Commercial Vehicle Trip List
Zonal Land Use Data
Worker Traveler Generation
Vehicle Assignment
Starting Time Assignment
Post-Processor
Next Stop Purpose Choice
Model
Next Stop Location
Choice Model
Dynamic Activity Pattern Generation
Commercial Vehicle Trip List
Microsimulation Examplesbull Ohiobull Oregon
Source Ohio Department of Transportation
Statewide Model Current Practicebull Meetings of the TRB Subcommittee on Statewide Travel Forecasting Models
bull NCHRP Report 735 Long-Distance and Rural Transferable Parameters for Statewide Travel Forecasting Models 2012
bull NCHRP Validation and Sensitivity Considerations for Statewide Models September 2010
bull White Paper on Statewide Modeling presented at TRB Statewide Planning Conference September 2008
bull National Travel Demand Forecasting Model Phase I Final Scope 2008
bull Forecasting Statewide Freight Toolkit NCHRP Report 606 2008
bull TRB Annual Meeting Sessions in 2004 2006 and 2008 web page
bull Statewide Travel Forecasting Models NCHRP Synthesis 358 2006
bull Statewide Travel Demand Modeling Peer Exchange Transportation Research Circular E-C075 September 2004
bull Guidebook on Statewide Travel Forecasting and Irvine Conference Transportation Research Circular E-C011 1999
Topic ndash Statewide Model Theory
bull Common themes in statewide model development
bull Alternative structures Trip table estimation (2 step) Highway model structure (3 step) Fully multimodal (4 step) Beyond 4 step
bull Deciding on best approach
bull Data development challengesSource Massachusetts Highway DeptCambridge Systematics Inc
Common Themes in Review of Statewide Models
bull Most statewide models use similar methodologies to metropolitan models but there are many variations
bull The largest problems relate to issues of scale
bull Models from different states vary greatly in complexity cost and development time
bull Models are most successful when addressing statewide priorities
bull There are major deficiencies in available data about long-distance and rural passenger travelhellip NCHRP Report 735 has helped
bull Statewide models are more compatible with secondary freight data sources such as the Commodity Flow Survey than are metropolitan models
bull Interest is high among states in progress in deploying models
Source A Horowitz White Paper on Statewide Modeling presented at TRB Conference on Statewide Planning 2008
Alternative Statewide Model Structures
bull Trip table estimation (2 step) Estimate initial ldquoseed matrixrdquo
Roadside intercept surveys American Travel Survey National HH Travel SurveyAdd-ons Census Journey-to-Work flows
Matrix estimation program Inputs ndash seed matrix highway
network traffic counts Adjusts origin-destination flows
between zones to optimize matching network traffic counts
Fratar factors applied to synthetic trip table in estimating future growth
bull Trip table estimation (2 step) Pros
Limited need to prepare aggregate socioeconomic input data
Simplicity of model generally means short execution times
Forecasting needs are limited to zone or district growth factors
Cons Less explanatory power than most
MPOregional models Fixed distribution pattern Difficulties in assessing benefits of
adding new roadway capacity
Alternative Statewide Model Structures (Contrsquod)
bull Highway model structure (3 step) Trip generation Trip distribution Highway assignment
bull Mode choice excluded Either generate vehicle trips or Apply auto occupancy factors to person trip
tables
bull Pros Benefits of trip redistribution across range of
alternatives Greater sensitivity to roadway capacity and
congestion
bull Cons Likely reliance on MPOsothers to provide
socioeconomic input data Canrsquot fully test transportation strategies that
would alter mode split
Source Florida Dept of TransportationCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)
bull Freight considerations (3 step) Typical freight model steps
Trip generation ndash tons by commodity group
Trip distribution Mode choice ndash freight is
apportioned to transport modes
bull Joint highway assignment (trucks and passenger trips)
Source Florida Dept of TransportationCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)bull Full multi-modal structure (4 step)
Trip generation Trip distribution Mode choice Highway assignment
bull More common with mega-regions Focused on fixed guideway systems Intercity rail is important component
bull Pros Potential to model new intercity rail
including high speed rail Test greater range of potential
transportation solutions
bull Cons Requires coding of transit networks and
mode choice estimation Modeling of air travel might be needed
especially if intercity rail is a completely new mode of travel
Source California HSR AuthorityCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)
Beyond 4 Stepbull Potential components
Implement destination choice Tour-basedtrip chains Activity-based travel model Activity micro-simulation Traffic micro-simulation Integration with land use models Integration with economic models
bull Pros Ability to test a more complete set of
transportation strategies and solutions
bull Cons Extensive data requirements Cost to maintain Complexity
Source Oregon Dept of TransportationPBUniversity of Calgary
Deciding on Best Statewide Model Approach
bull What are the motivations for building a statewide travel demand modelbull What are the emerging transportation and policy issues to addressbull How significant is the impact of through travel on your statebull Do you have many multi-state MPOsbull How reliable (and current) are your survey databull Does your state have major international ports or border crossings
StudyArea
UrbanModel
AreaStudyArea
UrbanModelArea 1
Study AreaWithin OneModel Area
UrbanModelArea 2
Study AreasWithin Two Model Areas
StudyArea
UrbanModelArea
StudyArea
Statewide Model
Study AreaOutside Urban
Model Areas
Statewide ModelStatewide Model
Source Florida Department of TransportationCambridge Systematics Inc
Figure 317 Subarea Study Area Examples
Study
Area
Urban
Model
Area
Study
Area
Urban
Model
Area 1
Study Area
Within One
Model Area
Urban
Model
Area 2
Study Areas
Within Two
Model Areas
Study
Area
Urban
Model
Area
Study
Area
Statewide Model
Study Area
Outside Urban
Model Areas
Statewide Model
Statewide Model
Statewide Model Data Development Challenges
bull Primary data deficiencies Long distance travelers Rural resident trip-making
bull Challenges in compiling required data Age of data Confidentiality Coordination with data providers Proper level of geography Sufficient sample size
bull Four categories of data for discussion Networks and zone system Demographic and employment data Freight and other economic data NHTS and other behavioral data
Source Florida Department of TransportationCambridge Systematics Inc
Data Challengeshellipwhat is ldquolong distancerdquo
bull gt50 miles (American Traveler Survey)
bull gt100 miles (National Household Travel Survey)
bull IE trips (several states)
bull Trips made by residents for business personal business
bull Recreationaltourist trips
Topic ndash Similarities amp Differences with Urban Models
bull Zone systems
bull Networks
bull Travel behavior assumptions
bull Trip purposesmarket segments
bull Modal issues
bull Truck forecasting
Source Delaware Dept of TransportationWRampA Inc
Statewide vs Urban Model Zone Systems
bull Usually larger zones in statewide models than MPOs
bull Block groups are good starting point for statewide model TAZs
bull Block groups might need splitting along major transportation corridors
bull In small states MPO TAZs could equal SWM zones
bull In most states MPO TAZ equivalency tables will likely be neededdesirable
bull FAF disaggregation approach used for statewide model TAZs
Source FHWA Office of Freight Management and Operations
Source Florida Dept of TransportationCambridge Systematics Inc
FAF zones
Block groups = TAZs
Too small for statewide zones
Too big for statewide zones
Statewide vs Urban Model Networksbull Statewide model networks are sparse in
comparison with MPO models
bull At a minimum statewide networks should include Limited access highways Arterials Major collectors
bull For proper circulation patterns lower class facilities might be added
bull Outside state a very sparse network is needed for major truck corridors
bull Consistent level of detail for networks and zones
Source Tennessee Dept of TransportationCambridge Systematics Inc
MPO network
statewide network
Statewide vs Urban Model Travel Behaviorbull Understanding urban travel behavior
MPO models are usually developed with local household travel survey data
MPOs without recent HH survey data can borrow parameters from other MPO models or guidance documents such as NCHRP 716 Transferable Urban Travel Parameters
bull Rural and long-distance travel behavior A limited number of states have conducted
statewide household travel surveys American Travel Survey was designed to target
long-distance trips but was discontinued long ago
National Household Travel Survey Add-ons can be structured for a sufficient sample of rural trip-makers
NCHRP 735 Rural and Long-Distance Travel Parameters recently published for use in statewide models
Map of 2009 Add-on Surveys Source FHWA
Statewide vs Urban Model Trip Purposes
bull Typical MPO trip purposes are usually collapsed into HBW HBOHBNW and NHB for statewide models
bull Long-distance purposes Business Tourist Personal business
bull Trucks Commodity group is often
surrogate for trip purpose
bull External trips External generally means trip from
outside state Might exclude ldquohalordquo area
surrounding state boundary
Trip PurposeFL
2000 IN LA MA MS RI
Home-based Work 1914 1995 1900 1841 1606 2000
Home-based Nonwork 5547 5208 4770 5206 4917 5700
Home-based Other 2792 5208 4770 1938 4917 5700
Home-based Shop 1576 ndash ndash 1173 ndash ndash
Home-based SocialRecreation
1179 ndash ndash 1395 ndash ndash
Home-base RecreationShop ndash ndash ndash ndash ndash ndash
Home-based School ndash ndash ndash 700 ndash ndash
Nonhome-based 2539 2686 3100 2468 3289 2300
Nonhome-based Work ndash ndash ndash 1027 ndash ndash
Nonhome-based Nonwork ndash ndash ndash 1441 ndash ndash
Other ndash ndash ndash ndash ndash
Truck ndash ndash 170 ndash ndash ndash
Long-Distance 000 111 060 417 022 000
Long-Distance Business ndash ndash 020 284 011 ndash
Long-Distance Personal Business
ndash ndash ndash ndash ndash ndash
Long-Distance Tourist ndash ndash 010 027 000 ndash
Long-Distance Work ndash ndash ndash 105 ndash ndash
Long-Distance NonworkOther
ndash ndash 030 ndash 011 ndash
Statewide vs Urban Model Trip Purposes (Contrsquod)
Number of person trips
Mode Pleasure Business Personal business Other Total Percent
Personal vehicle 12185 5708 3120 789 21802 908Commercial air 1012 746 151 24 1933 8Train 74 150 14 0 238 1Other 4 2 5 12 23 01Ship 19 0 0 0 19 01
Total 13294 6606 3290 825 24015 100Percent 554 275 137 34 100
Percent by trip purpose
Mode Pleasure Business Personal business Other
Personal vehicle 917 864 948 956Commercial air 76 113 46 29Train 06 23 04 0Other 0 0 02 15Ship 01 0 0 0
Trip purpose
Trip purpose
Statewide vs Urban Model Modal Issuesbull Most statewide models are limited to highway trips
bull Modeling of local bus routes is likely overkill
bull Mode choice is usually just a conversion from person trips to vehicle trips via auto occupancy rates
bull Transit components of statewide models are more common in regions with history of fixed guideway transit
bull Transit model is needed to test options such as high speedintercity rail
ATS 95 NHTS 01 MDOT 03-04 ODOT 02-03Personal vehicle 81 89 89 88Commercial air 16 8 10 8Bus 2 2 04 19Train 05 01 0 005Other 05 09 06 205
Survey and yearPrimary mode of transport
Statewide vs Urban Model Modal Issues (Contrsquod)
3727 24 22
32
29 36
10
21
28 22
10
22
12
1529
11
44 43
0
10
20
30
40
50
60
70
80
90
100
Personalvehicle
Bus Train Marine Commercialair
Perc
ent o
f per
son
trips
lt300 300-499 500-599 1000-1999 gt2000
68
85 85 82 85
30
12 11 15 13
0
10
20
30
40
50
60
70
80
90
100
Busi-ness
Plea-sure
pBusi-ness
Vaca-tion
Week-end
Perc
ent o
f per
son
trips
Personal vehicle Commercial air All other
Mode by Trip Purpose Mode by Distance (miles)
Statewide vs Urban Model Modal Issues (Contrsquod)
bull Most statewide assignment models separate trucks from passenger cars Preloading process commonly used for trucks Takes into account more limited route options for truck travel
Source Ohio Dept of Transportation
Statewide vs Urban Model Truck Forecasting
bull Urban modelsndash Freight movements are partially
ldquoexternalrdquo to model (truncated at external zone)
ndash More general categories of truckcommercial trips used
ndash Strong reliance on Quick Response Freight Manual
bull Statewide modelsndash Rooted in commodity flows ndash Often use econometric
approaches to modelingndash Generally model entire length of
freight travel
Freight Model Type by Peer Exchange Participants and
NonParticipants
0123456789
None
Traditio
nal
Commod
ity
Econo
metric
Unkno
wn
NotPeer Exchange
Source TRB Electronic Circular E-C075 Statewide Travel Demand Modeling ndash A Peer Exchange
Topic ndash Data Needs and Considerations
bull Impact of model complexity on data needs
bull Overview of data sources
bull Long-Distance travel gt different focus than MPOregional models
Sources NHTS Weight calculation Pitfalls Key statistics European surveys Lessons learned
Impact of Model Complexity on Data Needsbull Trip table estimation (2 step) ndash highway network
zone system trip table
bull Highway model structure (3 step) ndash 2 step plusbehavioral data for trip rates friction factors auto occupancy rates
bull Freight considerations (3 step) ndash 3 step pluscommodity flows economic data tonnages by mode truck counts
bull Full multi-modal structure (4 step) ndash 3 step plustransit networks and paths logit mode choice model
bull Beyond 4 Step ndash synthetic household generation operational characteristics etc
Source Georgia DOTCambridge Systematics Inc
Statewide Model Data Sources A-Zbull American Travel Survey (ATS)
bull Bureau of Economic Analysis (BEA)
bull Census Journey-to-Work (JTW) data
bull Census Public Use Microsample (PUMS)
bull Census Transportation Planning Package (CTPP)
bull Claritas
bull Commodity Flow Survey (CFS)
Source US Census Bureau
Statewide Model Data Sources (Contrsquod)bull Dun amp Bradstreet (DampB) employment data
bull Employment Security data (ES-202)
bull Freight Analysis Framework (FAF)
bull Highway Performance Monitoring System (HPMS)
bull InfoUSA employment data
bull Metropolitan Planning Organizations (MPO)
bull NCHRP 365 Manual
bull National Highway Planning Network (NHPN)
Source FAF2 Provisional Commodity Origin-Destination Estimates Revised Methodology Report BattelleFHWA
Statewide Model Data Sources (Contrsquod)
bull National Household Travel Survey (NHTS)
bull Origin-destination surveys
bull Regional Economic Models Inc (REMI)
bull State Departments of Transportation
bull TRANSEARCH data
bull Vehicle Use and Identification Survey (VIUS)
bull Woods amp Poole employment forecasts
Source Bureau of Transportation Statistics NHTS
Data Sources on Long-Distance Travel
bull Census Journey-to-Work databull Long distance component
American Travel Survey (77 95) National Household Travel Survey (01) Ohio statewide surveys (02-04) Michigan statewide survey (04-05) California (10-12)
bull Surveys focused on long distance Oregon (96-98 07-08)
bull Others BTS Omnibus surveys Tourism surveys European practice NCHRP Report 735
NHTS and Long-Distance Travelbull Evolution of US national surveys
NPTS (69 77 83 90 95) ATS (77 95)
bull 2001 NHTS ndash Total of 60282 long distance person trips Data collected March 01 to May 02 22204 (37 percent) before 911 38078 (63 percent) afterwards
bull Long distance trip is longer than 50 miles one-way (vs 100 miles in 1995 ATS)
bull 28-day retrospective survey (including travel day)
bull Not your typical travelers Low frequency Distinct segment of the population
Weight Calculation for Long-Distance Travelers
bull Use to ensure valid population-level estimates Accounts for non-response under-coverage multiple phones etc
bull Multi-stage process Match national margins Adjust for data splitting (force sumexpfac = 277208169) Raking (match independent controls) race ethnicity race by
month ethnicity by month sex by age census region MSA status month by day of week
Trim outlier expansion factors max expfac = 4 times mean Replication (correct for sampling error)
bull At least two passes through the data
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Freight Model Types | |||||||||||||||||||
State | Type | Peer Exc | |||||||||||||||||
OR | 4 | 1 | Count of Type | Count of Type | Peer Exc | ||||||||||||||
WI | 3 | 1 | Type | Total | Type | 0 | 1 | Grand Total | |||||||||||
MO | 1 | 1 | 1 | 2 | 1 | 2 | 2 | 4 | |||||||||||
LA | 3 | 1 | 2 | 2 | 2 | 2 | 2 | 4 | |||||||||||
IN | 3 | 1 | 3 | 6 | 3 | 2 | 6 | 8 | |||||||||||
OH | 4 | 1 | 4 | 2 | 4 | 2 | 2 | ||||||||||||
KY | 3 | 1 | Grand Total | 12 | 3 | 3 | |||||||||||||
FL | 3 | 1 | Grand Total | 9 | 12 | 21 | |||||||||||||
VA | 3 | 1 | None | 2 | Not | Peer Exchange | |||||||||||||
NJ | 2 | 1 | Traditional | 2 | None | 2 | 2 | ||||||||||||
MA | 2 | 1 | Commodity | 6 | Traditional | 2 | 2 | ||||||||||||
NH | 1 | 1 | Econometric | 2 | Commodity | 2 | 6 | ||||||||||||
CA | 0 | Econometric | 2 | ||||||||||||||||
TX | 3 | 0 | Unknown | 3 | |||||||||||||||
MI | 3 | 0 | |||||||||||||||||
GA | 0 | ||||||||||||||||||
SC | 1 | 0 | |||||||||||||||||
DE | 2 | 0 | |||||||||||||||||
CT | 0 | ||||||||||||||||||
VT | 2 | 0 | |||||||||||||||||
ME | 1 | 0 | |||||||||||||||||
MN | 0 |
None | None | ||
Traditional | Traditional | ||
Commodity | Commodity | ||
Econometric | Econometric | ||
Unknown | Unknown |
Statewide Model Topicsbull Why do we need Statewide Models
bull Statewide model theory and typical steps
bull Similarities and differences with urban models
bull Data needs and considerations
bull Validation and reasonableness checking transferability
bull Interfacing statewide and urban models
bull Shortcomings and areas for improvement
Topic ndash Why do we need Statewide Models
bull Rationale
bull Unique features
bull Current implementation
bull Contextunique issues
bull Typical uses
bull Examples of statewide models
bull Potential for micro-simulation
bull Documenting current practice
Source TRB presentation on Wisconsin Statewide Model WISDOT Cambridge Systematics Inc
Rationale ndash Why Statewide Modelsbull Consistency in statewide planning
ndash policy issues
bull Often motivated by specific project requirements
bull Issues or projects affecting entire state or beyond single MPO
bull Travel analyses for rural areas outside MPOs
bull Forecasting MPO external trips
bull Examining freight issues
Source University of Connecticut Connecticut Dept of Transportation
Unique Features of Statewide Models
bull Geographic scale and resolution In-state traffic analysis zones ldquoHalordquo or ldquoborderrdquo set of detailed
zones in adjacent state(s) where significant activity lies across the border
ldquoExternalrdquo zones for passenger travel at state line or halo border
Other states or groups of states for freight modeling (see next example map)
Source Florida Department of Transportation 2005 Statewide Model
Unique Features of Statewide Models (Contrsquod)
bull More aggregate than MPO models (networkszones)
bull Simplified approaches used at times (not necessarily recommended) Three step structure Synthetic or borrowed model structures and
parameters
bull Explicit treatment of visitors and long-distance travel
bull Freight a larger emphasis
bull Potential for inconsistencies with urban models
bull Some linkages with land use and economic models
Source Florida Department of Transportation 2005 Statewide Model
Current Status of Statewide Models
Operational (28)
Dormant (3)
No model (11)
Developing (8)
Source A Horowitz Statewide Travel Forecasting Models NCHRP Synthesis 358 (2006) with recent updates
Statewide ModelsContext for statewide modeling
MacroscopicMegascopic Mesoscopic Microscopic
Urban travel models (tour-
based or sequential)
Regional model of economic
trade and land use trends
Dynamic traffic assignment to
planning network
Detailed traffic analysis of key facilities and
corridors
Transport costs disutilities and
travel times
Multimodal person and freight tours
Time-sliced demand flows by mode
Congestion indices link and
node delay
Economic triggers for freight flows
Accessibilities and transport costs
Source Federal Highway Administration Pilot Course on Statewide Travel Forecasting 2007
Issues unique to statewide models
bull Explicit linkages to economic models
bull Scale issues Modeling states at an urban scale Expensive to build and maintain
bull Need for interfaces with urban models Consistency Data exchange
bull Lack of standard practicebull Unique components
Freight (commodities rarr vehicles) Long distance travel Economic impacts
Data and knowledge gaps
Source Michigan Department of Transportation
Typical uses of Statewide ModelsLess common usesbull Air quality conformity analysesbull Freight and intermodal
planningbull Traffic impact studiesbull Economic development studiesbull Long-term investment studiesbull Emergency evacuation studiesbull Multi-state corridors (possibly
with multi-state models)
Most commonly citedbull Corridor planning (including
community bypasses)bull Forecasting freighttruck
flowsbull External trip forecasts for
MPO amp regional models bull Expansion of MPO amp regional
models
Source Pennsylvania Department of Transportation TRB presentation on 2006-2030 Statewide Model
Louisianarsquos and Virginiarsquos statewide modelsMarket segments 2000 source amp estimation procedure Forecast source amp estimation
procedure
Macrolevel
1 Interstate and interparishtruck trip table
2000 macro zone level Transearch data some matrix estimation
2030 macro zone forecast of 2000 Transearch data interpolation and extrapolation to forecast other years
2 Long distance interstate business tourist and other auto trip tables
1995 American Travel Survey data disaggregated from state and MSA to county level and Fratar balanced to 2000 using household and employment growth
Fratar balance year 2000 county-level trip tables to forecast year using household and employment growth
3 Long distance intrastate business tourist and other auto trip tables
4 Short distance Interstate Louisiana HBW trip table
1990 Census county level JTW data Fratar balanced to 2000 using household and employment growth
Microlevel
5 Short distance intraparish truck trip table
Synthesized from 2000 intraparish Transearch data
Fratar balance 2000 intraparish trip table to forecast year using household and employment growth6 Short distance intrastate
Louisiana HBW trip table1990 tract-level JTW data Fratar balanced to 2000 using household and employment growth
7 Short distance intrastate HBO and NHB trip tables
Urban model style household-based trip generation and gravity model distribution
Urban style household-based trip generation and distribution models log-linear regression used to estimate external trip ends
Source Virginia Department of Transportation
MicrosimulationActivity-based Approach
bull Models dynamic evolution and actions of discrete agents across time and space
bull Observe emergent behavior Interaction Cooperation Coordination
bull Primary emphasis on understanding the system
bull Maybe forecast it
Between autonomous perceptive and deliberative agents
Source Ohio Department of Transportation
Key Differences ndash Traditional vs MicrosimulationldquoTraditional modelsrdquobull Closed form equilibriumbull Well understoodbull Typically deterministicbull Presumes central tendencies
in underlying databull Well-behavedbull ldquoManageablerdquo variationrdquobull Computationally efficient
Microsimulationbull Disequilibriumbull Large systems NWUbull Stochasticbull Diverse behaviorbull Handles wide variancesbull Accommodates weak to strong
understandingbull Computationally heavy
Zonal Land Use Data
Worker Traveler Generation
Vehicle Assignment
Starting Time Assignment
Post-Processor
Next Stop Purpose Choice
Model
Next Stop Location
Choice Model
Dynamic Activity Pattern Generation
Commercial Vehicle Trip List
Zonal Land Use Data
Worker Traveler Generation
Vehicle Assignment
Starting Time Assignment
Post-Processor
Next Stop Purpose Choice
Model
Next Stop Location
Choice Model
Dynamic Activity Pattern Generation
Commercial Vehicle Trip List
Microsimulation Examplesbull Ohiobull Oregon
Source Ohio Department of Transportation
Statewide Model Current Practicebull Meetings of the TRB Subcommittee on Statewide Travel Forecasting Models
bull NCHRP Report 735 Long-Distance and Rural Transferable Parameters for Statewide Travel Forecasting Models 2012
bull NCHRP Validation and Sensitivity Considerations for Statewide Models September 2010
bull White Paper on Statewide Modeling presented at TRB Statewide Planning Conference September 2008
bull National Travel Demand Forecasting Model Phase I Final Scope 2008
bull Forecasting Statewide Freight Toolkit NCHRP Report 606 2008
bull TRB Annual Meeting Sessions in 2004 2006 and 2008 web page
bull Statewide Travel Forecasting Models NCHRP Synthesis 358 2006
bull Statewide Travel Demand Modeling Peer Exchange Transportation Research Circular E-C075 September 2004
bull Guidebook on Statewide Travel Forecasting and Irvine Conference Transportation Research Circular E-C011 1999
Topic ndash Statewide Model Theory
bull Common themes in statewide model development
bull Alternative structures Trip table estimation (2 step) Highway model structure (3 step) Fully multimodal (4 step) Beyond 4 step
bull Deciding on best approach
bull Data development challengesSource Massachusetts Highway DeptCambridge Systematics Inc
Common Themes in Review of Statewide Models
bull Most statewide models use similar methodologies to metropolitan models but there are many variations
bull The largest problems relate to issues of scale
bull Models from different states vary greatly in complexity cost and development time
bull Models are most successful when addressing statewide priorities
bull There are major deficiencies in available data about long-distance and rural passenger travelhellip NCHRP Report 735 has helped
bull Statewide models are more compatible with secondary freight data sources such as the Commodity Flow Survey than are metropolitan models
bull Interest is high among states in progress in deploying models
Source A Horowitz White Paper on Statewide Modeling presented at TRB Conference on Statewide Planning 2008
Alternative Statewide Model Structures
bull Trip table estimation (2 step) Estimate initial ldquoseed matrixrdquo
Roadside intercept surveys American Travel Survey National HH Travel SurveyAdd-ons Census Journey-to-Work flows
Matrix estimation program Inputs ndash seed matrix highway
network traffic counts Adjusts origin-destination flows
between zones to optimize matching network traffic counts
Fratar factors applied to synthetic trip table in estimating future growth
bull Trip table estimation (2 step) Pros
Limited need to prepare aggregate socioeconomic input data
Simplicity of model generally means short execution times
Forecasting needs are limited to zone or district growth factors
Cons Less explanatory power than most
MPOregional models Fixed distribution pattern Difficulties in assessing benefits of
adding new roadway capacity
Alternative Statewide Model Structures (Contrsquod)
bull Highway model structure (3 step) Trip generation Trip distribution Highway assignment
bull Mode choice excluded Either generate vehicle trips or Apply auto occupancy factors to person trip
tables
bull Pros Benefits of trip redistribution across range of
alternatives Greater sensitivity to roadway capacity and
congestion
bull Cons Likely reliance on MPOsothers to provide
socioeconomic input data Canrsquot fully test transportation strategies that
would alter mode split
Source Florida Dept of TransportationCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)
bull Freight considerations (3 step) Typical freight model steps
Trip generation ndash tons by commodity group
Trip distribution Mode choice ndash freight is
apportioned to transport modes
bull Joint highway assignment (trucks and passenger trips)
Source Florida Dept of TransportationCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)bull Full multi-modal structure (4 step)
Trip generation Trip distribution Mode choice Highway assignment
bull More common with mega-regions Focused on fixed guideway systems Intercity rail is important component
bull Pros Potential to model new intercity rail
including high speed rail Test greater range of potential
transportation solutions
bull Cons Requires coding of transit networks and
mode choice estimation Modeling of air travel might be needed
especially if intercity rail is a completely new mode of travel
Source California HSR AuthorityCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)
Beyond 4 Stepbull Potential components
Implement destination choice Tour-basedtrip chains Activity-based travel model Activity micro-simulation Traffic micro-simulation Integration with land use models Integration with economic models
bull Pros Ability to test a more complete set of
transportation strategies and solutions
bull Cons Extensive data requirements Cost to maintain Complexity
Source Oregon Dept of TransportationPBUniversity of Calgary
Deciding on Best Statewide Model Approach
bull What are the motivations for building a statewide travel demand modelbull What are the emerging transportation and policy issues to addressbull How significant is the impact of through travel on your statebull Do you have many multi-state MPOsbull How reliable (and current) are your survey databull Does your state have major international ports or border crossings
StudyArea
UrbanModel
AreaStudyArea
UrbanModelArea 1
Study AreaWithin OneModel Area
UrbanModelArea 2
Study AreasWithin Two Model Areas
StudyArea
UrbanModelArea
StudyArea
Statewide Model
Study AreaOutside Urban
Model Areas
Statewide ModelStatewide Model
Source Florida Department of TransportationCambridge Systematics Inc
Figure 317 Subarea Study Area Examples
Study
Area
Urban
Model
Area
Study
Area
Urban
Model
Area 1
Study Area
Within One
Model Area
Urban
Model
Area 2
Study Areas
Within Two
Model Areas
Study
Area
Urban
Model
Area
Study
Area
Statewide Model
Study Area
Outside Urban
Model Areas
Statewide Model
Statewide Model
Statewide Model Data Development Challenges
bull Primary data deficiencies Long distance travelers Rural resident trip-making
bull Challenges in compiling required data Age of data Confidentiality Coordination with data providers Proper level of geography Sufficient sample size
bull Four categories of data for discussion Networks and zone system Demographic and employment data Freight and other economic data NHTS and other behavioral data
Source Florida Department of TransportationCambridge Systematics Inc
Data Challengeshellipwhat is ldquolong distancerdquo
bull gt50 miles (American Traveler Survey)
bull gt100 miles (National Household Travel Survey)
bull IE trips (several states)
bull Trips made by residents for business personal business
bull Recreationaltourist trips
Topic ndash Similarities amp Differences with Urban Models
bull Zone systems
bull Networks
bull Travel behavior assumptions
bull Trip purposesmarket segments
bull Modal issues
bull Truck forecasting
Source Delaware Dept of TransportationWRampA Inc
Statewide vs Urban Model Zone Systems
bull Usually larger zones in statewide models than MPOs
bull Block groups are good starting point for statewide model TAZs
bull Block groups might need splitting along major transportation corridors
bull In small states MPO TAZs could equal SWM zones
bull In most states MPO TAZ equivalency tables will likely be neededdesirable
bull FAF disaggregation approach used for statewide model TAZs
Source FHWA Office of Freight Management and Operations
Source Florida Dept of TransportationCambridge Systematics Inc
FAF zones
Block groups = TAZs
Too small for statewide zones
Too big for statewide zones
Statewide vs Urban Model Networksbull Statewide model networks are sparse in
comparison with MPO models
bull At a minimum statewide networks should include Limited access highways Arterials Major collectors
bull For proper circulation patterns lower class facilities might be added
bull Outside state a very sparse network is needed for major truck corridors
bull Consistent level of detail for networks and zones
Source Tennessee Dept of TransportationCambridge Systematics Inc
MPO network
statewide network
Statewide vs Urban Model Travel Behaviorbull Understanding urban travel behavior
MPO models are usually developed with local household travel survey data
MPOs without recent HH survey data can borrow parameters from other MPO models or guidance documents such as NCHRP 716 Transferable Urban Travel Parameters
bull Rural and long-distance travel behavior A limited number of states have conducted
statewide household travel surveys American Travel Survey was designed to target
long-distance trips but was discontinued long ago
National Household Travel Survey Add-ons can be structured for a sufficient sample of rural trip-makers
NCHRP 735 Rural and Long-Distance Travel Parameters recently published for use in statewide models
Map of 2009 Add-on Surveys Source FHWA
Statewide vs Urban Model Trip Purposes
bull Typical MPO trip purposes are usually collapsed into HBW HBOHBNW and NHB for statewide models
bull Long-distance purposes Business Tourist Personal business
bull Trucks Commodity group is often
surrogate for trip purpose
bull External trips External generally means trip from
outside state Might exclude ldquohalordquo area
surrounding state boundary
Trip PurposeFL
2000 IN LA MA MS RI
Home-based Work 1914 1995 1900 1841 1606 2000
Home-based Nonwork 5547 5208 4770 5206 4917 5700
Home-based Other 2792 5208 4770 1938 4917 5700
Home-based Shop 1576 ndash ndash 1173 ndash ndash
Home-based SocialRecreation
1179 ndash ndash 1395 ndash ndash
Home-base RecreationShop ndash ndash ndash ndash ndash ndash
Home-based School ndash ndash ndash 700 ndash ndash
Nonhome-based 2539 2686 3100 2468 3289 2300
Nonhome-based Work ndash ndash ndash 1027 ndash ndash
Nonhome-based Nonwork ndash ndash ndash 1441 ndash ndash
Other ndash ndash ndash ndash ndash
Truck ndash ndash 170 ndash ndash ndash
Long-Distance 000 111 060 417 022 000
Long-Distance Business ndash ndash 020 284 011 ndash
Long-Distance Personal Business
ndash ndash ndash ndash ndash ndash
Long-Distance Tourist ndash ndash 010 027 000 ndash
Long-Distance Work ndash ndash ndash 105 ndash ndash
Long-Distance NonworkOther
ndash ndash 030 ndash 011 ndash
Statewide vs Urban Model Trip Purposes (Contrsquod)
Number of person trips
Mode Pleasure Business Personal business Other Total Percent
Personal vehicle 12185 5708 3120 789 21802 908Commercial air 1012 746 151 24 1933 8Train 74 150 14 0 238 1Other 4 2 5 12 23 01Ship 19 0 0 0 19 01
Total 13294 6606 3290 825 24015 100Percent 554 275 137 34 100
Percent by trip purpose
Mode Pleasure Business Personal business Other
Personal vehicle 917 864 948 956Commercial air 76 113 46 29Train 06 23 04 0Other 0 0 02 15Ship 01 0 0 0
Trip purpose
Trip purpose
Statewide vs Urban Model Modal Issuesbull Most statewide models are limited to highway trips
bull Modeling of local bus routes is likely overkill
bull Mode choice is usually just a conversion from person trips to vehicle trips via auto occupancy rates
bull Transit components of statewide models are more common in regions with history of fixed guideway transit
bull Transit model is needed to test options such as high speedintercity rail
ATS 95 NHTS 01 MDOT 03-04 ODOT 02-03Personal vehicle 81 89 89 88Commercial air 16 8 10 8Bus 2 2 04 19Train 05 01 0 005Other 05 09 06 205
Survey and yearPrimary mode of transport
Statewide vs Urban Model Modal Issues (Contrsquod)
3727 24 22
32
29 36
10
21
28 22
10
22
12
1529
11
44 43
0
10
20
30
40
50
60
70
80
90
100
Personalvehicle
Bus Train Marine Commercialair
Perc
ent o
f per
son
trips
lt300 300-499 500-599 1000-1999 gt2000
68
85 85 82 85
30
12 11 15 13
0
10
20
30
40
50
60
70
80
90
100
Busi-ness
Plea-sure
pBusi-ness
Vaca-tion
Week-end
Perc
ent o
f per
son
trips
Personal vehicle Commercial air All other
Mode by Trip Purpose Mode by Distance (miles)
Statewide vs Urban Model Modal Issues (Contrsquod)
bull Most statewide assignment models separate trucks from passenger cars Preloading process commonly used for trucks Takes into account more limited route options for truck travel
Source Ohio Dept of Transportation
Statewide vs Urban Model Truck Forecasting
bull Urban modelsndash Freight movements are partially
ldquoexternalrdquo to model (truncated at external zone)
ndash More general categories of truckcommercial trips used
ndash Strong reliance on Quick Response Freight Manual
bull Statewide modelsndash Rooted in commodity flows ndash Often use econometric
approaches to modelingndash Generally model entire length of
freight travel
Freight Model Type by Peer Exchange Participants and
NonParticipants
0123456789
None
Traditio
nal
Commod
ity
Econo
metric
Unkno
wn
NotPeer Exchange
Source TRB Electronic Circular E-C075 Statewide Travel Demand Modeling ndash A Peer Exchange
Topic ndash Data Needs and Considerations
bull Impact of model complexity on data needs
bull Overview of data sources
bull Long-Distance travel gt different focus than MPOregional models
Sources NHTS Weight calculation Pitfalls Key statistics European surveys Lessons learned
Impact of Model Complexity on Data Needsbull Trip table estimation (2 step) ndash highway network
zone system trip table
bull Highway model structure (3 step) ndash 2 step plusbehavioral data for trip rates friction factors auto occupancy rates
bull Freight considerations (3 step) ndash 3 step pluscommodity flows economic data tonnages by mode truck counts
bull Full multi-modal structure (4 step) ndash 3 step plustransit networks and paths logit mode choice model
bull Beyond 4 Step ndash synthetic household generation operational characteristics etc
Source Georgia DOTCambridge Systematics Inc
Statewide Model Data Sources A-Zbull American Travel Survey (ATS)
bull Bureau of Economic Analysis (BEA)
bull Census Journey-to-Work (JTW) data
bull Census Public Use Microsample (PUMS)
bull Census Transportation Planning Package (CTPP)
bull Claritas
bull Commodity Flow Survey (CFS)
Source US Census Bureau
Statewide Model Data Sources (Contrsquod)bull Dun amp Bradstreet (DampB) employment data
bull Employment Security data (ES-202)
bull Freight Analysis Framework (FAF)
bull Highway Performance Monitoring System (HPMS)
bull InfoUSA employment data
bull Metropolitan Planning Organizations (MPO)
bull NCHRP 365 Manual
bull National Highway Planning Network (NHPN)
Source FAF2 Provisional Commodity Origin-Destination Estimates Revised Methodology Report BattelleFHWA
Statewide Model Data Sources (Contrsquod)
bull National Household Travel Survey (NHTS)
bull Origin-destination surveys
bull Regional Economic Models Inc (REMI)
bull State Departments of Transportation
bull TRANSEARCH data
bull Vehicle Use and Identification Survey (VIUS)
bull Woods amp Poole employment forecasts
Source Bureau of Transportation Statistics NHTS
Data Sources on Long-Distance Travel
bull Census Journey-to-Work databull Long distance component
American Travel Survey (77 95) National Household Travel Survey (01) Ohio statewide surveys (02-04) Michigan statewide survey (04-05) California (10-12)
bull Surveys focused on long distance Oregon (96-98 07-08)
bull Others BTS Omnibus surveys Tourism surveys European practice NCHRP Report 735
NHTS and Long-Distance Travelbull Evolution of US national surveys
NPTS (69 77 83 90 95) ATS (77 95)
bull 2001 NHTS ndash Total of 60282 long distance person trips Data collected March 01 to May 02 22204 (37 percent) before 911 38078 (63 percent) afterwards
bull Long distance trip is longer than 50 miles one-way (vs 100 miles in 1995 ATS)
bull 28-day retrospective survey (including travel day)
bull Not your typical travelers Low frequency Distinct segment of the population
Weight Calculation for Long-Distance Travelers
bull Use to ensure valid population-level estimates Accounts for non-response under-coverage multiple phones etc
bull Multi-stage process Match national margins Adjust for data splitting (force sumexpfac = 277208169) Raking (match independent controls) race ethnicity race by
month ethnicity by month sex by age census region MSA status month by day of week
Trim outlier expansion factors max expfac = 4 times mean Replication (correct for sampling error)
bull At least two passes through the data
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Freight Model Types | |||||||||||||||||||
State | Type | Peer Exc | |||||||||||||||||
OR | 4 | 1 | Count of Type | Count of Type | Peer Exc | ||||||||||||||
WI | 3 | 1 | Type | Total | Type | 0 | 1 | Grand Total | |||||||||||
MO | 1 | 1 | 1 | 2 | 1 | 2 | 2 | 4 | |||||||||||
LA | 3 | 1 | 2 | 2 | 2 | 2 | 2 | 4 | |||||||||||
IN | 3 | 1 | 3 | 6 | 3 | 2 | 6 | 8 | |||||||||||
OH | 4 | 1 | 4 | 2 | 4 | 2 | 2 | ||||||||||||
KY | 3 | 1 | Grand Total | 12 | 3 | 3 | |||||||||||||
FL | 3 | 1 | Grand Total | 9 | 12 | 21 | |||||||||||||
VA | 3 | 1 | None | 2 | Not | Peer Exchange | |||||||||||||
NJ | 2 | 1 | Traditional | 2 | None | 2 | 2 | ||||||||||||
MA | 2 | 1 | Commodity | 6 | Traditional | 2 | 2 | ||||||||||||
NH | 1 | 1 | Econometric | 2 | Commodity | 2 | 6 | ||||||||||||
CA | 0 | Econometric | 2 | ||||||||||||||||
TX | 3 | 0 | Unknown | 3 | |||||||||||||||
MI | 3 | 0 | |||||||||||||||||
GA | 0 | ||||||||||||||||||
SC | 1 | 0 | |||||||||||||||||
DE | 2 | 0 | |||||||||||||||||
CT | 0 | ||||||||||||||||||
VT | 2 | 0 | |||||||||||||||||
ME | 1 | 0 | |||||||||||||||||
MN | 0 |
None | None | ||
Traditional | Traditional | ||
Commodity | Commodity | ||
Econometric | Econometric | ||
Unknown | Unknown |
Topic ndash Why do we need Statewide Models
bull Rationale
bull Unique features
bull Current implementation
bull Contextunique issues
bull Typical uses
bull Examples of statewide models
bull Potential for micro-simulation
bull Documenting current practice
Source TRB presentation on Wisconsin Statewide Model WISDOT Cambridge Systematics Inc
Rationale ndash Why Statewide Modelsbull Consistency in statewide planning
ndash policy issues
bull Often motivated by specific project requirements
bull Issues or projects affecting entire state or beyond single MPO
bull Travel analyses for rural areas outside MPOs
bull Forecasting MPO external trips
bull Examining freight issues
Source University of Connecticut Connecticut Dept of Transportation
Unique Features of Statewide Models
bull Geographic scale and resolution In-state traffic analysis zones ldquoHalordquo or ldquoborderrdquo set of detailed
zones in adjacent state(s) where significant activity lies across the border
ldquoExternalrdquo zones for passenger travel at state line or halo border
Other states or groups of states for freight modeling (see next example map)
Source Florida Department of Transportation 2005 Statewide Model
Unique Features of Statewide Models (Contrsquod)
bull More aggregate than MPO models (networkszones)
bull Simplified approaches used at times (not necessarily recommended) Three step structure Synthetic or borrowed model structures and
parameters
bull Explicit treatment of visitors and long-distance travel
bull Freight a larger emphasis
bull Potential for inconsistencies with urban models
bull Some linkages with land use and economic models
Source Florida Department of Transportation 2005 Statewide Model
Current Status of Statewide Models
Operational (28)
Dormant (3)
No model (11)
Developing (8)
Source A Horowitz Statewide Travel Forecasting Models NCHRP Synthesis 358 (2006) with recent updates
Statewide ModelsContext for statewide modeling
MacroscopicMegascopic Mesoscopic Microscopic
Urban travel models (tour-
based or sequential)
Regional model of economic
trade and land use trends
Dynamic traffic assignment to
planning network
Detailed traffic analysis of key facilities and
corridors
Transport costs disutilities and
travel times
Multimodal person and freight tours
Time-sliced demand flows by mode
Congestion indices link and
node delay
Economic triggers for freight flows
Accessibilities and transport costs
Source Federal Highway Administration Pilot Course on Statewide Travel Forecasting 2007
Issues unique to statewide models
bull Explicit linkages to economic models
bull Scale issues Modeling states at an urban scale Expensive to build and maintain
bull Need for interfaces with urban models Consistency Data exchange
bull Lack of standard practicebull Unique components
Freight (commodities rarr vehicles) Long distance travel Economic impacts
Data and knowledge gaps
Source Michigan Department of Transportation
Typical uses of Statewide ModelsLess common usesbull Air quality conformity analysesbull Freight and intermodal
planningbull Traffic impact studiesbull Economic development studiesbull Long-term investment studiesbull Emergency evacuation studiesbull Multi-state corridors (possibly
with multi-state models)
Most commonly citedbull Corridor planning (including
community bypasses)bull Forecasting freighttruck
flowsbull External trip forecasts for
MPO amp regional models bull Expansion of MPO amp regional
models
Source Pennsylvania Department of Transportation TRB presentation on 2006-2030 Statewide Model
Louisianarsquos and Virginiarsquos statewide modelsMarket segments 2000 source amp estimation procedure Forecast source amp estimation
procedure
Macrolevel
1 Interstate and interparishtruck trip table
2000 macro zone level Transearch data some matrix estimation
2030 macro zone forecast of 2000 Transearch data interpolation and extrapolation to forecast other years
2 Long distance interstate business tourist and other auto trip tables
1995 American Travel Survey data disaggregated from state and MSA to county level and Fratar balanced to 2000 using household and employment growth
Fratar balance year 2000 county-level trip tables to forecast year using household and employment growth
3 Long distance intrastate business tourist and other auto trip tables
4 Short distance Interstate Louisiana HBW trip table
1990 Census county level JTW data Fratar balanced to 2000 using household and employment growth
Microlevel
5 Short distance intraparish truck trip table
Synthesized from 2000 intraparish Transearch data
Fratar balance 2000 intraparish trip table to forecast year using household and employment growth6 Short distance intrastate
Louisiana HBW trip table1990 tract-level JTW data Fratar balanced to 2000 using household and employment growth
7 Short distance intrastate HBO and NHB trip tables
Urban model style household-based trip generation and gravity model distribution
Urban style household-based trip generation and distribution models log-linear regression used to estimate external trip ends
Source Virginia Department of Transportation
MicrosimulationActivity-based Approach
bull Models dynamic evolution and actions of discrete agents across time and space
bull Observe emergent behavior Interaction Cooperation Coordination
bull Primary emphasis on understanding the system
bull Maybe forecast it
Between autonomous perceptive and deliberative agents
Source Ohio Department of Transportation
Key Differences ndash Traditional vs MicrosimulationldquoTraditional modelsrdquobull Closed form equilibriumbull Well understoodbull Typically deterministicbull Presumes central tendencies
in underlying databull Well-behavedbull ldquoManageablerdquo variationrdquobull Computationally efficient
Microsimulationbull Disequilibriumbull Large systems NWUbull Stochasticbull Diverse behaviorbull Handles wide variancesbull Accommodates weak to strong
understandingbull Computationally heavy
Zonal Land Use Data
Worker Traveler Generation
Vehicle Assignment
Starting Time Assignment
Post-Processor
Next Stop Purpose Choice
Model
Next Stop Location
Choice Model
Dynamic Activity Pattern Generation
Commercial Vehicle Trip List
Zonal Land Use Data
Worker Traveler Generation
Vehicle Assignment
Starting Time Assignment
Post-Processor
Next Stop Purpose Choice
Model
Next Stop Location
Choice Model
Dynamic Activity Pattern Generation
Commercial Vehicle Trip List
Microsimulation Examplesbull Ohiobull Oregon
Source Ohio Department of Transportation
Statewide Model Current Practicebull Meetings of the TRB Subcommittee on Statewide Travel Forecasting Models
bull NCHRP Report 735 Long-Distance and Rural Transferable Parameters for Statewide Travel Forecasting Models 2012
bull NCHRP Validation and Sensitivity Considerations for Statewide Models September 2010
bull White Paper on Statewide Modeling presented at TRB Statewide Planning Conference September 2008
bull National Travel Demand Forecasting Model Phase I Final Scope 2008
bull Forecasting Statewide Freight Toolkit NCHRP Report 606 2008
bull TRB Annual Meeting Sessions in 2004 2006 and 2008 web page
bull Statewide Travel Forecasting Models NCHRP Synthesis 358 2006
bull Statewide Travel Demand Modeling Peer Exchange Transportation Research Circular E-C075 September 2004
bull Guidebook on Statewide Travel Forecasting and Irvine Conference Transportation Research Circular E-C011 1999
Topic ndash Statewide Model Theory
bull Common themes in statewide model development
bull Alternative structures Trip table estimation (2 step) Highway model structure (3 step) Fully multimodal (4 step) Beyond 4 step
bull Deciding on best approach
bull Data development challengesSource Massachusetts Highway DeptCambridge Systematics Inc
Common Themes in Review of Statewide Models
bull Most statewide models use similar methodologies to metropolitan models but there are many variations
bull The largest problems relate to issues of scale
bull Models from different states vary greatly in complexity cost and development time
bull Models are most successful when addressing statewide priorities
bull There are major deficiencies in available data about long-distance and rural passenger travelhellip NCHRP Report 735 has helped
bull Statewide models are more compatible with secondary freight data sources such as the Commodity Flow Survey than are metropolitan models
bull Interest is high among states in progress in deploying models
Source A Horowitz White Paper on Statewide Modeling presented at TRB Conference on Statewide Planning 2008
Alternative Statewide Model Structures
bull Trip table estimation (2 step) Estimate initial ldquoseed matrixrdquo
Roadside intercept surveys American Travel Survey National HH Travel SurveyAdd-ons Census Journey-to-Work flows
Matrix estimation program Inputs ndash seed matrix highway
network traffic counts Adjusts origin-destination flows
between zones to optimize matching network traffic counts
Fratar factors applied to synthetic trip table in estimating future growth
bull Trip table estimation (2 step) Pros
Limited need to prepare aggregate socioeconomic input data
Simplicity of model generally means short execution times
Forecasting needs are limited to zone or district growth factors
Cons Less explanatory power than most
MPOregional models Fixed distribution pattern Difficulties in assessing benefits of
adding new roadway capacity
Alternative Statewide Model Structures (Contrsquod)
bull Highway model structure (3 step) Trip generation Trip distribution Highway assignment
bull Mode choice excluded Either generate vehicle trips or Apply auto occupancy factors to person trip
tables
bull Pros Benefits of trip redistribution across range of
alternatives Greater sensitivity to roadway capacity and
congestion
bull Cons Likely reliance on MPOsothers to provide
socioeconomic input data Canrsquot fully test transportation strategies that
would alter mode split
Source Florida Dept of TransportationCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)
bull Freight considerations (3 step) Typical freight model steps
Trip generation ndash tons by commodity group
Trip distribution Mode choice ndash freight is
apportioned to transport modes
bull Joint highway assignment (trucks and passenger trips)
Source Florida Dept of TransportationCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)bull Full multi-modal structure (4 step)
Trip generation Trip distribution Mode choice Highway assignment
bull More common with mega-regions Focused on fixed guideway systems Intercity rail is important component
bull Pros Potential to model new intercity rail
including high speed rail Test greater range of potential
transportation solutions
bull Cons Requires coding of transit networks and
mode choice estimation Modeling of air travel might be needed
especially if intercity rail is a completely new mode of travel
Source California HSR AuthorityCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)
Beyond 4 Stepbull Potential components
Implement destination choice Tour-basedtrip chains Activity-based travel model Activity micro-simulation Traffic micro-simulation Integration with land use models Integration with economic models
bull Pros Ability to test a more complete set of
transportation strategies and solutions
bull Cons Extensive data requirements Cost to maintain Complexity
Source Oregon Dept of TransportationPBUniversity of Calgary
Deciding on Best Statewide Model Approach
bull What are the motivations for building a statewide travel demand modelbull What are the emerging transportation and policy issues to addressbull How significant is the impact of through travel on your statebull Do you have many multi-state MPOsbull How reliable (and current) are your survey databull Does your state have major international ports or border crossings
StudyArea
UrbanModel
AreaStudyArea
UrbanModelArea 1
Study AreaWithin OneModel Area
UrbanModelArea 2
Study AreasWithin Two Model Areas
StudyArea
UrbanModelArea
StudyArea
Statewide Model
Study AreaOutside Urban
Model Areas
Statewide ModelStatewide Model
Source Florida Department of TransportationCambridge Systematics Inc
Figure 317 Subarea Study Area Examples
Study
Area
Urban
Model
Area
Study
Area
Urban
Model
Area 1
Study Area
Within One
Model Area
Urban
Model
Area 2
Study Areas
Within Two
Model Areas
Study
Area
Urban
Model
Area
Study
Area
Statewide Model
Study Area
Outside Urban
Model Areas
Statewide Model
Statewide Model
Statewide Model Data Development Challenges
bull Primary data deficiencies Long distance travelers Rural resident trip-making
bull Challenges in compiling required data Age of data Confidentiality Coordination with data providers Proper level of geography Sufficient sample size
bull Four categories of data for discussion Networks and zone system Demographic and employment data Freight and other economic data NHTS and other behavioral data
Source Florida Department of TransportationCambridge Systematics Inc
Data Challengeshellipwhat is ldquolong distancerdquo
bull gt50 miles (American Traveler Survey)
bull gt100 miles (National Household Travel Survey)
bull IE trips (several states)
bull Trips made by residents for business personal business
bull Recreationaltourist trips
Topic ndash Similarities amp Differences with Urban Models
bull Zone systems
bull Networks
bull Travel behavior assumptions
bull Trip purposesmarket segments
bull Modal issues
bull Truck forecasting
Source Delaware Dept of TransportationWRampA Inc
Statewide vs Urban Model Zone Systems
bull Usually larger zones in statewide models than MPOs
bull Block groups are good starting point for statewide model TAZs
bull Block groups might need splitting along major transportation corridors
bull In small states MPO TAZs could equal SWM zones
bull In most states MPO TAZ equivalency tables will likely be neededdesirable
bull FAF disaggregation approach used for statewide model TAZs
Source FHWA Office of Freight Management and Operations
Source Florida Dept of TransportationCambridge Systematics Inc
FAF zones
Block groups = TAZs
Too small for statewide zones
Too big for statewide zones
Statewide vs Urban Model Networksbull Statewide model networks are sparse in
comparison with MPO models
bull At a minimum statewide networks should include Limited access highways Arterials Major collectors
bull For proper circulation patterns lower class facilities might be added
bull Outside state a very sparse network is needed for major truck corridors
bull Consistent level of detail for networks and zones
Source Tennessee Dept of TransportationCambridge Systematics Inc
MPO network
statewide network
Statewide vs Urban Model Travel Behaviorbull Understanding urban travel behavior
MPO models are usually developed with local household travel survey data
MPOs without recent HH survey data can borrow parameters from other MPO models or guidance documents such as NCHRP 716 Transferable Urban Travel Parameters
bull Rural and long-distance travel behavior A limited number of states have conducted
statewide household travel surveys American Travel Survey was designed to target
long-distance trips but was discontinued long ago
National Household Travel Survey Add-ons can be structured for a sufficient sample of rural trip-makers
NCHRP 735 Rural and Long-Distance Travel Parameters recently published for use in statewide models
Map of 2009 Add-on Surveys Source FHWA
Statewide vs Urban Model Trip Purposes
bull Typical MPO trip purposes are usually collapsed into HBW HBOHBNW and NHB for statewide models
bull Long-distance purposes Business Tourist Personal business
bull Trucks Commodity group is often
surrogate for trip purpose
bull External trips External generally means trip from
outside state Might exclude ldquohalordquo area
surrounding state boundary
Trip PurposeFL
2000 IN LA MA MS RI
Home-based Work 1914 1995 1900 1841 1606 2000
Home-based Nonwork 5547 5208 4770 5206 4917 5700
Home-based Other 2792 5208 4770 1938 4917 5700
Home-based Shop 1576 ndash ndash 1173 ndash ndash
Home-based SocialRecreation
1179 ndash ndash 1395 ndash ndash
Home-base RecreationShop ndash ndash ndash ndash ndash ndash
Home-based School ndash ndash ndash 700 ndash ndash
Nonhome-based 2539 2686 3100 2468 3289 2300
Nonhome-based Work ndash ndash ndash 1027 ndash ndash
Nonhome-based Nonwork ndash ndash ndash 1441 ndash ndash
Other ndash ndash ndash ndash ndash
Truck ndash ndash 170 ndash ndash ndash
Long-Distance 000 111 060 417 022 000
Long-Distance Business ndash ndash 020 284 011 ndash
Long-Distance Personal Business
ndash ndash ndash ndash ndash ndash
Long-Distance Tourist ndash ndash 010 027 000 ndash
Long-Distance Work ndash ndash ndash 105 ndash ndash
Long-Distance NonworkOther
ndash ndash 030 ndash 011 ndash
Statewide vs Urban Model Trip Purposes (Contrsquod)
Number of person trips
Mode Pleasure Business Personal business Other Total Percent
Personal vehicle 12185 5708 3120 789 21802 908Commercial air 1012 746 151 24 1933 8Train 74 150 14 0 238 1Other 4 2 5 12 23 01Ship 19 0 0 0 19 01
Total 13294 6606 3290 825 24015 100Percent 554 275 137 34 100
Percent by trip purpose
Mode Pleasure Business Personal business Other
Personal vehicle 917 864 948 956Commercial air 76 113 46 29Train 06 23 04 0Other 0 0 02 15Ship 01 0 0 0
Trip purpose
Trip purpose
Statewide vs Urban Model Modal Issuesbull Most statewide models are limited to highway trips
bull Modeling of local bus routes is likely overkill
bull Mode choice is usually just a conversion from person trips to vehicle trips via auto occupancy rates
bull Transit components of statewide models are more common in regions with history of fixed guideway transit
bull Transit model is needed to test options such as high speedintercity rail
ATS 95 NHTS 01 MDOT 03-04 ODOT 02-03Personal vehicle 81 89 89 88Commercial air 16 8 10 8Bus 2 2 04 19Train 05 01 0 005Other 05 09 06 205
Survey and yearPrimary mode of transport
Statewide vs Urban Model Modal Issues (Contrsquod)
3727 24 22
32
29 36
10
21
28 22
10
22
12
1529
11
44 43
0
10
20
30
40
50
60
70
80
90
100
Personalvehicle
Bus Train Marine Commercialair
Perc
ent o
f per
son
trips
lt300 300-499 500-599 1000-1999 gt2000
68
85 85 82 85
30
12 11 15 13
0
10
20
30
40
50
60
70
80
90
100
Busi-ness
Plea-sure
pBusi-ness
Vaca-tion
Week-end
Perc
ent o
f per
son
trips
Personal vehicle Commercial air All other
Mode by Trip Purpose Mode by Distance (miles)
Statewide vs Urban Model Modal Issues (Contrsquod)
bull Most statewide assignment models separate trucks from passenger cars Preloading process commonly used for trucks Takes into account more limited route options for truck travel
Source Ohio Dept of Transportation
Statewide vs Urban Model Truck Forecasting
bull Urban modelsndash Freight movements are partially
ldquoexternalrdquo to model (truncated at external zone)
ndash More general categories of truckcommercial trips used
ndash Strong reliance on Quick Response Freight Manual
bull Statewide modelsndash Rooted in commodity flows ndash Often use econometric
approaches to modelingndash Generally model entire length of
freight travel
Freight Model Type by Peer Exchange Participants and
NonParticipants
0123456789
None
Traditio
nal
Commod
ity
Econo
metric
Unkno
wn
NotPeer Exchange
Source TRB Electronic Circular E-C075 Statewide Travel Demand Modeling ndash A Peer Exchange
Topic ndash Data Needs and Considerations
bull Impact of model complexity on data needs
bull Overview of data sources
bull Long-Distance travel gt different focus than MPOregional models
Sources NHTS Weight calculation Pitfalls Key statistics European surveys Lessons learned
Impact of Model Complexity on Data Needsbull Trip table estimation (2 step) ndash highway network
zone system trip table
bull Highway model structure (3 step) ndash 2 step plusbehavioral data for trip rates friction factors auto occupancy rates
bull Freight considerations (3 step) ndash 3 step pluscommodity flows economic data tonnages by mode truck counts
bull Full multi-modal structure (4 step) ndash 3 step plustransit networks and paths logit mode choice model
bull Beyond 4 Step ndash synthetic household generation operational characteristics etc
Source Georgia DOTCambridge Systematics Inc
Statewide Model Data Sources A-Zbull American Travel Survey (ATS)
bull Bureau of Economic Analysis (BEA)
bull Census Journey-to-Work (JTW) data
bull Census Public Use Microsample (PUMS)
bull Census Transportation Planning Package (CTPP)
bull Claritas
bull Commodity Flow Survey (CFS)
Source US Census Bureau
Statewide Model Data Sources (Contrsquod)bull Dun amp Bradstreet (DampB) employment data
bull Employment Security data (ES-202)
bull Freight Analysis Framework (FAF)
bull Highway Performance Monitoring System (HPMS)
bull InfoUSA employment data
bull Metropolitan Planning Organizations (MPO)
bull NCHRP 365 Manual
bull National Highway Planning Network (NHPN)
Source FAF2 Provisional Commodity Origin-Destination Estimates Revised Methodology Report BattelleFHWA
Statewide Model Data Sources (Contrsquod)
bull National Household Travel Survey (NHTS)
bull Origin-destination surveys
bull Regional Economic Models Inc (REMI)
bull State Departments of Transportation
bull TRANSEARCH data
bull Vehicle Use and Identification Survey (VIUS)
bull Woods amp Poole employment forecasts
Source Bureau of Transportation Statistics NHTS
Data Sources on Long-Distance Travel
bull Census Journey-to-Work databull Long distance component
American Travel Survey (77 95) National Household Travel Survey (01) Ohio statewide surveys (02-04) Michigan statewide survey (04-05) California (10-12)
bull Surveys focused on long distance Oregon (96-98 07-08)
bull Others BTS Omnibus surveys Tourism surveys European practice NCHRP Report 735
NHTS and Long-Distance Travelbull Evolution of US national surveys
NPTS (69 77 83 90 95) ATS (77 95)
bull 2001 NHTS ndash Total of 60282 long distance person trips Data collected March 01 to May 02 22204 (37 percent) before 911 38078 (63 percent) afterwards
bull Long distance trip is longer than 50 miles one-way (vs 100 miles in 1995 ATS)
bull 28-day retrospective survey (including travel day)
bull Not your typical travelers Low frequency Distinct segment of the population
Weight Calculation for Long-Distance Travelers
bull Use to ensure valid population-level estimates Accounts for non-response under-coverage multiple phones etc
bull Multi-stage process Match national margins Adjust for data splitting (force sumexpfac = 277208169) Raking (match independent controls) race ethnicity race by
month ethnicity by month sex by age census region MSA status month by day of week
Trim outlier expansion factors max expfac = 4 times mean Replication (correct for sampling error)
bull At least two passes through the data
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Freight Model Types | |||||||||||||||||||
State | Type | Peer Exc | |||||||||||||||||
OR | 4 | 1 | Count of Type | Count of Type | Peer Exc | ||||||||||||||
WI | 3 | 1 | Type | Total | Type | 0 | 1 | Grand Total | |||||||||||
MO | 1 | 1 | 1 | 2 | 1 | 2 | 2 | 4 | |||||||||||
LA | 3 | 1 | 2 | 2 | 2 | 2 | 2 | 4 | |||||||||||
IN | 3 | 1 | 3 | 6 | 3 | 2 | 6 | 8 | |||||||||||
OH | 4 | 1 | 4 | 2 | 4 | 2 | 2 | ||||||||||||
KY | 3 | 1 | Grand Total | 12 | 3 | 3 | |||||||||||||
FL | 3 | 1 | Grand Total | 9 | 12 | 21 | |||||||||||||
VA | 3 | 1 | None | 2 | Not | Peer Exchange | |||||||||||||
NJ | 2 | 1 | Traditional | 2 | None | 2 | 2 | ||||||||||||
MA | 2 | 1 | Commodity | 6 | Traditional | 2 | 2 | ||||||||||||
NH | 1 | 1 | Econometric | 2 | Commodity | 2 | 6 | ||||||||||||
CA | 0 | Econometric | 2 | ||||||||||||||||
TX | 3 | 0 | Unknown | 3 | |||||||||||||||
MI | 3 | 0 | |||||||||||||||||
GA | 0 | ||||||||||||||||||
SC | 1 | 0 | |||||||||||||||||
DE | 2 | 0 | |||||||||||||||||
CT | 0 | ||||||||||||||||||
VT | 2 | 0 | |||||||||||||||||
ME | 1 | 0 | |||||||||||||||||
MN | 0 |
None | None | ||
Traditional | Traditional | ||
Commodity | Commodity | ||
Econometric | Econometric | ||
Unknown | Unknown |
Rationale ndash Why Statewide Modelsbull Consistency in statewide planning
ndash policy issues
bull Often motivated by specific project requirements
bull Issues or projects affecting entire state or beyond single MPO
bull Travel analyses for rural areas outside MPOs
bull Forecasting MPO external trips
bull Examining freight issues
Source University of Connecticut Connecticut Dept of Transportation
Unique Features of Statewide Models
bull Geographic scale and resolution In-state traffic analysis zones ldquoHalordquo or ldquoborderrdquo set of detailed
zones in adjacent state(s) where significant activity lies across the border
ldquoExternalrdquo zones for passenger travel at state line or halo border
Other states or groups of states for freight modeling (see next example map)
Source Florida Department of Transportation 2005 Statewide Model
Unique Features of Statewide Models (Contrsquod)
bull More aggregate than MPO models (networkszones)
bull Simplified approaches used at times (not necessarily recommended) Three step structure Synthetic or borrowed model structures and
parameters
bull Explicit treatment of visitors and long-distance travel
bull Freight a larger emphasis
bull Potential for inconsistencies with urban models
bull Some linkages with land use and economic models
Source Florida Department of Transportation 2005 Statewide Model
Current Status of Statewide Models
Operational (28)
Dormant (3)
No model (11)
Developing (8)
Source A Horowitz Statewide Travel Forecasting Models NCHRP Synthesis 358 (2006) with recent updates
Statewide ModelsContext for statewide modeling
MacroscopicMegascopic Mesoscopic Microscopic
Urban travel models (tour-
based or sequential)
Regional model of economic
trade and land use trends
Dynamic traffic assignment to
planning network
Detailed traffic analysis of key facilities and
corridors
Transport costs disutilities and
travel times
Multimodal person and freight tours
Time-sliced demand flows by mode
Congestion indices link and
node delay
Economic triggers for freight flows
Accessibilities and transport costs
Source Federal Highway Administration Pilot Course on Statewide Travel Forecasting 2007
Issues unique to statewide models
bull Explicit linkages to economic models
bull Scale issues Modeling states at an urban scale Expensive to build and maintain
bull Need for interfaces with urban models Consistency Data exchange
bull Lack of standard practicebull Unique components
Freight (commodities rarr vehicles) Long distance travel Economic impacts
Data and knowledge gaps
Source Michigan Department of Transportation
Typical uses of Statewide ModelsLess common usesbull Air quality conformity analysesbull Freight and intermodal
planningbull Traffic impact studiesbull Economic development studiesbull Long-term investment studiesbull Emergency evacuation studiesbull Multi-state corridors (possibly
with multi-state models)
Most commonly citedbull Corridor planning (including
community bypasses)bull Forecasting freighttruck
flowsbull External trip forecasts for
MPO amp regional models bull Expansion of MPO amp regional
models
Source Pennsylvania Department of Transportation TRB presentation on 2006-2030 Statewide Model
Louisianarsquos and Virginiarsquos statewide modelsMarket segments 2000 source amp estimation procedure Forecast source amp estimation
procedure
Macrolevel
1 Interstate and interparishtruck trip table
2000 macro zone level Transearch data some matrix estimation
2030 macro zone forecast of 2000 Transearch data interpolation and extrapolation to forecast other years
2 Long distance interstate business tourist and other auto trip tables
1995 American Travel Survey data disaggregated from state and MSA to county level and Fratar balanced to 2000 using household and employment growth
Fratar balance year 2000 county-level trip tables to forecast year using household and employment growth
3 Long distance intrastate business tourist and other auto trip tables
4 Short distance Interstate Louisiana HBW trip table
1990 Census county level JTW data Fratar balanced to 2000 using household and employment growth
Microlevel
5 Short distance intraparish truck trip table
Synthesized from 2000 intraparish Transearch data
Fratar balance 2000 intraparish trip table to forecast year using household and employment growth6 Short distance intrastate
Louisiana HBW trip table1990 tract-level JTW data Fratar balanced to 2000 using household and employment growth
7 Short distance intrastate HBO and NHB trip tables
Urban model style household-based trip generation and gravity model distribution
Urban style household-based trip generation and distribution models log-linear regression used to estimate external trip ends
Source Virginia Department of Transportation
MicrosimulationActivity-based Approach
bull Models dynamic evolution and actions of discrete agents across time and space
bull Observe emergent behavior Interaction Cooperation Coordination
bull Primary emphasis on understanding the system
bull Maybe forecast it
Between autonomous perceptive and deliberative agents
Source Ohio Department of Transportation
Key Differences ndash Traditional vs MicrosimulationldquoTraditional modelsrdquobull Closed form equilibriumbull Well understoodbull Typically deterministicbull Presumes central tendencies
in underlying databull Well-behavedbull ldquoManageablerdquo variationrdquobull Computationally efficient
Microsimulationbull Disequilibriumbull Large systems NWUbull Stochasticbull Diverse behaviorbull Handles wide variancesbull Accommodates weak to strong
understandingbull Computationally heavy
Zonal Land Use Data
Worker Traveler Generation
Vehicle Assignment
Starting Time Assignment
Post-Processor
Next Stop Purpose Choice
Model
Next Stop Location
Choice Model
Dynamic Activity Pattern Generation
Commercial Vehicle Trip List
Zonal Land Use Data
Worker Traveler Generation
Vehicle Assignment
Starting Time Assignment
Post-Processor
Next Stop Purpose Choice
Model
Next Stop Location
Choice Model
Dynamic Activity Pattern Generation
Commercial Vehicle Trip List
Microsimulation Examplesbull Ohiobull Oregon
Source Ohio Department of Transportation
Statewide Model Current Practicebull Meetings of the TRB Subcommittee on Statewide Travel Forecasting Models
bull NCHRP Report 735 Long-Distance and Rural Transferable Parameters for Statewide Travel Forecasting Models 2012
bull NCHRP Validation and Sensitivity Considerations for Statewide Models September 2010
bull White Paper on Statewide Modeling presented at TRB Statewide Planning Conference September 2008
bull National Travel Demand Forecasting Model Phase I Final Scope 2008
bull Forecasting Statewide Freight Toolkit NCHRP Report 606 2008
bull TRB Annual Meeting Sessions in 2004 2006 and 2008 web page
bull Statewide Travel Forecasting Models NCHRP Synthesis 358 2006
bull Statewide Travel Demand Modeling Peer Exchange Transportation Research Circular E-C075 September 2004
bull Guidebook on Statewide Travel Forecasting and Irvine Conference Transportation Research Circular E-C011 1999
Topic ndash Statewide Model Theory
bull Common themes in statewide model development
bull Alternative structures Trip table estimation (2 step) Highway model structure (3 step) Fully multimodal (4 step) Beyond 4 step
bull Deciding on best approach
bull Data development challengesSource Massachusetts Highway DeptCambridge Systematics Inc
Common Themes in Review of Statewide Models
bull Most statewide models use similar methodologies to metropolitan models but there are many variations
bull The largest problems relate to issues of scale
bull Models from different states vary greatly in complexity cost and development time
bull Models are most successful when addressing statewide priorities
bull There are major deficiencies in available data about long-distance and rural passenger travelhellip NCHRP Report 735 has helped
bull Statewide models are more compatible with secondary freight data sources such as the Commodity Flow Survey than are metropolitan models
bull Interest is high among states in progress in deploying models
Source A Horowitz White Paper on Statewide Modeling presented at TRB Conference on Statewide Planning 2008
Alternative Statewide Model Structures
bull Trip table estimation (2 step) Estimate initial ldquoseed matrixrdquo
Roadside intercept surveys American Travel Survey National HH Travel SurveyAdd-ons Census Journey-to-Work flows
Matrix estimation program Inputs ndash seed matrix highway
network traffic counts Adjusts origin-destination flows
between zones to optimize matching network traffic counts
Fratar factors applied to synthetic trip table in estimating future growth
bull Trip table estimation (2 step) Pros
Limited need to prepare aggregate socioeconomic input data
Simplicity of model generally means short execution times
Forecasting needs are limited to zone or district growth factors
Cons Less explanatory power than most
MPOregional models Fixed distribution pattern Difficulties in assessing benefits of
adding new roadway capacity
Alternative Statewide Model Structures (Contrsquod)
bull Highway model structure (3 step) Trip generation Trip distribution Highway assignment
bull Mode choice excluded Either generate vehicle trips or Apply auto occupancy factors to person trip
tables
bull Pros Benefits of trip redistribution across range of
alternatives Greater sensitivity to roadway capacity and
congestion
bull Cons Likely reliance on MPOsothers to provide
socioeconomic input data Canrsquot fully test transportation strategies that
would alter mode split
Source Florida Dept of TransportationCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)
bull Freight considerations (3 step) Typical freight model steps
Trip generation ndash tons by commodity group
Trip distribution Mode choice ndash freight is
apportioned to transport modes
bull Joint highway assignment (trucks and passenger trips)
Source Florida Dept of TransportationCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)bull Full multi-modal structure (4 step)
Trip generation Trip distribution Mode choice Highway assignment
bull More common with mega-regions Focused on fixed guideway systems Intercity rail is important component
bull Pros Potential to model new intercity rail
including high speed rail Test greater range of potential
transportation solutions
bull Cons Requires coding of transit networks and
mode choice estimation Modeling of air travel might be needed
especially if intercity rail is a completely new mode of travel
Source California HSR AuthorityCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)
Beyond 4 Stepbull Potential components
Implement destination choice Tour-basedtrip chains Activity-based travel model Activity micro-simulation Traffic micro-simulation Integration with land use models Integration with economic models
bull Pros Ability to test a more complete set of
transportation strategies and solutions
bull Cons Extensive data requirements Cost to maintain Complexity
Source Oregon Dept of TransportationPBUniversity of Calgary
Deciding on Best Statewide Model Approach
bull What are the motivations for building a statewide travel demand modelbull What are the emerging transportation and policy issues to addressbull How significant is the impact of through travel on your statebull Do you have many multi-state MPOsbull How reliable (and current) are your survey databull Does your state have major international ports or border crossings
StudyArea
UrbanModel
AreaStudyArea
UrbanModelArea 1
Study AreaWithin OneModel Area
UrbanModelArea 2
Study AreasWithin Two Model Areas
StudyArea
UrbanModelArea
StudyArea
Statewide Model
Study AreaOutside Urban
Model Areas
Statewide ModelStatewide Model
Source Florida Department of TransportationCambridge Systematics Inc
Figure 317 Subarea Study Area Examples
Study
Area
Urban
Model
Area
Study
Area
Urban
Model
Area 1
Study Area
Within One
Model Area
Urban
Model
Area 2
Study Areas
Within Two
Model Areas
Study
Area
Urban
Model
Area
Study
Area
Statewide Model
Study Area
Outside Urban
Model Areas
Statewide Model
Statewide Model
Statewide Model Data Development Challenges
bull Primary data deficiencies Long distance travelers Rural resident trip-making
bull Challenges in compiling required data Age of data Confidentiality Coordination with data providers Proper level of geography Sufficient sample size
bull Four categories of data for discussion Networks and zone system Demographic and employment data Freight and other economic data NHTS and other behavioral data
Source Florida Department of TransportationCambridge Systematics Inc
Data Challengeshellipwhat is ldquolong distancerdquo
bull gt50 miles (American Traveler Survey)
bull gt100 miles (National Household Travel Survey)
bull IE trips (several states)
bull Trips made by residents for business personal business
bull Recreationaltourist trips
Topic ndash Similarities amp Differences with Urban Models
bull Zone systems
bull Networks
bull Travel behavior assumptions
bull Trip purposesmarket segments
bull Modal issues
bull Truck forecasting
Source Delaware Dept of TransportationWRampA Inc
Statewide vs Urban Model Zone Systems
bull Usually larger zones in statewide models than MPOs
bull Block groups are good starting point for statewide model TAZs
bull Block groups might need splitting along major transportation corridors
bull In small states MPO TAZs could equal SWM zones
bull In most states MPO TAZ equivalency tables will likely be neededdesirable
bull FAF disaggregation approach used for statewide model TAZs
Source FHWA Office of Freight Management and Operations
Source Florida Dept of TransportationCambridge Systematics Inc
FAF zones
Block groups = TAZs
Too small for statewide zones
Too big for statewide zones
Statewide vs Urban Model Networksbull Statewide model networks are sparse in
comparison with MPO models
bull At a minimum statewide networks should include Limited access highways Arterials Major collectors
bull For proper circulation patterns lower class facilities might be added
bull Outside state a very sparse network is needed for major truck corridors
bull Consistent level of detail for networks and zones
Source Tennessee Dept of TransportationCambridge Systematics Inc
MPO network
statewide network
Statewide vs Urban Model Travel Behaviorbull Understanding urban travel behavior
MPO models are usually developed with local household travel survey data
MPOs without recent HH survey data can borrow parameters from other MPO models or guidance documents such as NCHRP 716 Transferable Urban Travel Parameters
bull Rural and long-distance travel behavior A limited number of states have conducted
statewide household travel surveys American Travel Survey was designed to target
long-distance trips but was discontinued long ago
National Household Travel Survey Add-ons can be structured for a sufficient sample of rural trip-makers
NCHRP 735 Rural and Long-Distance Travel Parameters recently published for use in statewide models
Map of 2009 Add-on Surveys Source FHWA
Statewide vs Urban Model Trip Purposes
bull Typical MPO trip purposes are usually collapsed into HBW HBOHBNW and NHB for statewide models
bull Long-distance purposes Business Tourist Personal business
bull Trucks Commodity group is often
surrogate for trip purpose
bull External trips External generally means trip from
outside state Might exclude ldquohalordquo area
surrounding state boundary
Trip PurposeFL
2000 IN LA MA MS RI
Home-based Work 1914 1995 1900 1841 1606 2000
Home-based Nonwork 5547 5208 4770 5206 4917 5700
Home-based Other 2792 5208 4770 1938 4917 5700
Home-based Shop 1576 ndash ndash 1173 ndash ndash
Home-based SocialRecreation
1179 ndash ndash 1395 ndash ndash
Home-base RecreationShop ndash ndash ndash ndash ndash ndash
Home-based School ndash ndash ndash 700 ndash ndash
Nonhome-based 2539 2686 3100 2468 3289 2300
Nonhome-based Work ndash ndash ndash 1027 ndash ndash
Nonhome-based Nonwork ndash ndash ndash 1441 ndash ndash
Other ndash ndash ndash ndash ndash
Truck ndash ndash 170 ndash ndash ndash
Long-Distance 000 111 060 417 022 000
Long-Distance Business ndash ndash 020 284 011 ndash
Long-Distance Personal Business
ndash ndash ndash ndash ndash ndash
Long-Distance Tourist ndash ndash 010 027 000 ndash
Long-Distance Work ndash ndash ndash 105 ndash ndash
Long-Distance NonworkOther
ndash ndash 030 ndash 011 ndash
Statewide vs Urban Model Trip Purposes (Contrsquod)
Number of person trips
Mode Pleasure Business Personal business Other Total Percent
Personal vehicle 12185 5708 3120 789 21802 908Commercial air 1012 746 151 24 1933 8Train 74 150 14 0 238 1Other 4 2 5 12 23 01Ship 19 0 0 0 19 01
Total 13294 6606 3290 825 24015 100Percent 554 275 137 34 100
Percent by trip purpose
Mode Pleasure Business Personal business Other
Personal vehicle 917 864 948 956Commercial air 76 113 46 29Train 06 23 04 0Other 0 0 02 15Ship 01 0 0 0
Trip purpose
Trip purpose
Statewide vs Urban Model Modal Issuesbull Most statewide models are limited to highway trips
bull Modeling of local bus routes is likely overkill
bull Mode choice is usually just a conversion from person trips to vehicle trips via auto occupancy rates
bull Transit components of statewide models are more common in regions with history of fixed guideway transit
bull Transit model is needed to test options such as high speedintercity rail
ATS 95 NHTS 01 MDOT 03-04 ODOT 02-03Personal vehicle 81 89 89 88Commercial air 16 8 10 8Bus 2 2 04 19Train 05 01 0 005Other 05 09 06 205
Survey and yearPrimary mode of transport
Statewide vs Urban Model Modal Issues (Contrsquod)
3727 24 22
32
29 36
10
21
28 22
10
22
12
1529
11
44 43
0
10
20
30
40
50
60
70
80
90
100
Personalvehicle
Bus Train Marine Commercialair
Perc
ent o
f per
son
trips
lt300 300-499 500-599 1000-1999 gt2000
68
85 85 82 85
30
12 11 15 13
0
10
20
30
40
50
60
70
80
90
100
Busi-ness
Plea-sure
pBusi-ness
Vaca-tion
Week-end
Perc
ent o
f per
son
trips
Personal vehicle Commercial air All other
Mode by Trip Purpose Mode by Distance (miles)
Statewide vs Urban Model Modal Issues (Contrsquod)
bull Most statewide assignment models separate trucks from passenger cars Preloading process commonly used for trucks Takes into account more limited route options for truck travel
Source Ohio Dept of Transportation
Statewide vs Urban Model Truck Forecasting
bull Urban modelsndash Freight movements are partially
ldquoexternalrdquo to model (truncated at external zone)
ndash More general categories of truckcommercial trips used
ndash Strong reliance on Quick Response Freight Manual
bull Statewide modelsndash Rooted in commodity flows ndash Often use econometric
approaches to modelingndash Generally model entire length of
freight travel
Freight Model Type by Peer Exchange Participants and
NonParticipants
0123456789
None
Traditio
nal
Commod
ity
Econo
metric
Unkno
wn
NotPeer Exchange
Source TRB Electronic Circular E-C075 Statewide Travel Demand Modeling ndash A Peer Exchange
Topic ndash Data Needs and Considerations
bull Impact of model complexity on data needs
bull Overview of data sources
bull Long-Distance travel gt different focus than MPOregional models
Sources NHTS Weight calculation Pitfalls Key statistics European surveys Lessons learned
Impact of Model Complexity on Data Needsbull Trip table estimation (2 step) ndash highway network
zone system trip table
bull Highway model structure (3 step) ndash 2 step plusbehavioral data for trip rates friction factors auto occupancy rates
bull Freight considerations (3 step) ndash 3 step pluscommodity flows economic data tonnages by mode truck counts
bull Full multi-modal structure (4 step) ndash 3 step plustransit networks and paths logit mode choice model
bull Beyond 4 Step ndash synthetic household generation operational characteristics etc
Source Georgia DOTCambridge Systematics Inc
Statewide Model Data Sources A-Zbull American Travel Survey (ATS)
bull Bureau of Economic Analysis (BEA)
bull Census Journey-to-Work (JTW) data
bull Census Public Use Microsample (PUMS)
bull Census Transportation Planning Package (CTPP)
bull Claritas
bull Commodity Flow Survey (CFS)
Source US Census Bureau
Statewide Model Data Sources (Contrsquod)bull Dun amp Bradstreet (DampB) employment data
bull Employment Security data (ES-202)
bull Freight Analysis Framework (FAF)
bull Highway Performance Monitoring System (HPMS)
bull InfoUSA employment data
bull Metropolitan Planning Organizations (MPO)
bull NCHRP 365 Manual
bull National Highway Planning Network (NHPN)
Source FAF2 Provisional Commodity Origin-Destination Estimates Revised Methodology Report BattelleFHWA
Statewide Model Data Sources (Contrsquod)
bull National Household Travel Survey (NHTS)
bull Origin-destination surveys
bull Regional Economic Models Inc (REMI)
bull State Departments of Transportation
bull TRANSEARCH data
bull Vehicle Use and Identification Survey (VIUS)
bull Woods amp Poole employment forecasts
Source Bureau of Transportation Statistics NHTS
Data Sources on Long-Distance Travel
bull Census Journey-to-Work databull Long distance component
American Travel Survey (77 95) National Household Travel Survey (01) Ohio statewide surveys (02-04) Michigan statewide survey (04-05) California (10-12)
bull Surveys focused on long distance Oregon (96-98 07-08)
bull Others BTS Omnibus surveys Tourism surveys European practice NCHRP Report 735
NHTS and Long-Distance Travelbull Evolution of US national surveys
NPTS (69 77 83 90 95) ATS (77 95)
bull 2001 NHTS ndash Total of 60282 long distance person trips Data collected March 01 to May 02 22204 (37 percent) before 911 38078 (63 percent) afterwards
bull Long distance trip is longer than 50 miles one-way (vs 100 miles in 1995 ATS)
bull 28-day retrospective survey (including travel day)
bull Not your typical travelers Low frequency Distinct segment of the population
Weight Calculation for Long-Distance Travelers
bull Use to ensure valid population-level estimates Accounts for non-response under-coverage multiple phones etc
bull Multi-stage process Match national margins Adjust for data splitting (force sumexpfac = 277208169) Raking (match independent controls) race ethnicity race by
month ethnicity by month sex by age census region MSA status month by day of week
Trim outlier expansion factors max expfac = 4 times mean Replication (correct for sampling error)
bull At least two passes through the data
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Freight Model Types | |||||||||||||||||||
State | Type | Peer Exc | |||||||||||||||||
OR | 4 | 1 | Count of Type | Count of Type | Peer Exc | ||||||||||||||
WI | 3 | 1 | Type | Total | Type | 0 | 1 | Grand Total | |||||||||||
MO | 1 | 1 | 1 | 2 | 1 | 2 | 2 | 4 | |||||||||||
LA | 3 | 1 | 2 | 2 | 2 | 2 | 2 | 4 | |||||||||||
IN | 3 | 1 | 3 | 6 | 3 | 2 | 6 | 8 | |||||||||||
OH | 4 | 1 | 4 | 2 | 4 | 2 | 2 | ||||||||||||
KY | 3 | 1 | Grand Total | 12 | 3 | 3 | |||||||||||||
FL | 3 | 1 | Grand Total | 9 | 12 | 21 | |||||||||||||
VA | 3 | 1 | None | 2 | Not | Peer Exchange | |||||||||||||
NJ | 2 | 1 | Traditional | 2 | None | 2 | 2 | ||||||||||||
MA | 2 | 1 | Commodity | 6 | Traditional | 2 | 2 | ||||||||||||
NH | 1 | 1 | Econometric | 2 | Commodity | 2 | 6 | ||||||||||||
CA | 0 | Econometric | 2 | ||||||||||||||||
TX | 3 | 0 | Unknown | 3 | |||||||||||||||
MI | 3 | 0 | |||||||||||||||||
GA | 0 | ||||||||||||||||||
SC | 1 | 0 | |||||||||||||||||
DE | 2 | 0 | |||||||||||||||||
CT | 0 | ||||||||||||||||||
VT | 2 | 0 | |||||||||||||||||
ME | 1 | 0 | |||||||||||||||||
MN | 0 |
None | None | ||
Traditional | Traditional | ||
Commodity | Commodity | ||
Econometric | Econometric | ||
Unknown | Unknown |
Unique Features of Statewide Models
bull Geographic scale and resolution In-state traffic analysis zones ldquoHalordquo or ldquoborderrdquo set of detailed
zones in adjacent state(s) where significant activity lies across the border
ldquoExternalrdquo zones for passenger travel at state line or halo border
Other states or groups of states for freight modeling (see next example map)
Source Florida Department of Transportation 2005 Statewide Model
Unique Features of Statewide Models (Contrsquod)
bull More aggregate than MPO models (networkszones)
bull Simplified approaches used at times (not necessarily recommended) Three step structure Synthetic or borrowed model structures and
parameters
bull Explicit treatment of visitors and long-distance travel
bull Freight a larger emphasis
bull Potential for inconsistencies with urban models
bull Some linkages with land use and economic models
Source Florida Department of Transportation 2005 Statewide Model
Current Status of Statewide Models
Operational (28)
Dormant (3)
No model (11)
Developing (8)
Source A Horowitz Statewide Travel Forecasting Models NCHRP Synthesis 358 (2006) with recent updates
Statewide ModelsContext for statewide modeling
MacroscopicMegascopic Mesoscopic Microscopic
Urban travel models (tour-
based or sequential)
Regional model of economic
trade and land use trends
Dynamic traffic assignment to
planning network
Detailed traffic analysis of key facilities and
corridors
Transport costs disutilities and
travel times
Multimodal person and freight tours
Time-sliced demand flows by mode
Congestion indices link and
node delay
Economic triggers for freight flows
Accessibilities and transport costs
Source Federal Highway Administration Pilot Course on Statewide Travel Forecasting 2007
Issues unique to statewide models
bull Explicit linkages to economic models
bull Scale issues Modeling states at an urban scale Expensive to build and maintain
bull Need for interfaces with urban models Consistency Data exchange
bull Lack of standard practicebull Unique components
Freight (commodities rarr vehicles) Long distance travel Economic impacts
Data and knowledge gaps
Source Michigan Department of Transportation
Typical uses of Statewide ModelsLess common usesbull Air quality conformity analysesbull Freight and intermodal
planningbull Traffic impact studiesbull Economic development studiesbull Long-term investment studiesbull Emergency evacuation studiesbull Multi-state corridors (possibly
with multi-state models)
Most commonly citedbull Corridor planning (including
community bypasses)bull Forecasting freighttruck
flowsbull External trip forecasts for
MPO amp regional models bull Expansion of MPO amp regional
models
Source Pennsylvania Department of Transportation TRB presentation on 2006-2030 Statewide Model
Louisianarsquos and Virginiarsquos statewide modelsMarket segments 2000 source amp estimation procedure Forecast source amp estimation
procedure
Macrolevel
1 Interstate and interparishtruck trip table
2000 macro zone level Transearch data some matrix estimation
2030 macro zone forecast of 2000 Transearch data interpolation and extrapolation to forecast other years
2 Long distance interstate business tourist and other auto trip tables
1995 American Travel Survey data disaggregated from state and MSA to county level and Fratar balanced to 2000 using household and employment growth
Fratar balance year 2000 county-level trip tables to forecast year using household and employment growth
3 Long distance intrastate business tourist and other auto trip tables
4 Short distance Interstate Louisiana HBW trip table
1990 Census county level JTW data Fratar balanced to 2000 using household and employment growth
Microlevel
5 Short distance intraparish truck trip table
Synthesized from 2000 intraparish Transearch data
Fratar balance 2000 intraparish trip table to forecast year using household and employment growth6 Short distance intrastate
Louisiana HBW trip table1990 tract-level JTW data Fratar balanced to 2000 using household and employment growth
7 Short distance intrastate HBO and NHB trip tables
Urban model style household-based trip generation and gravity model distribution
Urban style household-based trip generation and distribution models log-linear regression used to estimate external trip ends
Source Virginia Department of Transportation
MicrosimulationActivity-based Approach
bull Models dynamic evolution and actions of discrete agents across time and space
bull Observe emergent behavior Interaction Cooperation Coordination
bull Primary emphasis on understanding the system
bull Maybe forecast it
Between autonomous perceptive and deliberative agents
Source Ohio Department of Transportation
Key Differences ndash Traditional vs MicrosimulationldquoTraditional modelsrdquobull Closed form equilibriumbull Well understoodbull Typically deterministicbull Presumes central tendencies
in underlying databull Well-behavedbull ldquoManageablerdquo variationrdquobull Computationally efficient
Microsimulationbull Disequilibriumbull Large systems NWUbull Stochasticbull Diverse behaviorbull Handles wide variancesbull Accommodates weak to strong
understandingbull Computationally heavy
Zonal Land Use Data
Worker Traveler Generation
Vehicle Assignment
Starting Time Assignment
Post-Processor
Next Stop Purpose Choice
Model
Next Stop Location
Choice Model
Dynamic Activity Pattern Generation
Commercial Vehicle Trip List
Zonal Land Use Data
Worker Traveler Generation
Vehicle Assignment
Starting Time Assignment
Post-Processor
Next Stop Purpose Choice
Model
Next Stop Location
Choice Model
Dynamic Activity Pattern Generation
Commercial Vehicle Trip List
Microsimulation Examplesbull Ohiobull Oregon
Source Ohio Department of Transportation
Statewide Model Current Practicebull Meetings of the TRB Subcommittee on Statewide Travel Forecasting Models
bull NCHRP Report 735 Long-Distance and Rural Transferable Parameters for Statewide Travel Forecasting Models 2012
bull NCHRP Validation and Sensitivity Considerations for Statewide Models September 2010
bull White Paper on Statewide Modeling presented at TRB Statewide Planning Conference September 2008
bull National Travel Demand Forecasting Model Phase I Final Scope 2008
bull Forecasting Statewide Freight Toolkit NCHRP Report 606 2008
bull TRB Annual Meeting Sessions in 2004 2006 and 2008 web page
bull Statewide Travel Forecasting Models NCHRP Synthesis 358 2006
bull Statewide Travel Demand Modeling Peer Exchange Transportation Research Circular E-C075 September 2004
bull Guidebook on Statewide Travel Forecasting and Irvine Conference Transportation Research Circular E-C011 1999
Topic ndash Statewide Model Theory
bull Common themes in statewide model development
bull Alternative structures Trip table estimation (2 step) Highway model structure (3 step) Fully multimodal (4 step) Beyond 4 step
bull Deciding on best approach
bull Data development challengesSource Massachusetts Highway DeptCambridge Systematics Inc
Common Themes in Review of Statewide Models
bull Most statewide models use similar methodologies to metropolitan models but there are many variations
bull The largest problems relate to issues of scale
bull Models from different states vary greatly in complexity cost and development time
bull Models are most successful when addressing statewide priorities
bull There are major deficiencies in available data about long-distance and rural passenger travelhellip NCHRP Report 735 has helped
bull Statewide models are more compatible with secondary freight data sources such as the Commodity Flow Survey than are metropolitan models
bull Interest is high among states in progress in deploying models
Source A Horowitz White Paper on Statewide Modeling presented at TRB Conference on Statewide Planning 2008
Alternative Statewide Model Structures
bull Trip table estimation (2 step) Estimate initial ldquoseed matrixrdquo
Roadside intercept surveys American Travel Survey National HH Travel SurveyAdd-ons Census Journey-to-Work flows
Matrix estimation program Inputs ndash seed matrix highway
network traffic counts Adjusts origin-destination flows
between zones to optimize matching network traffic counts
Fratar factors applied to synthetic trip table in estimating future growth
bull Trip table estimation (2 step) Pros
Limited need to prepare aggregate socioeconomic input data
Simplicity of model generally means short execution times
Forecasting needs are limited to zone or district growth factors
Cons Less explanatory power than most
MPOregional models Fixed distribution pattern Difficulties in assessing benefits of
adding new roadway capacity
Alternative Statewide Model Structures (Contrsquod)
bull Highway model structure (3 step) Trip generation Trip distribution Highway assignment
bull Mode choice excluded Either generate vehicle trips or Apply auto occupancy factors to person trip
tables
bull Pros Benefits of trip redistribution across range of
alternatives Greater sensitivity to roadway capacity and
congestion
bull Cons Likely reliance on MPOsothers to provide
socioeconomic input data Canrsquot fully test transportation strategies that
would alter mode split
Source Florida Dept of TransportationCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)
bull Freight considerations (3 step) Typical freight model steps
Trip generation ndash tons by commodity group
Trip distribution Mode choice ndash freight is
apportioned to transport modes
bull Joint highway assignment (trucks and passenger trips)
Source Florida Dept of TransportationCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)bull Full multi-modal structure (4 step)
Trip generation Trip distribution Mode choice Highway assignment
bull More common with mega-regions Focused on fixed guideway systems Intercity rail is important component
bull Pros Potential to model new intercity rail
including high speed rail Test greater range of potential
transportation solutions
bull Cons Requires coding of transit networks and
mode choice estimation Modeling of air travel might be needed
especially if intercity rail is a completely new mode of travel
Source California HSR AuthorityCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)
Beyond 4 Stepbull Potential components
Implement destination choice Tour-basedtrip chains Activity-based travel model Activity micro-simulation Traffic micro-simulation Integration with land use models Integration with economic models
bull Pros Ability to test a more complete set of
transportation strategies and solutions
bull Cons Extensive data requirements Cost to maintain Complexity
Source Oregon Dept of TransportationPBUniversity of Calgary
Deciding on Best Statewide Model Approach
bull What are the motivations for building a statewide travel demand modelbull What are the emerging transportation and policy issues to addressbull How significant is the impact of through travel on your statebull Do you have many multi-state MPOsbull How reliable (and current) are your survey databull Does your state have major international ports or border crossings
StudyArea
UrbanModel
AreaStudyArea
UrbanModelArea 1
Study AreaWithin OneModel Area
UrbanModelArea 2
Study AreasWithin Two Model Areas
StudyArea
UrbanModelArea
StudyArea
Statewide Model
Study AreaOutside Urban
Model Areas
Statewide ModelStatewide Model
Source Florida Department of TransportationCambridge Systematics Inc
Figure 317 Subarea Study Area Examples
Study
Area
Urban
Model
Area
Study
Area
Urban
Model
Area 1
Study Area
Within One
Model Area
Urban
Model
Area 2
Study Areas
Within Two
Model Areas
Study
Area
Urban
Model
Area
Study
Area
Statewide Model
Study Area
Outside Urban
Model Areas
Statewide Model
Statewide Model
Statewide Model Data Development Challenges
bull Primary data deficiencies Long distance travelers Rural resident trip-making
bull Challenges in compiling required data Age of data Confidentiality Coordination with data providers Proper level of geography Sufficient sample size
bull Four categories of data for discussion Networks and zone system Demographic and employment data Freight and other economic data NHTS and other behavioral data
Source Florida Department of TransportationCambridge Systematics Inc
Data Challengeshellipwhat is ldquolong distancerdquo
bull gt50 miles (American Traveler Survey)
bull gt100 miles (National Household Travel Survey)
bull IE trips (several states)
bull Trips made by residents for business personal business
bull Recreationaltourist trips
Topic ndash Similarities amp Differences with Urban Models
bull Zone systems
bull Networks
bull Travel behavior assumptions
bull Trip purposesmarket segments
bull Modal issues
bull Truck forecasting
Source Delaware Dept of TransportationWRampA Inc
Statewide vs Urban Model Zone Systems
bull Usually larger zones in statewide models than MPOs
bull Block groups are good starting point for statewide model TAZs
bull Block groups might need splitting along major transportation corridors
bull In small states MPO TAZs could equal SWM zones
bull In most states MPO TAZ equivalency tables will likely be neededdesirable
bull FAF disaggregation approach used for statewide model TAZs
Source FHWA Office of Freight Management and Operations
Source Florida Dept of TransportationCambridge Systematics Inc
FAF zones
Block groups = TAZs
Too small for statewide zones
Too big for statewide zones
Statewide vs Urban Model Networksbull Statewide model networks are sparse in
comparison with MPO models
bull At a minimum statewide networks should include Limited access highways Arterials Major collectors
bull For proper circulation patterns lower class facilities might be added
bull Outside state a very sparse network is needed for major truck corridors
bull Consistent level of detail for networks and zones
Source Tennessee Dept of TransportationCambridge Systematics Inc
MPO network
statewide network
Statewide vs Urban Model Travel Behaviorbull Understanding urban travel behavior
MPO models are usually developed with local household travel survey data
MPOs without recent HH survey data can borrow parameters from other MPO models or guidance documents such as NCHRP 716 Transferable Urban Travel Parameters
bull Rural and long-distance travel behavior A limited number of states have conducted
statewide household travel surveys American Travel Survey was designed to target
long-distance trips but was discontinued long ago
National Household Travel Survey Add-ons can be structured for a sufficient sample of rural trip-makers
NCHRP 735 Rural and Long-Distance Travel Parameters recently published for use in statewide models
Map of 2009 Add-on Surveys Source FHWA
Statewide vs Urban Model Trip Purposes
bull Typical MPO trip purposes are usually collapsed into HBW HBOHBNW and NHB for statewide models
bull Long-distance purposes Business Tourist Personal business
bull Trucks Commodity group is often
surrogate for trip purpose
bull External trips External generally means trip from
outside state Might exclude ldquohalordquo area
surrounding state boundary
Trip PurposeFL
2000 IN LA MA MS RI
Home-based Work 1914 1995 1900 1841 1606 2000
Home-based Nonwork 5547 5208 4770 5206 4917 5700
Home-based Other 2792 5208 4770 1938 4917 5700
Home-based Shop 1576 ndash ndash 1173 ndash ndash
Home-based SocialRecreation
1179 ndash ndash 1395 ndash ndash
Home-base RecreationShop ndash ndash ndash ndash ndash ndash
Home-based School ndash ndash ndash 700 ndash ndash
Nonhome-based 2539 2686 3100 2468 3289 2300
Nonhome-based Work ndash ndash ndash 1027 ndash ndash
Nonhome-based Nonwork ndash ndash ndash 1441 ndash ndash
Other ndash ndash ndash ndash ndash
Truck ndash ndash 170 ndash ndash ndash
Long-Distance 000 111 060 417 022 000
Long-Distance Business ndash ndash 020 284 011 ndash
Long-Distance Personal Business
ndash ndash ndash ndash ndash ndash
Long-Distance Tourist ndash ndash 010 027 000 ndash
Long-Distance Work ndash ndash ndash 105 ndash ndash
Long-Distance NonworkOther
ndash ndash 030 ndash 011 ndash
Statewide vs Urban Model Trip Purposes (Contrsquod)
Number of person trips
Mode Pleasure Business Personal business Other Total Percent
Personal vehicle 12185 5708 3120 789 21802 908Commercial air 1012 746 151 24 1933 8Train 74 150 14 0 238 1Other 4 2 5 12 23 01Ship 19 0 0 0 19 01
Total 13294 6606 3290 825 24015 100Percent 554 275 137 34 100
Percent by trip purpose
Mode Pleasure Business Personal business Other
Personal vehicle 917 864 948 956Commercial air 76 113 46 29Train 06 23 04 0Other 0 0 02 15Ship 01 0 0 0
Trip purpose
Trip purpose
Statewide vs Urban Model Modal Issuesbull Most statewide models are limited to highway trips
bull Modeling of local bus routes is likely overkill
bull Mode choice is usually just a conversion from person trips to vehicle trips via auto occupancy rates
bull Transit components of statewide models are more common in regions with history of fixed guideway transit
bull Transit model is needed to test options such as high speedintercity rail
ATS 95 NHTS 01 MDOT 03-04 ODOT 02-03Personal vehicle 81 89 89 88Commercial air 16 8 10 8Bus 2 2 04 19Train 05 01 0 005Other 05 09 06 205
Survey and yearPrimary mode of transport
Statewide vs Urban Model Modal Issues (Contrsquod)
3727 24 22
32
29 36
10
21
28 22
10
22
12
1529
11
44 43
0
10
20
30
40
50
60
70
80
90
100
Personalvehicle
Bus Train Marine Commercialair
Perc
ent o
f per
son
trips
lt300 300-499 500-599 1000-1999 gt2000
68
85 85 82 85
30
12 11 15 13
0
10
20
30
40
50
60
70
80
90
100
Busi-ness
Plea-sure
pBusi-ness
Vaca-tion
Week-end
Perc
ent o
f per
son
trips
Personal vehicle Commercial air All other
Mode by Trip Purpose Mode by Distance (miles)
Statewide vs Urban Model Modal Issues (Contrsquod)
bull Most statewide assignment models separate trucks from passenger cars Preloading process commonly used for trucks Takes into account more limited route options for truck travel
Source Ohio Dept of Transportation
Statewide vs Urban Model Truck Forecasting
bull Urban modelsndash Freight movements are partially
ldquoexternalrdquo to model (truncated at external zone)
ndash More general categories of truckcommercial trips used
ndash Strong reliance on Quick Response Freight Manual
bull Statewide modelsndash Rooted in commodity flows ndash Often use econometric
approaches to modelingndash Generally model entire length of
freight travel
Freight Model Type by Peer Exchange Participants and
NonParticipants
0123456789
None
Traditio
nal
Commod
ity
Econo
metric
Unkno
wn
NotPeer Exchange
Source TRB Electronic Circular E-C075 Statewide Travel Demand Modeling ndash A Peer Exchange
Topic ndash Data Needs and Considerations
bull Impact of model complexity on data needs
bull Overview of data sources
bull Long-Distance travel gt different focus than MPOregional models
Sources NHTS Weight calculation Pitfalls Key statistics European surveys Lessons learned
Impact of Model Complexity on Data Needsbull Trip table estimation (2 step) ndash highway network
zone system trip table
bull Highway model structure (3 step) ndash 2 step plusbehavioral data for trip rates friction factors auto occupancy rates
bull Freight considerations (3 step) ndash 3 step pluscommodity flows economic data tonnages by mode truck counts
bull Full multi-modal structure (4 step) ndash 3 step plustransit networks and paths logit mode choice model
bull Beyond 4 Step ndash synthetic household generation operational characteristics etc
Source Georgia DOTCambridge Systematics Inc
Statewide Model Data Sources A-Zbull American Travel Survey (ATS)
bull Bureau of Economic Analysis (BEA)
bull Census Journey-to-Work (JTW) data
bull Census Public Use Microsample (PUMS)
bull Census Transportation Planning Package (CTPP)
bull Claritas
bull Commodity Flow Survey (CFS)
Source US Census Bureau
Statewide Model Data Sources (Contrsquod)bull Dun amp Bradstreet (DampB) employment data
bull Employment Security data (ES-202)
bull Freight Analysis Framework (FAF)
bull Highway Performance Monitoring System (HPMS)
bull InfoUSA employment data
bull Metropolitan Planning Organizations (MPO)
bull NCHRP 365 Manual
bull National Highway Planning Network (NHPN)
Source FAF2 Provisional Commodity Origin-Destination Estimates Revised Methodology Report BattelleFHWA
Statewide Model Data Sources (Contrsquod)
bull National Household Travel Survey (NHTS)
bull Origin-destination surveys
bull Regional Economic Models Inc (REMI)
bull State Departments of Transportation
bull TRANSEARCH data
bull Vehicle Use and Identification Survey (VIUS)
bull Woods amp Poole employment forecasts
Source Bureau of Transportation Statistics NHTS
Data Sources on Long-Distance Travel
bull Census Journey-to-Work databull Long distance component
American Travel Survey (77 95) National Household Travel Survey (01) Ohio statewide surveys (02-04) Michigan statewide survey (04-05) California (10-12)
bull Surveys focused on long distance Oregon (96-98 07-08)
bull Others BTS Omnibus surveys Tourism surveys European practice NCHRP Report 735
NHTS and Long-Distance Travelbull Evolution of US national surveys
NPTS (69 77 83 90 95) ATS (77 95)
bull 2001 NHTS ndash Total of 60282 long distance person trips Data collected March 01 to May 02 22204 (37 percent) before 911 38078 (63 percent) afterwards
bull Long distance trip is longer than 50 miles one-way (vs 100 miles in 1995 ATS)
bull 28-day retrospective survey (including travel day)
bull Not your typical travelers Low frequency Distinct segment of the population
Weight Calculation for Long-Distance Travelers
bull Use to ensure valid population-level estimates Accounts for non-response under-coverage multiple phones etc
bull Multi-stage process Match national margins Adjust for data splitting (force sumexpfac = 277208169) Raking (match independent controls) race ethnicity race by
month ethnicity by month sex by age census region MSA status month by day of week
Trim outlier expansion factors max expfac = 4 times mean Replication (correct for sampling error)
bull At least two passes through the data
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Freight Model Types | |||||||||||||||||||
State | Type | Peer Exc | |||||||||||||||||
OR | 4 | 1 | Count of Type | Count of Type | Peer Exc | ||||||||||||||
WI | 3 | 1 | Type | Total | Type | 0 | 1 | Grand Total | |||||||||||
MO | 1 | 1 | 1 | 2 | 1 | 2 | 2 | 4 | |||||||||||
LA | 3 | 1 | 2 | 2 | 2 | 2 | 2 | 4 | |||||||||||
IN | 3 | 1 | 3 | 6 | 3 | 2 | 6 | 8 | |||||||||||
OH | 4 | 1 | 4 | 2 | 4 | 2 | 2 | ||||||||||||
KY | 3 | 1 | Grand Total | 12 | 3 | 3 | |||||||||||||
FL | 3 | 1 | Grand Total | 9 | 12 | 21 | |||||||||||||
VA | 3 | 1 | None | 2 | Not | Peer Exchange | |||||||||||||
NJ | 2 | 1 | Traditional | 2 | None | 2 | 2 | ||||||||||||
MA | 2 | 1 | Commodity | 6 | Traditional | 2 | 2 | ||||||||||||
NH | 1 | 1 | Econometric | 2 | Commodity | 2 | 6 | ||||||||||||
CA | 0 | Econometric | 2 | ||||||||||||||||
TX | 3 | 0 | Unknown | 3 | |||||||||||||||
MI | 3 | 0 | |||||||||||||||||
GA | 0 | ||||||||||||||||||
SC | 1 | 0 | |||||||||||||||||
DE | 2 | 0 | |||||||||||||||||
CT | 0 | ||||||||||||||||||
VT | 2 | 0 | |||||||||||||||||
ME | 1 | 0 | |||||||||||||||||
MN | 0 |
None | None | ||
Traditional | Traditional | ||
Commodity | Commodity | ||
Econometric | Econometric | ||
Unknown | Unknown |
Unique Features of Statewide Models (Contrsquod)
bull More aggregate than MPO models (networkszones)
bull Simplified approaches used at times (not necessarily recommended) Three step structure Synthetic or borrowed model structures and
parameters
bull Explicit treatment of visitors and long-distance travel
bull Freight a larger emphasis
bull Potential for inconsistencies with urban models
bull Some linkages with land use and economic models
Source Florida Department of Transportation 2005 Statewide Model
Current Status of Statewide Models
Operational (28)
Dormant (3)
No model (11)
Developing (8)
Source A Horowitz Statewide Travel Forecasting Models NCHRP Synthesis 358 (2006) with recent updates
Statewide ModelsContext for statewide modeling
MacroscopicMegascopic Mesoscopic Microscopic
Urban travel models (tour-
based or sequential)
Regional model of economic
trade and land use trends
Dynamic traffic assignment to
planning network
Detailed traffic analysis of key facilities and
corridors
Transport costs disutilities and
travel times
Multimodal person and freight tours
Time-sliced demand flows by mode
Congestion indices link and
node delay
Economic triggers for freight flows
Accessibilities and transport costs
Source Federal Highway Administration Pilot Course on Statewide Travel Forecasting 2007
Issues unique to statewide models
bull Explicit linkages to economic models
bull Scale issues Modeling states at an urban scale Expensive to build and maintain
bull Need for interfaces with urban models Consistency Data exchange
bull Lack of standard practicebull Unique components
Freight (commodities rarr vehicles) Long distance travel Economic impacts
Data and knowledge gaps
Source Michigan Department of Transportation
Typical uses of Statewide ModelsLess common usesbull Air quality conformity analysesbull Freight and intermodal
planningbull Traffic impact studiesbull Economic development studiesbull Long-term investment studiesbull Emergency evacuation studiesbull Multi-state corridors (possibly
with multi-state models)
Most commonly citedbull Corridor planning (including
community bypasses)bull Forecasting freighttruck
flowsbull External trip forecasts for
MPO amp regional models bull Expansion of MPO amp regional
models
Source Pennsylvania Department of Transportation TRB presentation on 2006-2030 Statewide Model
Louisianarsquos and Virginiarsquos statewide modelsMarket segments 2000 source amp estimation procedure Forecast source amp estimation
procedure
Macrolevel
1 Interstate and interparishtruck trip table
2000 macro zone level Transearch data some matrix estimation
2030 macro zone forecast of 2000 Transearch data interpolation and extrapolation to forecast other years
2 Long distance interstate business tourist and other auto trip tables
1995 American Travel Survey data disaggregated from state and MSA to county level and Fratar balanced to 2000 using household and employment growth
Fratar balance year 2000 county-level trip tables to forecast year using household and employment growth
3 Long distance intrastate business tourist and other auto trip tables
4 Short distance Interstate Louisiana HBW trip table
1990 Census county level JTW data Fratar balanced to 2000 using household and employment growth
Microlevel
5 Short distance intraparish truck trip table
Synthesized from 2000 intraparish Transearch data
Fratar balance 2000 intraparish trip table to forecast year using household and employment growth6 Short distance intrastate
Louisiana HBW trip table1990 tract-level JTW data Fratar balanced to 2000 using household and employment growth
7 Short distance intrastate HBO and NHB trip tables
Urban model style household-based trip generation and gravity model distribution
Urban style household-based trip generation and distribution models log-linear regression used to estimate external trip ends
Source Virginia Department of Transportation
MicrosimulationActivity-based Approach
bull Models dynamic evolution and actions of discrete agents across time and space
bull Observe emergent behavior Interaction Cooperation Coordination
bull Primary emphasis on understanding the system
bull Maybe forecast it
Between autonomous perceptive and deliberative agents
Source Ohio Department of Transportation
Key Differences ndash Traditional vs MicrosimulationldquoTraditional modelsrdquobull Closed form equilibriumbull Well understoodbull Typically deterministicbull Presumes central tendencies
in underlying databull Well-behavedbull ldquoManageablerdquo variationrdquobull Computationally efficient
Microsimulationbull Disequilibriumbull Large systems NWUbull Stochasticbull Diverse behaviorbull Handles wide variancesbull Accommodates weak to strong
understandingbull Computationally heavy
Zonal Land Use Data
Worker Traveler Generation
Vehicle Assignment
Starting Time Assignment
Post-Processor
Next Stop Purpose Choice
Model
Next Stop Location
Choice Model
Dynamic Activity Pattern Generation
Commercial Vehicle Trip List
Zonal Land Use Data
Worker Traveler Generation
Vehicle Assignment
Starting Time Assignment
Post-Processor
Next Stop Purpose Choice
Model
Next Stop Location
Choice Model
Dynamic Activity Pattern Generation
Commercial Vehicle Trip List
Microsimulation Examplesbull Ohiobull Oregon
Source Ohio Department of Transportation
Statewide Model Current Practicebull Meetings of the TRB Subcommittee on Statewide Travel Forecasting Models
bull NCHRP Report 735 Long-Distance and Rural Transferable Parameters for Statewide Travel Forecasting Models 2012
bull NCHRP Validation and Sensitivity Considerations for Statewide Models September 2010
bull White Paper on Statewide Modeling presented at TRB Statewide Planning Conference September 2008
bull National Travel Demand Forecasting Model Phase I Final Scope 2008
bull Forecasting Statewide Freight Toolkit NCHRP Report 606 2008
bull TRB Annual Meeting Sessions in 2004 2006 and 2008 web page
bull Statewide Travel Forecasting Models NCHRP Synthesis 358 2006
bull Statewide Travel Demand Modeling Peer Exchange Transportation Research Circular E-C075 September 2004
bull Guidebook on Statewide Travel Forecasting and Irvine Conference Transportation Research Circular E-C011 1999
Topic ndash Statewide Model Theory
bull Common themes in statewide model development
bull Alternative structures Trip table estimation (2 step) Highway model structure (3 step) Fully multimodal (4 step) Beyond 4 step
bull Deciding on best approach
bull Data development challengesSource Massachusetts Highway DeptCambridge Systematics Inc
Common Themes in Review of Statewide Models
bull Most statewide models use similar methodologies to metropolitan models but there are many variations
bull The largest problems relate to issues of scale
bull Models from different states vary greatly in complexity cost and development time
bull Models are most successful when addressing statewide priorities
bull There are major deficiencies in available data about long-distance and rural passenger travelhellip NCHRP Report 735 has helped
bull Statewide models are more compatible with secondary freight data sources such as the Commodity Flow Survey than are metropolitan models
bull Interest is high among states in progress in deploying models
Source A Horowitz White Paper on Statewide Modeling presented at TRB Conference on Statewide Planning 2008
Alternative Statewide Model Structures
bull Trip table estimation (2 step) Estimate initial ldquoseed matrixrdquo
Roadside intercept surveys American Travel Survey National HH Travel SurveyAdd-ons Census Journey-to-Work flows
Matrix estimation program Inputs ndash seed matrix highway
network traffic counts Adjusts origin-destination flows
between zones to optimize matching network traffic counts
Fratar factors applied to synthetic trip table in estimating future growth
bull Trip table estimation (2 step) Pros
Limited need to prepare aggregate socioeconomic input data
Simplicity of model generally means short execution times
Forecasting needs are limited to zone or district growth factors
Cons Less explanatory power than most
MPOregional models Fixed distribution pattern Difficulties in assessing benefits of
adding new roadway capacity
Alternative Statewide Model Structures (Contrsquod)
bull Highway model structure (3 step) Trip generation Trip distribution Highway assignment
bull Mode choice excluded Either generate vehicle trips or Apply auto occupancy factors to person trip
tables
bull Pros Benefits of trip redistribution across range of
alternatives Greater sensitivity to roadway capacity and
congestion
bull Cons Likely reliance on MPOsothers to provide
socioeconomic input data Canrsquot fully test transportation strategies that
would alter mode split
Source Florida Dept of TransportationCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)
bull Freight considerations (3 step) Typical freight model steps
Trip generation ndash tons by commodity group
Trip distribution Mode choice ndash freight is
apportioned to transport modes
bull Joint highway assignment (trucks and passenger trips)
Source Florida Dept of TransportationCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)bull Full multi-modal structure (4 step)
Trip generation Trip distribution Mode choice Highway assignment
bull More common with mega-regions Focused on fixed guideway systems Intercity rail is important component
bull Pros Potential to model new intercity rail
including high speed rail Test greater range of potential
transportation solutions
bull Cons Requires coding of transit networks and
mode choice estimation Modeling of air travel might be needed
especially if intercity rail is a completely new mode of travel
Source California HSR AuthorityCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)
Beyond 4 Stepbull Potential components
Implement destination choice Tour-basedtrip chains Activity-based travel model Activity micro-simulation Traffic micro-simulation Integration with land use models Integration with economic models
bull Pros Ability to test a more complete set of
transportation strategies and solutions
bull Cons Extensive data requirements Cost to maintain Complexity
Source Oregon Dept of TransportationPBUniversity of Calgary
Deciding on Best Statewide Model Approach
bull What are the motivations for building a statewide travel demand modelbull What are the emerging transportation and policy issues to addressbull How significant is the impact of through travel on your statebull Do you have many multi-state MPOsbull How reliable (and current) are your survey databull Does your state have major international ports or border crossings
StudyArea
UrbanModel
AreaStudyArea
UrbanModelArea 1
Study AreaWithin OneModel Area
UrbanModelArea 2
Study AreasWithin Two Model Areas
StudyArea
UrbanModelArea
StudyArea
Statewide Model
Study AreaOutside Urban
Model Areas
Statewide ModelStatewide Model
Source Florida Department of TransportationCambridge Systematics Inc
Figure 317 Subarea Study Area Examples
Study
Area
Urban
Model
Area
Study
Area
Urban
Model
Area 1
Study Area
Within One
Model Area
Urban
Model
Area 2
Study Areas
Within Two
Model Areas
Study
Area
Urban
Model
Area
Study
Area
Statewide Model
Study Area
Outside Urban
Model Areas
Statewide Model
Statewide Model
Statewide Model Data Development Challenges
bull Primary data deficiencies Long distance travelers Rural resident trip-making
bull Challenges in compiling required data Age of data Confidentiality Coordination with data providers Proper level of geography Sufficient sample size
bull Four categories of data for discussion Networks and zone system Demographic and employment data Freight and other economic data NHTS and other behavioral data
Source Florida Department of TransportationCambridge Systematics Inc
Data Challengeshellipwhat is ldquolong distancerdquo
bull gt50 miles (American Traveler Survey)
bull gt100 miles (National Household Travel Survey)
bull IE trips (several states)
bull Trips made by residents for business personal business
bull Recreationaltourist trips
Topic ndash Similarities amp Differences with Urban Models
bull Zone systems
bull Networks
bull Travel behavior assumptions
bull Trip purposesmarket segments
bull Modal issues
bull Truck forecasting
Source Delaware Dept of TransportationWRampA Inc
Statewide vs Urban Model Zone Systems
bull Usually larger zones in statewide models than MPOs
bull Block groups are good starting point for statewide model TAZs
bull Block groups might need splitting along major transportation corridors
bull In small states MPO TAZs could equal SWM zones
bull In most states MPO TAZ equivalency tables will likely be neededdesirable
bull FAF disaggregation approach used for statewide model TAZs
Source FHWA Office of Freight Management and Operations
Source Florida Dept of TransportationCambridge Systematics Inc
FAF zones
Block groups = TAZs
Too small for statewide zones
Too big for statewide zones
Statewide vs Urban Model Networksbull Statewide model networks are sparse in
comparison with MPO models
bull At a minimum statewide networks should include Limited access highways Arterials Major collectors
bull For proper circulation patterns lower class facilities might be added
bull Outside state a very sparse network is needed for major truck corridors
bull Consistent level of detail for networks and zones
Source Tennessee Dept of TransportationCambridge Systematics Inc
MPO network
statewide network
Statewide vs Urban Model Travel Behaviorbull Understanding urban travel behavior
MPO models are usually developed with local household travel survey data
MPOs without recent HH survey data can borrow parameters from other MPO models or guidance documents such as NCHRP 716 Transferable Urban Travel Parameters
bull Rural and long-distance travel behavior A limited number of states have conducted
statewide household travel surveys American Travel Survey was designed to target
long-distance trips but was discontinued long ago
National Household Travel Survey Add-ons can be structured for a sufficient sample of rural trip-makers
NCHRP 735 Rural and Long-Distance Travel Parameters recently published for use in statewide models
Map of 2009 Add-on Surveys Source FHWA
Statewide vs Urban Model Trip Purposes
bull Typical MPO trip purposes are usually collapsed into HBW HBOHBNW and NHB for statewide models
bull Long-distance purposes Business Tourist Personal business
bull Trucks Commodity group is often
surrogate for trip purpose
bull External trips External generally means trip from
outside state Might exclude ldquohalordquo area
surrounding state boundary
Trip PurposeFL
2000 IN LA MA MS RI
Home-based Work 1914 1995 1900 1841 1606 2000
Home-based Nonwork 5547 5208 4770 5206 4917 5700
Home-based Other 2792 5208 4770 1938 4917 5700
Home-based Shop 1576 ndash ndash 1173 ndash ndash
Home-based SocialRecreation
1179 ndash ndash 1395 ndash ndash
Home-base RecreationShop ndash ndash ndash ndash ndash ndash
Home-based School ndash ndash ndash 700 ndash ndash
Nonhome-based 2539 2686 3100 2468 3289 2300
Nonhome-based Work ndash ndash ndash 1027 ndash ndash
Nonhome-based Nonwork ndash ndash ndash 1441 ndash ndash
Other ndash ndash ndash ndash ndash
Truck ndash ndash 170 ndash ndash ndash
Long-Distance 000 111 060 417 022 000
Long-Distance Business ndash ndash 020 284 011 ndash
Long-Distance Personal Business
ndash ndash ndash ndash ndash ndash
Long-Distance Tourist ndash ndash 010 027 000 ndash
Long-Distance Work ndash ndash ndash 105 ndash ndash
Long-Distance NonworkOther
ndash ndash 030 ndash 011 ndash
Statewide vs Urban Model Trip Purposes (Contrsquod)
Number of person trips
Mode Pleasure Business Personal business Other Total Percent
Personal vehicle 12185 5708 3120 789 21802 908Commercial air 1012 746 151 24 1933 8Train 74 150 14 0 238 1Other 4 2 5 12 23 01Ship 19 0 0 0 19 01
Total 13294 6606 3290 825 24015 100Percent 554 275 137 34 100
Percent by trip purpose
Mode Pleasure Business Personal business Other
Personal vehicle 917 864 948 956Commercial air 76 113 46 29Train 06 23 04 0Other 0 0 02 15Ship 01 0 0 0
Trip purpose
Trip purpose
Statewide vs Urban Model Modal Issuesbull Most statewide models are limited to highway trips
bull Modeling of local bus routes is likely overkill
bull Mode choice is usually just a conversion from person trips to vehicle trips via auto occupancy rates
bull Transit components of statewide models are more common in regions with history of fixed guideway transit
bull Transit model is needed to test options such as high speedintercity rail
ATS 95 NHTS 01 MDOT 03-04 ODOT 02-03Personal vehicle 81 89 89 88Commercial air 16 8 10 8Bus 2 2 04 19Train 05 01 0 005Other 05 09 06 205
Survey and yearPrimary mode of transport
Statewide vs Urban Model Modal Issues (Contrsquod)
3727 24 22
32
29 36
10
21
28 22
10
22
12
1529
11
44 43
0
10
20
30
40
50
60
70
80
90
100
Personalvehicle
Bus Train Marine Commercialair
Perc
ent o
f per
son
trips
lt300 300-499 500-599 1000-1999 gt2000
68
85 85 82 85
30
12 11 15 13
0
10
20
30
40
50
60
70
80
90
100
Busi-ness
Plea-sure
pBusi-ness
Vaca-tion
Week-end
Perc
ent o
f per
son
trips
Personal vehicle Commercial air All other
Mode by Trip Purpose Mode by Distance (miles)
Statewide vs Urban Model Modal Issues (Contrsquod)
bull Most statewide assignment models separate trucks from passenger cars Preloading process commonly used for trucks Takes into account more limited route options for truck travel
Source Ohio Dept of Transportation
Statewide vs Urban Model Truck Forecasting
bull Urban modelsndash Freight movements are partially
ldquoexternalrdquo to model (truncated at external zone)
ndash More general categories of truckcommercial trips used
ndash Strong reliance on Quick Response Freight Manual
bull Statewide modelsndash Rooted in commodity flows ndash Often use econometric
approaches to modelingndash Generally model entire length of
freight travel
Freight Model Type by Peer Exchange Participants and
NonParticipants
0123456789
None
Traditio
nal
Commod
ity
Econo
metric
Unkno
wn
NotPeer Exchange
Source TRB Electronic Circular E-C075 Statewide Travel Demand Modeling ndash A Peer Exchange
Topic ndash Data Needs and Considerations
bull Impact of model complexity on data needs
bull Overview of data sources
bull Long-Distance travel gt different focus than MPOregional models
Sources NHTS Weight calculation Pitfalls Key statistics European surveys Lessons learned
Impact of Model Complexity on Data Needsbull Trip table estimation (2 step) ndash highway network
zone system trip table
bull Highway model structure (3 step) ndash 2 step plusbehavioral data for trip rates friction factors auto occupancy rates
bull Freight considerations (3 step) ndash 3 step pluscommodity flows economic data tonnages by mode truck counts
bull Full multi-modal structure (4 step) ndash 3 step plustransit networks and paths logit mode choice model
bull Beyond 4 Step ndash synthetic household generation operational characteristics etc
Source Georgia DOTCambridge Systematics Inc
Statewide Model Data Sources A-Zbull American Travel Survey (ATS)
bull Bureau of Economic Analysis (BEA)
bull Census Journey-to-Work (JTW) data
bull Census Public Use Microsample (PUMS)
bull Census Transportation Planning Package (CTPP)
bull Claritas
bull Commodity Flow Survey (CFS)
Source US Census Bureau
Statewide Model Data Sources (Contrsquod)bull Dun amp Bradstreet (DampB) employment data
bull Employment Security data (ES-202)
bull Freight Analysis Framework (FAF)
bull Highway Performance Monitoring System (HPMS)
bull InfoUSA employment data
bull Metropolitan Planning Organizations (MPO)
bull NCHRP 365 Manual
bull National Highway Planning Network (NHPN)
Source FAF2 Provisional Commodity Origin-Destination Estimates Revised Methodology Report BattelleFHWA
Statewide Model Data Sources (Contrsquod)
bull National Household Travel Survey (NHTS)
bull Origin-destination surveys
bull Regional Economic Models Inc (REMI)
bull State Departments of Transportation
bull TRANSEARCH data
bull Vehicle Use and Identification Survey (VIUS)
bull Woods amp Poole employment forecasts
Source Bureau of Transportation Statistics NHTS
Data Sources on Long-Distance Travel
bull Census Journey-to-Work databull Long distance component
American Travel Survey (77 95) National Household Travel Survey (01) Ohio statewide surveys (02-04) Michigan statewide survey (04-05) California (10-12)
bull Surveys focused on long distance Oregon (96-98 07-08)
bull Others BTS Omnibus surveys Tourism surveys European practice NCHRP Report 735
NHTS and Long-Distance Travelbull Evolution of US national surveys
NPTS (69 77 83 90 95) ATS (77 95)
bull 2001 NHTS ndash Total of 60282 long distance person trips Data collected March 01 to May 02 22204 (37 percent) before 911 38078 (63 percent) afterwards
bull Long distance trip is longer than 50 miles one-way (vs 100 miles in 1995 ATS)
bull 28-day retrospective survey (including travel day)
bull Not your typical travelers Low frequency Distinct segment of the population
Weight Calculation for Long-Distance Travelers
bull Use to ensure valid population-level estimates Accounts for non-response under-coverage multiple phones etc
bull Multi-stage process Match national margins Adjust for data splitting (force sumexpfac = 277208169) Raking (match independent controls) race ethnicity race by
month ethnicity by month sex by age census region MSA status month by day of week
Trim outlier expansion factors max expfac = 4 times mean Replication (correct for sampling error)
bull At least two passes through the data
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Freight Model Types | |||||||||||||||||||
State | Type | Peer Exc | |||||||||||||||||
OR | 4 | 1 | Count of Type | Count of Type | Peer Exc | ||||||||||||||
WI | 3 | 1 | Type | Total | Type | 0 | 1 | Grand Total | |||||||||||
MO | 1 | 1 | 1 | 2 | 1 | 2 | 2 | 4 | |||||||||||
LA | 3 | 1 | 2 | 2 | 2 | 2 | 2 | 4 | |||||||||||
IN | 3 | 1 | 3 | 6 | 3 | 2 | 6 | 8 | |||||||||||
OH | 4 | 1 | 4 | 2 | 4 | 2 | 2 | ||||||||||||
KY | 3 | 1 | Grand Total | 12 | 3 | 3 | |||||||||||||
FL | 3 | 1 | Grand Total | 9 | 12 | 21 | |||||||||||||
VA | 3 | 1 | None | 2 | Not | Peer Exchange | |||||||||||||
NJ | 2 | 1 | Traditional | 2 | None | 2 | 2 | ||||||||||||
MA | 2 | 1 | Commodity | 6 | Traditional | 2 | 2 | ||||||||||||
NH | 1 | 1 | Econometric | 2 | Commodity | 2 | 6 | ||||||||||||
CA | 0 | Econometric | 2 | ||||||||||||||||
TX | 3 | 0 | Unknown | 3 | |||||||||||||||
MI | 3 | 0 | |||||||||||||||||
GA | 0 | ||||||||||||||||||
SC | 1 | 0 | |||||||||||||||||
DE | 2 | 0 | |||||||||||||||||
CT | 0 | ||||||||||||||||||
VT | 2 | 0 | |||||||||||||||||
ME | 1 | 0 | |||||||||||||||||
MN | 0 |
None | None | ||
Traditional | Traditional | ||
Commodity | Commodity | ||
Econometric | Econometric | ||
Unknown | Unknown |
Current Status of Statewide Models
Operational (28)
Dormant (3)
No model (11)
Developing (8)
Source A Horowitz Statewide Travel Forecasting Models NCHRP Synthesis 358 (2006) with recent updates
Statewide ModelsContext for statewide modeling
MacroscopicMegascopic Mesoscopic Microscopic
Urban travel models (tour-
based or sequential)
Regional model of economic
trade and land use trends
Dynamic traffic assignment to
planning network
Detailed traffic analysis of key facilities and
corridors
Transport costs disutilities and
travel times
Multimodal person and freight tours
Time-sliced demand flows by mode
Congestion indices link and
node delay
Economic triggers for freight flows
Accessibilities and transport costs
Source Federal Highway Administration Pilot Course on Statewide Travel Forecasting 2007
Issues unique to statewide models
bull Explicit linkages to economic models
bull Scale issues Modeling states at an urban scale Expensive to build and maintain
bull Need for interfaces with urban models Consistency Data exchange
bull Lack of standard practicebull Unique components
Freight (commodities rarr vehicles) Long distance travel Economic impacts
Data and knowledge gaps
Source Michigan Department of Transportation
Typical uses of Statewide ModelsLess common usesbull Air quality conformity analysesbull Freight and intermodal
planningbull Traffic impact studiesbull Economic development studiesbull Long-term investment studiesbull Emergency evacuation studiesbull Multi-state corridors (possibly
with multi-state models)
Most commonly citedbull Corridor planning (including
community bypasses)bull Forecasting freighttruck
flowsbull External trip forecasts for
MPO amp regional models bull Expansion of MPO amp regional
models
Source Pennsylvania Department of Transportation TRB presentation on 2006-2030 Statewide Model
Louisianarsquos and Virginiarsquos statewide modelsMarket segments 2000 source amp estimation procedure Forecast source amp estimation
procedure
Macrolevel
1 Interstate and interparishtruck trip table
2000 macro zone level Transearch data some matrix estimation
2030 macro zone forecast of 2000 Transearch data interpolation and extrapolation to forecast other years
2 Long distance interstate business tourist and other auto trip tables
1995 American Travel Survey data disaggregated from state and MSA to county level and Fratar balanced to 2000 using household and employment growth
Fratar balance year 2000 county-level trip tables to forecast year using household and employment growth
3 Long distance intrastate business tourist and other auto trip tables
4 Short distance Interstate Louisiana HBW trip table
1990 Census county level JTW data Fratar balanced to 2000 using household and employment growth
Microlevel
5 Short distance intraparish truck trip table
Synthesized from 2000 intraparish Transearch data
Fratar balance 2000 intraparish trip table to forecast year using household and employment growth6 Short distance intrastate
Louisiana HBW trip table1990 tract-level JTW data Fratar balanced to 2000 using household and employment growth
7 Short distance intrastate HBO and NHB trip tables
Urban model style household-based trip generation and gravity model distribution
Urban style household-based trip generation and distribution models log-linear regression used to estimate external trip ends
Source Virginia Department of Transportation
MicrosimulationActivity-based Approach
bull Models dynamic evolution and actions of discrete agents across time and space
bull Observe emergent behavior Interaction Cooperation Coordination
bull Primary emphasis on understanding the system
bull Maybe forecast it
Between autonomous perceptive and deliberative agents
Source Ohio Department of Transportation
Key Differences ndash Traditional vs MicrosimulationldquoTraditional modelsrdquobull Closed form equilibriumbull Well understoodbull Typically deterministicbull Presumes central tendencies
in underlying databull Well-behavedbull ldquoManageablerdquo variationrdquobull Computationally efficient
Microsimulationbull Disequilibriumbull Large systems NWUbull Stochasticbull Diverse behaviorbull Handles wide variancesbull Accommodates weak to strong
understandingbull Computationally heavy
Zonal Land Use Data
Worker Traveler Generation
Vehicle Assignment
Starting Time Assignment
Post-Processor
Next Stop Purpose Choice
Model
Next Stop Location
Choice Model
Dynamic Activity Pattern Generation
Commercial Vehicle Trip List
Zonal Land Use Data
Worker Traveler Generation
Vehicle Assignment
Starting Time Assignment
Post-Processor
Next Stop Purpose Choice
Model
Next Stop Location
Choice Model
Dynamic Activity Pattern Generation
Commercial Vehicle Trip List
Microsimulation Examplesbull Ohiobull Oregon
Source Ohio Department of Transportation
Statewide Model Current Practicebull Meetings of the TRB Subcommittee on Statewide Travel Forecasting Models
bull NCHRP Report 735 Long-Distance and Rural Transferable Parameters for Statewide Travel Forecasting Models 2012
bull NCHRP Validation and Sensitivity Considerations for Statewide Models September 2010
bull White Paper on Statewide Modeling presented at TRB Statewide Planning Conference September 2008
bull National Travel Demand Forecasting Model Phase I Final Scope 2008
bull Forecasting Statewide Freight Toolkit NCHRP Report 606 2008
bull TRB Annual Meeting Sessions in 2004 2006 and 2008 web page
bull Statewide Travel Forecasting Models NCHRP Synthesis 358 2006
bull Statewide Travel Demand Modeling Peer Exchange Transportation Research Circular E-C075 September 2004
bull Guidebook on Statewide Travel Forecasting and Irvine Conference Transportation Research Circular E-C011 1999
Topic ndash Statewide Model Theory
bull Common themes in statewide model development
bull Alternative structures Trip table estimation (2 step) Highway model structure (3 step) Fully multimodal (4 step) Beyond 4 step
bull Deciding on best approach
bull Data development challengesSource Massachusetts Highway DeptCambridge Systematics Inc
Common Themes in Review of Statewide Models
bull Most statewide models use similar methodologies to metropolitan models but there are many variations
bull The largest problems relate to issues of scale
bull Models from different states vary greatly in complexity cost and development time
bull Models are most successful when addressing statewide priorities
bull There are major deficiencies in available data about long-distance and rural passenger travelhellip NCHRP Report 735 has helped
bull Statewide models are more compatible with secondary freight data sources such as the Commodity Flow Survey than are metropolitan models
bull Interest is high among states in progress in deploying models
Source A Horowitz White Paper on Statewide Modeling presented at TRB Conference on Statewide Planning 2008
Alternative Statewide Model Structures
bull Trip table estimation (2 step) Estimate initial ldquoseed matrixrdquo
Roadside intercept surveys American Travel Survey National HH Travel SurveyAdd-ons Census Journey-to-Work flows
Matrix estimation program Inputs ndash seed matrix highway
network traffic counts Adjusts origin-destination flows
between zones to optimize matching network traffic counts
Fratar factors applied to synthetic trip table in estimating future growth
bull Trip table estimation (2 step) Pros
Limited need to prepare aggregate socioeconomic input data
Simplicity of model generally means short execution times
Forecasting needs are limited to zone or district growth factors
Cons Less explanatory power than most
MPOregional models Fixed distribution pattern Difficulties in assessing benefits of
adding new roadway capacity
Alternative Statewide Model Structures (Contrsquod)
bull Highway model structure (3 step) Trip generation Trip distribution Highway assignment
bull Mode choice excluded Either generate vehicle trips or Apply auto occupancy factors to person trip
tables
bull Pros Benefits of trip redistribution across range of
alternatives Greater sensitivity to roadway capacity and
congestion
bull Cons Likely reliance on MPOsothers to provide
socioeconomic input data Canrsquot fully test transportation strategies that
would alter mode split
Source Florida Dept of TransportationCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)
bull Freight considerations (3 step) Typical freight model steps
Trip generation ndash tons by commodity group
Trip distribution Mode choice ndash freight is
apportioned to transport modes
bull Joint highway assignment (trucks and passenger trips)
Source Florida Dept of TransportationCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)bull Full multi-modal structure (4 step)
Trip generation Trip distribution Mode choice Highway assignment
bull More common with mega-regions Focused on fixed guideway systems Intercity rail is important component
bull Pros Potential to model new intercity rail
including high speed rail Test greater range of potential
transportation solutions
bull Cons Requires coding of transit networks and
mode choice estimation Modeling of air travel might be needed
especially if intercity rail is a completely new mode of travel
Source California HSR AuthorityCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)
Beyond 4 Stepbull Potential components
Implement destination choice Tour-basedtrip chains Activity-based travel model Activity micro-simulation Traffic micro-simulation Integration with land use models Integration with economic models
bull Pros Ability to test a more complete set of
transportation strategies and solutions
bull Cons Extensive data requirements Cost to maintain Complexity
Source Oregon Dept of TransportationPBUniversity of Calgary
Deciding on Best Statewide Model Approach
bull What are the motivations for building a statewide travel demand modelbull What are the emerging transportation and policy issues to addressbull How significant is the impact of through travel on your statebull Do you have many multi-state MPOsbull How reliable (and current) are your survey databull Does your state have major international ports or border crossings
StudyArea
UrbanModel
AreaStudyArea
UrbanModelArea 1
Study AreaWithin OneModel Area
UrbanModelArea 2
Study AreasWithin Two Model Areas
StudyArea
UrbanModelArea
StudyArea
Statewide Model
Study AreaOutside Urban
Model Areas
Statewide ModelStatewide Model
Source Florida Department of TransportationCambridge Systematics Inc
Figure 317 Subarea Study Area Examples
Study
Area
Urban
Model
Area
Study
Area
Urban
Model
Area 1
Study Area
Within One
Model Area
Urban
Model
Area 2
Study Areas
Within Two
Model Areas
Study
Area
Urban
Model
Area
Study
Area
Statewide Model
Study Area
Outside Urban
Model Areas
Statewide Model
Statewide Model
Statewide Model Data Development Challenges
bull Primary data deficiencies Long distance travelers Rural resident trip-making
bull Challenges in compiling required data Age of data Confidentiality Coordination with data providers Proper level of geography Sufficient sample size
bull Four categories of data for discussion Networks and zone system Demographic and employment data Freight and other economic data NHTS and other behavioral data
Source Florida Department of TransportationCambridge Systematics Inc
Data Challengeshellipwhat is ldquolong distancerdquo
bull gt50 miles (American Traveler Survey)
bull gt100 miles (National Household Travel Survey)
bull IE trips (several states)
bull Trips made by residents for business personal business
bull Recreationaltourist trips
Topic ndash Similarities amp Differences with Urban Models
bull Zone systems
bull Networks
bull Travel behavior assumptions
bull Trip purposesmarket segments
bull Modal issues
bull Truck forecasting
Source Delaware Dept of TransportationWRampA Inc
Statewide vs Urban Model Zone Systems
bull Usually larger zones in statewide models than MPOs
bull Block groups are good starting point for statewide model TAZs
bull Block groups might need splitting along major transportation corridors
bull In small states MPO TAZs could equal SWM zones
bull In most states MPO TAZ equivalency tables will likely be neededdesirable
bull FAF disaggregation approach used for statewide model TAZs
Source FHWA Office of Freight Management and Operations
Source Florida Dept of TransportationCambridge Systematics Inc
FAF zones
Block groups = TAZs
Too small for statewide zones
Too big for statewide zones
Statewide vs Urban Model Networksbull Statewide model networks are sparse in
comparison with MPO models
bull At a minimum statewide networks should include Limited access highways Arterials Major collectors
bull For proper circulation patterns lower class facilities might be added
bull Outside state a very sparse network is needed for major truck corridors
bull Consistent level of detail for networks and zones
Source Tennessee Dept of TransportationCambridge Systematics Inc
MPO network
statewide network
Statewide vs Urban Model Travel Behaviorbull Understanding urban travel behavior
MPO models are usually developed with local household travel survey data
MPOs without recent HH survey data can borrow parameters from other MPO models or guidance documents such as NCHRP 716 Transferable Urban Travel Parameters
bull Rural and long-distance travel behavior A limited number of states have conducted
statewide household travel surveys American Travel Survey was designed to target
long-distance trips but was discontinued long ago
National Household Travel Survey Add-ons can be structured for a sufficient sample of rural trip-makers
NCHRP 735 Rural and Long-Distance Travel Parameters recently published for use in statewide models
Map of 2009 Add-on Surveys Source FHWA
Statewide vs Urban Model Trip Purposes
bull Typical MPO trip purposes are usually collapsed into HBW HBOHBNW and NHB for statewide models
bull Long-distance purposes Business Tourist Personal business
bull Trucks Commodity group is often
surrogate for trip purpose
bull External trips External generally means trip from
outside state Might exclude ldquohalordquo area
surrounding state boundary
Trip PurposeFL
2000 IN LA MA MS RI
Home-based Work 1914 1995 1900 1841 1606 2000
Home-based Nonwork 5547 5208 4770 5206 4917 5700
Home-based Other 2792 5208 4770 1938 4917 5700
Home-based Shop 1576 ndash ndash 1173 ndash ndash
Home-based SocialRecreation
1179 ndash ndash 1395 ndash ndash
Home-base RecreationShop ndash ndash ndash ndash ndash ndash
Home-based School ndash ndash ndash 700 ndash ndash
Nonhome-based 2539 2686 3100 2468 3289 2300
Nonhome-based Work ndash ndash ndash 1027 ndash ndash
Nonhome-based Nonwork ndash ndash ndash 1441 ndash ndash
Other ndash ndash ndash ndash ndash
Truck ndash ndash 170 ndash ndash ndash
Long-Distance 000 111 060 417 022 000
Long-Distance Business ndash ndash 020 284 011 ndash
Long-Distance Personal Business
ndash ndash ndash ndash ndash ndash
Long-Distance Tourist ndash ndash 010 027 000 ndash
Long-Distance Work ndash ndash ndash 105 ndash ndash
Long-Distance NonworkOther
ndash ndash 030 ndash 011 ndash
Statewide vs Urban Model Trip Purposes (Contrsquod)
Number of person trips
Mode Pleasure Business Personal business Other Total Percent
Personal vehicle 12185 5708 3120 789 21802 908Commercial air 1012 746 151 24 1933 8Train 74 150 14 0 238 1Other 4 2 5 12 23 01Ship 19 0 0 0 19 01
Total 13294 6606 3290 825 24015 100Percent 554 275 137 34 100
Percent by trip purpose
Mode Pleasure Business Personal business Other
Personal vehicle 917 864 948 956Commercial air 76 113 46 29Train 06 23 04 0Other 0 0 02 15Ship 01 0 0 0
Trip purpose
Trip purpose
Statewide vs Urban Model Modal Issuesbull Most statewide models are limited to highway trips
bull Modeling of local bus routes is likely overkill
bull Mode choice is usually just a conversion from person trips to vehicle trips via auto occupancy rates
bull Transit components of statewide models are more common in regions with history of fixed guideway transit
bull Transit model is needed to test options such as high speedintercity rail
ATS 95 NHTS 01 MDOT 03-04 ODOT 02-03Personal vehicle 81 89 89 88Commercial air 16 8 10 8Bus 2 2 04 19Train 05 01 0 005Other 05 09 06 205
Survey and yearPrimary mode of transport
Statewide vs Urban Model Modal Issues (Contrsquod)
3727 24 22
32
29 36
10
21
28 22
10
22
12
1529
11
44 43
0
10
20
30
40
50
60
70
80
90
100
Personalvehicle
Bus Train Marine Commercialair
Perc
ent o
f per
son
trips
lt300 300-499 500-599 1000-1999 gt2000
68
85 85 82 85
30
12 11 15 13
0
10
20
30
40
50
60
70
80
90
100
Busi-ness
Plea-sure
pBusi-ness
Vaca-tion
Week-end
Perc
ent o
f per
son
trips
Personal vehicle Commercial air All other
Mode by Trip Purpose Mode by Distance (miles)
Statewide vs Urban Model Modal Issues (Contrsquod)
bull Most statewide assignment models separate trucks from passenger cars Preloading process commonly used for trucks Takes into account more limited route options for truck travel
Source Ohio Dept of Transportation
Statewide vs Urban Model Truck Forecasting
bull Urban modelsndash Freight movements are partially
ldquoexternalrdquo to model (truncated at external zone)
ndash More general categories of truckcommercial trips used
ndash Strong reliance on Quick Response Freight Manual
bull Statewide modelsndash Rooted in commodity flows ndash Often use econometric
approaches to modelingndash Generally model entire length of
freight travel
Freight Model Type by Peer Exchange Participants and
NonParticipants
0123456789
None
Traditio
nal
Commod
ity
Econo
metric
Unkno
wn
NotPeer Exchange
Source TRB Electronic Circular E-C075 Statewide Travel Demand Modeling ndash A Peer Exchange
Topic ndash Data Needs and Considerations
bull Impact of model complexity on data needs
bull Overview of data sources
bull Long-Distance travel gt different focus than MPOregional models
Sources NHTS Weight calculation Pitfalls Key statistics European surveys Lessons learned
Impact of Model Complexity on Data Needsbull Trip table estimation (2 step) ndash highway network
zone system trip table
bull Highway model structure (3 step) ndash 2 step plusbehavioral data for trip rates friction factors auto occupancy rates
bull Freight considerations (3 step) ndash 3 step pluscommodity flows economic data tonnages by mode truck counts
bull Full multi-modal structure (4 step) ndash 3 step plustransit networks and paths logit mode choice model
bull Beyond 4 Step ndash synthetic household generation operational characteristics etc
Source Georgia DOTCambridge Systematics Inc
Statewide Model Data Sources A-Zbull American Travel Survey (ATS)
bull Bureau of Economic Analysis (BEA)
bull Census Journey-to-Work (JTW) data
bull Census Public Use Microsample (PUMS)
bull Census Transportation Planning Package (CTPP)
bull Claritas
bull Commodity Flow Survey (CFS)
Source US Census Bureau
Statewide Model Data Sources (Contrsquod)bull Dun amp Bradstreet (DampB) employment data
bull Employment Security data (ES-202)
bull Freight Analysis Framework (FAF)
bull Highway Performance Monitoring System (HPMS)
bull InfoUSA employment data
bull Metropolitan Planning Organizations (MPO)
bull NCHRP 365 Manual
bull National Highway Planning Network (NHPN)
Source FAF2 Provisional Commodity Origin-Destination Estimates Revised Methodology Report BattelleFHWA
Statewide Model Data Sources (Contrsquod)
bull National Household Travel Survey (NHTS)
bull Origin-destination surveys
bull Regional Economic Models Inc (REMI)
bull State Departments of Transportation
bull TRANSEARCH data
bull Vehicle Use and Identification Survey (VIUS)
bull Woods amp Poole employment forecasts
Source Bureau of Transportation Statistics NHTS
Data Sources on Long-Distance Travel
bull Census Journey-to-Work databull Long distance component
American Travel Survey (77 95) National Household Travel Survey (01) Ohio statewide surveys (02-04) Michigan statewide survey (04-05) California (10-12)
bull Surveys focused on long distance Oregon (96-98 07-08)
bull Others BTS Omnibus surveys Tourism surveys European practice NCHRP Report 735
NHTS and Long-Distance Travelbull Evolution of US national surveys
NPTS (69 77 83 90 95) ATS (77 95)
bull 2001 NHTS ndash Total of 60282 long distance person trips Data collected March 01 to May 02 22204 (37 percent) before 911 38078 (63 percent) afterwards
bull Long distance trip is longer than 50 miles one-way (vs 100 miles in 1995 ATS)
bull 28-day retrospective survey (including travel day)
bull Not your typical travelers Low frequency Distinct segment of the population
Weight Calculation for Long-Distance Travelers
bull Use to ensure valid population-level estimates Accounts for non-response under-coverage multiple phones etc
bull Multi-stage process Match national margins Adjust for data splitting (force sumexpfac = 277208169) Raking (match independent controls) race ethnicity race by
month ethnicity by month sex by age census region MSA status month by day of week
Trim outlier expansion factors max expfac = 4 times mean Replication (correct for sampling error)
bull At least two passes through the data
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Freight Model Types | |||||||||||||||||||
State | Type | Peer Exc | |||||||||||||||||
OR | 4 | 1 | Count of Type | Count of Type | Peer Exc | ||||||||||||||
WI | 3 | 1 | Type | Total | Type | 0 | 1 | Grand Total | |||||||||||
MO | 1 | 1 | 1 | 2 | 1 | 2 | 2 | 4 | |||||||||||
LA | 3 | 1 | 2 | 2 | 2 | 2 | 2 | 4 | |||||||||||
IN | 3 | 1 | 3 | 6 | 3 | 2 | 6 | 8 | |||||||||||
OH | 4 | 1 | 4 | 2 | 4 | 2 | 2 | ||||||||||||
KY | 3 | 1 | Grand Total | 12 | 3 | 3 | |||||||||||||
FL | 3 | 1 | Grand Total | 9 | 12 | 21 | |||||||||||||
VA | 3 | 1 | None | 2 | Not | Peer Exchange | |||||||||||||
NJ | 2 | 1 | Traditional | 2 | None | 2 | 2 | ||||||||||||
MA | 2 | 1 | Commodity | 6 | Traditional | 2 | 2 | ||||||||||||
NH | 1 | 1 | Econometric | 2 | Commodity | 2 | 6 | ||||||||||||
CA | 0 | Econometric | 2 | ||||||||||||||||
TX | 3 | 0 | Unknown | 3 | |||||||||||||||
MI | 3 | 0 | |||||||||||||||||
GA | 0 | ||||||||||||||||||
SC | 1 | 0 | |||||||||||||||||
DE | 2 | 0 | |||||||||||||||||
CT | 0 | ||||||||||||||||||
VT | 2 | 0 | |||||||||||||||||
ME | 1 | 0 | |||||||||||||||||
MN | 0 |
None | None | ||
Traditional | Traditional | ||
Commodity | Commodity | ||
Econometric | Econometric | ||
Unknown | Unknown |
Statewide ModelsContext for statewide modeling
MacroscopicMegascopic Mesoscopic Microscopic
Urban travel models (tour-
based or sequential)
Regional model of economic
trade and land use trends
Dynamic traffic assignment to
planning network
Detailed traffic analysis of key facilities and
corridors
Transport costs disutilities and
travel times
Multimodal person and freight tours
Time-sliced demand flows by mode
Congestion indices link and
node delay
Economic triggers for freight flows
Accessibilities and transport costs
Source Federal Highway Administration Pilot Course on Statewide Travel Forecasting 2007
Issues unique to statewide models
bull Explicit linkages to economic models
bull Scale issues Modeling states at an urban scale Expensive to build and maintain
bull Need for interfaces with urban models Consistency Data exchange
bull Lack of standard practicebull Unique components
Freight (commodities rarr vehicles) Long distance travel Economic impacts
Data and knowledge gaps
Source Michigan Department of Transportation
Typical uses of Statewide ModelsLess common usesbull Air quality conformity analysesbull Freight and intermodal
planningbull Traffic impact studiesbull Economic development studiesbull Long-term investment studiesbull Emergency evacuation studiesbull Multi-state corridors (possibly
with multi-state models)
Most commonly citedbull Corridor planning (including
community bypasses)bull Forecasting freighttruck
flowsbull External trip forecasts for
MPO amp regional models bull Expansion of MPO amp regional
models
Source Pennsylvania Department of Transportation TRB presentation on 2006-2030 Statewide Model
Louisianarsquos and Virginiarsquos statewide modelsMarket segments 2000 source amp estimation procedure Forecast source amp estimation
procedure
Macrolevel
1 Interstate and interparishtruck trip table
2000 macro zone level Transearch data some matrix estimation
2030 macro zone forecast of 2000 Transearch data interpolation and extrapolation to forecast other years
2 Long distance interstate business tourist and other auto trip tables
1995 American Travel Survey data disaggregated from state and MSA to county level and Fratar balanced to 2000 using household and employment growth
Fratar balance year 2000 county-level trip tables to forecast year using household and employment growth
3 Long distance intrastate business tourist and other auto trip tables
4 Short distance Interstate Louisiana HBW trip table
1990 Census county level JTW data Fratar balanced to 2000 using household and employment growth
Microlevel
5 Short distance intraparish truck trip table
Synthesized from 2000 intraparish Transearch data
Fratar balance 2000 intraparish trip table to forecast year using household and employment growth6 Short distance intrastate
Louisiana HBW trip table1990 tract-level JTW data Fratar balanced to 2000 using household and employment growth
7 Short distance intrastate HBO and NHB trip tables
Urban model style household-based trip generation and gravity model distribution
Urban style household-based trip generation and distribution models log-linear regression used to estimate external trip ends
Source Virginia Department of Transportation
MicrosimulationActivity-based Approach
bull Models dynamic evolution and actions of discrete agents across time and space
bull Observe emergent behavior Interaction Cooperation Coordination
bull Primary emphasis on understanding the system
bull Maybe forecast it
Between autonomous perceptive and deliberative agents
Source Ohio Department of Transportation
Key Differences ndash Traditional vs MicrosimulationldquoTraditional modelsrdquobull Closed form equilibriumbull Well understoodbull Typically deterministicbull Presumes central tendencies
in underlying databull Well-behavedbull ldquoManageablerdquo variationrdquobull Computationally efficient
Microsimulationbull Disequilibriumbull Large systems NWUbull Stochasticbull Diverse behaviorbull Handles wide variancesbull Accommodates weak to strong
understandingbull Computationally heavy
Zonal Land Use Data
Worker Traveler Generation
Vehicle Assignment
Starting Time Assignment
Post-Processor
Next Stop Purpose Choice
Model
Next Stop Location
Choice Model
Dynamic Activity Pattern Generation
Commercial Vehicle Trip List
Zonal Land Use Data
Worker Traveler Generation
Vehicle Assignment
Starting Time Assignment
Post-Processor
Next Stop Purpose Choice
Model
Next Stop Location
Choice Model
Dynamic Activity Pattern Generation
Commercial Vehicle Trip List
Microsimulation Examplesbull Ohiobull Oregon
Source Ohio Department of Transportation
Statewide Model Current Practicebull Meetings of the TRB Subcommittee on Statewide Travel Forecasting Models
bull NCHRP Report 735 Long-Distance and Rural Transferable Parameters for Statewide Travel Forecasting Models 2012
bull NCHRP Validation and Sensitivity Considerations for Statewide Models September 2010
bull White Paper on Statewide Modeling presented at TRB Statewide Planning Conference September 2008
bull National Travel Demand Forecasting Model Phase I Final Scope 2008
bull Forecasting Statewide Freight Toolkit NCHRP Report 606 2008
bull TRB Annual Meeting Sessions in 2004 2006 and 2008 web page
bull Statewide Travel Forecasting Models NCHRP Synthesis 358 2006
bull Statewide Travel Demand Modeling Peer Exchange Transportation Research Circular E-C075 September 2004
bull Guidebook on Statewide Travel Forecasting and Irvine Conference Transportation Research Circular E-C011 1999
Topic ndash Statewide Model Theory
bull Common themes in statewide model development
bull Alternative structures Trip table estimation (2 step) Highway model structure (3 step) Fully multimodal (4 step) Beyond 4 step
bull Deciding on best approach
bull Data development challengesSource Massachusetts Highway DeptCambridge Systematics Inc
Common Themes in Review of Statewide Models
bull Most statewide models use similar methodologies to metropolitan models but there are many variations
bull The largest problems relate to issues of scale
bull Models from different states vary greatly in complexity cost and development time
bull Models are most successful when addressing statewide priorities
bull There are major deficiencies in available data about long-distance and rural passenger travelhellip NCHRP Report 735 has helped
bull Statewide models are more compatible with secondary freight data sources such as the Commodity Flow Survey than are metropolitan models
bull Interest is high among states in progress in deploying models
Source A Horowitz White Paper on Statewide Modeling presented at TRB Conference on Statewide Planning 2008
Alternative Statewide Model Structures
bull Trip table estimation (2 step) Estimate initial ldquoseed matrixrdquo
Roadside intercept surveys American Travel Survey National HH Travel SurveyAdd-ons Census Journey-to-Work flows
Matrix estimation program Inputs ndash seed matrix highway
network traffic counts Adjusts origin-destination flows
between zones to optimize matching network traffic counts
Fratar factors applied to synthetic trip table in estimating future growth
bull Trip table estimation (2 step) Pros
Limited need to prepare aggregate socioeconomic input data
Simplicity of model generally means short execution times
Forecasting needs are limited to zone or district growth factors
Cons Less explanatory power than most
MPOregional models Fixed distribution pattern Difficulties in assessing benefits of
adding new roadway capacity
Alternative Statewide Model Structures (Contrsquod)
bull Highway model structure (3 step) Trip generation Trip distribution Highway assignment
bull Mode choice excluded Either generate vehicle trips or Apply auto occupancy factors to person trip
tables
bull Pros Benefits of trip redistribution across range of
alternatives Greater sensitivity to roadway capacity and
congestion
bull Cons Likely reliance on MPOsothers to provide
socioeconomic input data Canrsquot fully test transportation strategies that
would alter mode split
Source Florida Dept of TransportationCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)
bull Freight considerations (3 step) Typical freight model steps
Trip generation ndash tons by commodity group
Trip distribution Mode choice ndash freight is
apportioned to transport modes
bull Joint highway assignment (trucks and passenger trips)
Source Florida Dept of TransportationCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)bull Full multi-modal structure (4 step)
Trip generation Trip distribution Mode choice Highway assignment
bull More common with mega-regions Focused on fixed guideway systems Intercity rail is important component
bull Pros Potential to model new intercity rail
including high speed rail Test greater range of potential
transportation solutions
bull Cons Requires coding of transit networks and
mode choice estimation Modeling of air travel might be needed
especially if intercity rail is a completely new mode of travel
Source California HSR AuthorityCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)
Beyond 4 Stepbull Potential components
Implement destination choice Tour-basedtrip chains Activity-based travel model Activity micro-simulation Traffic micro-simulation Integration with land use models Integration with economic models
bull Pros Ability to test a more complete set of
transportation strategies and solutions
bull Cons Extensive data requirements Cost to maintain Complexity
Source Oregon Dept of TransportationPBUniversity of Calgary
Deciding on Best Statewide Model Approach
bull What are the motivations for building a statewide travel demand modelbull What are the emerging transportation and policy issues to addressbull How significant is the impact of through travel on your statebull Do you have many multi-state MPOsbull How reliable (and current) are your survey databull Does your state have major international ports or border crossings
StudyArea
UrbanModel
AreaStudyArea
UrbanModelArea 1
Study AreaWithin OneModel Area
UrbanModelArea 2
Study AreasWithin Two Model Areas
StudyArea
UrbanModelArea
StudyArea
Statewide Model
Study AreaOutside Urban
Model Areas
Statewide ModelStatewide Model
Source Florida Department of TransportationCambridge Systematics Inc
Figure 317 Subarea Study Area Examples
Study
Area
Urban
Model
Area
Study
Area
Urban
Model
Area 1
Study Area
Within One
Model Area
Urban
Model
Area 2
Study Areas
Within Two
Model Areas
Study
Area
Urban
Model
Area
Study
Area
Statewide Model
Study Area
Outside Urban
Model Areas
Statewide Model
Statewide Model
Statewide Model Data Development Challenges
bull Primary data deficiencies Long distance travelers Rural resident trip-making
bull Challenges in compiling required data Age of data Confidentiality Coordination with data providers Proper level of geography Sufficient sample size
bull Four categories of data for discussion Networks and zone system Demographic and employment data Freight and other economic data NHTS and other behavioral data
Source Florida Department of TransportationCambridge Systematics Inc
Data Challengeshellipwhat is ldquolong distancerdquo
bull gt50 miles (American Traveler Survey)
bull gt100 miles (National Household Travel Survey)
bull IE trips (several states)
bull Trips made by residents for business personal business
bull Recreationaltourist trips
Topic ndash Similarities amp Differences with Urban Models
bull Zone systems
bull Networks
bull Travel behavior assumptions
bull Trip purposesmarket segments
bull Modal issues
bull Truck forecasting
Source Delaware Dept of TransportationWRampA Inc
Statewide vs Urban Model Zone Systems
bull Usually larger zones in statewide models than MPOs
bull Block groups are good starting point for statewide model TAZs
bull Block groups might need splitting along major transportation corridors
bull In small states MPO TAZs could equal SWM zones
bull In most states MPO TAZ equivalency tables will likely be neededdesirable
bull FAF disaggregation approach used for statewide model TAZs
Source FHWA Office of Freight Management and Operations
Source Florida Dept of TransportationCambridge Systematics Inc
FAF zones
Block groups = TAZs
Too small for statewide zones
Too big for statewide zones
Statewide vs Urban Model Networksbull Statewide model networks are sparse in
comparison with MPO models
bull At a minimum statewide networks should include Limited access highways Arterials Major collectors
bull For proper circulation patterns lower class facilities might be added
bull Outside state a very sparse network is needed for major truck corridors
bull Consistent level of detail for networks and zones
Source Tennessee Dept of TransportationCambridge Systematics Inc
MPO network
statewide network
Statewide vs Urban Model Travel Behaviorbull Understanding urban travel behavior
MPO models are usually developed with local household travel survey data
MPOs without recent HH survey data can borrow parameters from other MPO models or guidance documents such as NCHRP 716 Transferable Urban Travel Parameters
bull Rural and long-distance travel behavior A limited number of states have conducted
statewide household travel surveys American Travel Survey was designed to target
long-distance trips but was discontinued long ago
National Household Travel Survey Add-ons can be structured for a sufficient sample of rural trip-makers
NCHRP 735 Rural and Long-Distance Travel Parameters recently published for use in statewide models
Map of 2009 Add-on Surveys Source FHWA
Statewide vs Urban Model Trip Purposes
bull Typical MPO trip purposes are usually collapsed into HBW HBOHBNW and NHB for statewide models
bull Long-distance purposes Business Tourist Personal business
bull Trucks Commodity group is often
surrogate for trip purpose
bull External trips External generally means trip from
outside state Might exclude ldquohalordquo area
surrounding state boundary
Trip PurposeFL
2000 IN LA MA MS RI
Home-based Work 1914 1995 1900 1841 1606 2000
Home-based Nonwork 5547 5208 4770 5206 4917 5700
Home-based Other 2792 5208 4770 1938 4917 5700
Home-based Shop 1576 ndash ndash 1173 ndash ndash
Home-based SocialRecreation
1179 ndash ndash 1395 ndash ndash
Home-base RecreationShop ndash ndash ndash ndash ndash ndash
Home-based School ndash ndash ndash 700 ndash ndash
Nonhome-based 2539 2686 3100 2468 3289 2300
Nonhome-based Work ndash ndash ndash 1027 ndash ndash
Nonhome-based Nonwork ndash ndash ndash 1441 ndash ndash
Other ndash ndash ndash ndash ndash
Truck ndash ndash 170 ndash ndash ndash
Long-Distance 000 111 060 417 022 000
Long-Distance Business ndash ndash 020 284 011 ndash
Long-Distance Personal Business
ndash ndash ndash ndash ndash ndash
Long-Distance Tourist ndash ndash 010 027 000 ndash
Long-Distance Work ndash ndash ndash 105 ndash ndash
Long-Distance NonworkOther
ndash ndash 030 ndash 011 ndash
Statewide vs Urban Model Trip Purposes (Contrsquod)
Number of person trips
Mode Pleasure Business Personal business Other Total Percent
Personal vehicle 12185 5708 3120 789 21802 908Commercial air 1012 746 151 24 1933 8Train 74 150 14 0 238 1Other 4 2 5 12 23 01Ship 19 0 0 0 19 01
Total 13294 6606 3290 825 24015 100Percent 554 275 137 34 100
Percent by trip purpose
Mode Pleasure Business Personal business Other
Personal vehicle 917 864 948 956Commercial air 76 113 46 29Train 06 23 04 0Other 0 0 02 15Ship 01 0 0 0
Trip purpose
Trip purpose
Statewide vs Urban Model Modal Issuesbull Most statewide models are limited to highway trips
bull Modeling of local bus routes is likely overkill
bull Mode choice is usually just a conversion from person trips to vehicle trips via auto occupancy rates
bull Transit components of statewide models are more common in regions with history of fixed guideway transit
bull Transit model is needed to test options such as high speedintercity rail
ATS 95 NHTS 01 MDOT 03-04 ODOT 02-03Personal vehicle 81 89 89 88Commercial air 16 8 10 8Bus 2 2 04 19Train 05 01 0 005Other 05 09 06 205
Survey and yearPrimary mode of transport
Statewide vs Urban Model Modal Issues (Contrsquod)
3727 24 22
32
29 36
10
21
28 22
10
22
12
1529
11
44 43
0
10
20
30
40
50
60
70
80
90
100
Personalvehicle
Bus Train Marine Commercialair
Perc
ent o
f per
son
trips
lt300 300-499 500-599 1000-1999 gt2000
68
85 85 82 85
30
12 11 15 13
0
10
20
30
40
50
60
70
80
90
100
Busi-ness
Plea-sure
pBusi-ness
Vaca-tion
Week-end
Perc
ent o
f per
son
trips
Personal vehicle Commercial air All other
Mode by Trip Purpose Mode by Distance (miles)
Statewide vs Urban Model Modal Issues (Contrsquod)
bull Most statewide assignment models separate trucks from passenger cars Preloading process commonly used for trucks Takes into account more limited route options for truck travel
Source Ohio Dept of Transportation
Statewide vs Urban Model Truck Forecasting
bull Urban modelsndash Freight movements are partially
ldquoexternalrdquo to model (truncated at external zone)
ndash More general categories of truckcommercial trips used
ndash Strong reliance on Quick Response Freight Manual
bull Statewide modelsndash Rooted in commodity flows ndash Often use econometric
approaches to modelingndash Generally model entire length of
freight travel
Freight Model Type by Peer Exchange Participants and
NonParticipants
0123456789
None
Traditio
nal
Commod
ity
Econo
metric
Unkno
wn
NotPeer Exchange
Source TRB Electronic Circular E-C075 Statewide Travel Demand Modeling ndash A Peer Exchange
Topic ndash Data Needs and Considerations
bull Impact of model complexity on data needs
bull Overview of data sources
bull Long-Distance travel gt different focus than MPOregional models
Sources NHTS Weight calculation Pitfalls Key statistics European surveys Lessons learned
Impact of Model Complexity on Data Needsbull Trip table estimation (2 step) ndash highway network
zone system trip table
bull Highway model structure (3 step) ndash 2 step plusbehavioral data for trip rates friction factors auto occupancy rates
bull Freight considerations (3 step) ndash 3 step pluscommodity flows economic data tonnages by mode truck counts
bull Full multi-modal structure (4 step) ndash 3 step plustransit networks and paths logit mode choice model
bull Beyond 4 Step ndash synthetic household generation operational characteristics etc
Source Georgia DOTCambridge Systematics Inc
Statewide Model Data Sources A-Zbull American Travel Survey (ATS)
bull Bureau of Economic Analysis (BEA)
bull Census Journey-to-Work (JTW) data
bull Census Public Use Microsample (PUMS)
bull Census Transportation Planning Package (CTPP)
bull Claritas
bull Commodity Flow Survey (CFS)
Source US Census Bureau
Statewide Model Data Sources (Contrsquod)bull Dun amp Bradstreet (DampB) employment data
bull Employment Security data (ES-202)
bull Freight Analysis Framework (FAF)
bull Highway Performance Monitoring System (HPMS)
bull InfoUSA employment data
bull Metropolitan Planning Organizations (MPO)
bull NCHRP 365 Manual
bull National Highway Planning Network (NHPN)
Source FAF2 Provisional Commodity Origin-Destination Estimates Revised Methodology Report BattelleFHWA
Statewide Model Data Sources (Contrsquod)
bull National Household Travel Survey (NHTS)
bull Origin-destination surveys
bull Regional Economic Models Inc (REMI)
bull State Departments of Transportation
bull TRANSEARCH data
bull Vehicle Use and Identification Survey (VIUS)
bull Woods amp Poole employment forecasts
Source Bureau of Transportation Statistics NHTS
Data Sources on Long-Distance Travel
bull Census Journey-to-Work databull Long distance component
American Travel Survey (77 95) National Household Travel Survey (01) Ohio statewide surveys (02-04) Michigan statewide survey (04-05) California (10-12)
bull Surveys focused on long distance Oregon (96-98 07-08)
bull Others BTS Omnibus surveys Tourism surveys European practice NCHRP Report 735
NHTS and Long-Distance Travelbull Evolution of US national surveys
NPTS (69 77 83 90 95) ATS (77 95)
bull 2001 NHTS ndash Total of 60282 long distance person trips Data collected March 01 to May 02 22204 (37 percent) before 911 38078 (63 percent) afterwards
bull Long distance trip is longer than 50 miles one-way (vs 100 miles in 1995 ATS)
bull 28-day retrospective survey (including travel day)
bull Not your typical travelers Low frequency Distinct segment of the population
Weight Calculation for Long-Distance Travelers
bull Use to ensure valid population-level estimates Accounts for non-response under-coverage multiple phones etc
bull Multi-stage process Match national margins Adjust for data splitting (force sumexpfac = 277208169) Raking (match independent controls) race ethnicity race by
month ethnicity by month sex by age census region MSA status month by day of week
Trim outlier expansion factors max expfac = 4 times mean Replication (correct for sampling error)
bull At least two passes through the data
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Freight Model Types | |||||||||||||||||||
State | Type | Peer Exc | |||||||||||||||||
OR | 4 | 1 | Count of Type | Count of Type | Peer Exc | ||||||||||||||
WI | 3 | 1 | Type | Total | Type | 0 | 1 | Grand Total | |||||||||||
MO | 1 | 1 | 1 | 2 | 1 | 2 | 2 | 4 | |||||||||||
LA | 3 | 1 | 2 | 2 | 2 | 2 | 2 | 4 | |||||||||||
IN | 3 | 1 | 3 | 6 | 3 | 2 | 6 | 8 | |||||||||||
OH | 4 | 1 | 4 | 2 | 4 | 2 | 2 | ||||||||||||
KY | 3 | 1 | Grand Total | 12 | 3 | 3 | |||||||||||||
FL | 3 | 1 | Grand Total | 9 | 12 | 21 | |||||||||||||
VA | 3 | 1 | None | 2 | Not | Peer Exchange | |||||||||||||
NJ | 2 | 1 | Traditional | 2 | None | 2 | 2 | ||||||||||||
MA | 2 | 1 | Commodity | 6 | Traditional | 2 | 2 | ||||||||||||
NH | 1 | 1 | Econometric | 2 | Commodity | 2 | 6 | ||||||||||||
CA | 0 | Econometric | 2 | ||||||||||||||||
TX | 3 | 0 | Unknown | 3 | |||||||||||||||
MI | 3 | 0 | |||||||||||||||||
GA | 0 | ||||||||||||||||||
SC | 1 | 0 | |||||||||||||||||
DE | 2 | 0 | |||||||||||||||||
CT | 0 | ||||||||||||||||||
VT | 2 | 0 | |||||||||||||||||
ME | 1 | 0 | |||||||||||||||||
MN | 0 |
None | None | ||
Traditional | Traditional | ||
Commodity | Commodity | ||
Econometric | Econometric | ||
Unknown | Unknown |
Issues unique to statewide models
bull Explicit linkages to economic models
bull Scale issues Modeling states at an urban scale Expensive to build and maintain
bull Need for interfaces with urban models Consistency Data exchange
bull Lack of standard practicebull Unique components
Freight (commodities rarr vehicles) Long distance travel Economic impacts
Data and knowledge gaps
Source Michigan Department of Transportation
Typical uses of Statewide ModelsLess common usesbull Air quality conformity analysesbull Freight and intermodal
planningbull Traffic impact studiesbull Economic development studiesbull Long-term investment studiesbull Emergency evacuation studiesbull Multi-state corridors (possibly
with multi-state models)
Most commonly citedbull Corridor planning (including
community bypasses)bull Forecasting freighttruck
flowsbull External trip forecasts for
MPO amp regional models bull Expansion of MPO amp regional
models
Source Pennsylvania Department of Transportation TRB presentation on 2006-2030 Statewide Model
Louisianarsquos and Virginiarsquos statewide modelsMarket segments 2000 source amp estimation procedure Forecast source amp estimation
procedure
Macrolevel
1 Interstate and interparishtruck trip table
2000 macro zone level Transearch data some matrix estimation
2030 macro zone forecast of 2000 Transearch data interpolation and extrapolation to forecast other years
2 Long distance interstate business tourist and other auto trip tables
1995 American Travel Survey data disaggregated from state and MSA to county level and Fratar balanced to 2000 using household and employment growth
Fratar balance year 2000 county-level trip tables to forecast year using household and employment growth
3 Long distance intrastate business tourist and other auto trip tables
4 Short distance Interstate Louisiana HBW trip table
1990 Census county level JTW data Fratar balanced to 2000 using household and employment growth
Microlevel
5 Short distance intraparish truck trip table
Synthesized from 2000 intraparish Transearch data
Fratar balance 2000 intraparish trip table to forecast year using household and employment growth6 Short distance intrastate
Louisiana HBW trip table1990 tract-level JTW data Fratar balanced to 2000 using household and employment growth
7 Short distance intrastate HBO and NHB trip tables
Urban model style household-based trip generation and gravity model distribution
Urban style household-based trip generation and distribution models log-linear regression used to estimate external trip ends
Source Virginia Department of Transportation
MicrosimulationActivity-based Approach
bull Models dynamic evolution and actions of discrete agents across time and space
bull Observe emergent behavior Interaction Cooperation Coordination
bull Primary emphasis on understanding the system
bull Maybe forecast it
Between autonomous perceptive and deliberative agents
Source Ohio Department of Transportation
Key Differences ndash Traditional vs MicrosimulationldquoTraditional modelsrdquobull Closed form equilibriumbull Well understoodbull Typically deterministicbull Presumes central tendencies
in underlying databull Well-behavedbull ldquoManageablerdquo variationrdquobull Computationally efficient
Microsimulationbull Disequilibriumbull Large systems NWUbull Stochasticbull Diverse behaviorbull Handles wide variancesbull Accommodates weak to strong
understandingbull Computationally heavy
Zonal Land Use Data
Worker Traveler Generation
Vehicle Assignment
Starting Time Assignment
Post-Processor
Next Stop Purpose Choice
Model
Next Stop Location
Choice Model
Dynamic Activity Pattern Generation
Commercial Vehicle Trip List
Zonal Land Use Data
Worker Traveler Generation
Vehicle Assignment
Starting Time Assignment
Post-Processor
Next Stop Purpose Choice
Model
Next Stop Location
Choice Model
Dynamic Activity Pattern Generation
Commercial Vehicle Trip List
Microsimulation Examplesbull Ohiobull Oregon
Source Ohio Department of Transportation
Statewide Model Current Practicebull Meetings of the TRB Subcommittee on Statewide Travel Forecasting Models
bull NCHRP Report 735 Long-Distance and Rural Transferable Parameters for Statewide Travel Forecasting Models 2012
bull NCHRP Validation and Sensitivity Considerations for Statewide Models September 2010
bull White Paper on Statewide Modeling presented at TRB Statewide Planning Conference September 2008
bull National Travel Demand Forecasting Model Phase I Final Scope 2008
bull Forecasting Statewide Freight Toolkit NCHRP Report 606 2008
bull TRB Annual Meeting Sessions in 2004 2006 and 2008 web page
bull Statewide Travel Forecasting Models NCHRP Synthesis 358 2006
bull Statewide Travel Demand Modeling Peer Exchange Transportation Research Circular E-C075 September 2004
bull Guidebook on Statewide Travel Forecasting and Irvine Conference Transportation Research Circular E-C011 1999
Topic ndash Statewide Model Theory
bull Common themes in statewide model development
bull Alternative structures Trip table estimation (2 step) Highway model structure (3 step) Fully multimodal (4 step) Beyond 4 step
bull Deciding on best approach
bull Data development challengesSource Massachusetts Highway DeptCambridge Systematics Inc
Common Themes in Review of Statewide Models
bull Most statewide models use similar methodologies to metropolitan models but there are many variations
bull The largest problems relate to issues of scale
bull Models from different states vary greatly in complexity cost and development time
bull Models are most successful when addressing statewide priorities
bull There are major deficiencies in available data about long-distance and rural passenger travelhellip NCHRP Report 735 has helped
bull Statewide models are more compatible with secondary freight data sources such as the Commodity Flow Survey than are metropolitan models
bull Interest is high among states in progress in deploying models
Source A Horowitz White Paper on Statewide Modeling presented at TRB Conference on Statewide Planning 2008
Alternative Statewide Model Structures
bull Trip table estimation (2 step) Estimate initial ldquoseed matrixrdquo
Roadside intercept surveys American Travel Survey National HH Travel SurveyAdd-ons Census Journey-to-Work flows
Matrix estimation program Inputs ndash seed matrix highway
network traffic counts Adjusts origin-destination flows
between zones to optimize matching network traffic counts
Fratar factors applied to synthetic trip table in estimating future growth
bull Trip table estimation (2 step) Pros
Limited need to prepare aggregate socioeconomic input data
Simplicity of model generally means short execution times
Forecasting needs are limited to zone or district growth factors
Cons Less explanatory power than most
MPOregional models Fixed distribution pattern Difficulties in assessing benefits of
adding new roadway capacity
Alternative Statewide Model Structures (Contrsquod)
bull Highway model structure (3 step) Trip generation Trip distribution Highway assignment
bull Mode choice excluded Either generate vehicle trips or Apply auto occupancy factors to person trip
tables
bull Pros Benefits of trip redistribution across range of
alternatives Greater sensitivity to roadway capacity and
congestion
bull Cons Likely reliance on MPOsothers to provide
socioeconomic input data Canrsquot fully test transportation strategies that
would alter mode split
Source Florida Dept of TransportationCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)
bull Freight considerations (3 step) Typical freight model steps
Trip generation ndash tons by commodity group
Trip distribution Mode choice ndash freight is
apportioned to transport modes
bull Joint highway assignment (trucks and passenger trips)
Source Florida Dept of TransportationCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)bull Full multi-modal structure (4 step)
Trip generation Trip distribution Mode choice Highway assignment
bull More common with mega-regions Focused on fixed guideway systems Intercity rail is important component
bull Pros Potential to model new intercity rail
including high speed rail Test greater range of potential
transportation solutions
bull Cons Requires coding of transit networks and
mode choice estimation Modeling of air travel might be needed
especially if intercity rail is a completely new mode of travel
Source California HSR AuthorityCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)
Beyond 4 Stepbull Potential components
Implement destination choice Tour-basedtrip chains Activity-based travel model Activity micro-simulation Traffic micro-simulation Integration with land use models Integration with economic models
bull Pros Ability to test a more complete set of
transportation strategies and solutions
bull Cons Extensive data requirements Cost to maintain Complexity
Source Oregon Dept of TransportationPBUniversity of Calgary
Deciding on Best Statewide Model Approach
bull What are the motivations for building a statewide travel demand modelbull What are the emerging transportation and policy issues to addressbull How significant is the impact of through travel on your statebull Do you have many multi-state MPOsbull How reliable (and current) are your survey databull Does your state have major international ports or border crossings
StudyArea
UrbanModel
AreaStudyArea
UrbanModelArea 1
Study AreaWithin OneModel Area
UrbanModelArea 2
Study AreasWithin Two Model Areas
StudyArea
UrbanModelArea
StudyArea
Statewide Model
Study AreaOutside Urban
Model Areas
Statewide ModelStatewide Model
Source Florida Department of TransportationCambridge Systematics Inc
Figure 317 Subarea Study Area Examples
Study
Area
Urban
Model
Area
Study
Area
Urban
Model
Area 1
Study Area
Within One
Model Area
Urban
Model
Area 2
Study Areas
Within Two
Model Areas
Study
Area
Urban
Model
Area
Study
Area
Statewide Model
Study Area
Outside Urban
Model Areas
Statewide Model
Statewide Model
Statewide Model Data Development Challenges
bull Primary data deficiencies Long distance travelers Rural resident trip-making
bull Challenges in compiling required data Age of data Confidentiality Coordination with data providers Proper level of geography Sufficient sample size
bull Four categories of data for discussion Networks and zone system Demographic and employment data Freight and other economic data NHTS and other behavioral data
Source Florida Department of TransportationCambridge Systematics Inc
Data Challengeshellipwhat is ldquolong distancerdquo
bull gt50 miles (American Traveler Survey)
bull gt100 miles (National Household Travel Survey)
bull IE trips (several states)
bull Trips made by residents for business personal business
bull Recreationaltourist trips
Topic ndash Similarities amp Differences with Urban Models
bull Zone systems
bull Networks
bull Travel behavior assumptions
bull Trip purposesmarket segments
bull Modal issues
bull Truck forecasting
Source Delaware Dept of TransportationWRampA Inc
Statewide vs Urban Model Zone Systems
bull Usually larger zones in statewide models than MPOs
bull Block groups are good starting point for statewide model TAZs
bull Block groups might need splitting along major transportation corridors
bull In small states MPO TAZs could equal SWM zones
bull In most states MPO TAZ equivalency tables will likely be neededdesirable
bull FAF disaggregation approach used for statewide model TAZs
Source FHWA Office of Freight Management and Operations
Source Florida Dept of TransportationCambridge Systematics Inc
FAF zones
Block groups = TAZs
Too small for statewide zones
Too big for statewide zones
Statewide vs Urban Model Networksbull Statewide model networks are sparse in
comparison with MPO models
bull At a minimum statewide networks should include Limited access highways Arterials Major collectors
bull For proper circulation patterns lower class facilities might be added
bull Outside state a very sparse network is needed for major truck corridors
bull Consistent level of detail for networks and zones
Source Tennessee Dept of TransportationCambridge Systematics Inc
MPO network
statewide network
Statewide vs Urban Model Travel Behaviorbull Understanding urban travel behavior
MPO models are usually developed with local household travel survey data
MPOs without recent HH survey data can borrow parameters from other MPO models or guidance documents such as NCHRP 716 Transferable Urban Travel Parameters
bull Rural and long-distance travel behavior A limited number of states have conducted
statewide household travel surveys American Travel Survey was designed to target
long-distance trips but was discontinued long ago
National Household Travel Survey Add-ons can be structured for a sufficient sample of rural trip-makers
NCHRP 735 Rural and Long-Distance Travel Parameters recently published for use in statewide models
Map of 2009 Add-on Surveys Source FHWA
Statewide vs Urban Model Trip Purposes
bull Typical MPO trip purposes are usually collapsed into HBW HBOHBNW and NHB for statewide models
bull Long-distance purposes Business Tourist Personal business
bull Trucks Commodity group is often
surrogate for trip purpose
bull External trips External generally means trip from
outside state Might exclude ldquohalordquo area
surrounding state boundary
Trip PurposeFL
2000 IN LA MA MS RI
Home-based Work 1914 1995 1900 1841 1606 2000
Home-based Nonwork 5547 5208 4770 5206 4917 5700
Home-based Other 2792 5208 4770 1938 4917 5700
Home-based Shop 1576 ndash ndash 1173 ndash ndash
Home-based SocialRecreation
1179 ndash ndash 1395 ndash ndash
Home-base RecreationShop ndash ndash ndash ndash ndash ndash
Home-based School ndash ndash ndash 700 ndash ndash
Nonhome-based 2539 2686 3100 2468 3289 2300
Nonhome-based Work ndash ndash ndash 1027 ndash ndash
Nonhome-based Nonwork ndash ndash ndash 1441 ndash ndash
Other ndash ndash ndash ndash ndash
Truck ndash ndash 170 ndash ndash ndash
Long-Distance 000 111 060 417 022 000
Long-Distance Business ndash ndash 020 284 011 ndash
Long-Distance Personal Business
ndash ndash ndash ndash ndash ndash
Long-Distance Tourist ndash ndash 010 027 000 ndash
Long-Distance Work ndash ndash ndash 105 ndash ndash
Long-Distance NonworkOther
ndash ndash 030 ndash 011 ndash
Statewide vs Urban Model Trip Purposes (Contrsquod)
Number of person trips
Mode Pleasure Business Personal business Other Total Percent
Personal vehicle 12185 5708 3120 789 21802 908Commercial air 1012 746 151 24 1933 8Train 74 150 14 0 238 1Other 4 2 5 12 23 01Ship 19 0 0 0 19 01
Total 13294 6606 3290 825 24015 100Percent 554 275 137 34 100
Percent by trip purpose
Mode Pleasure Business Personal business Other
Personal vehicle 917 864 948 956Commercial air 76 113 46 29Train 06 23 04 0Other 0 0 02 15Ship 01 0 0 0
Trip purpose
Trip purpose
Statewide vs Urban Model Modal Issuesbull Most statewide models are limited to highway trips
bull Modeling of local bus routes is likely overkill
bull Mode choice is usually just a conversion from person trips to vehicle trips via auto occupancy rates
bull Transit components of statewide models are more common in regions with history of fixed guideway transit
bull Transit model is needed to test options such as high speedintercity rail
ATS 95 NHTS 01 MDOT 03-04 ODOT 02-03Personal vehicle 81 89 89 88Commercial air 16 8 10 8Bus 2 2 04 19Train 05 01 0 005Other 05 09 06 205
Survey and yearPrimary mode of transport
Statewide vs Urban Model Modal Issues (Contrsquod)
3727 24 22
32
29 36
10
21
28 22
10
22
12
1529
11
44 43
0
10
20
30
40
50
60
70
80
90
100
Personalvehicle
Bus Train Marine Commercialair
Perc
ent o
f per
son
trips
lt300 300-499 500-599 1000-1999 gt2000
68
85 85 82 85
30
12 11 15 13
0
10
20
30
40
50
60
70
80
90
100
Busi-ness
Plea-sure
pBusi-ness
Vaca-tion
Week-end
Perc
ent o
f per
son
trips
Personal vehicle Commercial air All other
Mode by Trip Purpose Mode by Distance (miles)
Statewide vs Urban Model Modal Issues (Contrsquod)
bull Most statewide assignment models separate trucks from passenger cars Preloading process commonly used for trucks Takes into account more limited route options for truck travel
Source Ohio Dept of Transportation
Statewide vs Urban Model Truck Forecasting
bull Urban modelsndash Freight movements are partially
ldquoexternalrdquo to model (truncated at external zone)
ndash More general categories of truckcommercial trips used
ndash Strong reliance on Quick Response Freight Manual
bull Statewide modelsndash Rooted in commodity flows ndash Often use econometric
approaches to modelingndash Generally model entire length of
freight travel
Freight Model Type by Peer Exchange Participants and
NonParticipants
0123456789
None
Traditio
nal
Commod
ity
Econo
metric
Unkno
wn
NotPeer Exchange
Source TRB Electronic Circular E-C075 Statewide Travel Demand Modeling ndash A Peer Exchange
Topic ndash Data Needs and Considerations
bull Impact of model complexity on data needs
bull Overview of data sources
bull Long-Distance travel gt different focus than MPOregional models
Sources NHTS Weight calculation Pitfalls Key statistics European surveys Lessons learned
Impact of Model Complexity on Data Needsbull Trip table estimation (2 step) ndash highway network
zone system trip table
bull Highway model structure (3 step) ndash 2 step plusbehavioral data for trip rates friction factors auto occupancy rates
bull Freight considerations (3 step) ndash 3 step pluscommodity flows economic data tonnages by mode truck counts
bull Full multi-modal structure (4 step) ndash 3 step plustransit networks and paths logit mode choice model
bull Beyond 4 Step ndash synthetic household generation operational characteristics etc
Source Georgia DOTCambridge Systematics Inc
Statewide Model Data Sources A-Zbull American Travel Survey (ATS)
bull Bureau of Economic Analysis (BEA)
bull Census Journey-to-Work (JTW) data
bull Census Public Use Microsample (PUMS)
bull Census Transportation Planning Package (CTPP)
bull Claritas
bull Commodity Flow Survey (CFS)
Source US Census Bureau
Statewide Model Data Sources (Contrsquod)bull Dun amp Bradstreet (DampB) employment data
bull Employment Security data (ES-202)
bull Freight Analysis Framework (FAF)
bull Highway Performance Monitoring System (HPMS)
bull InfoUSA employment data
bull Metropolitan Planning Organizations (MPO)
bull NCHRP 365 Manual
bull National Highway Planning Network (NHPN)
Source FAF2 Provisional Commodity Origin-Destination Estimates Revised Methodology Report BattelleFHWA
Statewide Model Data Sources (Contrsquod)
bull National Household Travel Survey (NHTS)
bull Origin-destination surveys
bull Regional Economic Models Inc (REMI)
bull State Departments of Transportation
bull TRANSEARCH data
bull Vehicle Use and Identification Survey (VIUS)
bull Woods amp Poole employment forecasts
Source Bureau of Transportation Statistics NHTS
Data Sources on Long-Distance Travel
bull Census Journey-to-Work databull Long distance component
American Travel Survey (77 95) National Household Travel Survey (01) Ohio statewide surveys (02-04) Michigan statewide survey (04-05) California (10-12)
bull Surveys focused on long distance Oregon (96-98 07-08)
bull Others BTS Omnibus surveys Tourism surveys European practice NCHRP Report 735
NHTS and Long-Distance Travelbull Evolution of US national surveys
NPTS (69 77 83 90 95) ATS (77 95)
bull 2001 NHTS ndash Total of 60282 long distance person trips Data collected March 01 to May 02 22204 (37 percent) before 911 38078 (63 percent) afterwards
bull Long distance trip is longer than 50 miles one-way (vs 100 miles in 1995 ATS)
bull 28-day retrospective survey (including travel day)
bull Not your typical travelers Low frequency Distinct segment of the population
Weight Calculation for Long-Distance Travelers
bull Use to ensure valid population-level estimates Accounts for non-response under-coverage multiple phones etc
bull Multi-stage process Match national margins Adjust for data splitting (force sumexpfac = 277208169) Raking (match independent controls) race ethnicity race by
month ethnicity by month sex by age census region MSA status month by day of week
Trim outlier expansion factors max expfac = 4 times mean Replication (correct for sampling error)
bull At least two passes through the data
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Freight Model Types | |||||||||||||||||||
State | Type | Peer Exc | |||||||||||||||||
OR | 4 | 1 | Count of Type | Count of Type | Peer Exc | ||||||||||||||
WI | 3 | 1 | Type | Total | Type | 0 | 1 | Grand Total | |||||||||||
MO | 1 | 1 | 1 | 2 | 1 | 2 | 2 | 4 | |||||||||||
LA | 3 | 1 | 2 | 2 | 2 | 2 | 2 | 4 | |||||||||||
IN | 3 | 1 | 3 | 6 | 3 | 2 | 6 | 8 | |||||||||||
OH | 4 | 1 | 4 | 2 | 4 | 2 | 2 | ||||||||||||
KY | 3 | 1 | Grand Total | 12 | 3 | 3 | |||||||||||||
FL | 3 | 1 | Grand Total | 9 | 12 | 21 | |||||||||||||
VA | 3 | 1 | None | 2 | Not | Peer Exchange | |||||||||||||
NJ | 2 | 1 | Traditional | 2 | None | 2 | 2 | ||||||||||||
MA | 2 | 1 | Commodity | 6 | Traditional | 2 | 2 | ||||||||||||
NH | 1 | 1 | Econometric | 2 | Commodity | 2 | 6 | ||||||||||||
CA | 0 | Econometric | 2 | ||||||||||||||||
TX | 3 | 0 | Unknown | 3 | |||||||||||||||
MI | 3 | 0 | |||||||||||||||||
GA | 0 | ||||||||||||||||||
SC | 1 | 0 | |||||||||||||||||
DE | 2 | 0 | |||||||||||||||||
CT | 0 | ||||||||||||||||||
VT | 2 | 0 | |||||||||||||||||
ME | 1 | 0 | |||||||||||||||||
MN | 0 |
None | None | ||
Traditional | Traditional | ||
Commodity | Commodity | ||
Econometric | Econometric | ||
Unknown | Unknown |
Typical uses of Statewide ModelsLess common usesbull Air quality conformity analysesbull Freight and intermodal
planningbull Traffic impact studiesbull Economic development studiesbull Long-term investment studiesbull Emergency evacuation studiesbull Multi-state corridors (possibly
with multi-state models)
Most commonly citedbull Corridor planning (including
community bypasses)bull Forecasting freighttruck
flowsbull External trip forecasts for
MPO amp regional models bull Expansion of MPO amp regional
models
Source Pennsylvania Department of Transportation TRB presentation on 2006-2030 Statewide Model
Louisianarsquos and Virginiarsquos statewide modelsMarket segments 2000 source amp estimation procedure Forecast source amp estimation
procedure
Macrolevel
1 Interstate and interparishtruck trip table
2000 macro zone level Transearch data some matrix estimation
2030 macro zone forecast of 2000 Transearch data interpolation and extrapolation to forecast other years
2 Long distance interstate business tourist and other auto trip tables
1995 American Travel Survey data disaggregated from state and MSA to county level and Fratar balanced to 2000 using household and employment growth
Fratar balance year 2000 county-level trip tables to forecast year using household and employment growth
3 Long distance intrastate business tourist and other auto trip tables
4 Short distance Interstate Louisiana HBW trip table
1990 Census county level JTW data Fratar balanced to 2000 using household and employment growth
Microlevel
5 Short distance intraparish truck trip table
Synthesized from 2000 intraparish Transearch data
Fratar balance 2000 intraparish trip table to forecast year using household and employment growth6 Short distance intrastate
Louisiana HBW trip table1990 tract-level JTW data Fratar balanced to 2000 using household and employment growth
7 Short distance intrastate HBO and NHB trip tables
Urban model style household-based trip generation and gravity model distribution
Urban style household-based trip generation and distribution models log-linear regression used to estimate external trip ends
Source Virginia Department of Transportation
MicrosimulationActivity-based Approach
bull Models dynamic evolution and actions of discrete agents across time and space
bull Observe emergent behavior Interaction Cooperation Coordination
bull Primary emphasis on understanding the system
bull Maybe forecast it
Between autonomous perceptive and deliberative agents
Source Ohio Department of Transportation
Key Differences ndash Traditional vs MicrosimulationldquoTraditional modelsrdquobull Closed form equilibriumbull Well understoodbull Typically deterministicbull Presumes central tendencies
in underlying databull Well-behavedbull ldquoManageablerdquo variationrdquobull Computationally efficient
Microsimulationbull Disequilibriumbull Large systems NWUbull Stochasticbull Diverse behaviorbull Handles wide variancesbull Accommodates weak to strong
understandingbull Computationally heavy
Zonal Land Use Data
Worker Traveler Generation
Vehicle Assignment
Starting Time Assignment
Post-Processor
Next Stop Purpose Choice
Model
Next Stop Location
Choice Model
Dynamic Activity Pattern Generation
Commercial Vehicle Trip List
Zonal Land Use Data
Worker Traveler Generation
Vehicle Assignment
Starting Time Assignment
Post-Processor
Next Stop Purpose Choice
Model
Next Stop Location
Choice Model
Dynamic Activity Pattern Generation
Commercial Vehicle Trip List
Microsimulation Examplesbull Ohiobull Oregon
Source Ohio Department of Transportation
Statewide Model Current Practicebull Meetings of the TRB Subcommittee on Statewide Travel Forecasting Models
bull NCHRP Report 735 Long-Distance and Rural Transferable Parameters for Statewide Travel Forecasting Models 2012
bull NCHRP Validation and Sensitivity Considerations for Statewide Models September 2010
bull White Paper on Statewide Modeling presented at TRB Statewide Planning Conference September 2008
bull National Travel Demand Forecasting Model Phase I Final Scope 2008
bull Forecasting Statewide Freight Toolkit NCHRP Report 606 2008
bull TRB Annual Meeting Sessions in 2004 2006 and 2008 web page
bull Statewide Travel Forecasting Models NCHRP Synthesis 358 2006
bull Statewide Travel Demand Modeling Peer Exchange Transportation Research Circular E-C075 September 2004
bull Guidebook on Statewide Travel Forecasting and Irvine Conference Transportation Research Circular E-C011 1999
Topic ndash Statewide Model Theory
bull Common themes in statewide model development
bull Alternative structures Trip table estimation (2 step) Highway model structure (3 step) Fully multimodal (4 step) Beyond 4 step
bull Deciding on best approach
bull Data development challengesSource Massachusetts Highway DeptCambridge Systematics Inc
Common Themes in Review of Statewide Models
bull Most statewide models use similar methodologies to metropolitan models but there are many variations
bull The largest problems relate to issues of scale
bull Models from different states vary greatly in complexity cost and development time
bull Models are most successful when addressing statewide priorities
bull There are major deficiencies in available data about long-distance and rural passenger travelhellip NCHRP Report 735 has helped
bull Statewide models are more compatible with secondary freight data sources such as the Commodity Flow Survey than are metropolitan models
bull Interest is high among states in progress in deploying models
Source A Horowitz White Paper on Statewide Modeling presented at TRB Conference on Statewide Planning 2008
Alternative Statewide Model Structures
bull Trip table estimation (2 step) Estimate initial ldquoseed matrixrdquo
Roadside intercept surveys American Travel Survey National HH Travel SurveyAdd-ons Census Journey-to-Work flows
Matrix estimation program Inputs ndash seed matrix highway
network traffic counts Adjusts origin-destination flows
between zones to optimize matching network traffic counts
Fratar factors applied to synthetic trip table in estimating future growth
bull Trip table estimation (2 step) Pros
Limited need to prepare aggregate socioeconomic input data
Simplicity of model generally means short execution times
Forecasting needs are limited to zone or district growth factors
Cons Less explanatory power than most
MPOregional models Fixed distribution pattern Difficulties in assessing benefits of
adding new roadway capacity
Alternative Statewide Model Structures (Contrsquod)
bull Highway model structure (3 step) Trip generation Trip distribution Highway assignment
bull Mode choice excluded Either generate vehicle trips or Apply auto occupancy factors to person trip
tables
bull Pros Benefits of trip redistribution across range of
alternatives Greater sensitivity to roadway capacity and
congestion
bull Cons Likely reliance on MPOsothers to provide
socioeconomic input data Canrsquot fully test transportation strategies that
would alter mode split
Source Florida Dept of TransportationCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)
bull Freight considerations (3 step) Typical freight model steps
Trip generation ndash tons by commodity group
Trip distribution Mode choice ndash freight is
apportioned to transport modes
bull Joint highway assignment (trucks and passenger trips)
Source Florida Dept of TransportationCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)bull Full multi-modal structure (4 step)
Trip generation Trip distribution Mode choice Highway assignment
bull More common with mega-regions Focused on fixed guideway systems Intercity rail is important component
bull Pros Potential to model new intercity rail
including high speed rail Test greater range of potential
transportation solutions
bull Cons Requires coding of transit networks and
mode choice estimation Modeling of air travel might be needed
especially if intercity rail is a completely new mode of travel
Source California HSR AuthorityCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)
Beyond 4 Stepbull Potential components
Implement destination choice Tour-basedtrip chains Activity-based travel model Activity micro-simulation Traffic micro-simulation Integration with land use models Integration with economic models
bull Pros Ability to test a more complete set of
transportation strategies and solutions
bull Cons Extensive data requirements Cost to maintain Complexity
Source Oregon Dept of TransportationPBUniversity of Calgary
Deciding on Best Statewide Model Approach
bull What are the motivations for building a statewide travel demand modelbull What are the emerging transportation and policy issues to addressbull How significant is the impact of through travel on your statebull Do you have many multi-state MPOsbull How reliable (and current) are your survey databull Does your state have major international ports or border crossings
StudyArea
UrbanModel
AreaStudyArea
UrbanModelArea 1
Study AreaWithin OneModel Area
UrbanModelArea 2
Study AreasWithin Two Model Areas
StudyArea
UrbanModelArea
StudyArea
Statewide Model
Study AreaOutside Urban
Model Areas
Statewide ModelStatewide Model
Source Florida Department of TransportationCambridge Systematics Inc
Figure 317 Subarea Study Area Examples
Study
Area
Urban
Model
Area
Study
Area
Urban
Model
Area 1
Study Area
Within One
Model Area
Urban
Model
Area 2
Study Areas
Within Two
Model Areas
Study
Area
Urban
Model
Area
Study
Area
Statewide Model
Study Area
Outside Urban
Model Areas
Statewide Model
Statewide Model
Statewide Model Data Development Challenges
bull Primary data deficiencies Long distance travelers Rural resident trip-making
bull Challenges in compiling required data Age of data Confidentiality Coordination with data providers Proper level of geography Sufficient sample size
bull Four categories of data for discussion Networks and zone system Demographic and employment data Freight and other economic data NHTS and other behavioral data
Source Florida Department of TransportationCambridge Systematics Inc
Data Challengeshellipwhat is ldquolong distancerdquo
bull gt50 miles (American Traveler Survey)
bull gt100 miles (National Household Travel Survey)
bull IE trips (several states)
bull Trips made by residents for business personal business
bull Recreationaltourist trips
Topic ndash Similarities amp Differences with Urban Models
bull Zone systems
bull Networks
bull Travel behavior assumptions
bull Trip purposesmarket segments
bull Modal issues
bull Truck forecasting
Source Delaware Dept of TransportationWRampA Inc
Statewide vs Urban Model Zone Systems
bull Usually larger zones in statewide models than MPOs
bull Block groups are good starting point for statewide model TAZs
bull Block groups might need splitting along major transportation corridors
bull In small states MPO TAZs could equal SWM zones
bull In most states MPO TAZ equivalency tables will likely be neededdesirable
bull FAF disaggregation approach used for statewide model TAZs
Source FHWA Office of Freight Management and Operations
Source Florida Dept of TransportationCambridge Systematics Inc
FAF zones
Block groups = TAZs
Too small for statewide zones
Too big for statewide zones
Statewide vs Urban Model Networksbull Statewide model networks are sparse in
comparison with MPO models
bull At a minimum statewide networks should include Limited access highways Arterials Major collectors
bull For proper circulation patterns lower class facilities might be added
bull Outside state a very sparse network is needed for major truck corridors
bull Consistent level of detail for networks and zones
Source Tennessee Dept of TransportationCambridge Systematics Inc
MPO network
statewide network
Statewide vs Urban Model Travel Behaviorbull Understanding urban travel behavior
MPO models are usually developed with local household travel survey data
MPOs without recent HH survey data can borrow parameters from other MPO models or guidance documents such as NCHRP 716 Transferable Urban Travel Parameters
bull Rural and long-distance travel behavior A limited number of states have conducted
statewide household travel surveys American Travel Survey was designed to target
long-distance trips but was discontinued long ago
National Household Travel Survey Add-ons can be structured for a sufficient sample of rural trip-makers
NCHRP 735 Rural and Long-Distance Travel Parameters recently published for use in statewide models
Map of 2009 Add-on Surveys Source FHWA
Statewide vs Urban Model Trip Purposes
bull Typical MPO trip purposes are usually collapsed into HBW HBOHBNW and NHB for statewide models
bull Long-distance purposes Business Tourist Personal business
bull Trucks Commodity group is often
surrogate for trip purpose
bull External trips External generally means trip from
outside state Might exclude ldquohalordquo area
surrounding state boundary
Trip PurposeFL
2000 IN LA MA MS RI
Home-based Work 1914 1995 1900 1841 1606 2000
Home-based Nonwork 5547 5208 4770 5206 4917 5700
Home-based Other 2792 5208 4770 1938 4917 5700
Home-based Shop 1576 ndash ndash 1173 ndash ndash
Home-based SocialRecreation
1179 ndash ndash 1395 ndash ndash
Home-base RecreationShop ndash ndash ndash ndash ndash ndash
Home-based School ndash ndash ndash 700 ndash ndash
Nonhome-based 2539 2686 3100 2468 3289 2300
Nonhome-based Work ndash ndash ndash 1027 ndash ndash
Nonhome-based Nonwork ndash ndash ndash 1441 ndash ndash
Other ndash ndash ndash ndash ndash
Truck ndash ndash 170 ndash ndash ndash
Long-Distance 000 111 060 417 022 000
Long-Distance Business ndash ndash 020 284 011 ndash
Long-Distance Personal Business
ndash ndash ndash ndash ndash ndash
Long-Distance Tourist ndash ndash 010 027 000 ndash
Long-Distance Work ndash ndash ndash 105 ndash ndash
Long-Distance NonworkOther
ndash ndash 030 ndash 011 ndash
Statewide vs Urban Model Trip Purposes (Contrsquod)
Number of person trips
Mode Pleasure Business Personal business Other Total Percent
Personal vehicle 12185 5708 3120 789 21802 908Commercial air 1012 746 151 24 1933 8Train 74 150 14 0 238 1Other 4 2 5 12 23 01Ship 19 0 0 0 19 01
Total 13294 6606 3290 825 24015 100Percent 554 275 137 34 100
Percent by trip purpose
Mode Pleasure Business Personal business Other
Personal vehicle 917 864 948 956Commercial air 76 113 46 29Train 06 23 04 0Other 0 0 02 15Ship 01 0 0 0
Trip purpose
Trip purpose
Statewide vs Urban Model Modal Issuesbull Most statewide models are limited to highway trips
bull Modeling of local bus routes is likely overkill
bull Mode choice is usually just a conversion from person trips to vehicle trips via auto occupancy rates
bull Transit components of statewide models are more common in regions with history of fixed guideway transit
bull Transit model is needed to test options such as high speedintercity rail
ATS 95 NHTS 01 MDOT 03-04 ODOT 02-03Personal vehicle 81 89 89 88Commercial air 16 8 10 8Bus 2 2 04 19Train 05 01 0 005Other 05 09 06 205
Survey and yearPrimary mode of transport
Statewide vs Urban Model Modal Issues (Contrsquod)
3727 24 22
32
29 36
10
21
28 22
10
22
12
1529
11
44 43
0
10
20
30
40
50
60
70
80
90
100
Personalvehicle
Bus Train Marine Commercialair
Perc
ent o
f per
son
trips
lt300 300-499 500-599 1000-1999 gt2000
68
85 85 82 85
30
12 11 15 13
0
10
20
30
40
50
60
70
80
90
100
Busi-ness
Plea-sure
pBusi-ness
Vaca-tion
Week-end
Perc
ent o
f per
son
trips
Personal vehicle Commercial air All other
Mode by Trip Purpose Mode by Distance (miles)
Statewide vs Urban Model Modal Issues (Contrsquod)
bull Most statewide assignment models separate trucks from passenger cars Preloading process commonly used for trucks Takes into account more limited route options for truck travel
Source Ohio Dept of Transportation
Statewide vs Urban Model Truck Forecasting
bull Urban modelsndash Freight movements are partially
ldquoexternalrdquo to model (truncated at external zone)
ndash More general categories of truckcommercial trips used
ndash Strong reliance on Quick Response Freight Manual
bull Statewide modelsndash Rooted in commodity flows ndash Often use econometric
approaches to modelingndash Generally model entire length of
freight travel
Freight Model Type by Peer Exchange Participants and
NonParticipants
0123456789
None
Traditio
nal
Commod
ity
Econo
metric
Unkno
wn
NotPeer Exchange
Source TRB Electronic Circular E-C075 Statewide Travel Demand Modeling ndash A Peer Exchange
Topic ndash Data Needs and Considerations
bull Impact of model complexity on data needs
bull Overview of data sources
bull Long-Distance travel gt different focus than MPOregional models
Sources NHTS Weight calculation Pitfalls Key statistics European surveys Lessons learned
Impact of Model Complexity on Data Needsbull Trip table estimation (2 step) ndash highway network
zone system trip table
bull Highway model structure (3 step) ndash 2 step plusbehavioral data for trip rates friction factors auto occupancy rates
bull Freight considerations (3 step) ndash 3 step pluscommodity flows economic data tonnages by mode truck counts
bull Full multi-modal structure (4 step) ndash 3 step plustransit networks and paths logit mode choice model
bull Beyond 4 Step ndash synthetic household generation operational characteristics etc
Source Georgia DOTCambridge Systematics Inc
Statewide Model Data Sources A-Zbull American Travel Survey (ATS)
bull Bureau of Economic Analysis (BEA)
bull Census Journey-to-Work (JTW) data
bull Census Public Use Microsample (PUMS)
bull Census Transportation Planning Package (CTPP)
bull Claritas
bull Commodity Flow Survey (CFS)
Source US Census Bureau
Statewide Model Data Sources (Contrsquod)bull Dun amp Bradstreet (DampB) employment data
bull Employment Security data (ES-202)
bull Freight Analysis Framework (FAF)
bull Highway Performance Monitoring System (HPMS)
bull InfoUSA employment data
bull Metropolitan Planning Organizations (MPO)
bull NCHRP 365 Manual
bull National Highway Planning Network (NHPN)
Source FAF2 Provisional Commodity Origin-Destination Estimates Revised Methodology Report BattelleFHWA
Statewide Model Data Sources (Contrsquod)
bull National Household Travel Survey (NHTS)
bull Origin-destination surveys
bull Regional Economic Models Inc (REMI)
bull State Departments of Transportation
bull TRANSEARCH data
bull Vehicle Use and Identification Survey (VIUS)
bull Woods amp Poole employment forecasts
Source Bureau of Transportation Statistics NHTS
Data Sources on Long-Distance Travel
bull Census Journey-to-Work databull Long distance component
American Travel Survey (77 95) National Household Travel Survey (01) Ohio statewide surveys (02-04) Michigan statewide survey (04-05) California (10-12)
bull Surveys focused on long distance Oregon (96-98 07-08)
bull Others BTS Omnibus surveys Tourism surveys European practice NCHRP Report 735
NHTS and Long-Distance Travelbull Evolution of US national surveys
NPTS (69 77 83 90 95) ATS (77 95)
bull 2001 NHTS ndash Total of 60282 long distance person trips Data collected March 01 to May 02 22204 (37 percent) before 911 38078 (63 percent) afterwards
bull Long distance trip is longer than 50 miles one-way (vs 100 miles in 1995 ATS)
bull 28-day retrospective survey (including travel day)
bull Not your typical travelers Low frequency Distinct segment of the population
Weight Calculation for Long-Distance Travelers
bull Use to ensure valid population-level estimates Accounts for non-response under-coverage multiple phones etc
bull Multi-stage process Match national margins Adjust for data splitting (force sumexpfac = 277208169) Raking (match independent controls) race ethnicity race by
month ethnicity by month sex by age census region MSA status month by day of week
Trim outlier expansion factors max expfac = 4 times mean Replication (correct for sampling error)
bull At least two passes through the data
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Freight Model Types | |||||||||||||||||||
State | Type | Peer Exc | |||||||||||||||||
OR | 4 | 1 | Count of Type | Count of Type | Peer Exc | ||||||||||||||
WI | 3 | 1 | Type | Total | Type | 0 | 1 | Grand Total | |||||||||||
MO | 1 | 1 | 1 | 2 | 1 | 2 | 2 | 4 | |||||||||||
LA | 3 | 1 | 2 | 2 | 2 | 2 | 2 | 4 | |||||||||||
IN | 3 | 1 | 3 | 6 | 3 | 2 | 6 | 8 | |||||||||||
OH | 4 | 1 | 4 | 2 | 4 | 2 | 2 | ||||||||||||
KY | 3 | 1 | Grand Total | 12 | 3 | 3 | |||||||||||||
FL | 3 | 1 | Grand Total | 9 | 12 | 21 | |||||||||||||
VA | 3 | 1 | None | 2 | Not | Peer Exchange | |||||||||||||
NJ | 2 | 1 | Traditional | 2 | None | 2 | 2 | ||||||||||||
MA | 2 | 1 | Commodity | 6 | Traditional | 2 | 2 | ||||||||||||
NH | 1 | 1 | Econometric | 2 | Commodity | 2 | 6 | ||||||||||||
CA | 0 | Econometric | 2 | ||||||||||||||||
TX | 3 | 0 | Unknown | 3 | |||||||||||||||
MI | 3 | 0 | |||||||||||||||||
GA | 0 | ||||||||||||||||||
SC | 1 | 0 | |||||||||||||||||
DE | 2 | 0 | |||||||||||||||||
CT | 0 | ||||||||||||||||||
VT | 2 | 0 | |||||||||||||||||
ME | 1 | 0 | |||||||||||||||||
MN | 0 |
None | None | ||
Traditional | Traditional | ||
Commodity | Commodity | ||
Econometric | Econometric | ||
Unknown | Unknown |
Louisianarsquos and Virginiarsquos statewide modelsMarket segments 2000 source amp estimation procedure Forecast source amp estimation
procedure
Macrolevel
1 Interstate and interparishtruck trip table
2000 macro zone level Transearch data some matrix estimation
2030 macro zone forecast of 2000 Transearch data interpolation and extrapolation to forecast other years
2 Long distance interstate business tourist and other auto trip tables
1995 American Travel Survey data disaggregated from state and MSA to county level and Fratar balanced to 2000 using household and employment growth
Fratar balance year 2000 county-level trip tables to forecast year using household and employment growth
3 Long distance intrastate business tourist and other auto trip tables
4 Short distance Interstate Louisiana HBW trip table
1990 Census county level JTW data Fratar balanced to 2000 using household and employment growth
Microlevel
5 Short distance intraparish truck trip table
Synthesized from 2000 intraparish Transearch data
Fratar balance 2000 intraparish trip table to forecast year using household and employment growth6 Short distance intrastate
Louisiana HBW trip table1990 tract-level JTW data Fratar balanced to 2000 using household and employment growth
7 Short distance intrastate HBO and NHB trip tables
Urban model style household-based trip generation and gravity model distribution
Urban style household-based trip generation and distribution models log-linear regression used to estimate external trip ends
Source Virginia Department of Transportation
MicrosimulationActivity-based Approach
bull Models dynamic evolution and actions of discrete agents across time and space
bull Observe emergent behavior Interaction Cooperation Coordination
bull Primary emphasis on understanding the system
bull Maybe forecast it
Between autonomous perceptive and deliberative agents
Source Ohio Department of Transportation
Key Differences ndash Traditional vs MicrosimulationldquoTraditional modelsrdquobull Closed form equilibriumbull Well understoodbull Typically deterministicbull Presumes central tendencies
in underlying databull Well-behavedbull ldquoManageablerdquo variationrdquobull Computationally efficient
Microsimulationbull Disequilibriumbull Large systems NWUbull Stochasticbull Diverse behaviorbull Handles wide variancesbull Accommodates weak to strong
understandingbull Computationally heavy
Zonal Land Use Data
Worker Traveler Generation
Vehicle Assignment
Starting Time Assignment
Post-Processor
Next Stop Purpose Choice
Model
Next Stop Location
Choice Model
Dynamic Activity Pattern Generation
Commercial Vehicle Trip List
Zonal Land Use Data
Worker Traveler Generation
Vehicle Assignment
Starting Time Assignment
Post-Processor
Next Stop Purpose Choice
Model
Next Stop Location
Choice Model
Dynamic Activity Pattern Generation
Commercial Vehicle Trip List
Microsimulation Examplesbull Ohiobull Oregon
Source Ohio Department of Transportation
Statewide Model Current Practicebull Meetings of the TRB Subcommittee on Statewide Travel Forecasting Models
bull NCHRP Report 735 Long-Distance and Rural Transferable Parameters for Statewide Travel Forecasting Models 2012
bull NCHRP Validation and Sensitivity Considerations for Statewide Models September 2010
bull White Paper on Statewide Modeling presented at TRB Statewide Planning Conference September 2008
bull National Travel Demand Forecasting Model Phase I Final Scope 2008
bull Forecasting Statewide Freight Toolkit NCHRP Report 606 2008
bull TRB Annual Meeting Sessions in 2004 2006 and 2008 web page
bull Statewide Travel Forecasting Models NCHRP Synthesis 358 2006
bull Statewide Travel Demand Modeling Peer Exchange Transportation Research Circular E-C075 September 2004
bull Guidebook on Statewide Travel Forecasting and Irvine Conference Transportation Research Circular E-C011 1999
Topic ndash Statewide Model Theory
bull Common themes in statewide model development
bull Alternative structures Trip table estimation (2 step) Highway model structure (3 step) Fully multimodal (4 step) Beyond 4 step
bull Deciding on best approach
bull Data development challengesSource Massachusetts Highway DeptCambridge Systematics Inc
Common Themes in Review of Statewide Models
bull Most statewide models use similar methodologies to metropolitan models but there are many variations
bull The largest problems relate to issues of scale
bull Models from different states vary greatly in complexity cost and development time
bull Models are most successful when addressing statewide priorities
bull There are major deficiencies in available data about long-distance and rural passenger travelhellip NCHRP Report 735 has helped
bull Statewide models are more compatible with secondary freight data sources such as the Commodity Flow Survey than are metropolitan models
bull Interest is high among states in progress in deploying models
Source A Horowitz White Paper on Statewide Modeling presented at TRB Conference on Statewide Planning 2008
Alternative Statewide Model Structures
bull Trip table estimation (2 step) Estimate initial ldquoseed matrixrdquo
Roadside intercept surveys American Travel Survey National HH Travel SurveyAdd-ons Census Journey-to-Work flows
Matrix estimation program Inputs ndash seed matrix highway
network traffic counts Adjusts origin-destination flows
between zones to optimize matching network traffic counts
Fratar factors applied to synthetic trip table in estimating future growth
bull Trip table estimation (2 step) Pros
Limited need to prepare aggregate socioeconomic input data
Simplicity of model generally means short execution times
Forecasting needs are limited to zone or district growth factors
Cons Less explanatory power than most
MPOregional models Fixed distribution pattern Difficulties in assessing benefits of
adding new roadway capacity
Alternative Statewide Model Structures (Contrsquod)
bull Highway model structure (3 step) Trip generation Trip distribution Highway assignment
bull Mode choice excluded Either generate vehicle trips or Apply auto occupancy factors to person trip
tables
bull Pros Benefits of trip redistribution across range of
alternatives Greater sensitivity to roadway capacity and
congestion
bull Cons Likely reliance on MPOsothers to provide
socioeconomic input data Canrsquot fully test transportation strategies that
would alter mode split
Source Florida Dept of TransportationCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)
bull Freight considerations (3 step) Typical freight model steps
Trip generation ndash tons by commodity group
Trip distribution Mode choice ndash freight is
apportioned to transport modes
bull Joint highway assignment (trucks and passenger trips)
Source Florida Dept of TransportationCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)bull Full multi-modal structure (4 step)
Trip generation Trip distribution Mode choice Highway assignment
bull More common with mega-regions Focused on fixed guideway systems Intercity rail is important component
bull Pros Potential to model new intercity rail
including high speed rail Test greater range of potential
transportation solutions
bull Cons Requires coding of transit networks and
mode choice estimation Modeling of air travel might be needed
especially if intercity rail is a completely new mode of travel
Source California HSR AuthorityCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)
Beyond 4 Stepbull Potential components
Implement destination choice Tour-basedtrip chains Activity-based travel model Activity micro-simulation Traffic micro-simulation Integration with land use models Integration with economic models
bull Pros Ability to test a more complete set of
transportation strategies and solutions
bull Cons Extensive data requirements Cost to maintain Complexity
Source Oregon Dept of TransportationPBUniversity of Calgary
Deciding on Best Statewide Model Approach
bull What are the motivations for building a statewide travel demand modelbull What are the emerging transportation and policy issues to addressbull How significant is the impact of through travel on your statebull Do you have many multi-state MPOsbull How reliable (and current) are your survey databull Does your state have major international ports or border crossings
StudyArea
UrbanModel
AreaStudyArea
UrbanModelArea 1
Study AreaWithin OneModel Area
UrbanModelArea 2
Study AreasWithin Two Model Areas
StudyArea
UrbanModelArea
StudyArea
Statewide Model
Study AreaOutside Urban
Model Areas
Statewide ModelStatewide Model
Source Florida Department of TransportationCambridge Systematics Inc
Figure 317 Subarea Study Area Examples
Study
Area
Urban
Model
Area
Study
Area
Urban
Model
Area 1
Study Area
Within One
Model Area
Urban
Model
Area 2
Study Areas
Within Two
Model Areas
Study
Area
Urban
Model
Area
Study
Area
Statewide Model
Study Area
Outside Urban
Model Areas
Statewide Model
Statewide Model
Statewide Model Data Development Challenges
bull Primary data deficiencies Long distance travelers Rural resident trip-making
bull Challenges in compiling required data Age of data Confidentiality Coordination with data providers Proper level of geography Sufficient sample size
bull Four categories of data for discussion Networks and zone system Demographic and employment data Freight and other economic data NHTS and other behavioral data
Source Florida Department of TransportationCambridge Systematics Inc
Data Challengeshellipwhat is ldquolong distancerdquo
bull gt50 miles (American Traveler Survey)
bull gt100 miles (National Household Travel Survey)
bull IE trips (several states)
bull Trips made by residents for business personal business
bull Recreationaltourist trips
Topic ndash Similarities amp Differences with Urban Models
bull Zone systems
bull Networks
bull Travel behavior assumptions
bull Trip purposesmarket segments
bull Modal issues
bull Truck forecasting
Source Delaware Dept of TransportationWRampA Inc
Statewide vs Urban Model Zone Systems
bull Usually larger zones in statewide models than MPOs
bull Block groups are good starting point for statewide model TAZs
bull Block groups might need splitting along major transportation corridors
bull In small states MPO TAZs could equal SWM zones
bull In most states MPO TAZ equivalency tables will likely be neededdesirable
bull FAF disaggregation approach used for statewide model TAZs
Source FHWA Office of Freight Management and Operations
Source Florida Dept of TransportationCambridge Systematics Inc
FAF zones
Block groups = TAZs
Too small for statewide zones
Too big for statewide zones
Statewide vs Urban Model Networksbull Statewide model networks are sparse in
comparison with MPO models
bull At a minimum statewide networks should include Limited access highways Arterials Major collectors
bull For proper circulation patterns lower class facilities might be added
bull Outside state a very sparse network is needed for major truck corridors
bull Consistent level of detail for networks and zones
Source Tennessee Dept of TransportationCambridge Systematics Inc
MPO network
statewide network
Statewide vs Urban Model Travel Behaviorbull Understanding urban travel behavior
MPO models are usually developed with local household travel survey data
MPOs without recent HH survey data can borrow parameters from other MPO models or guidance documents such as NCHRP 716 Transferable Urban Travel Parameters
bull Rural and long-distance travel behavior A limited number of states have conducted
statewide household travel surveys American Travel Survey was designed to target
long-distance trips but was discontinued long ago
National Household Travel Survey Add-ons can be structured for a sufficient sample of rural trip-makers
NCHRP 735 Rural and Long-Distance Travel Parameters recently published for use in statewide models
Map of 2009 Add-on Surveys Source FHWA
Statewide vs Urban Model Trip Purposes
bull Typical MPO trip purposes are usually collapsed into HBW HBOHBNW and NHB for statewide models
bull Long-distance purposes Business Tourist Personal business
bull Trucks Commodity group is often
surrogate for trip purpose
bull External trips External generally means trip from
outside state Might exclude ldquohalordquo area
surrounding state boundary
Trip PurposeFL
2000 IN LA MA MS RI
Home-based Work 1914 1995 1900 1841 1606 2000
Home-based Nonwork 5547 5208 4770 5206 4917 5700
Home-based Other 2792 5208 4770 1938 4917 5700
Home-based Shop 1576 ndash ndash 1173 ndash ndash
Home-based SocialRecreation
1179 ndash ndash 1395 ndash ndash
Home-base RecreationShop ndash ndash ndash ndash ndash ndash
Home-based School ndash ndash ndash 700 ndash ndash
Nonhome-based 2539 2686 3100 2468 3289 2300
Nonhome-based Work ndash ndash ndash 1027 ndash ndash
Nonhome-based Nonwork ndash ndash ndash 1441 ndash ndash
Other ndash ndash ndash ndash ndash
Truck ndash ndash 170 ndash ndash ndash
Long-Distance 000 111 060 417 022 000
Long-Distance Business ndash ndash 020 284 011 ndash
Long-Distance Personal Business
ndash ndash ndash ndash ndash ndash
Long-Distance Tourist ndash ndash 010 027 000 ndash
Long-Distance Work ndash ndash ndash 105 ndash ndash
Long-Distance NonworkOther
ndash ndash 030 ndash 011 ndash
Statewide vs Urban Model Trip Purposes (Contrsquod)
Number of person trips
Mode Pleasure Business Personal business Other Total Percent
Personal vehicle 12185 5708 3120 789 21802 908Commercial air 1012 746 151 24 1933 8Train 74 150 14 0 238 1Other 4 2 5 12 23 01Ship 19 0 0 0 19 01
Total 13294 6606 3290 825 24015 100Percent 554 275 137 34 100
Percent by trip purpose
Mode Pleasure Business Personal business Other
Personal vehicle 917 864 948 956Commercial air 76 113 46 29Train 06 23 04 0Other 0 0 02 15Ship 01 0 0 0
Trip purpose
Trip purpose
Statewide vs Urban Model Modal Issuesbull Most statewide models are limited to highway trips
bull Modeling of local bus routes is likely overkill
bull Mode choice is usually just a conversion from person trips to vehicle trips via auto occupancy rates
bull Transit components of statewide models are more common in regions with history of fixed guideway transit
bull Transit model is needed to test options such as high speedintercity rail
ATS 95 NHTS 01 MDOT 03-04 ODOT 02-03Personal vehicle 81 89 89 88Commercial air 16 8 10 8Bus 2 2 04 19Train 05 01 0 005Other 05 09 06 205
Survey and yearPrimary mode of transport
Statewide vs Urban Model Modal Issues (Contrsquod)
3727 24 22
32
29 36
10
21
28 22
10
22
12
1529
11
44 43
0
10
20
30
40
50
60
70
80
90
100
Personalvehicle
Bus Train Marine Commercialair
Perc
ent o
f per
son
trips
lt300 300-499 500-599 1000-1999 gt2000
68
85 85 82 85
30
12 11 15 13
0
10
20
30
40
50
60
70
80
90
100
Busi-ness
Plea-sure
pBusi-ness
Vaca-tion
Week-end
Perc
ent o
f per
son
trips
Personal vehicle Commercial air All other
Mode by Trip Purpose Mode by Distance (miles)
Statewide vs Urban Model Modal Issues (Contrsquod)
bull Most statewide assignment models separate trucks from passenger cars Preloading process commonly used for trucks Takes into account more limited route options for truck travel
Source Ohio Dept of Transportation
Statewide vs Urban Model Truck Forecasting
bull Urban modelsndash Freight movements are partially
ldquoexternalrdquo to model (truncated at external zone)
ndash More general categories of truckcommercial trips used
ndash Strong reliance on Quick Response Freight Manual
bull Statewide modelsndash Rooted in commodity flows ndash Often use econometric
approaches to modelingndash Generally model entire length of
freight travel
Freight Model Type by Peer Exchange Participants and
NonParticipants
0123456789
None
Traditio
nal
Commod
ity
Econo
metric
Unkno
wn
NotPeer Exchange
Source TRB Electronic Circular E-C075 Statewide Travel Demand Modeling ndash A Peer Exchange
Topic ndash Data Needs and Considerations
bull Impact of model complexity on data needs
bull Overview of data sources
bull Long-Distance travel gt different focus than MPOregional models
Sources NHTS Weight calculation Pitfalls Key statistics European surveys Lessons learned
Impact of Model Complexity on Data Needsbull Trip table estimation (2 step) ndash highway network
zone system trip table
bull Highway model structure (3 step) ndash 2 step plusbehavioral data for trip rates friction factors auto occupancy rates
bull Freight considerations (3 step) ndash 3 step pluscommodity flows economic data tonnages by mode truck counts
bull Full multi-modal structure (4 step) ndash 3 step plustransit networks and paths logit mode choice model
bull Beyond 4 Step ndash synthetic household generation operational characteristics etc
Source Georgia DOTCambridge Systematics Inc
Statewide Model Data Sources A-Zbull American Travel Survey (ATS)
bull Bureau of Economic Analysis (BEA)
bull Census Journey-to-Work (JTW) data
bull Census Public Use Microsample (PUMS)
bull Census Transportation Planning Package (CTPP)
bull Claritas
bull Commodity Flow Survey (CFS)
Source US Census Bureau
Statewide Model Data Sources (Contrsquod)bull Dun amp Bradstreet (DampB) employment data
bull Employment Security data (ES-202)
bull Freight Analysis Framework (FAF)
bull Highway Performance Monitoring System (HPMS)
bull InfoUSA employment data
bull Metropolitan Planning Organizations (MPO)
bull NCHRP 365 Manual
bull National Highway Planning Network (NHPN)
Source FAF2 Provisional Commodity Origin-Destination Estimates Revised Methodology Report BattelleFHWA
Statewide Model Data Sources (Contrsquod)
bull National Household Travel Survey (NHTS)
bull Origin-destination surveys
bull Regional Economic Models Inc (REMI)
bull State Departments of Transportation
bull TRANSEARCH data
bull Vehicle Use and Identification Survey (VIUS)
bull Woods amp Poole employment forecasts
Source Bureau of Transportation Statistics NHTS
Data Sources on Long-Distance Travel
bull Census Journey-to-Work databull Long distance component
American Travel Survey (77 95) National Household Travel Survey (01) Ohio statewide surveys (02-04) Michigan statewide survey (04-05) California (10-12)
bull Surveys focused on long distance Oregon (96-98 07-08)
bull Others BTS Omnibus surveys Tourism surveys European practice NCHRP Report 735
NHTS and Long-Distance Travelbull Evolution of US national surveys
NPTS (69 77 83 90 95) ATS (77 95)
bull 2001 NHTS ndash Total of 60282 long distance person trips Data collected March 01 to May 02 22204 (37 percent) before 911 38078 (63 percent) afterwards
bull Long distance trip is longer than 50 miles one-way (vs 100 miles in 1995 ATS)
bull 28-day retrospective survey (including travel day)
bull Not your typical travelers Low frequency Distinct segment of the population
Weight Calculation for Long-Distance Travelers
bull Use to ensure valid population-level estimates Accounts for non-response under-coverage multiple phones etc
bull Multi-stage process Match national margins Adjust for data splitting (force sumexpfac = 277208169) Raking (match independent controls) race ethnicity race by
month ethnicity by month sex by age census region MSA status month by day of week
Trim outlier expansion factors max expfac = 4 times mean Replication (correct for sampling error)
bull At least two passes through the data
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Freight Model Types | |||||||||||||||||||
State | Type | Peer Exc | |||||||||||||||||
OR | 4 | 1 | Count of Type | Count of Type | Peer Exc | ||||||||||||||
WI | 3 | 1 | Type | Total | Type | 0 | 1 | Grand Total | |||||||||||
MO | 1 | 1 | 1 | 2 | 1 | 2 | 2 | 4 | |||||||||||
LA | 3 | 1 | 2 | 2 | 2 | 2 | 2 | 4 | |||||||||||
IN | 3 | 1 | 3 | 6 | 3 | 2 | 6 | 8 | |||||||||||
OH | 4 | 1 | 4 | 2 | 4 | 2 | 2 | ||||||||||||
KY | 3 | 1 | Grand Total | 12 | 3 | 3 | |||||||||||||
FL | 3 | 1 | Grand Total | 9 | 12 | 21 | |||||||||||||
VA | 3 | 1 | None | 2 | Not | Peer Exchange | |||||||||||||
NJ | 2 | 1 | Traditional | 2 | None | 2 | 2 | ||||||||||||
MA | 2 | 1 | Commodity | 6 | Traditional | 2 | 2 | ||||||||||||
NH | 1 | 1 | Econometric | 2 | Commodity | 2 | 6 | ||||||||||||
CA | 0 | Econometric | 2 | ||||||||||||||||
TX | 3 | 0 | Unknown | 3 | |||||||||||||||
MI | 3 | 0 | |||||||||||||||||
GA | 0 | ||||||||||||||||||
SC | 1 | 0 | |||||||||||||||||
DE | 2 | 0 | |||||||||||||||||
CT | 0 | ||||||||||||||||||
VT | 2 | 0 | |||||||||||||||||
ME | 1 | 0 | |||||||||||||||||
MN | 0 |
None | None | ||
Traditional | Traditional | ||
Commodity | Commodity | ||
Econometric | Econometric | ||
Unknown | Unknown |
MicrosimulationActivity-based Approach
bull Models dynamic evolution and actions of discrete agents across time and space
bull Observe emergent behavior Interaction Cooperation Coordination
bull Primary emphasis on understanding the system
bull Maybe forecast it
Between autonomous perceptive and deliberative agents
Source Ohio Department of Transportation
Key Differences ndash Traditional vs MicrosimulationldquoTraditional modelsrdquobull Closed form equilibriumbull Well understoodbull Typically deterministicbull Presumes central tendencies
in underlying databull Well-behavedbull ldquoManageablerdquo variationrdquobull Computationally efficient
Microsimulationbull Disequilibriumbull Large systems NWUbull Stochasticbull Diverse behaviorbull Handles wide variancesbull Accommodates weak to strong
understandingbull Computationally heavy
Zonal Land Use Data
Worker Traveler Generation
Vehicle Assignment
Starting Time Assignment
Post-Processor
Next Stop Purpose Choice
Model
Next Stop Location
Choice Model
Dynamic Activity Pattern Generation
Commercial Vehicle Trip List
Zonal Land Use Data
Worker Traveler Generation
Vehicle Assignment
Starting Time Assignment
Post-Processor
Next Stop Purpose Choice
Model
Next Stop Location
Choice Model
Dynamic Activity Pattern Generation
Commercial Vehicle Trip List
Microsimulation Examplesbull Ohiobull Oregon
Source Ohio Department of Transportation
Statewide Model Current Practicebull Meetings of the TRB Subcommittee on Statewide Travel Forecasting Models
bull NCHRP Report 735 Long-Distance and Rural Transferable Parameters for Statewide Travel Forecasting Models 2012
bull NCHRP Validation and Sensitivity Considerations for Statewide Models September 2010
bull White Paper on Statewide Modeling presented at TRB Statewide Planning Conference September 2008
bull National Travel Demand Forecasting Model Phase I Final Scope 2008
bull Forecasting Statewide Freight Toolkit NCHRP Report 606 2008
bull TRB Annual Meeting Sessions in 2004 2006 and 2008 web page
bull Statewide Travel Forecasting Models NCHRP Synthesis 358 2006
bull Statewide Travel Demand Modeling Peer Exchange Transportation Research Circular E-C075 September 2004
bull Guidebook on Statewide Travel Forecasting and Irvine Conference Transportation Research Circular E-C011 1999
Topic ndash Statewide Model Theory
bull Common themes in statewide model development
bull Alternative structures Trip table estimation (2 step) Highway model structure (3 step) Fully multimodal (4 step) Beyond 4 step
bull Deciding on best approach
bull Data development challengesSource Massachusetts Highway DeptCambridge Systematics Inc
Common Themes in Review of Statewide Models
bull Most statewide models use similar methodologies to metropolitan models but there are many variations
bull The largest problems relate to issues of scale
bull Models from different states vary greatly in complexity cost and development time
bull Models are most successful when addressing statewide priorities
bull There are major deficiencies in available data about long-distance and rural passenger travelhellip NCHRP Report 735 has helped
bull Statewide models are more compatible with secondary freight data sources such as the Commodity Flow Survey than are metropolitan models
bull Interest is high among states in progress in deploying models
Source A Horowitz White Paper on Statewide Modeling presented at TRB Conference on Statewide Planning 2008
Alternative Statewide Model Structures
bull Trip table estimation (2 step) Estimate initial ldquoseed matrixrdquo
Roadside intercept surveys American Travel Survey National HH Travel SurveyAdd-ons Census Journey-to-Work flows
Matrix estimation program Inputs ndash seed matrix highway
network traffic counts Adjusts origin-destination flows
between zones to optimize matching network traffic counts
Fratar factors applied to synthetic trip table in estimating future growth
bull Trip table estimation (2 step) Pros
Limited need to prepare aggregate socioeconomic input data
Simplicity of model generally means short execution times
Forecasting needs are limited to zone or district growth factors
Cons Less explanatory power than most
MPOregional models Fixed distribution pattern Difficulties in assessing benefits of
adding new roadway capacity
Alternative Statewide Model Structures (Contrsquod)
bull Highway model structure (3 step) Trip generation Trip distribution Highway assignment
bull Mode choice excluded Either generate vehicle trips or Apply auto occupancy factors to person trip
tables
bull Pros Benefits of trip redistribution across range of
alternatives Greater sensitivity to roadway capacity and
congestion
bull Cons Likely reliance on MPOsothers to provide
socioeconomic input data Canrsquot fully test transportation strategies that
would alter mode split
Source Florida Dept of TransportationCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)
bull Freight considerations (3 step) Typical freight model steps
Trip generation ndash tons by commodity group
Trip distribution Mode choice ndash freight is
apportioned to transport modes
bull Joint highway assignment (trucks and passenger trips)
Source Florida Dept of TransportationCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)bull Full multi-modal structure (4 step)
Trip generation Trip distribution Mode choice Highway assignment
bull More common with mega-regions Focused on fixed guideway systems Intercity rail is important component
bull Pros Potential to model new intercity rail
including high speed rail Test greater range of potential
transportation solutions
bull Cons Requires coding of transit networks and
mode choice estimation Modeling of air travel might be needed
especially if intercity rail is a completely new mode of travel
Source California HSR AuthorityCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)
Beyond 4 Stepbull Potential components
Implement destination choice Tour-basedtrip chains Activity-based travel model Activity micro-simulation Traffic micro-simulation Integration with land use models Integration with economic models
bull Pros Ability to test a more complete set of
transportation strategies and solutions
bull Cons Extensive data requirements Cost to maintain Complexity
Source Oregon Dept of TransportationPBUniversity of Calgary
Deciding on Best Statewide Model Approach
bull What are the motivations for building a statewide travel demand modelbull What are the emerging transportation and policy issues to addressbull How significant is the impact of through travel on your statebull Do you have many multi-state MPOsbull How reliable (and current) are your survey databull Does your state have major international ports or border crossings
StudyArea
UrbanModel
AreaStudyArea
UrbanModelArea 1
Study AreaWithin OneModel Area
UrbanModelArea 2
Study AreasWithin Two Model Areas
StudyArea
UrbanModelArea
StudyArea
Statewide Model
Study AreaOutside Urban
Model Areas
Statewide ModelStatewide Model
Source Florida Department of TransportationCambridge Systematics Inc
Figure 317 Subarea Study Area Examples
Study
Area
Urban
Model
Area
Study
Area
Urban
Model
Area 1
Study Area
Within One
Model Area
Urban
Model
Area 2
Study Areas
Within Two
Model Areas
Study
Area
Urban
Model
Area
Study
Area
Statewide Model
Study Area
Outside Urban
Model Areas
Statewide Model
Statewide Model
Statewide Model Data Development Challenges
bull Primary data deficiencies Long distance travelers Rural resident trip-making
bull Challenges in compiling required data Age of data Confidentiality Coordination with data providers Proper level of geography Sufficient sample size
bull Four categories of data for discussion Networks and zone system Demographic and employment data Freight and other economic data NHTS and other behavioral data
Source Florida Department of TransportationCambridge Systematics Inc
Data Challengeshellipwhat is ldquolong distancerdquo
bull gt50 miles (American Traveler Survey)
bull gt100 miles (National Household Travel Survey)
bull IE trips (several states)
bull Trips made by residents for business personal business
bull Recreationaltourist trips
Topic ndash Similarities amp Differences with Urban Models
bull Zone systems
bull Networks
bull Travel behavior assumptions
bull Trip purposesmarket segments
bull Modal issues
bull Truck forecasting
Source Delaware Dept of TransportationWRampA Inc
Statewide vs Urban Model Zone Systems
bull Usually larger zones in statewide models than MPOs
bull Block groups are good starting point for statewide model TAZs
bull Block groups might need splitting along major transportation corridors
bull In small states MPO TAZs could equal SWM zones
bull In most states MPO TAZ equivalency tables will likely be neededdesirable
bull FAF disaggregation approach used for statewide model TAZs
Source FHWA Office of Freight Management and Operations
Source Florida Dept of TransportationCambridge Systematics Inc
FAF zones
Block groups = TAZs
Too small for statewide zones
Too big for statewide zones
Statewide vs Urban Model Networksbull Statewide model networks are sparse in
comparison with MPO models
bull At a minimum statewide networks should include Limited access highways Arterials Major collectors
bull For proper circulation patterns lower class facilities might be added
bull Outside state a very sparse network is needed for major truck corridors
bull Consistent level of detail for networks and zones
Source Tennessee Dept of TransportationCambridge Systematics Inc
MPO network
statewide network
Statewide vs Urban Model Travel Behaviorbull Understanding urban travel behavior
MPO models are usually developed with local household travel survey data
MPOs without recent HH survey data can borrow parameters from other MPO models or guidance documents such as NCHRP 716 Transferable Urban Travel Parameters
bull Rural and long-distance travel behavior A limited number of states have conducted
statewide household travel surveys American Travel Survey was designed to target
long-distance trips but was discontinued long ago
National Household Travel Survey Add-ons can be structured for a sufficient sample of rural trip-makers
NCHRP 735 Rural and Long-Distance Travel Parameters recently published for use in statewide models
Map of 2009 Add-on Surveys Source FHWA
Statewide vs Urban Model Trip Purposes
bull Typical MPO trip purposes are usually collapsed into HBW HBOHBNW and NHB for statewide models
bull Long-distance purposes Business Tourist Personal business
bull Trucks Commodity group is often
surrogate for trip purpose
bull External trips External generally means trip from
outside state Might exclude ldquohalordquo area
surrounding state boundary
Trip PurposeFL
2000 IN LA MA MS RI
Home-based Work 1914 1995 1900 1841 1606 2000
Home-based Nonwork 5547 5208 4770 5206 4917 5700
Home-based Other 2792 5208 4770 1938 4917 5700
Home-based Shop 1576 ndash ndash 1173 ndash ndash
Home-based SocialRecreation
1179 ndash ndash 1395 ndash ndash
Home-base RecreationShop ndash ndash ndash ndash ndash ndash
Home-based School ndash ndash ndash 700 ndash ndash
Nonhome-based 2539 2686 3100 2468 3289 2300
Nonhome-based Work ndash ndash ndash 1027 ndash ndash
Nonhome-based Nonwork ndash ndash ndash 1441 ndash ndash
Other ndash ndash ndash ndash ndash
Truck ndash ndash 170 ndash ndash ndash
Long-Distance 000 111 060 417 022 000
Long-Distance Business ndash ndash 020 284 011 ndash
Long-Distance Personal Business
ndash ndash ndash ndash ndash ndash
Long-Distance Tourist ndash ndash 010 027 000 ndash
Long-Distance Work ndash ndash ndash 105 ndash ndash
Long-Distance NonworkOther
ndash ndash 030 ndash 011 ndash
Statewide vs Urban Model Trip Purposes (Contrsquod)
Number of person trips
Mode Pleasure Business Personal business Other Total Percent
Personal vehicle 12185 5708 3120 789 21802 908Commercial air 1012 746 151 24 1933 8Train 74 150 14 0 238 1Other 4 2 5 12 23 01Ship 19 0 0 0 19 01
Total 13294 6606 3290 825 24015 100Percent 554 275 137 34 100
Percent by trip purpose
Mode Pleasure Business Personal business Other
Personal vehicle 917 864 948 956Commercial air 76 113 46 29Train 06 23 04 0Other 0 0 02 15Ship 01 0 0 0
Trip purpose
Trip purpose
Statewide vs Urban Model Modal Issuesbull Most statewide models are limited to highway trips
bull Modeling of local bus routes is likely overkill
bull Mode choice is usually just a conversion from person trips to vehicle trips via auto occupancy rates
bull Transit components of statewide models are more common in regions with history of fixed guideway transit
bull Transit model is needed to test options such as high speedintercity rail
ATS 95 NHTS 01 MDOT 03-04 ODOT 02-03Personal vehicle 81 89 89 88Commercial air 16 8 10 8Bus 2 2 04 19Train 05 01 0 005Other 05 09 06 205
Survey and yearPrimary mode of transport
Statewide vs Urban Model Modal Issues (Contrsquod)
3727 24 22
32
29 36
10
21
28 22
10
22
12
1529
11
44 43
0
10
20
30
40
50
60
70
80
90
100
Personalvehicle
Bus Train Marine Commercialair
Perc
ent o
f per
son
trips
lt300 300-499 500-599 1000-1999 gt2000
68
85 85 82 85
30
12 11 15 13
0
10
20
30
40
50
60
70
80
90
100
Busi-ness
Plea-sure
pBusi-ness
Vaca-tion
Week-end
Perc
ent o
f per
son
trips
Personal vehicle Commercial air All other
Mode by Trip Purpose Mode by Distance (miles)
Statewide vs Urban Model Modal Issues (Contrsquod)
bull Most statewide assignment models separate trucks from passenger cars Preloading process commonly used for trucks Takes into account more limited route options for truck travel
Source Ohio Dept of Transportation
Statewide vs Urban Model Truck Forecasting
bull Urban modelsndash Freight movements are partially
ldquoexternalrdquo to model (truncated at external zone)
ndash More general categories of truckcommercial trips used
ndash Strong reliance on Quick Response Freight Manual
bull Statewide modelsndash Rooted in commodity flows ndash Often use econometric
approaches to modelingndash Generally model entire length of
freight travel
Freight Model Type by Peer Exchange Participants and
NonParticipants
0123456789
None
Traditio
nal
Commod
ity
Econo
metric
Unkno
wn
NotPeer Exchange
Source TRB Electronic Circular E-C075 Statewide Travel Demand Modeling ndash A Peer Exchange
Topic ndash Data Needs and Considerations
bull Impact of model complexity on data needs
bull Overview of data sources
bull Long-Distance travel gt different focus than MPOregional models
Sources NHTS Weight calculation Pitfalls Key statistics European surveys Lessons learned
Impact of Model Complexity on Data Needsbull Trip table estimation (2 step) ndash highway network
zone system trip table
bull Highway model structure (3 step) ndash 2 step plusbehavioral data for trip rates friction factors auto occupancy rates
bull Freight considerations (3 step) ndash 3 step pluscommodity flows economic data tonnages by mode truck counts
bull Full multi-modal structure (4 step) ndash 3 step plustransit networks and paths logit mode choice model
bull Beyond 4 Step ndash synthetic household generation operational characteristics etc
Source Georgia DOTCambridge Systematics Inc
Statewide Model Data Sources A-Zbull American Travel Survey (ATS)
bull Bureau of Economic Analysis (BEA)
bull Census Journey-to-Work (JTW) data
bull Census Public Use Microsample (PUMS)
bull Census Transportation Planning Package (CTPP)
bull Claritas
bull Commodity Flow Survey (CFS)
Source US Census Bureau
Statewide Model Data Sources (Contrsquod)bull Dun amp Bradstreet (DampB) employment data
bull Employment Security data (ES-202)
bull Freight Analysis Framework (FAF)
bull Highway Performance Monitoring System (HPMS)
bull InfoUSA employment data
bull Metropolitan Planning Organizations (MPO)
bull NCHRP 365 Manual
bull National Highway Planning Network (NHPN)
Source FAF2 Provisional Commodity Origin-Destination Estimates Revised Methodology Report BattelleFHWA
Statewide Model Data Sources (Contrsquod)
bull National Household Travel Survey (NHTS)
bull Origin-destination surveys
bull Regional Economic Models Inc (REMI)
bull State Departments of Transportation
bull TRANSEARCH data
bull Vehicle Use and Identification Survey (VIUS)
bull Woods amp Poole employment forecasts
Source Bureau of Transportation Statistics NHTS
Data Sources on Long-Distance Travel
bull Census Journey-to-Work databull Long distance component
American Travel Survey (77 95) National Household Travel Survey (01) Ohio statewide surveys (02-04) Michigan statewide survey (04-05) California (10-12)
bull Surveys focused on long distance Oregon (96-98 07-08)
bull Others BTS Omnibus surveys Tourism surveys European practice NCHRP Report 735
NHTS and Long-Distance Travelbull Evolution of US national surveys
NPTS (69 77 83 90 95) ATS (77 95)
bull 2001 NHTS ndash Total of 60282 long distance person trips Data collected March 01 to May 02 22204 (37 percent) before 911 38078 (63 percent) afterwards
bull Long distance trip is longer than 50 miles one-way (vs 100 miles in 1995 ATS)
bull 28-day retrospective survey (including travel day)
bull Not your typical travelers Low frequency Distinct segment of the population
Weight Calculation for Long-Distance Travelers
bull Use to ensure valid population-level estimates Accounts for non-response under-coverage multiple phones etc
bull Multi-stage process Match national margins Adjust for data splitting (force sumexpfac = 277208169) Raking (match independent controls) race ethnicity race by
month ethnicity by month sex by age census region MSA status month by day of week
Trim outlier expansion factors max expfac = 4 times mean Replication (correct for sampling error)
bull At least two passes through the data
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Freight Model Types | |||||||||||||||||||
State | Type | Peer Exc | |||||||||||||||||
OR | 4 | 1 | Count of Type | Count of Type | Peer Exc | ||||||||||||||
WI | 3 | 1 | Type | Total | Type | 0 | 1 | Grand Total | |||||||||||
MO | 1 | 1 | 1 | 2 | 1 | 2 | 2 | 4 | |||||||||||
LA | 3 | 1 | 2 | 2 | 2 | 2 | 2 | 4 | |||||||||||
IN | 3 | 1 | 3 | 6 | 3 | 2 | 6 | 8 | |||||||||||
OH | 4 | 1 | 4 | 2 | 4 | 2 | 2 | ||||||||||||
KY | 3 | 1 | Grand Total | 12 | 3 | 3 | |||||||||||||
FL | 3 | 1 | Grand Total | 9 | 12 | 21 | |||||||||||||
VA | 3 | 1 | None | 2 | Not | Peer Exchange | |||||||||||||
NJ | 2 | 1 | Traditional | 2 | None | 2 | 2 | ||||||||||||
MA | 2 | 1 | Commodity | 6 | Traditional | 2 | 2 | ||||||||||||
NH | 1 | 1 | Econometric | 2 | Commodity | 2 | 6 | ||||||||||||
CA | 0 | Econometric | 2 | ||||||||||||||||
TX | 3 | 0 | Unknown | 3 | |||||||||||||||
MI | 3 | 0 | |||||||||||||||||
GA | 0 | ||||||||||||||||||
SC | 1 | 0 | |||||||||||||||||
DE | 2 | 0 | |||||||||||||||||
CT | 0 | ||||||||||||||||||
VT | 2 | 0 | |||||||||||||||||
ME | 1 | 0 | |||||||||||||||||
MN | 0 |
None | None | ||
Traditional | Traditional | ||
Commodity | Commodity | ||
Econometric | Econometric | ||
Unknown | Unknown |
Key Differences ndash Traditional vs MicrosimulationldquoTraditional modelsrdquobull Closed form equilibriumbull Well understoodbull Typically deterministicbull Presumes central tendencies
in underlying databull Well-behavedbull ldquoManageablerdquo variationrdquobull Computationally efficient
Microsimulationbull Disequilibriumbull Large systems NWUbull Stochasticbull Diverse behaviorbull Handles wide variancesbull Accommodates weak to strong
understandingbull Computationally heavy
Zonal Land Use Data
Worker Traveler Generation
Vehicle Assignment
Starting Time Assignment
Post-Processor
Next Stop Purpose Choice
Model
Next Stop Location
Choice Model
Dynamic Activity Pattern Generation
Commercial Vehicle Trip List
Zonal Land Use Data
Worker Traveler Generation
Vehicle Assignment
Starting Time Assignment
Post-Processor
Next Stop Purpose Choice
Model
Next Stop Location
Choice Model
Dynamic Activity Pattern Generation
Commercial Vehicle Trip List
Microsimulation Examplesbull Ohiobull Oregon
Source Ohio Department of Transportation
Statewide Model Current Practicebull Meetings of the TRB Subcommittee on Statewide Travel Forecasting Models
bull NCHRP Report 735 Long-Distance and Rural Transferable Parameters for Statewide Travel Forecasting Models 2012
bull NCHRP Validation and Sensitivity Considerations for Statewide Models September 2010
bull White Paper on Statewide Modeling presented at TRB Statewide Planning Conference September 2008
bull National Travel Demand Forecasting Model Phase I Final Scope 2008
bull Forecasting Statewide Freight Toolkit NCHRP Report 606 2008
bull TRB Annual Meeting Sessions in 2004 2006 and 2008 web page
bull Statewide Travel Forecasting Models NCHRP Synthesis 358 2006
bull Statewide Travel Demand Modeling Peer Exchange Transportation Research Circular E-C075 September 2004
bull Guidebook on Statewide Travel Forecasting and Irvine Conference Transportation Research Circular E-C011 1999
Topic ndash Statewide Model Theory
bull Common themes in statewide model development
bull Alternative structures Trip table estimation (2 step) Highway model structure (3 step) Fully multimodal (4 step) Beyond 4 step
bull Deciding on best approach
bull Data development challengesSource Massachusetts Highway DeptCambridge Systematics Inc
Common Themes in Review of Statewide Models
bull Most statewide models use similar methodologies to metropolitan models but there are many variations
bull The largest problems relate to issues of scale
bull Models from different states vary greatly in complexity cost and development time
bull Models are most successful when addressing statewide priorities
bull There are major deficiencies in available data about long-distance and rural passenger travelhellip NCHRP Report 735 has helped
bull Statewide models are more compatible with secondary freight data sources such as the Commodity Flow Survey than are metropolitan models
bull Interest is high among states in progress in deploying models
Source A Horowitz White Paper on Statewide Modeling presented at TRB Conference on Statewide Planning 2008
Alternative Statewide Model Structures
bull Trip table estimation (2 step) Estimate initial ldquoseed matrixrdquo
Roadside intercept surveys American Travel Survey National HH Travel SurveyAdd-ons Census Journey-to-Work flows
Matrix estimation program Inputs ndash seed matrix highway
network traffic counts Adjusts origin-destination flows
between zones to optimize matching network traffic counts
Fratar factors applied to synthetic trip table in estimating future growth
bull Trip table estimation (2 step) Pros
Limited need to prepare aggregate socioeconomic input data
Simplicity of model generally means short execution times
Forecasting needs are limited to zone or district growth factors
Cons Less explanatory power than most
MPOregional models Fixed distribution pattern Difficulties in assessing benefits of
adding new roadway capacity
Alternative Statewide Model Structures (Contrsquod)
bull Highway model structure (3 step) Trip generation Trip distribution Highway assignment
bull Mode choice excluded Either generate vehicle trips or Apply auto occupancy factors to person trip
tables
bull Pros Benefits of trip redistribution across range of
alternatives Greater sensitivity to roadway capacity and
congestion
bull Cons Likely reliance on MPOsothers to provide
socioeconomic input data Canrsquot fully test transportation strategies that
would alter mode split
Source Florida Dept of TransportationCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)
bull Freight considerations (3 step) Typical freight model steps
Trip generation ndash tons by commodity group
Trip distribution Mode choice ndash freight is
apportioned to transport modes
bull Joint highway assignment (trucks and passenger trips)
Source Florida Dept of TransportationCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)bull Full multi-modal structure (4 step)
Trip generation Trip distribution Mode choice Highway assignment
bull More common with mega-regions Focused on fixed guideway systems Intercity rail is important component
bull Pros Potential to model new intercity rail
including high speed rail Test greater range of potential
transportation solutions
bull Cons Requires coding of transit networks and
mode choice estimation Modeling of air travel might be needed
especially if intercity rail is a completely new mode of travel
Source California HSR AuthorityCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)
Beyond 4 Stepbull Potential components
Implement destination choice Tour-basedtrip chains Activity-based travel model Activity micro-simulation Traffic micro-simulation Integration with land use models Integration with economic models
bull Pros Ability to test a more complete set of
transportation strategies and solutions
bull Cons Extensive data requirements Cost to maintain Complexity
Source Oregon Dept of TransportationPBUniversity of Calgary
Deciding on Best Statewide Model Approach
bull What are the motivations for building a statewide travel demand modelbull What are the emerging transportation and policy issues to addressbull How significant is the impact of through travel on your statebull Do you have many multi-state MPOsbull How reliable (and current) are your survey databull Does your state have major international ports or border crossings
StudyArea
UrbanModel
AreaStudyArea
UrbanModelArea 1
Study AreaWithin OneModel Area
UrbanModelArea 2
Study AreasWithin Two Model Areas
StudyArea
UrbanModelArea
StudyArea
Statewide Model
Study AreaOutside Urban
Model Areas
Statewide ModelStatewide Model
Source Florida Department of TransportationCambridge Systematics Inc
Figure 317 Subarea Study Area Examples
Study
Area
Urban
Model
Area
Study
Area
Urban
Model
Area 1
Study Area
Within One
Model Area
Urban
Model
Area 2
Study Areas
Within Two
Model Areas
Study
Area
Urban
Model
Area
Study
Area
Statewide Model
Study Area
Outside Urban
Model Areas
Statewide Model
Statewide Model
Statewide Model Data Development Challenges
bull Primary data deficiencies Long distance travelers Rural resident trip-making
bull Challenges in compiling required data Age of data Confidentiality Coordination with data providers Proper level of geography Sufficient sample size
bull Four categories of data for discussion Networks and zone system Demographic and employment data Freight and other economic data NHTS and other behavioral data
Source Florida Department of TransportationCambridge Systematics Inc
Data Challengeshellipwhat is ldquolong distancerdquo
bull gt50 miles (American Traveler Survey)
bull gt100 miles (National Household Travel Survey)
bull IE trips (several states)
bull Trips made by residents for business personal business
bull Recreationaltourist trips
Topic ndash Similarities amp Differences with Urban Models
bull Zone systems
bull Networks
bull Travel behavior assumptions
bull Trip purposesmarket segments
bull Modal issues
bull Truck forecasting
Source Delaware Dept of TransportationWRampA Inc
Statewide vs Urban Model Zone Systems
bull Usually larger zones in statewide models than MPOs
bull Block groups are good starting point for statewide model TAZs
bull Block groups might need splitting along major transportation corridors
bull In small states MPO TAZs could equal SWM zones
bull In most states MPO TAZ equivalency tables will likely be neededdesirable
bull FAF disaggregation approach used for statewide model TAZs
Source FHWA Office of Freight Management and Operations
Source Florida Dept of TransportationCambridge Systematics Inc
FAF zones
Block groups = TAZs
Too small for statewide zones
Too big for statewide zones
Statewide vs Urban Model Networksbull Statewide model networks are sparse in
comparison with MPO models
bull At a minimum statewide networks should include Limited access highways Arterials Major collectors
bull For proper circulation patterns lower class facilities might be added
bull Outside state a very sparse network is needed for major truck corridors
bull Consistent level of detail for networks and zones
Source Tennessee Dept of TransportationCambridge Systematics Inc
MPO network
statewide network
Statewide vs Urban Model Travel Behaviorbull Understanding urban travel behavior
MPO models are usually developed with local household travel survey data
MPOs without recent HH survey data can borrow parameters from other MPO models or guidance documents such as NCHRP 716 Transferable Urban Travel Parameters
bull Rural and long-distance travel behavior A limited number of states have conducted
statewide household travel surveys American Travel Survey was designed to target
long-distance trips but was discontinued long ago
National Household Travel Survey Add-ons can be structured for a sufficient sample of rural trip-makers
NCHRP 735 Rural and Long-Distance Travel Parameters recently published for use in statewide models
Map of 2009 Add-on Surveys Source FHWA
Statewide vs Urban Model Trip Purposes
bull Typical MPO trip purposes are usually collapsed into HBW HBOHBNW and NHB for statewide models
bull Long-distance purposes Business Tourist Personal business
bull Trucks Commodity group is often
surrogate for trip purpose
bull External trips External generally means trip from
outside state Might exclude ldquohalordquo area
surrounding state boundary
Trip PurposeFL
2000 IN LA MA MS RI
Home-based Work 1914 1995 1900 1841 1606 2000
Home-based Nonwork 5547 5208 4770 5206 4917 5700
Home-based Other 2792 5208 4770 1938 4917 5700
Home-based Shop 1576 ndash ndash 1173 ndash ndash
Home-based SocialRecreation
1179 ndash ndash 1395 ndash ndash
Home-base RecreationShop ndash ndash ndash ndash ndash ndash
Home-based School ndash ndash ndash 700 ndash ndash
Nonhome-based 2539 2686 3100 2468 3289 2300
Nonhome-based Work ndash ndash ndash 1027 ndash ndash
Nonhome-based Nonwork ndash ndash ndash 1441 ndash ndash
Other ndash ndash ndash ndash ndash
Truck ndash ndash 170 ndash ndash ndash
Long-Distance 000 111 060 417 022 000
Long-Distance Business ndash ndash 020 284 011 ndash
Long-Distance Personal Business
ndash ndash ndash ndash ndash ndash
Long-Distance Tourist ndash ndash 010 027 000 ndash
Long-Distance Work ndash ndash ndash 105 ndash ndash
Long-Distance NonworkOther
ndash ndash 030 ndash 011 ndash
Statewide vs Urban Model Trip Purposes (Contrsquod)
Number of person trips
Mode Pleasure Business Personal business Other Total Percent
Personal vehicle 12185 5708 3120 789 21802 908Commercial air 1012 746 151 24 1933 8Train 74 150 14 0 238 1Other 4 2 5 12 23 01Ship 19 0 0 0 19 01
Total 13294 6606 3290 825 24015 100Percent 554 275 137 34 100
Percent by trip purpose
Mode Pleasure Business Personal business Other
Personal vehicle 917 864 948 956Commercial air 76 113 46 29Train 06 23 04 0Other 0 0 02 15Ship 01 0 0 0
Trip purpose
Trip purpose
Statewide vs Urban Model Modal Issuesbull Most statewide models are limited to highway trips
bull Modeling of local bus routes is likely overkill
bull Mode choice is usually just a conversion from person trips to vehicle trips via auto occupancy rates
bull Transit components of statewide models are more common in regions with history of fixed guideway transit
bull Transit model is needed to test options such as high speedintercity rail
ATS 95 NHTS 01 MDOT 03-04 ODOT 02-03Personal vehicle 81 89 89 88Commercial air 16 8 10 8Bus 2 2 04 19Train 05 01 0 005Other 05 09 06 205
Survey and yearPrimary mode of transport
Statewide vs Urban Model Modal Issues (Contrsquod)
3727 24 22
32
29 36
10
21
28 22
10
22
12
1529
11
44 43
0
10
20
30
40
50
60
70
80
90
100
Personalvehicle
Bus Train Marine Commercialair
Perc
ent o
f per
son
trips
lt300 300-499 500-599 1000-1999 gt2000
68
85 85 82 85
30
12 11 15 13
0
10
20
30
40
50
60
70
80
90
100
Busi-ness
Plea-sure
pBusi-ness
Vaca-tion
Week-end
Perc
ent o
f per
son
trips
Personal vehicle Commercial air All other
Mode by Trip Purpose Mode by Distance (miles)
Statewide vs Urban Model Modal Issues (Contrsquod)
bull Most statewide assignment models separate trucks from passenger cars Preloading process commonly used for trucks Takes into account more limited route options for truck travel
Source Ohio Dept of Transportation
Statewide vs Urban Model Truck Forecasting
bull Urban modelsndash Freight movements are partially
ldquoexternalrdquo to model (truncated at external zone)
ndash More general categories of truckcommercial trips used
ndash Strong reliance on Quick Response Freight Manual
bull Statewide modelsndash Rooted in commodity flows ndash Often use econometric
approaches to modelingndash Generally model entire length of
freight travel
Freight Model Type by Peer Exchange Participants and
NonParticipants
0123456789
None
Traditio
nal
Commod
ity
Econo
metric
Unkno
wn
NotPeer Exchange
Source TRB Electronic Circular E-C075 Statewide Travel Demand Modeling ndash A Peer Exchange
Topic ndash Data Needs and Considerations
bull Impact of model complexity on data needs
bull Overview of data sources
bull Long-Distance travel gt different focus than MPOregional models
Sources NHTS Weight calculation Pitfalls Key statistics European surveys Lessons learned
Impact of Model Complexity on Data Needsbull Trip table estimation (2 step) ndash highway network
zone system trip table
bull Highway model structure (3 step) ndash 2 step plusbehavioral data for trip rates friction factors auto occupancy rates
bull Freight considerations (3 step) ndash 3 step pluscommodity flows economic data tonnages by mode truck counts
bull Full multi-modal structure (4 step) ndash 3 step plustransit networks and paths logit mode choice model
bull Beyond 4 Step ndash synthetic household generation operational characteristics etc
Source Georgia DOTCambridge Systematics Inc
Statewide Model Data Sources A-Zbull American Travel Survey (ATS)
bull Bureau of Economic Analysis (BEA)
bull Census Journey-to-Work (JTW) data
bull Census Public Use Microsample (PUMS)
bull Census Transportation Planning Package (CTPP)
bull Claritas
bull Commodity Flow Survey (CFS)
Source US Census Bureau
Statewide Model Data Sources (Contrsquod)bull Dun amp Bradstreet (DampB) employment data
bull Employment Security data (ES-202)
bull Freight Analysis Framework (FAF)
bull Highway Performance Monitoring System (HPMS)
bull InfoUSA employment data
bull Metropolitan Planning Organizations (MPO)
bull NCHRP 365 Manual
bull National Highway Planning Network (NHPN)
Source FAF2 Provisional Commodity Origin-Destination Estimates Revised Methodology Report BattelleFHWA
Statewide Model Data Sources (Contrsquod)
bull National Household Travel Survey (NHTS)
bull Origin-destination surveys
bull Regional Economic Models Inc (REMI)
bull State Departments of Transportation
bull TRANSEARCH data
bull Vehicle Use and Identification Survey (VIUS)
bull Woods amp Poole employment forecasts
Source Bureau of Transportation Statistics NHTS
Data Sources on Long-Distance Travel
bull Census Journey-to-Work databull Long distance component
American Travel Survey (77 95) National Household Travel Survey (01) Ohio statewide surveys (02-04) Michigan statewide survey (04-05) California (10-12)
bull Surveys focused on long distance Oregon (96-98 07-08)
bull Others BTS Omnibus surveys Tourism surveys European practice NCHRP Report 735
NHTS and Long-Distance Travelbull Evolution of US national surveys
NPTS (69 77 83 90 95) ATS (77 95)
bull 2001 NHTS ndash Total of 60282 long distance person trips Data collected March 01 to May 02 22204 (37 percent) before 911 38078 (63 percent) afterwards
bull Long distance trip is longer than 50 miles one-way (vs 100 miles in 1995 ATS)
bull 28-day retrospective survey (including travel day)
bull Not your typical travelers Low frequency Distinct segment of the population
Weight Calculation for Long-Distance Travelers
bull Use to ensure valid population-level estimates Accounts for non-response under-coverage multiple phones etc
bull Multi-stage process Match national margins Adjust for data splitting (force sumexpfac = 277208169) Raking (match independent controls) race ethnicity race by
month ethnicity by month sex by age census region MSA status month by day of week
Trim outlier expansion factors max expfac = 4 times mean Replication (correct for sampling error)
bull At least two passes through the data
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Freight Model Types | |||||||||||||||||||
State | Type | Peer Exc | |||||||||||||||||
OR | 4 | 1 | Count of Type | Count of Type | Peer Exc | ||||||||||||||
WI | 3 | 1 | Type | Total | Type | 0 | 1 | Grand Total | |||||||||||
MO | 1 | 1 | 1 | 2 | 1 | 2 | 2 | 4 | |||||||||||
LA | 3 | 1 | 2 | 2 | 2 | 2 | 2 | 4 | |||||||||||
IN | 3 | 1 | 3 | 6 | 3 | 2 | 6 | 8 | |||||||||||
OH | 4 | 1 | 4 | 2 | 4 | 2 | 2 | ||||||||||||
KY | 3 | 1 | Grand Total | 12 | 3 | 3 | |||||||||||||
FL | 3 | 1 | Grand Total | 9 | 12 | 21 | |||||||||||||
VA | 3 | 1 | None | 2 | Not | Peer Exchange | |||||||||||||
NJ | 2 | 1 | Traditional | 2 | None | 2 | 2 | ||||||||||||
MA | 2 | 1 | Commodity | 6 | Traditional | 2 | 2 | ||||||||||||
NH | 1 | 1 | Econometric | 2 | Commodity | 2 | 6 | ||||||||||||
CA | 0 | Econometric | 2 | ||||||||||||||||
TX | 3 | 0 | Unknown | 3 | |||||||||||||||
MI | 3 | 0 | |||||||||||||||||
GA | 0 | ||||||||||||||||||
SC | 1 | 0 | |||||||||||||||||
DE | 2 | 0 | |||||||||||||||||
CT | 0 | ||||||||||||||||||
VT | 2 | 0 | |||||||||||||||||
ME | 1 | 0 | |||||||||||||||||
MN | 0 |
None | None | ||
Traditional | Traditional | ||
Commodity | Commodity | ||
Econometric | Econometric | ||
Unknown | Unknown |
Statewide Model Current Practicebull Meetings of the TRB Subcommittee on Statewide Travel Forecasting Models
bull NCHRP Report 735 Long-Distance and Rural Transferable Parameters for Statewide Travel Forecasting Models 2012
bull NCHRP Validation and Sensitivity Considerations for Statewide Models September 2010
bull White Paper on Statewide Modeling presented at TRB Statewide Planning Conference September 2008
bull National Travel Demand Forecasting Model Phase I Final Scope 2008
bull Forecasting Statewide Freight Toolkit NCHRP Report 606 2008
bull TRB Annual Meeting Sessions in 2004 2006 and 2008 web page
bull Statewide Travel Forecasting Models NCHRP Synthesis 358 2006
bull Statewide Travel Demand Modeling Peer Exchange Transportation Research Circular E-C075 September 2004
bull Guidebook on Statewide Travel Forecasting and Irvine Conference Transportation Research Circular E-C011 1999
Topic ndash Statewide Model Theory
bull Common themes in statewide model development
bull Alternative structures Trip table estimation (2 step) Highway model structure (3 step) Fully multimodal (4 step) Beyond 4 step
bull Deciding on best approach
bull Data development challengesSource Massachusetts Highway DeptCambridge Systematics Inc
Common Themes in Review of Statewide Models
bull Most statewide models use similar methodologies to metropolitan models but there are many variations
bull The largest problems relate to issues of scale
bull Models from different states vary greatly in complexity cost and development time
bull Models are most successful when addressing statewide priorities
bull There are major deficiencies in available data about long-distance and rural passenger travelhellip NCHRP Report 735 has helped
bull Statewide models are more compatible with secondary freight data sources such as the Commodity Flow Survey than are metropolitan models
bull Interest is high among states in progress in deploying models
Source A Horowitz White Paper on Statewide Modeling presented at TRB Conference on Statewide Planning 2008
Alternative Statewide Model Structures
bull Trip table estimation (2 step) Estimate initial ldquoseed matrixrdquo
Roadside intercept surveys American Travel Survey National HH Travel SurveyAdd-ons Census Journey-to-Work flows
Matrix estimation program Inputs ndash seed matrix highway
network traffic counts Adjusts origin-destination flows
between zones to optimize matching network traffic counts
Fratar factors applied to synthetic trip table in estimating future growth
bull Trip table estimation (2 step) Pros
Limited need to prepare aggregate socioeconomic input data
Simplicity of model generally means short execution times
Forecasting needs are limited to zone or district growth factors
Cons Less explanatory power than most
MPOregional models Fixed distribution pattern Difficulties in assessing benefits of
adding new roadway capacity
Alternative Statewide Model Structures (Contrsquod)
bull Highway model structure (3 step) Trip generation Trip distribution Highway assignment
bull Mode choice excluded Either generate vehicle trips or Apply auto occupancy factors to person trip
tables
bull Pros Benefits of trip redistribution across range of
alternatives Greater sensitivity to roadway capacity and
congestion
bull Cons Likely reliance on MPOsothers to provide
socioeconomic input data Canrsquot fully test transportation strategies that
would alter mode split
Source Florida Dept of TransportationCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)
bull Freight considerations (3 step) Typical freight model steps
Trip generation ndash tons by commodity group
Trip distribution Mode choice ndash freight is
apportioned to transport modes
bull Joint highway assignment (trucks and passenger trips)
Source Florida Dept of TransportationCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)bull Full multi-modal structure (4 step)
Trip generation Trip distribution Mode choice Highway assignment
bull More common with mega-regions Focused on fixed guideway systems Intercity rail is important component
bull Pros Potential to model new intercity rail
including high speed rail Test greater range of potential
transportation solutions
bull Cons Requires coding of transit networks and
mode choice estimation Modeling of air travel might be needed
especially if intercity rail is a completely new mode of travel
Source California HSR AuthorityCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)
Beyond 4 Stepbull Potential components
Implement destination choice Tour-basedtrip chains Activity-based travel model Activity micro-simulation Traffic micro-simulation Integration with land use models Integration with economic models
bull Pros Ability to test a more complete set of
transportation strategies and solutions
bull Cons Extensive data requirements Cost to maintain Complexity
Source Oregon Dept of TransportationPBUniversity of Calgary
Deciding on Best Statewide Model Approach
bull What are the motivations for building a statewide travel demand modelbull What are the emerging transportation and policy issues to addressbull How significant is the impact of through travel on your statebull Do you have many multi-state MPOsbull How reliable (and current) are your survey databull Does your state have major international ports or border crossings
StudyArea
UrbanModel
AreaStudyArea
UrbanModelArea 1
Study AreaWithin OneModel Area
UrbanModelArea 2
Study AreasWithin Two Model Areas
StudyArea
UrbanModelArea
StudyArea
Statewide Model
Study AreaOutside Urban
Model Areas
Statewide ModelStatewide Model
Source Florida Department of TransportationCambridge Systematics Inc
Figure 317 Subarea Study Area Examples
Study
Area
Urban
Model
Area
Study
Area
Urban
Model
Area 1
Study Area
Within One
Model Area
Urban
Model
Area 2
Study Areas
Within Two
Model Areas
Study
Area
Urban
Model
Area
Study
Area
Statewide Model
Study Area
Outside Urban
Model Areas
Statewide Model
Statewide Model
Statewide Model Data Development Challenges
bull Primary data deficiencies Long distance travelers Rural resident trip-making
bull Challenges in compiling required data Age of data Confidentiality Coordination with data providers Proper level of geography Sufficient sample size
bull Four categories of data for discussion Networks and zone system Demographic and employment data Freight and other economic data NHTS and other behavioral data
Source Florida Department of TransportationCambridge Systematics Inc
Data Challengeshellipwhat is ldquolong distancerdquo
bull gt50 miles (American Traveler Survey)
bull gt100 miles (National Household Travel Survey)
bull IE trips (several states)
bull Trips made by residents for business personal business
bull Recreationaltourist trips
Topic ndash Similarities amp Differences with Urban Models
bull Zone systems
bull Networks
bull Travel behavior assumptions
bull Trip purposesmarket segments
bull Modal issues
bull Truck forecasting
Source Delaware Dept of TransportationWRampA Inc
Statewide vs Urban Model Zone Systems
bull Usually larger zones in statewide models than MPOs
bull Block groups are good starting point for statewide model TAZs
bull Block groups might need splitting along major transportation corridors
bull In small states MPO TAZs could equal SWM zones
bull In most states MPO TAZ equivalency tables will likely be neededdesirable
bull FAF disaggregation approach used for statewide model TAZs
Source FHWA Office of Freight Management and Operations
Source Florida Dept of TransportationCambridge Systematics Inc
FAF zones
Block groups = TAZs
Too small for statewide zones
Too big for statewide zones
Statewide vs Urban Model Networksbull Statewide model networks are sparse in
comparison with MPO models
bull At a minimum statewide networks should include Limited access highways Arterials Major collectors
bull For proper circulation patterns lower class facilities might be added
bull Outside state a very sparse network is needed for major truck corridors
bull Consistent level of detail for networks and zones
Source Tennessee Dept of TransportationCambridge Systematics Inc
MPO network
statewide network
Statewide vs Urban Model Travel Behaviorbull Understanding urban travel behavior
MPO models are usually developed with local household travel survey data
MPOs without recent HH survey data can borrow parameters from other MPO models or guidance documents such as NCHRP 716 Transferable Urban Travel Parameters
bull Rural and long-distance travel behavior A limited number of states have conducted
statewide household travel surveys American Travel Survey was designed to target
long-distance trips but was discontinued long ago
National Household Travel Survey Add-ons can be structured for a sufficient sample of rural trip-makers
NCHRP 735 Rural and Long-Distance Travel Parameters recently published for use in statewide models
Map of 2009 Add-on Surveys Source FHWA
Statewide vs Urban Model Trip Purposes
bull Typical MPO trip purposes are usually collapsed into HBW HBOHBNW and NHB for statewide models
bull Long-distance purposes Business Tourist Personal business
bull Trucks Commodity group is often
surrogate for trip purpose
bull External trips External generally means trip from
outside state Might exclude ldquohalordquo area
surrounding state boundary
Trip PurposeFL
2000 IN LA MA MS RI
Home-based Work 1914 1995 1900 1841 1606 2000
Home-based Nonwork 5547 5208 4770 5206 4917 5700
Home-based Other 2792 5208 4770 1938 4917 5700
Home-based Shop 1576 ndash ndash 1173 ndash ndash
Home-based SocialRecreation
1179 ndash ndash 1395 ndash ndash
Home-base RecreationShop ndash ndash ndash ndash ndash ndash
Home-based School ndash ndash ndash 700 ndash ndash
Nonhome-based 2539 2686 3100 2468 3289 2300
Nonhome-based Work ndash ndash ndash 1027 ndash ndash
Nonhome-based Nonwork ndash ndash ndash 1441 ndash ndash
Other ndash ndash ndash ndash ndash
Truck ndash ndash 170 ndash ndash ndash
Long-Distance 000 111 060 417 022 000
Long-Distance Business ndash ndash 020 284 011 ndash
Long-Distance Personal Business
ndash ndash ndash ndash ndash ndash
Long-Distance Tourist ndash ndash 010 027 000 ndash
Long-Distance Work ndash ndash ndash 105 ndash ndash
Long-Distance NonworkOther
ndash ndash 030 ndash 011 ndash
Statewide vs Urban Model Trip Purposes (Contrsquod)
Number of person trips
Mode Pleasure Business Personal business Other Total Percent
Personal vehicle 12185 5708 3120 789 21802 908Commercial air 1012 746 151 24 1933 8Train 74 150 14 0 238 1Other 4 2 5 12 23 01Ship 19 0 0 0 19 01
Total 13294 6606 3290 825 24015 100Percent 554 275 137 34 100
Percent by trip purpose
Mode Pleasure Business Personal business Other
Personal vehicle 917 864 948 956Commercial air 76 113 46 29Train 06 23 04 0Other 0 0 02 15Ship 01 0 0 0
Trip purpose
Trip purpose
Statewide vs Urban Model Modal Issuesbull Most statewide models are limited to highway trips
bull Modeling of local bus routes is likely overkill
bull Mode choice is usually just a conversion from person trips to vehicle trips via auto occupancy rates
bull Transit components of statewide models are more common in regions with history of fixed guideway transit
bull Transit model is needed to test options such as high speedintercity rail
ATS 95 NHTS 01 MDOT 03-04 ODOT 02-03Personal vehicle 81 89 89 88Commercial air 16 8 10 8Bus 2 2 04 19Train 05 01 0 005Other 05 09 06 205
Survey and yearPrimary mode of transport
Statewide vs Urban Model Modal Issues (Contrsquod)
3727 24 22
32
29 36
10
21
28 22
10
22
12
1529
11
44 43
0
10
20
30
40
50
60
70
80
90
100
Personalvehicle
Bus Train Marine Commercialair
Perc
ent o
f per
son
trips
lt300 300-499 500-599 1000-1999 gt2000
68
85 85 82 85
30
12 11 15 13
0
10
20
30
40
50
60
70
80
90
100
Busi-ness
Plea-sure
pBusi-ness
Vaca-tion
Week-end
Perc
ent o
f per
son
trips
Personal vehicle Commercial air All other
Mode by Trip Purpose Mode by Distance (miles)
Statewide vs Urban Model Modal Issues (Contrsquod)
bull Most statewide assignment models separate trucks from passenger cars Preloading process commonly used for trucks Takes into account more limited route options for truck travel
Source Ohio Dept of Transportation
Statewide vs Urban Model Truck Forecasting
bull Urban modelsndash Freight movements are partially
ldquoexternalrdquo to model (truncated at external zone)
ndash More general categories of truckcommercial trips used
ndash Strong reliance on Quick Response Freight Manual
bull Statewide modelsndash Rooted in commodity flows ndash Often use econometric
approaches to modelingndash Generally model entire length of
freight travel
Freight Model Type by Peer Exchange Participants and
NonParticipants
0123456789
None
Traditio
nal
Commod
ity
Econo
metric
Unkno
wn
NotPeer Exchange
Source TRB Electronic Circular E-C075 Statewide Travel Demand Modeling ndash A Peer Exchange
Topic ndash Data Needs and Considerations
bull Impact of model complexity on data needs
bull Overview of data sources
bull Long-Distance travel gt different focus than MPOregional models
Sources NHTS Weight calculation Pitfalls Key statistics European surveys Lessons learned
Impact of Model Complexity on Data Needsbull Trip table estimation (2 step) ndash highway network
zone system trip table
bull Highway model structure (3 step) ndash 2 step plusbehavioral data for trip rates friction factors auto occupancy rates
bull Freight considerations (3 step) ndash 3 step pluscommodity flows economic data tonnages by mode truck counts
bull Full multi-modal structure (4 step) ndash 3 step plustransit networks and paths logit mode choice model
bull Beyond 4 Step ndash synthetic household generation operational characteristics etc
Source Georgia DOTCambridge Systematics Inc
Statewide Model Data Sources A-Zbull American Travel Survey (ATS)
bull Bureau of Economic Analysis (BEA)
bull Census Journey-to-Work (JTW) data
bull Census Public Use Microsample (PUMS)
bull Census Transportation Planning Package (CTPP)
bull Claritas
bull Commodity Flow Survey (CFS)
Source US Census Bureau
Statewide Model Data Sources (Contrsquod)bull Dun amp Bradstreet (DampB) employment data
bull Employment Security data (ES-202)
bull Freight Analysis Framework (FAF)
bull Highway Performance Monitoring System (HPMS)
bull InfoUSA employment data
bull Metropolitan Planning Organizations (MPO)
bull NCHRP 365 Manual
bull National Highway Planning Network (NHPN)
Source FAF2 Provisional Commodity Origin-Destination Estimates Revised Methodology Report BattelleFHWA
Statewide Model Data Sources (Contrsquod)
bull National Household Travel Survey (NHTS)
bull Origin-destination surveys
bull Regional Economic Models Inc (REMI)
bull State Departments of Transportation
bull TRANSEARCH data
bull Vehicle Use and Identification Survey (VIUS)
bull Woods amp Poole employment forecasts
Source Bureau of Transportation Statistics NHTS
Data Sources on Long-Distance Travel
bull Census Journey-to-Work databull Long distance component
American Travel Survey (77 95) National Household Travel Survey (01) Ohio statewide surveys (02-04) Michigan statewide survey (04-05) California (10-12)
bull Surveys focused on long distance Oregon (96-98 07-08)
bull Others BTS Omnibus surveys Tourism surveys European practice NCHRP Report 735
NHTS and Long-Distance Travelbull Evolution of US national surveys
NPTS (69 77 83 90 95) ATS (77 95)
bull 2001 NHTS ndash Total of 60282 long distance person trips Data collected March 01 to May 02 22204 (37 percent) before 911 38078 (63 percent) afterwards
bull Long distance trip is longer than 50 miles one-way (vs 100 miles in 1995 ATS)
bull 28-day retrospective survey (including travel day)
bull Not your typical travelers Low frequency Distinct segment of the population
Weight Calculation for Long-Distance Travelers
bull Use to ensure valid population-level estimates Accounts for non-response under-coverage multiple phones etc
bull Multi-stage process Match national margins Adjust for data splitting (force sumexpfac = 277208169) Raking (match independent controls) race ethnicity race by
month ethnicity by month sex by age census region MSA status month by day of week
Trim outlier expansion factors max expfac = 4 times mean Replication (correct for sampling error)
bull At least two passes through the data
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Freight Model Types | |||||||||||||||||||
State | Type | Peer Exc | |||||||||||||||||
OR | 4 | 1 | Count of Type | Count of Type | Peer Exc | ||||||||||||||
WI | 3 | 1 | Type | Total | Type | 0 | 1 | Grand Total | |||||||||||
MO | 1 | 1 | 1 | 2 | 1 | 2 | 2 | 4 | |||||||||||
LA | 3 | 1 | 2 | 2 | 2 | 2 | 2 | 4 | |||||||||||
IN | 3 | 1 | 3 | 6 | 3 | 2 | 6 | 8 | |||||||||||
OH | 4 | 1 | 4 | 2 | 4 | 2 | 2 | ||||||||||||
KY | 3 | 1 | Grand Total | 12 | 3 | 3 | |||||||||||||
FL | 3 | 1 | Grand Total | 9 | 12 | 21 | |||||||||||||
VA | 3 | 1 | None | 2 | Not | Peer Exchange | |||||||||||||
NJ | 2 | 1 | Traditional | 2 | None | 2 | 2 | ||||||||||||
MA | 2 | 1 | Commodity | 6 | Traditional | 2 | 2 | ||||||||||||
NH | 1 | 1 | Econometric | 2 | Commodity | 2 | 6 | ||||||||||||
CA | 0 | Econometric | 2 | ||||||||||||||||
TX | 3 | 0 | Unknown | 3 | |||||||||||||||
MI | 3 | 0 | |||||||||||||||||
GA | 0 | ||||||||||||||||||
SC | 1 | 0 | |||||||||||||||||
DE | 2 | 0 | |||||||||||||||||
CT | 0 | ||||||||||||||||||
VT | 2 | 0 | |||||||||||||||||
ME | 1 | 0 | |||||||||||||||||
MN | 0 |
None | None | ||
Traditional | Traditional | ||
Commodity | Commodity | ||
Econometric | Econometric | ||
Unknown | Unknown |
Topic ndash Statewide Model Theory
bull Common themes in statewide model development
bull Alternative structures Trip table estimation (2 step) Highway model structure (3 step) Fully multimodal (4 step) Beyond 4 step
bull Deciding on best approach
bull Data development challengesSource Massachusetts Highway DeptCambridge Systematics Inc
Common Themes in Review of Statewide Models
bull Most statewide models use similar methodologies to metropolitan models but there are many variations
bull The largest problems relate to issues of scale
bull Models from different states vary greatly in complexity cost and development time
bull Models are most successful when addressing statewide priorities
bull There are major deficiencies in available data about long-distance and rural passenger travelhellip NCHRP Report 735 has helped
bull Statewide models are more compatible with secondary freight data sources such as the Commodity Flow Survey than are metropolitan models
bull Interest is high among states in progress in deploying models
Source A Horowitz White Paper on Statewide Modeling presented at TRB Conference on Statewide Planning 2008
Alternative Statewide Model Structures
bull Trip table estimation (2 step) Estimate initial ldquoseed matrixrdquo
Roadside intercept surveys American Travel Survey National HH Travel SurveyAdd-ons Census Journey-to-Work flows
Matrix estimation program Inputs ndash seed matrix highway
network traffic counts Adjusts origin-destination flows
between zones to optimize matching network traffic counts
Fratar factors applied to synthetic trip table in estimating future growth
bull Trip table estimation (2 step) Pros
Limited need to prepare aggregate socioeconomic input data
Simplicity of model generally means short execution times
Forecasting needs are limited to zone or district growth factors
Cons Less explanatory power than most
MPOregional models Fixed distribution pattern Difficulties in assessing benefits of
adding new roadway capacity
Alternative Statewide Model Structures (Contrsquod)
bull Highway model structure (3 step) Trip generation Trip distribution Highway assignment
bull Mode choice excluded Either generate vehicle trips or Apply auto occupancy factors to person trip
tables
bull Pros Benefits of trip redistribution across range of
alternatives Greater sensitivity to roadway capacity and
congestion
bull Cons Likely reliance on MPOsothers to provide
socioeconomic input data Canrsquot fully test transportation strategies that
would alter mode split
Source Florida Dept of TransportationCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)
bull Freight considerations (3 step) Typical freight model steps
Trip generation ndash tons by commodity group
Trip distribution Mode choice ndash freight is
apportioned to transport modes
bull Joint highway assignment (trucks and passenger trips)
Source Florida Dept of TransportationCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)bull Full multi-modal structure (4 step)
Trip generation Trip distribution Mode choice Highway assignment
bull More common with mega-regions Focused on fixed guideway systems Intercity rail is important component
bull Pros Potential to model new intercity rail
including high speed rail Test greater range of potential
transportation solutions
bull Cons Requires coding of transit networks and
mode choice estimation Modeling of air travel might be needed
especially if intercity rail is a completely new mode of travel
Source California HSR AuthorityCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)
Beyond 4 Stepbull Potential components
Implement destination choice Tour-basedtrip chains Activity-based travel model Activity micro-simulation Traffic micro-simulation Integration with land use models Integration with economic models
bull Pros Ability to test a more complete set of
transportation strategies and solutions
bull Cons Extensive data requirements Cost to maintain Complexity
Source Oregon Dept of TransportationPBUniversity of Calgary
Deciding on Best Statewide Model Approach
bull What are the motivations for building a statewide travel demand modelbull What are the emerging transportation and policy issues to addressbull How significant is the impact of through travel on your statebull Do you have many multi-state MPOsbull How reliable (and current) are your survey databull Does your state have major international ports or border crossings
StudyArea
UrbanModel
AreaStudyArea
UrbanModelArea 1
Study AreaWithin OneModel Area
UrbanModelArea 2
Study AreasWithin Two Model Areas
StudyArea
UrbanModelArea
StudyArea
Statewide Model
Study AreaOutside Urban
Model Areas
Statewide ModelStatewide Model
Source Florida Department of TransportationCambridge Systematics Inc
Figure 317 Subarea Study Area Examples
Study
Area
Urban
Model
Area
Study
Area
Urban
Model
Area 1
Study Area
Within One
Model Area
Urban
Model
Area 2
Study Areas
Within Two
Model Areas
Study
Area
Urban
Model
Area
Study
Area
Statewide Model
Study Area
Outside Urban
Model Areas
Statewide Model
Statewide Model
Statewide Model Data Development Challenges
bull Primary data deficiencies Long distance travelers Rural resident trip-making
bull Challenges in compiling required data Age of data Confidentiality Coordination with data providers Proper level of geography Sufficient sample size
bull Four categories of data for discussion Networks and zone system Demographic and employment data Freight and other economic data NHTS and other behavioral data
Source Florida Department of TransportationCambridge Systematics Inc
Data Challengeshellipwhat is ldquolong distancerdquo
bull gt50 miles (American Traveler Survey)
bull gt100 miles (National Household Travel Survey)
bull IE trips (several states)
bull Trips made by residents for business personal business
bull Recreationaltourist trips
Topic ndash Similarities amp Differences with Urban Models
bull Zone systems
bull Networks
bull Travel behavior assumptions
bull Trip purposesmarket segments
bull Modal issues
bull Truck forecasting
Source Delaware Dept of TransportationWRampA Inc
Statewide vs Urban Model Zone Systems
bull Usually larger zones in statewide models than MPOs
bull Block groups are good starting point for statewide model TAZs
bull Block groups might need splitting along major transportation corridors
bull In small states MPO TAZs could equal SWM zones
bull In most states MPO TAZ equivalency tables will likely be neededdesirable
bull FAF disaggregation approach used for statewide model TAZs
Source FHWA Office of Freight Management and Operations
Source Florida Dept of TransportationCambridge Systematics Inc
FAF zones
Block groups = TAZs
Too small for statewide zones
Too big for statewide zones
Statewide vs Urban Model Networksbull Statewide model networks are sparse in
comparison with MPO models
bull At a minimum statewide networks should include Limited access highways Arterials Major collectors
bull For proper circulation patterns lower class facilities might be added
bull Outside state a very sparse network is needed for major truck corridors
bull Consistent level of detail for networks and zones
Source Tennessee Dept of TransportationCambridge Systematics Inc
MPO network
statewide network
Statewide vs Urban Model Travel Behaviorbull Understanding urban travel behavior
MPO models are usually developed with local household travel survey data
MPOs without recent HH survey data can borrow parameters from other MPO models or guidance documents such as NCHRP 716 Transferable Urban Travel Parameters
bull Rural and long-distance travel behavior A limited number of states have conducted
statewide household travel surveys American Travel Survey was designed to target
long-distance trips but was discontinued long ago
National Household Travel Survey Add-ons can be structured for a sufficient sample of rural trip-makers
NCHRP 735 Rural and Long-Distance Travel Parameters recently published for use in statewide models
Map of 2009 Add-on Surveys Source FHWA
Statewide vs Urban Model Trip Purposes
bull Typical MPO trip purposes are usually collapsed into HBW HBOHBNW and NHB for statewide models
bull Long-distance purposes Business Tourist Personal business
bull Trucks Commodity group is often
surrogate for trip purpose
bull External trips External generally means trip from
outside state Might exclude ldquohalordquo area
surrounding state boundary
Trip PurposeFL
2000 IN LA MA MS RI
Home-based Work 1914 1995 1900 1841 1606 2000
Home-based Nonwork 5547 5208 4770 5206 4917 5700
Home-based Other 2792 5208 4770 1938 4917 5700
Home-based Shop 1576 ndash ndash 1173 ndash ndash
Home-based SocialRecreation
1179 ndash ndash 1395 ndash ndash
Home-base RecreationShop ndash ndash ndash ndash ndash ndash
Home-based School ndash ndash ndash 700 ndash ndash
Nonhome-based 2539 2686 3100 2468 3289 2300
Nonhome-based Work ndash ndash ndash 1027 ndash ndash
Nonhome-based Nonwork ndash ndash ndash 1441 ndash ndash
Other ndash ndash ndash ndash ndash
Truck ndash ndash 170 ndash ndash ndash
Long-Distance 000 111 060 417 022 000
Long-Distance Business ndash ndash 020 284 011 ndash
Long-Distance Personal Business
ndash ndash ndash ndash ndash ndash
Long-Distance Tourist ndash ndash 010 027 000 ndash
Long-Distance Work ndash ndash ndash 105 ndash ndash
Long-Distance NonworkOther
ndash ndash 030 ndash 011 ndash
Statewide vs Urban Model Trip Purposes (Contrsquod)
Number of person trips
Mode Pleasure Business Personal business Other Total Percent
Personal vehicle 12185 5708 3120 789 21802 908Commercial air 1012 746 151 24 1933 8Train 74 150 14 0 238 1Other 4 2 5 12 23 01Ship 19 0 0 0 19 01
Total 13294 6606 3290 825 24015 100Percent 554 275 137 34 100
Percent by trip purpose
Mode Pleasure Business Personal business Other
Personal vehicle 917 864 948 956Commercial air 76 113 46 29Train 06 23 04 0Other 0 0 02 15Ship 01 0 0 0
Trip purpose
Trip purpose
Statewide vs Urban Model Modal Issuesbull Most statewide models are limited to highway trips
bull Modeling of local bus routes is likely overkill
bull Mode choice is usually just a conversion from person trips to vehicle trips via auto occupancy rates
bull Transit components of statewide models are more common in regions with history of fixed guideway transit
bull Transit model is needed to test options such as high speedintercity rail
ATS 95 NHTS 01 MDOT 03-04 ODOT 02-03Personal vehicle 81 89 89 88Commercial air 16 8 10 8Bus 2 2 04 19Train 05 01 0 005Other 05 09 06 205
Survey and yearPrimary mode of transport
Statewide vs Urban Model Modal Issues (Contrsquod)
3727 24 22
32
29 36
10
21
28 22
10
22
12
1529
11
44 43
0
10
20
30
40
50
60
70
80
90
100
Personalvehicle
Bus Train Marine Commercialair
Perc
ent o
f per
son
trips
lt300 300-499 500-599 1000-1999 gt2000
68
85 85 82 85
30
12 11 15 13
0
10
20
30
40
50
60
70
80
90
100
Busi-ness
Plea-sure
pBusi-ness
Vaca-tion
Week-end
Perc
ent o
f per
son
trips
Personal vehicle Commercial air All other
Mode by Trip Purpose Mode by Distance (miles)
Statewide vs Urban Model Modal Issues (Contrsquod)
bull Most statewide assignment models separate trucks from passenger cars Preloading process commonly used for trucks Takes into account more limited route options for truck travel
Source Ohio Dept of Transportation
Statewide vs Urban Model Truck Forecasting
bull Urban modelsndash Freight movements are partially
ldquoexternalrdquo to model (truncated at external zone)
ndash More general categories of truckcommercial trips used
ndash Strong reliance on Quick Response Freight Manual
bull Statewide modelsndash Rooted in commodity flows ndash Often use econometric
approaches to modelingndash Generally model entire length of
freight travel
Freight Model Type by Peer Exchange Participants and
NonParticipants
0123456789
None
Traditio
nal
Commod
ity
Econo
metric
Unkno
wn
NotPeer Exchange
Source TRB Electronic Circular E-C075 Statewide Travel Demand Modeling ndash A Peer Exchange
Topic ndash Data Needs and Considerations
bull Impact of model complexity on data needs
bull Overview of data sources
bull Long-Distance travel gt different focus than MPOregional models
Sources NHTS Weight calculation Pitfalls Key statistics European surveys Lessons learned
Impact of Model Complexity on Data Needsbull Trip table estimation (2 step) ndash highway network
zone system trip table
bull Highway model structure (3 step) ndash 2 step plusbehavioral data for trip rates friction factors auto occupancy rates
bull Freight considerations (3 step) ndash 3 step pluscommodity flows economic data tonnages by mode truck counts
bull Full multi-modal structure (4 step) ndash 3 step plustransit networks and paths logit mode choice model
bull Beyond 4 Step ndash synthetic household generation operational characteristics etc
Source Georgia DOTCambridge Systematics Inc
Statewide Model Data Sources A-Zbull American Travel Survey (ATS)
bull Bureau of Economic Analysis (BEA)
bull Census Journey-to-Work (JTW) data
bull Census Public Use Microsample (PUMS)
bull Census Transportation Planning Package (CTPP)
bull Claritas
bull Commodity Flow Survey (CFS)
Source US Census Bureau
Statewide Model Data Sources (Contrsquod)bull Dun amp Bradstreet (DampB) employment data
bull Employment Security data (ES-202)
bull Freight Analysis Framework (FAF)
bull Highway Performance Monitoring System (HPMS)
bull InfoUSA employment data
bull Metropolitan Planning Organizations (MPO)
bull NCHRP 365 Manual
bull National Highway Planning Network (NHPN)
Source FAF2 Provisional Commodity Origin-Destination Estimates Revised Methodology Report BattelleFHWA
Statewide Model Data Sources (Contrsquod)
bull National Household Travel Survey (NHTS)
bull Origin-destination surveys
bull Regional Economic Models Inc (REMI)
bull State Departments of Transportation
bull TRANSEARCH data
bull Vehicle Use and Identification Survey (VIUS)
bull Woods amp Poole employment forecasts
Source Bureau of Transportation Statistics NHTS
Data Sources on Long-Distance Travel
bull Census Journey-to-Work databull Long distance component
American Travel Survey (77 95) National Household Travel Survey (01) Ohio statewide surveys (02-04) Michigan statewide survey (04-05) California (10-12)
bull Surveys focused on long distance Oregon (96-98 07-08)
bull Others BTS Omnibus surveys Tourism surveys European practice NCHRP Report 735
NHTS and Long-Distance Travelbull Evolution of US national surveys
NPTS (69 77 83 90 95) ATS (77 95)
bull 2001 NHTS ndash Total of 60282 long distance person trips Data collected March 01 to May 02 22204 (37 percent) before 911 38078 (63 percent) afterwards
bull Long distance trip is longer than 50 miles one-way (vs 100 miles in 1995 ATS)
bull 28-day retrospective survey (including travel day)
bull Not your typical travelers Low frequency Distinct segment of the population
Weight Calculation for Long-Distance Travelers
bull Use to ensure valid population-level estimates Accounts for non-response under-coverage multiple phones etc
bull Multi-stage process Match national margins Adjust for data splitting (force sumexpfac = 277208169) Raking (match independent controls) race ethnicity race by
month ethnicity by month sex by age census region MSA status month by day of week
Trim outlier expansion factors max expfac = 4 times mean Replication (correct for sampling error)
bull At least two passes through the data
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Freight Model Types | |||||||||||||||||||
State | Type | Peer Exc | |||||||||||||||||
OR | 4 | 1 | Count of Type | Count of Type | Peer Exc | ||||||||||||||
WI | 3 | 1 | Type | Total | Type | 0 | 1 | Grand Total | |||||||||||
MO | 1 | 1 | 1 | 2 | 1 | 2 | 2 | 4 | |||||||||||
LA | 3 | 1 | 2 | 2 | 2 | 2 | 2 | 4 | |||||||||||
IN | 3 | 1 | 3 | 6 | 3 | 2 | 6 | 8 | |||||||||||
OH | 4 | 1 | 4 | 2 | 4 | 2 | 2 | ||||||||||||
KY | 3 | 1 | Grand Total | 12 | 3 | 3 | |||||||||||||
FL | 3 | 1 | Grand Total | 9 | 12 | 21 | |||||||||||||
VA | 3 | 1 | None | 2 | Not | Peer Exchange | |||||||||||||
NJ | 2 | 1 | Traditional | 2 | None | 2 | 2 | ||||||||||||
MA | 2 | 1 | Commodity | 6 | Traditional | 2 | 2 | ||||||||||||
NH | 1 | 1 | Econometric | 2 | Commodity | 2 | 6 | ||||||||||||
CA | 0 | Econometric | 2 | ||||||||||||||||
TX | 3 | 0 | Unknown | 3 | |||||||||||||||
MI | 3 | 0 | |||||||||||||||||
GA | 0 | ||||||||||||||||||
SC | 1 | 0 | |||||||||||||||||
DE | 2 | 0 | |||||||||||||||||
CT | 0 | ||||||||||||||||||
VT | 2 | 0 | |||||||||||||||||
ME | 1 | 0 | |||||||||||||||||
MN | 0 |
None | None | ||
Traditional | Traditional | ||
Commodity | Commodity | ||
Econometric | Econometric | ||
Unknown | Unknown |
Common Themes in Review of Statewide Models
bull Most statewide models use similar methodologies to metropolitan models but there are many variations
bull The largest problems relate to issues of scale
bull Models from different states vary greatly in complexity cost and development time
bull Models are most successful when addressing statewide priorities
bull There are major deficiencies in available data about long-distance and rural passenger travelhellip NCHRP Report 735 has helped
bull Statewide models are more compatible with secondary freight data sources such as the Commodity Flow Survey than are metropolitan models
bull Interest is high among states in progress in deploying models
Source A Horowitz White Paper on Statewide Modeling presented at TRB Conference on Statewide Planning 2008
Alternative Statewide Model Structures
bull Trip table estimation (2 step) Estimate initial ldquoseed matrixrdquo
Roadside intercept surveys American Travel Survey National HH Travel SurveyAdd-ons Census Journey-to-Work flows
Matrix estimation program Inputs ndash seed matrix highway
network traffic counts Adjusts origin-destination flows
between zones to optimize matching network traffic counts
Fratar factors applied to synthetic trip table in estimating future growth
bull Trip table estimation (2 step) Pros
Limited need to prepare aggregate socioeconomic input data
Simplicity of model generally means short execution times
Forecasting needs are limited to zone or district growth factors
Cons Less explanatory power than most
MPOregional models Fixed distribution pattern Difficulties in assessing benefits of
adding new roadway capacity
Alternative Statewide Model Structures (Contrsquod)
bull Highway model structure (3 step) Trip generation Trip distribution Highway assignment
bull Mode choice excluded Either generate vehicle trips or Apply auto occupancy factors to person trip
tables
bull Pros Benefits of trip redistribution across range of
alternatives Greater sensitivity to roadway capacity and
congestion
bull Cons Likely reliance on MPOsothers to provide
socioeconomic input data Canrsquot fully test transportation strategies that
would alter mode split
Source Florida Dept of TransportationCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)
bull Freight considerations (3 step) Typical freight model steps
Trip generation ndash tons by commodity group
Trip distribution Mode choice ndash freight is
apportioned to transport modes
bull Joint highway assignment (trucks and passenger trips)
Source Florida Dept of TransportationCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)bull Full multi-modal structure (4 step)
Trip generation Trip distribution Mode choice Highway assignment
bull More common with mega-regions Focused on fixed guideway systems Intercity rail is important component
bull Pros Potential to model new intercity rail
including high speed rail Test greater range of potential
transportation solutions
bull Cons Requires coding of transit networks and
mode choice estimation Modeling of air travel might be needed
especially if intercity rail is a completely new mode of travel
Source California HSR AuthorityCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)
Beyond 4 Stepbull Potential components
Implement destination choice Tour-basedtrip chains Activity-based travel model Activity micro-simulation Traffic micro-simulation Integration with land use models Integration with economic models
bull Pros Ability to test a more complete set of
transportation strategies and solutions
bull Cons Extensive data requirements Cost to maintain Complexity
Source Oregon Dept of TransportationPBUniversity of Calgary
Deciding on Best Statewide Model Approach
bull What are the motivations for building a statewide travel demand modelbull What are the emerging transportation and policy issues to addressbull How significant is the impact of through travel on your statebull Do you have many multi-state MPOsbull How reliable (and current) are your survey databull Does your state have major international ports or border crossings
StudyArea
UrbanModel
AreaStudyArea
UrbanModelArea 1
Study AreaWithin OneModel Area
UrbanModelArea 2
Study AreasWithin Two Model Areas
StudyArea
UrbanModelArea
StudyArea
Statewide Model
Study AreaOutside Urban
Model Areas
Statewide ModelStatewide Model
Source Florida Department of TransportationCambridge Systematics Inc
Figure 317 Subarea Study Area Examples
Study
Area
Urban
Model
Area
Study
Area
Urban
Model
Area 1
Study Area
Within One
Model Area
Urban
Model
Area 2
Study Areas
Within Two
Model Areas
Study
Area
Urban
Model
Area
Study
Area
Statewide Model
Study Area
Outside Urban
Model Areas
Statewide Model
Statewide Model
Statewide Model Data Development Challenges
bull Primary data deficiencies Long distance travelers Rural resident trip-making
bull Challenges in compiling required data Age of data Confidentiality Coordination with data providers Proper level of geography Sufficient sample size
bull Four categories of data for discussion Networks and zone system Demographic and employment data Freight and other economic data NHTS and other behavioral data
Source Florida Department of TransportationCambridge Systematics Inc
Data Challengeshellipwhat is ldquolong distancerdquo
bull gt50 miles (American Traveler Survey)
bull gt100 miles (National Household Travel Survey)
bull IE trips (several states)
bull Trips made by residents for business personal business
bull Recreationaltourist trips
Topic ndash Similarities amp Differences with Urban Models
bull Zone systems
bull Networks
bull Travel behavior assumptions
bull Trip purposesmarket segments
bull Modal issues
bull Truck forecasting
Source Delaware Dept of TransportationWRampA Inc
Statewide vs Urban Model Zone Systems
bull Usually larger zones in statewide models than MPOs
bull Block groups are good starting point for statewide model TAZs
bull Block groups might need splitting along major transportation corridors
bull In small states MPO TAZs could equal SWM zones
bull In most states MPO TAZ equivalency tables will likely be neededdesirable
bull FAF disaggregation approach used for statewide model TAZs
Source FHWA Office of Freight Management and Operations
Source Florida Dept of TransportationCambridge Systematics Inc
FAF zones
Block groups = TAZs
Too small for statewide zones
Too big for statewide zones
Statewide vs Urban Model Networksbull Statewide model networks are sparse in
comparison with MPO models
bull At a minimum statewide networks should include Limited access highways Arterials Major collectors
bull For proper circulation patterns lower class facilities might be added
bull Outside state a very sparse network is needed for major truck corridors
bull Consistent level of detail for networks and zones
Source Tennessee Dept of TransportationCambridge Systematics Inc
MPO network
statewide network
Statewide vs Urban Model Travel Behaviorbull Understanding urban travel behavior
MPO models are usually developed with local household travel survey data
MPOs without recent HH survey data can borrow parameters from other MPO models or guidance documents such as NCHRP 716 Transferable Urban Travel Parameters
bull Rural and long-distance travel behavior A limited number of states have conducted
statewide household travel surveys American Travel Survey was designed to target
long-distance trips but was discontinued long ago
National Household Travel Survey Add-ons can be structured for a sufficient sample of rural trip-makers
NCHRP 735 Rural and Long-Distance Travel Parameters recently published for use in statewide models
Map of 2009 Add-on Surveys Source FHWA
Statewide vs Urban Model Trip Purposes
bull Typical MPO trip purposes are usually collapsed into HBW HBOHBNW and NHB for statewide models
bull Long-distance purposes Business Tourist Personal business
bull Trucks Commodity group is often
surrogate for trip purpose
bull External trips External generally means trip from
outside state Might exclude ldquohalordquo area
surrounding state boundary
Trip PurposeFL
2000 IN LA MA MS RI
Home-based Work 1914 1995 1900 1841 1606 2000
Home-based Nonwork 5547 5208 4770 5206 4917 5700
Home-based Other 2792 5208 4770 1938 4917 5700
Home-based Shop 1576 ndash ndash 1173 ndash ndash
Home-based SocialRecreation
1179 ndash ndash 1395 ndash ndash
Home-base RecreationShop ndash ndash ndash ndash ndash ndash
Home-based School ndash ndash ndash 700 ndash ndash
Nonhome-based 2539 2686 3100 2468 3289 2300
Nonhome-based Work ndash ndash ndash 1027 ndash ndash
Nonhome-based Nonwork ndash ndash ndash 1441 ndash ndash
Other ndash ndash ndash ndash ndash
Truck ndash ndash 170 ndash ndash ndash
Long-Distance 000 111 060 417 022 000
Long-Distance Business ndash ndash 020 284 011 ndash
Long-Distance Personal Business
ndash ndash ndash ndash ndash ndash
Long-Distance Tourist ndash ndash 010 027 000 ndash
Long-Distance Work ndash ndash ndash 105 ndash ndash
Long-Distance NonworkOther
ndash ndash 030 ndash 011 ndash
Statewide vs Urban Model Trip Purposes (Contrsquod)
Number of person trips
Mode Pleasure Business Personal business Other Total Percent
Personal vehicle 12185 5708 3120 789 21802 908Commercial air 1012 746 151 24 1933 8Train 74 150 14 0 238 1Other 4 2 5 12 23 01Ship 19 0 0 0 19 01
Total 13294 6606 3290 825 24015 100Percent 554 275 137 34 100
Percent by trip purpose
Mode Pleasure Business Personal business Other
Personal vehicle 917 864 948 956Commercial air 76 113 46 29Train 06 23 04 0Other 0 0 02 15Ship 01 0 0 0
Trip purpose
Trip purpose
Statewide vs Urban Model Modal Issuesbull Most statewide models are limited to highway trips
bull Modeling of local bus routes is likely overkill
bull Mode choice is usually just a conversion from person trips to vehicle trips via auto occupancy rates
bull Transit components of statewide models are more common in regions with history of fixed guideway transit
bull Transit model is needed to test options such as high speedintercity rail
ATS 95 NHTS 01 MDOT 03-04 ODOT 02-03Personal vehicle 81 89 89 88Commercial air 16 8 10 8Bus 2 2 04 19Train 05 01 0 005Other 05 09 06 205
Survey and yearPrimary mode of transport
Statewide vs Urban Model Modal Issues (Contrsquod)
3727 24 22
32
29 36
10
21
28 22
10
22
12
1529
11
44 43
0
10
20
30
40
50
60
70
80
90
100
Personalvehicle
Bus Train Marine Commercialair
Perc
ent o
f per
son
trips
lt300 300-499 500-599 1000-1999 gt2000
68
85 85 82 85
30
12 11 15 13
0
10
20
30
40
50
60
70
80
90
100
Busi-ness
Plea-sure
pBusi-ness
Vaca-tion
Week-end
Perc
ent o
f per
son
trips
Personal vehicle Commercial air All other
Mode by Trip Purpose Mode by Distance (miles)
Statewide vs Urban Model Modal Issues (Contrsquod)
bull Most statewide assignment models separate trucks from passenger cars Preloading process commonly used for trucks Takes into account more limited route options for truck travel
Source Ohio Dept of Transportation
Statewide vs Urban Model Truck Forecasting
bull Urban modelsndash Freight movements are partially
ldquoexternalrdquo to model (truncated at external zone)
ndash More general categories of truckcommercial trips used
ndash Strong reliance on Quick Response Freight Manual
bull Statewide modelsndash Rooted in commodity flows ndash Often use econometric
approaches to modelingndash Generally model entire length of
freight travel
Freight Model Type by Peer Exchange Participants and
NonParticipants
0123456789
None
Traditio
nal
Commod
ity
Econo
metric
Unkno
wn
NotPeer Exchange
Source TRB Electronic Circular E-C075 Statewide Travel Demand Modeling ndash A Peer Exchange
Topic ndash Data Needs and Considerations
bull Impact of model complexity on data needs
bull Overview of data sources
bull Long-Distance travel gt different focus than MPOregional models
Sources NHTS Weight calculation Pitfalls Key statistics European surveys Lessons learned
Impact of Model Complexity on Data Needsbull Trip table estimation (2 step) ndash highway network
zone system trip table
bull Highway model structure (3 step) ndash 2 step plusbehavioral data for trip rates friction factors auto occupancy rates
bull Freight considerations (3 step) ndash 3 step pluscommodity flows economic data tonnages by mode truck counts
bull Full multi-modal structure (4 step) ndash 3 step plustransit networks and paths logit mode choice model
bull Beyond 4 Step ndash synthetic household generation operational characteristics etc
Source Georgia DOTCambridge Systematics Inc
Statewide Model Data Sources A-Zbull American Travel Survey (ATS)
bull Bureau of Economic Analysis (BEA)
bull Census Journey-to-Work (JTW) data
bull Census Public Use Microsample (PUMS)
bull Census Transportation Planning Package (CTPP)
bull Claritas
bull Commodity Flow Survey (CFS)
Source US Census Bureau
Statewide Model Data Sources (Contrsquod)bull Dun amp Bradstreet (DampB) employment data
bull Employment Security data (ES-202)
bull Freight Analysis Framework (FAF)
bull Highway Performance Monitoring System (HPMS)
bull InfoUSA employment data
bull Metropolitan Planning Organizations (MPO)
bull NCHRP 365 Manual
bull National Highway Planning Network (NHPN)
Source FAF2 Provisional Commodity Origin-Destination Estimates Revised Methodology Report BattelleFHWA
Statewide Model Data Sources (Contrsquod)
bull National Household Travel Survey (NHTS)
bull Origin-destination surveys
bull Regional Economic Models Inc (REMI)
bull State Departments of Transportation
bull TRANSEARCH data
bull Vehicle Use and Identification Survey (VIUS)
bull Woods amp Poole employment forecasts
Source Bureau of Transportation Statistics NHTS
Data Sources on Long-Distance Travel
bull Census Journey-to-Work databull Long distance component
American Travel Survey (77 95) National Household Travel Survey (01) Ohio statewide surveys (02-04) Michigan statewide survey (04-05) California (10-12)
bull Surveys focused on long distance Oregon (96-98 07-08)
bull Others BTS Omnibus surveys Tourism surveys European practice NCHRP Report 735
NHTS and Long-Distance Travelbull Evolution of US national surveys
NPTS (69 77 83 90 95) ATS (77 95)
bull 2001 NHTS ndash Total of 60282 long distance person trips Data collected March 01 to May 02 22204 (37 percent) before 911 38078 (63 percent) afterwards
bull Long distance trip is longer than 50 miles one-way (vs 100 miles in 1995 ATS)
bull 28-day retrospective survey (including travel day)
bull Not your typical travelers Low frequency Distinct segment of the population
Weight Calculation for Long-Distance Travelers
bull Use to ensure valid population-level estimates Accounts for non-response under-coverage multiple phones etc
bull Multi-stage process Match national margins Adjust for data splitting (force sumexpfac = 277208169) Raking (match independent controls) race ethnicity race by
month ethnicity by month sex by age census region MSA status month by day of week
Trim outlier expansion factors max expfac = 4 times mean Replication (correct for sampling error)
bull At least two passes through the data
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Freight Model Types | |||||||||||||||||||
State | Type | Peer Exc | |||||||||||||||||
OR | 4 | 1 | Count of Type | Count of Type | Peer Exc | ||||||||||||||
WI | 3 | 1 | Type | Total | Type | 0 | 1 | Grand Total | |||||||||||
MO | 1 | 1 | 1 | 2 | 1 | 2 | 2 | 4 | |||||||||||
LA | 3 | 1 | 2 | 2 | 2 | 2 | 2 | 4 | |||||||||||
IN | 3 | 1 | 3 | 6 | 3 | 2 | 6 | 8 | |||||||||||
OH | 4 | 1 | 4 | 2 | 4 | 2 | 2 | ||||||||||||
KY | 3 | 1 | Grand Total | 12 | 3 | 3 | |||||||||||||
FL | 3 | 1 | Grand Total | 9 | 12 | 21 | |||||||||||||
VA | 3 | 1 | None | 2 | Not | Peer Exchange | |||||||||||||
NJ | 2 | 1 | Traditional | 2 | None | 2 | 2 | ||||||||||||
MA | 2 | 1 | Commodity | 6 | Traditional | 2 | 2 | ||||||||||||
NH | 1 | 1 | Econometric | 2 | Commodity | 2 | 6 | ||||||||||||
CA | 0 | Econometric | 2 | ||||||||||||||||
TX | 3 | 0 | Unknown | 3 | |||||||||||||||
MI | 3 | 0 | |||||||||||||||||
GA | 0 | ||||||||||||||||||
SC | 1 | 0 | |||||||||||||||||
DE | 2 | 0 | |||||||||||||||||
CT | 0 | ||||||||||||||||||
VT | 2 | 0 | |||||||||||||||||
ME | 1 | 0 | |||||||||||||||||
MN | 0 |
None | None | ||
Traditional | Traditional | ||
Commodity | Commodity | ||
Econometric | Econometric | ||
Unknown | Unknown |
Alternative Statewide Model Structures
bull Trip table estimation (2 step) Estimate initial ldquoseed matrixrdquo
Roadside intercept surveys American Travel Survey National HH Travel SurveyAdd-ons Census Journey-to-Work flows
Matrix estimation program Inputs ndash seed matrix highway
network traffic counts Adjusts origin-destination flows
between zones to optimize matching network traffic counts
Fratar factors applied to synthetic trip table in estimating future growth
bull Trip table estimation (2 step) Pros
Limited need to prepare aggregate socioeconomic input data
Simplicity of model generally means short execution times
Forecasting needs are limited to zone or district growth factors
Cons Less explanatory power than most
MPOregional models Fixed distribution pattern Difficulties in assessing benefits of
adding new roadway capacity
Alternative Statewide Model Structures (Contrsquod)
bull Highway model structure (3 step) Trip generation Trip distribution Highway assignment
bull Mode choice excluded Either generate vehicle trips or Apply auto occupancy factors to person trip
tables
bull Pros Benefits of trip redistribution across range of
alternatives Greater sensitivity to roadway capacity and
congestion
bull Cons Likely reliance on MPOsothers to provide
socioeconomic input data Canrsquot fully test transportation strategies that
would alter mode split
Source Florida Dept of TransportationCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)
bull Freight considerations (3 step) Typical freight model steps
Trip generation ndash tons by commodity group
Trip distribution Mode choice ndash freight is
apportioned to transport modes
bull Joint highway assignment (trucks and passenger trips)
Source Florida Dept of TransportationCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)bull Full multi-modal structure (4 step)
Trip generation Trip distribution Mode choice Highway assignment
bull More common with mega-regions Focused on fixed guideway systems Intercity rail is important component
bull Pros Potential to model new intercity rail
including high speed rail Test greater range of potential
transportation solutions
bull Cons Requires coding of transit networks and
mode choice estimation Modeling of air travel might be needed
especially if intercity rail is a completely new mode of travel
Source California HSR AuthorityCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)
Beyond 4 Stepbull Potential components
Implement destination choice Tour-basedtrip chains Activity-based travel model Activity micro-simulation Traffic micro-simulation Integration with land use models Integration with economic models
bull Pros Ability to test a more complete set of
transportation strategies and solutions
bull Cons Extensive data requirements Cost to maintain Complexity
Source Oregon Dept of TransportationPBUniversity of Calgary
Deciding on Best Statewide Model Approach
bull What are the motivations for building a statewide travel demand modelbull What are the emerging transportation and policy issues to addressbull How significant is the impact of through travel on your statebull Do you have many multi-state MPOsbull How reliable (and current) are your survey databull Does your state have major international ports or border crossings
StudyArea
UrbanModel
AreaStudyArea
UrbanModelArea 1
Study AreaWithin OneModel Area
UrbanModelArea 2
Study AreasWithin Two Model Areas
StudyArea
UrbanModelArea
StudyArea
Statewide Model
Study AreaOutside Urban
Model Areas
Statewide ModelStatewide Model
Source Florida Department of TransportationCambridge Systematics Inc
Figure 317 Subarea Study Area Examples
Study
Area
Urban
Model
Area
Study
Area
Urban
Model
Area 1
Study Area
Within One
Model Area
Urban
Model
Area 2
Study Areas
Within Two
Model Areas
Study
Area
Urban
Model
Area
Study
Area
Statewide Model
Study Area
Outside Urban
Model Areas
Statewide Model
Statewide Model
Statewide Model Data Development Challenges
bull Primary data deficiencies Long distance travelers Rural resident trip-making
bull Challenges in compiling required data Age of data Confidentiality Coordination with data providers Proper level of geography Sufficient sample size
bull Four categories of data for discussion Networks and zone system Demographic and employment data Freight and other economic data NHTS and other behavioral data
Source Florida Department of TransportationCambridge Systematics Inc
Data Challengeshellipwhat is ldquolong distancerdquo
bull gt50 miles (American Traveler Survey)
bull gt100 miles (National Household Travel Survey)
bull IE trips (several states)
bull Trips made by residents for business personal business
bull Recreationaltourist trips
Topic ndash Similarities amp Differences with Urban Models
bull Zone systems
bull Networks
bull Travel behavior assumptions
bull Trip purposesmarket segments
bull Modal issues
bull Truck forecasting
Source Delaware Dept of TransportationWRampA Inc
Statewide vs Urban Model Zone Systems
bull Usually larger zones in statewide models than MPOs
bull Block groups are good starting point for statewide model TAZs
bull Block groups might need splitting along major transportation corridors
bull In small states MPO TAZs could equal SWM zones
bull In most states MPO TAZ equivalency tables will likely be neededdesirable
bull FAF disaggregation approach used for statewide model TAZs
Source FHWA Office of Freight Management and Operations
Source Florida Dept of TransportationCambridge Systematics Inc
FAF zones
Block groups = TAZs
Too small for statewide zones
Too big for statewide zones
Statewide vs Urban Model Networksbull Statewide model networks are sparse in
comparison with MPO models
bull At a minimum statewide networks should include Limited access highways Arterials Major collectors
bull For proper circulation patterns lower class facilities might be added
bull Outside state a very sparse network is needed for major truck corridors
bull Consistent level of detail for networks and zones
Source Tennessee Dept of TransportationCambridge Systematics Inc
MPO network
statewide network
Statewide vs Urban Model Travel Behaviorbull Understanding urban travel behavior
MPO models are usually developed with local household travel survey data
MPOs without recent HH survey data can borrow parameters from other MPO models or guidance documents such as NCHRP 716 Transferable Urban Travel Parameters
bull Rural and long-distance travel behavior A limited number of states have conducted
statewide household travel surveys American Travel Survey was designed to target
long-distance trips but was discontinued long ago
National Household Travel Survey Add-ons can be structured for a sufficient sample of rural trip-makers
NCHRP 735 Rural and Long-Distance Travel Parameters recently published for use in statewide models
Map of 2009 Add-on Surveys Source FHWA
Statewide vs Urban Model Trip Purposes
bull Typical MPO trip purposes are usually collapsed into HBW HBOHBNW and NHB for statewide models
bull Long-distance purposes Business Tourist Personal business
bull Trucks Commodity group is often
surrogate for trip purpose
bull External trips External generally means trip from
outside state Might exclude ldquohalordquo area
surrounding state boundary
Trip PurposeFL
2000 IN LA MA MS RI
Home-based Work 1914 1995 1900 1841 1606 2000
Home-based Nonwork 5547 5208 4770 5206 4917 5700
Home-based Other 2792 5208 4770 1938 4917 5700
Home-based Shop 1576 ndash ndash 1173 ndash ndash
Home-based SocialRecreation
1179 ndash ndash 1395 ndash ndash
Home-base RecreationShop ndash ndash ndash ndash ndash ndash
Home-based School ndash ndash ndash 700 ndash ndash
Nonhome-based 2539 2686 3100 2468 3289 2300
Nonhome-based Work ndash ndash ndash 1027 ndash ndash
Nonhome-based Nonwork ndash ndash ndash 1441 ndash ndash
Other ndash ndash ndash ndash ndash
Truck ndash ndash 170 ndash ndash ndash
Long-Distance 000 111 060 417 022 000
Long-Distance Business ndash ndash 020 284 011 ndash
Long-Distance Personal Business
ndash ndash ndash ndash ndash ndash
Long-Distance Tourist ndash ndash 010 027 000 ndash
Long-Distance Work ndash ndash ndash 105 ndash ndash
Long-Distance NonworkOther
ndash ndash 030 ndash 011 ndash
Statewide vs Urban Model Trip Purposes (Contrsquod)
Number of person trips
Mode Pleasure Business Personal business Other Total Percent
Personal vehicle 12185 5708 3120 789 21802 908Commercial air 1012 746 151 24 1933 8Train 74 150 14 0 238 1Other 4 2 5 12 23 01Ship 19 0 0 0 19 01
Total 13294 6606 3290 825 24015 100Percent 554 275 137 34 100
Percent by trip purpose
Mode Pleasure Business Personal business Other
Personal vehicle 917 864 948 956Commercial air 76 113 46 29Train 06 23 04 0Other 0 0 02 15Ship 01 0 0 0
Trip purpose
Trip purpose
Statewide vs Urban Model Modal Issuesbull Most statewide models are limited to highway trips
bull Modeling of local bus routes is likely overkill
bull Mode choice is usually just a conversion from person trips to vehicle trips via auto occupancy rates
bull Transit components of statewide models are more common in regions with history of fixed guideway transit
bull Transit model is needed to test options such as high speedintercity rail
ATS 95 NHTS 01 MDOT 03-04 ODOT 02-03Personal vehicle 81 89 89 88Commercial air 16 8 10 8Bus 2 2 04 19Train 05 01 0 005Other 05 09 06 205
Survey and yearPrimary mode of transport
Statewide vs Urban Model Modal Issues (Contrsquod)
3727 24 22
32
29 36
10
21
28 22
10
22
12
1529
11
44 43
0
10
20
30
40
50
60
70
80
90
100
Personalvehicle
Bus Train Marine Commercialair
Perc
ent o
f per
son
trips
lt300 300-499 500-599 1000-1999 gt2000
68
85 85 82 85
30
12 11 15 13
0
10
20
30
40
50
60
70
80
90
100
Busi-ness
Plea-sure
pBusi-ness
Vaca-tion
Week-end
Perc
ent o
f per
son
trips
Personal vehicle Commercial air All other
Mode by Trip Purpose Mode by Distance (miles)
Statewide vs Urban Model Modal Issues (Contrsquod)
bull Most statewide assignment models separate trucks from passenger cars Preloading process commonly used for trucks Takes into account more limited route options for truck travel
Source Ohio Dept of Transportation
Statewide vs Urban Model Truck Forecasting
bull Urban modelsndash Freight movements are partially
ldquoexternalrdquo to model (truncated at external zone)
ndash More general categories of truckcommercial trips used
ndash Strong reliance on Quick Response Freight Manual
bull Statewide modelsndash Rooted in commodity flows ndash Often use econometric
approaches to modelingndash Generally model entire length of
freight travel
Freight Model Type by Peer Exchange Participants and
NonParticipants
0123456789
None
Traditio
nal
Commod
ity
Econo
metric
Unkno
wn
NotPeer Exchange
Source TRB Electronic Circular E-C075 Statewide Travel Demand Modeling ndash A Peer Exchange
Topic ndash Data Needs and Considerations
bull Impact of model complexity on data needs
bull Overview of data sources
bull Long-Distance travel gt different focus than MPOregional models
Sources NHTS Weight calculation Pitfalls Key statistics European surveys Lessons learned
Impact of Model Complexity on Data Needsbull Trip table estimation (2 step) ndash highway network
zone system trip table
bull Highway model structure (3 step) ndash 2 step plusbehavioral data for trip rates friction factors auto occupancy rates
bull Freight considerations (3 step) ndash 3 step pluscommodity flows economic data tonnages by mode truck counts
bull Full multi-modal structure (4 step) ndash 3 step plustransit networks and paths logit mode choice model
bull Beyond 4 Step ndash synthetic household generation operational characteristics etc
Source Georgia DOTCambridge Systematics Inc
Statewide Model Data Sources A-Zbull American Travel Survey (ATS)
bull Bureau of Economic Analysis (BEA)
bull Census Journey-to-Work (JTW) data
bull Census Public Use Microsample (PUMS)
bull Census Transportation Planning Package (CTPP)
bull Claritas
bull Commodity Flow Survey (CFS)
Source US Census Bureau
Statewide Model Data Sources (Contrsquod)bull Dun amp Bradstreet (DampB) employment data
bull Employment Security data (ES-202)
bull Freight Analysis Framework (FAF)
bull Highway Performance Monitoring System (HPMS)
bull InfoUSA employment data
bull Metropolitan Planning Organizations (MPO)
bull NCHRP 365 Manual
bull National Highway Planning Network (NHPN)
Source FAF2 Provisional Commodity Origin-Destination Estimates Revised Methodology Report BattelleFHWA
Statewide Model Data Sources (Contrsquod)
bull National Household Travel Survey (NHTS)
bull Origin-destination surveys
bull Regional Economic Models Inc (REMI)
bull State Departments of Transportation
bull TRANSEARCH data
bull Vehicle Use and Identification Survey (VIUS)
bull Woods amp Poole employment forecasts
Source Bureau of Transportation Statistics NHTS
Data Sources on Long-Distance Travel
bull Census Journey-to-Work databull Long distance component
American Travel Survey (77 95) National Household Travel Survey (01) Ohio statewide surveys (02-04) Michigan statewide survey (04-05) California (10-12)
bull Surveys focused on long distance Oregon (96-98 07-08)
bull Others BTS Omnibus surveys Tourism surveys European practice NCHRP Report 735
NHTS and Long-Distance Travelbull Evolution of US national surveys
NPTS (69 77 83 90 95) ATS (77 95)
bull 2001 NHTS ndash Total of 60282 long distance person trips Data collected March 01 to May 02 22204 (37 percent) before 911 38078 (63 percent) afterwards
bull Long distance trip is longer than 50 miles one-way (vs 100 miles in 1995 ATS)
bull 28-day retrospective survey (including travel day)
bull Not your typical travelers Low frequency Distinct segment of the population
Weight Calculation for Long-Distance Travelers
bull Use to ensure valid population-level estimates Accounts for non-response under-coverage multiple phones etc
bull Multi-stage process Match national margins Adjust for data splitting (force sumexpfac = 277208169) Raking (match independent controls) race ethnicity race by
month ethnicity by month sex by age census region MSA status month by day of week
Trim outlier expansion factors max expfac = 4 times mean Replication (correct for sampling error)
bull At least two passes through the data
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Freight Model Types | |||||||||||||||||||
State | Type | Peer Exc | |||||||||||||||||
OR | 4 | 1 | Count of Type | Count of Type | Peer Exc | ||||||||||||||
WI | 3 | 1 | Type | Total | Type | 0 | 1 | Grand Total | |||||||||||
MO | 1 | 1 | 1 | 2 | 1 | 2 | 2 | 4 | |||||||||||
LA | 3 | 1 | 2 | 2 | 2 | 2 | 2 | 4 | |||||||||||
IN | 3 | 1 | 3 | 6 | 3 | 2 | 6 | 8 | |||||||||||
OH | 4 | 1 | 4 | 2 | 4 | 2 | 2 | ||||||||||||
KY | 3 | 1 | Grand Total | 12 | 3 | 3 | |||||||||||||
FL | 3 | 1 | Grand Total | 9 | 12 | 21 | |||||||||||||
VA | 3 | 1 | None | 2 | Not | Peer Exchange | |||||||||||||
NJ | 2 | 1 | Traditional | 2 | None | 2 | 2 | ||||||||||||
MA | 2 | 1 | Commodity | 6 | Traditional | 2 | 2 | ||||||||||||
NH | 1 | 1 | Econometric | 2 | Commodity | 2 | 6 | ||||||||||||
CA | 0 | Econometric | 2 | ||||||||||||||||
TX | 3 | 0 | Unknown | 3 | |||||||||||||||
MI | 3 | 0 | |||||||||||||||||
GA | 0 | ||||||||||||||||||
SC | 1 | 0 | |||||||||||||||||
DE | 2 | 0 | |||||||||||||||||
CT | 0 | ||||||||||||||||||
VT | 2 | 0 | |||||||||||||||||
ME | 1 | 0 | |||||||||||||||||
MN | 0 |
None | None | ||
Traditional | Traditional | ||
Commodity | Commodity | ||
Econometric | Econometric | ||
Unknown | Unknown |
Alternative Statewide Model Structures (Contrsquod)
bull Highway model structure (3 step) Trip generation Trip distribution Highway assignment
bull Mode choice excluded Either generate vehicle trips or Apply auto occupancy factors to person trip
tables
bull Pros Benefits of trip redistribution across range of
alternatives Greater sensitivity to roadway capacity and
congestion
bull Cons Likely reliance on MPOsothers to provide
socioeconomic input data Canrsquot fully test transportation strategies that
would alter mode split
Source Florida Dept of TransportationCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)
bull Freight considerations (3 step) Typical freight model steps
Trip generation ndash tons by commodity group
Trip distribution Mode choice ndash freight is
apportioned to transport modes
bull Joint highway assignment (trucks and passenger trips)
Source Florida Dept of TransportationCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)bull Full multi-modal structure (4 step)
Trip generation Trip distribution Mode choice Highway assignment
bull More common with mega-regions Focused on fixed guideway systems Intercity rail is important component
bull Pros Potential to model new intercity rail
including high speed rail Test greater range of potential
transportation solutions
bull Cons Requires coding of transit networks and
mode choice estimation Modeling of air travel might be needed
especially if intercity rail is a completely new mode of travel
Source California HSR AuthorityCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)
Beyond 4 Stepbull Potential components
Implement destination choice Tour-basedtrip chains Activity-based travel model Activity micro-simulation Traffic micro-simulation Integration with land use models Integration with economic models
bull Pros Ability to test a more complete set of
transportation strategies and solutions
bull Cons Extensive data requirements Cost to maintain Complexity
Source Oregon Dept of TransportationPBUniversity of Calgary
Deciding on Best Statewide Model Approach
bull What are the motivations for building a statewide travel demand modelbull What are the emerging transportation and policy issues to addressbull How significant is the impact of through travel on your statebull Do you have many multi-state MPOsbull How reliable (and current) are your survey databull Does your state have major international ports or border crossings
StudyArea
UrbanModel
AreaStudyArea
UrbanModelArea 1
Study AreaWithin OneModel Area
UrbanModelArea 2
Study AreasWithin Two Model Areas
StudyArea
UrbanModelArea
StudyArea
Statewide Model
Study AreaOutside Urban
Model Areas
Statewide ModelStatewide Model
Source Florida Department of TransportationCambridge Systematics Inc
Figure 317 Subarea Study Area Examples
Study
Area
Urban
Model
Area
Study
Area
Urban
Model
Area 1
Study Area
Within One
Model Area
Urban
Model
Area 2
Study Areas
Within Two
Model Areas
Study
Area
Urban
Model
Area
Study
Area
Statewide Model
Study Area
Outside Urban
Model Areas
Statewide Model
Statewide Model
Statewide Model Data Development Challenges
bull Primary data deficiencies Long distance travelers Rural resident trip-making
bull Challenges in compiling required data Age of data Confidentiality Coordination with data providers Proper level of geography Sufficient sample size
bull Four categories of data for discussion Networks and zone system Demographic and employment data Freight and other economic data NHTS and other behavioral data
Source Florida Department of TransportationCambridge Systematics Inc
Data Challengeshellipwhat is ldquolong distancerdquo
bull gt50 miles (American Traveler Survey)
bull gt100 miles (National Household Travel Survey)
bull IE trips (several states)
bull Trips made by residents for business personal business
bull Recreationaltourist trips
Topic ndash Similarities amp Differences with Urban Models
bull Zone systems
bull Networks
bull Travel behavior assumptions
bull Trip purposesmarket segments
bull Modal issues
bull Truck forecasting
Source Delaware Dept of TransportationWRampA Inc
Statewide vs Urban Model Zone Systems
bull Usually larger zones in statewide models than MPOs
bull Block groups are good starting point for statewide model TAZs
bull Block groups might need splitting along major transportation corridors
bull In small states MPO TAZs could equal SWM zones
bull In most states MPO TAZ equivalency tables will likely be neededdesirable
bull FAF disaggregation approach used for statewide model TAZs
Source FHWA Office of Freight Management and Operations
Source Florida Dept of TransportationCambridge Systematics Inc
FAF zones
Block groups = TAZs
Too small for statewide zones
Too big for statewide zones
Statewide vs Urban Model Networksbull Statewide model networks are sparse in
comparison with MPO models
bull At a minimum statewide networks should include Limited access highways Arterials Major collectors
bull For proper circulation patterns lower class facilities might be added
bull Outside state a very sparse network is needed for major truck corridors
bull Consistent level of detail for networks and zones
Source Tennessee Dept of TransportationCambridge Systematics Inc
MPO network
statewide network
Statewide vs Urban Model Travel Behaviorbull Understanding urban travel behavior
MPO models are usually developed with local household travel survey data
MPOs without recent HH survey data can borrow parameters from other MPO models or guidance documents such as NCHRP 716 Transferable Urban Travel Parameters
bull Rural and long-distance travel behavior A limited number of states have conducted
statewide household travel surveys American Travel Survey was designed to target
long-distance trips but was discontinued long ago
National Household Travel Survey Add-ons can be structured for a sufficient sample of rural trip-makers
NCHRP 735 Rural and Long-Distance Travel Parameters recently published for use in statewide models
Map of 2009 Add-on Surveys Source FHWA
Statewide vs Urban Model Trip Purposes
bull Typical MPO trip purposes are usually collapsed into HBW HBOHBNW and NHB for statewide models
bull Long-distance purposes Business Tourist Personal business
bull Trucks Commodity group is often
surrogate for trip purpose
bull External trips External generally means trip from
outside state Might exclude ldquohalordquo area
surrounding state boundary
Trip PurposeFL
2000 IN LA MA MS RI
Home-based Work 1914 1995 1900 1841 1606 2000
Home-based Nonwork 5547 5208 4770 5206 4917 5700
Home-based Other 2792 5208 4770 1938 4917 5700
Home-based Shop 1576 ndash ndash 1173 ndash ndash
Home-based SocialRecreation
1179 ndash ndash 1395 ndash ndash
Home-base RecreationShop ndash ndash ndash ndash ndash ndash
Home-based School ndash ndash ndash 700 ndash ndash
Nonhome-based 2539 2686 3100 2468 3289 2300
Nonhome-based Work ndash ndash ndash 1027 ndash ndash
Nonhome-based Nonwork ndash ndash ndash 1441 ndash ndash
Other ndash ndash ndash ndash ndash
Truck ndash ndash 170 ndash ndash ndash
Long-Distance 000 111 060 417 022 000
Long-Distance Business ndash ndash 020 284 011 ndash
Long-Distance Personal Business
ndash ndash ndash ndash ndash ndash
Long-Distance Tourist ndash ndash 010 027 000 ndash
Long-Distance Work ndash ndash ndash 105 ndash ndash
Long-Distance NonworkOther
ndash ndash 030 ndash 011 ndash
Statewide vs Urban Model Trip Purposes (Contrsquod)
Number of person trips
Mode Pleasure Business Personal business Other Total Percent
Personal vehicle 12185 5708 3120 789 21802 908Commercial air 1012 746 151 24 1933 8Train 74 150 14 0 238 1Other 4 2 5 12 23 01Ship 19 0 0 0 19 01
Total 13294 6606 3290 825 24015 100Percent 554 275 137 34 100
Percent by trip purpose
Mode Pleasure Business Personal business Other
Personal vehicle 917 864 948 956Commercial air 76 113 46 29Train 06 23 04 0Other 0 0 02 15Ship 01 0 0 0
Trip purpose
Trip purpose
Statewide vs Urban Model Modal Issuesbull Most statewide models are limited to highway trips
bull Modeling of local bus routes is likely overkill
bull Mode choice is usually just a conversion from person trips to vehicle trips via auto occupancy rates
bull Transit components of statewide models are more common in regions with history of fixed guideway transit
bull Transit model is needed to test options such as high speedintercity rail
ATS 95 NHTS 01 MDOT 03-04 ODOT 02-03Personal vehicle 81 89 89 88Commercial air 16 8 10 8Bus 2 2 04 19Train 05 01 0 005Other 05 09 06 205
Survey and yearPrimary mode of transport
Statewide vs Urban Model Modal Issues (Contrsquod)
3727 24 22
32
29 36
10
21
28 22
10
22
12
1529
11
44 43
0
10
20
30
40
50
60
70
80
90
100
Personalvehicle
Bus Train Marine Commercialair
Perc
ent o
f per
son
trips
lt300 300-499 500-599 1000-1999 gt2000
68
85 85 82 85
30
12 11 15 13
0
10
20
30
40
50
60
70
80
90
100
Busi-ness
Plea-sure
pBusi-ness
Vaca-tion
Week-end
Perc
ent o
f per
son
trips
Personal vehicle Commercial air All other
Mode by Trip Purpose Mode by Distance (miles)
Statewide vs Urban Model Modal Issues (Contrsquod)
bull Most statewide assignment models separate trucks from passenger cars Preloading process commonly used for trucks Takes into account more limited route options for truck travel
Source Ohio Dept of Transportation
Statewide vs Urban Model Truck Forecasting
bull Urban modelsndash Freight movements are partially
ldquoexternalrdquo to model (truncated at external zone)
ndash More general categories of truckcommercial trips used
ndash Strong reliance on Quick Response Freight Manual
bull Statewide modelsndash Rooted in commodity flows ndash Often use econometric
approaches to modelingndash Generally model entire length of
freight travel
Freight Model Type by Peer Exchange Participants and
NonParticipants
0123456789
None
Traditio
nal
Commod
ity
Econo
metric
Unkno
wn
NotPeer Exchange
Source TRB Electronic Circular E-C075 Statewide Travel Demand Modeling ndash A Peer Exchange
Topic ndash Data Needs and Considerations
bull Impact of model complexity on data needs
bull Overview of data sources
bull Long-Distance travel gt different focus than MPOregional models
Sources NHTS Weight calculation Pitfalls Key statistics European surveys Lessons learned
Impact of Model Complexity on Data Needsbull Trip table estimation (2 step) ndash highway network
zone system trip table
bull Highway model structure (3 step) ndash 2 step plusbehavioral data for trip rates friction factors auto occupancy rates
bull Freight considerations (3 step) ndash 3 step pluscommodity flows economic data tonnages by mode truck counts
bull Full multi-modal structure (4 step) ndash 3 step plustransit networks and paths logit mode choice model
bull Beyond 4 Step ndash synthetic household generation operational characteristics etc
Source Georgia DOTCambridge Systematics Inc
Statewide Model Data Sources A-Zbull American Travel Survey (ATS)
bull Bureau of Economic Analysis (BEA)
bull Census Journey-to-Work (JTW) data
bull Census Public Use Microsample (PUMS)
bull Census Transportation Planning Package (CTPP)
bull Claritas
bull Commodity Flow Survey (CFS)
Source US Census Bureau
Statewide Model Data Sources (Contrsquod)bull Dun amp Bradstreet (DampB) employment data
bull Employment Security data (ES-202)
bull Freight Analysis Framework (FAF)
bull Highway Performance Monitoring System (HPMS)
bull InfoUSA employment data
bull Metropolitan Planning Organizations (MPO)
bull NCHRP 365 Manual
bull National Highway Planning Network (NHPN)
Source FAF2 Provisional Commodity Origin-Destination Estimates Revised Methodology Report BattelleFHWA
Statewide Model Data Sources (Contrsquod)
bull National Household Travel Survey (NHTS)
bull Origin-destination surveys
bull Regional Economic Models Inc (REMI)
bull State Departments of Transportation
bull TRANSEARCH data
bull Vehicle Use and Identification Survey (VIUS)
bull Woods amp Poole employment forecasts
Source Bureau of Transportation Statistics NHTS
Data Sources on Long-Distance Travel
bull Census Journey-to-Work databull Long distance component
American Travel Survey (77 95) National Household Travel Survey (01) Ohio statewide surveys (02-04) Michigan statewide survey (04-05) California (10-12)
bull Surveys focused on long distance Oregon (96-98 07-08)
bull Others BTS Omnibus surveys Tourism surveys European practice NCHRP Report 735
NHTS and Long-Distance Travelbull Evolution of US national surveys
NPTS (69 77 83 90 95) ATS (77 95)
bull 2001 NHTS ndash Total of 60282 long distance person trips Data collected March 01 to May 02 22204 (37 percent) before 911 38078 (63 percent) afterwards
bull Long distance trip is longer than 50 miles one-way (vs 100 miles in 1995 ATS)
bull 28-day retrospective survey (including travel day)
bull Not your typical travelers Low frequency Distinct segment of the population
Weight Calculation for Long-Distance Travelers
bull Use to ensure valid population-level estimates Accounts for non-response under-coverage multiple phones etc
bull Multi-stage process Match national margins Adjust for data splitting (force sumexpfac = 277208169) Raking (match independent controls) race ethnicity race by
month ethnicity by month sex by age census region MSA status month by day of week
Trim outlier expansion factors max expfac = 4 times mean Replication (correct for sampling error)
bull At least two passes through the data
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Freight Model Types | |||||||||||||||||||
State | Type | Peer Exc | |||||||||||||||||
OR | 4 | 1 | Count of Type | Count of Type | Peer Exc | ||||||||||||||
WI | 3 | 1 | Type | Total | Type | 0 | 1 | Grand Total | |||||||||||
MO | 1 | 1 | 1 | 2 | 1 | 2 | 2 | 4 | |||||||||||
LA | 3 | 1 | 2 | 2 | 2 | 2 | 2 | 4 | |||||||||||
IN | 3 | 1 | 3 | 6 | 3 | 2 | 6 | 8 | |||||||||||
OH | 4 | 1 | 4 | 2 | 4 | 2 | 2 | ||||||||||||
KY | 3 | 1 | Grand Total | 12 | 3 | 3 | |||||||||||||
FL | 3 | 1 | Grand Total | 9 | 12 | 21 | |||||||||||||
VA | 3 | 1 | None | 2 | Not | Peer Exchange | |||||||||||||
NJ | 2 | 1 | Traditional | 2 | None | 2 | 2 | ||||||||||||
MA | 2 | 1 | Commodity | 6 | Traditional | 2 | 2 | ||||||||||||
NH | 1 | 1 | Econometric | 2 | Commodity | 2 | 6 | ||||||||||||
CA | 0 | Econometric | 2 | ||||||||||||||||
TX | 3 | 0 | Unknown | 3 | |||||||||||||||
MI | 3 | 0 | |||||||||||||||||
GA | 0 | ||||||||||||||||||
SC | 1 | 0 | |||||||||||||||||
DE | 2 | 0 | |||||||||||||||||
CT | 0 | ||||||||||||||||||
VT | 2 | 0 | |||||||||||||||||
ME | 1 | 0 | |||||||||||||||||
MN | 0 |
None | None | ||
Traditional | Traditional | ||
Commodity | Commodity | ||
Econometric | Econometric | ||
Unknown | Unknown |
Alternative Statewide Model Structures (Contrsquod)
bull Freight considerations (3 step) Typical freight model steps
Trip generation ndash tons by commodity group
Trip distribution Mode choice ndash freight is
apportioned to transport modes
bull Joint highway assignment (trucks and passenger trips)
Source Florida Dept of TransportationCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)bull Full multi-modal structure (4 step)
Trip generation Trip distribution Mode choice Highway assignment
bull More common with mega-regions Focused on fixed guideway systems Intercity rail is important component
bull Pros Potential to model new intercity rail
including high speed rail Test greater range of potential
transportation solutions
bull Cons Requires coding of transit networks and
mode choice estimation Modeling of air travel might be needed
especially if intercity rail is a completely new mode of travel
Source California HSR AuthorityCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)
Beyond 4 Stepbull Potential components
Implement destination choice Tour-basedtrip chains Activity-based travel model Activity micro-simulation Traffic micro-simulation Integration with land use models Integration with economic models
bull Pros Ability to test a more complete set of
transportation strategies and solutions
bull Cons Extensive data requirements Cost to maintain Complexity
Source Oregon Dept of TransportationPBUniversity of Calgary
Deciding on Best Statewide Model Approach
bull What are the motivations for building a statewide travel demand modelbull What are the emerging transportation and policy issues to addressbull How significant is the impact of through travel on your statebull Do you have many multi-state MPOsbull How reliable (and current) are your survey databull Does your state have major international ports or border crossings
StudyArea
UrbanModel
AreaStudyArea
UrbanModelArea 1
Study AreaWithin OneModel Area
UrbanModelArea 2
Study AreasWithin Two Model Areas
StudyArea
UrbanModelArea
StudyArea
Statewide Model
Study AreaOutside Urban
Model Areas
Statewide ModelStatewide Model
Source Florida Department of TransportationCambridge Systematics Inc
Figure 317 Subarea Study Area Examples
Study
Area
Urban
Model
Area
Study
Area
Urban
Model
Area 1
Study Area
Within One
Model Area
Urban
Model
Area 2
Study Areas
Within Two
Model Areas
Study
Area
Urban
Model
Area
Study
Area
Statewide Model
Study Area
Outside Urban
Model Areas
Statewide Model
Statewide Model
Statewide Model Data Development Challenges
bull Primary data deficiencies Long distance travelers Rural resident trip-making
bull Challenges in compiling required data Age of data Confidentiality Coordination with data providers Proper level of geography Sufficient sample size
bull Four categories of data for discussion Networks and zone system Demographic and employment data Freight and other economic data NHTS and other behavioral data
Source Florida Department of TransportationCambridge Systematics Inc
Data Challengeshellipwhat is ldquolong distancerdquo
bull gt50 miles (American Traveler Survey)
bull gt100 miles (National Household Travel Survey)
bull IE trips (several states)
bull Trips made by residents for business personal business
bull Recreationaltourist trips
Topic ndash Similarities amp Differences with Urban Models
bull Zone systems
bull Networks
bull Travel behavior assumptions
bull Trip purposesmarket segments
bull Modal issues
bull Truck forecasting
Source Delaware Dept of TransportationWRampA Inc
Statewide vs Urban Model Zone Systems
bull Usually larger zones in statewide models than MPOs
bull Block groups are good starting point for statewide model TAZs
bull Block groups might need splitting along major transportation corridors
bull In small states MPO TAZs could equal SWM zones
bull In most states MPO TAZ equivalency tables will likely be neededdesirable
bull FAF disaggregation approach used for statewide model TAZs
Source FHWA Office of Freight Management and Operations
Source Florida Dept of TransportationCambridge Systematics Inc
FAF zones
Block groups = TAZs
Too small for statewide zones
Too big for statewide zones
Statewide vs Urban Model Networksbull Statewide model networks are sparse in
comparison with MPO models
bull At a minimum statewide networks should include Limited access highways Arterials Major collectors
bull For proper circulation patterns lower class facilities might be added
bull Outside state a very sparse network is needed for major truck corridors
bull Consistent level of detail for networks and zones
Source Tennessee Dept of TransportationCambridge Systematics Inc
MPO network
statewide network
Statewide vs Urban Model Travel Behaviorbull Understanding urban travel behavior
MPO models are usually developed with local household travel survey data
MPOs without recent HH survey data can borrow parameters from other MPO models or guidance documents such as NCHRP 716 Transferable Urban Travel Parameters
bull Rural and long-distance travel behavior A limited number of states have conducted
statewide household travel surveys American Travel Survey was designed to target
long-distance trips but was discontinued long ago
National Household Travel Survey Add-ons can be structured for a sufficient sample of rural trip-makers
NCHRP 735 Rural and Long-Distance Travel Parameters recently published for use in statewide models
Map of 2009 Add-on Surveys Source FHWA
Statewide vs Urban Model Trip Purposes
bull Typical MPO trip purposes are usually collapsed into HBW HBOHBNW and NHB for statewide models
bull Long-distance purposes Business Tourist Personal business
bull Trucks Commodity group is often
surrogate for trip purpose
bull External trips External generally means trip from
outside state Might exclude ldquohalordquo area
surrounding state boundary
Trip PurposeFL
2000 IN LA MA MS RI
Home-based Work 1914 1995 1900 1841 1606 2000
Home-based Nonwork 5547 5208 4770 5206 4917 5700
Home-based Other 2792 5208 4770 1938 4917 5700
Home-based Shop 1576 ndash ndash 1173 ndash ndash
Home-based SocialRecreation
1179 ndash ndash 1395 ndash ndash
Home-base RecreationShop ndash ndash ndash ndash ndash ndash
Home-based School ndash ndash ndash 700 ndash ndash
Nonhome-based 2539 2686 3100 2468 3289 2300
Nonhome-based Work ndash ndash ndash 1027 ndash ndash
Nonhome-based Nonwork ndash ndash ndash 1441 ndash ndash
Other ndash ndash ndash ndash ndash
Truck ndash ndash 170 ndash ndash ndash
Long-Distance 000 111 060 417 022 000
Long-Distance Business ndash ndash 020 284 011 ndash
Long-Distance Personal Business
ndash ndash ndash ndash ndash ndash
Long-Distance Tourist ndash ndash 010 027 000 ndash
Long-Distance Work ndash ndash ndash 105 ndash ndash
Long-Distance NonworkOther
ndash ndash 030 ndash 011 ndash
Statewide vs Urban Model Trip Purposes (Contrsquod)
Number of person trips
Mode Pleasure Business Personal business Other Total Percent
Personal vehicle 12185 5708 3120 789 21802 908Commercial air 1012 746 151 24 1933 8Train 74 150 14 0 238 1Other 4 2 5 12 23 01Ship 19 0 0 0 19 01
Total 13294 6606 3290 825 24015 100Percent 554 275 137 34 100
Percent by trip purpose
Mode Pleasure Business Personal business Other
Personal vehicle 917 864 948 956Commercial air 76 113 46 29Train 06 23 04 0Other 0 0 02 15Ship 01 0 0 0
Trip purpose
Trip purpose
Statewide vs Urban Model Modal Issuesbull Most statewide models are limited to highway trips
bull Modeling of local bus routes is likely overkill
bull Mode choice is usually just a conversion from person trips to vehicle trips via auto occupancy rates
bull Transit components of statewide models are more common in regions with history of fixed guideway transit
bull Transit model is needed to test options such as high speedintercity rail
ATS 95 NHTS 01 MDOT 03-04 ODOT 02-03Personal vehicle 81 89 89 88Commercial air 16 8 10 8Bus 2 2 04 19Train 05 01 0 005Other 05 09 06 205
Survey and yearPrimary mode of transport
Statewide vs Urban Model Modal Issues (Contrsquod)
3727 24 22
32
29 36
10
21
28 22
10
22
12
1529
11
44 43
0
10
20
30
40
50
60
70
80
90
100
Personalvehicle
Bus Train Marine Commercialair
Perc
ent o
f per
son
trips
lt300 300-499 500-599 1000-1999 gt2000
68
85 85 82 85
30
12 11 15 13
0
10
20
30
40
50
60
70
80
90
100
Busi-ness
Plea-sure
pBusi-ness
Vaca-tion
Week-end
Perc
ent o
f per
son
trips
Personal vehicle Commercial air All other
Mode by Trip Purpose Mode by Distance (miles)
Statewide vs Urban Model Modal Issues (Contrsquod)
bull Most statewide assignment models separate trucks from passenger cars Preloading process commonly used for trucks Takes into account more limited route options for truck travel
Source Ohio Dept of Transportation
Statewide vs Urban Model Truck Forecasting
bull Urban modelsndash Freight movements are partially
ldquoexternalrdquo to model (truncated at external zone)
ndash More general categories of truckcommercial trips used
ndash Strong reliance on Quick Response Freight Manual
bull Statewide modelsndash Rooted in commodity flows ndash Often use econometric
approaches to modelingndash Generally model entire length of
freight travel
Freight Model Type by Peer Exchange Participants and
NonParticipants
0123456789
None
Traditio
nal
Commod
ity
Econo
metric
Unkno
wn
NotPeer Exchange
Source TRB Electronic Circular E-C075 Statewide Travel Demand Modeling ndash A Peer Exchange
Topic ndash Data Needs and Considerations
bull Impact of model complexity on data needs
bull Overview of data sources
bull Long-Distance travel gt different focus than MPOregional models
Sources NHTS Weight calculation Pitfalls Key statistics European surveys Lessons learned
Impact of Model Complexity on Data Needsbull Trip table estimation (2 step) ndash highway network
zone system trip table
bull Highway model structure (3 step) ndash 2 step plusbehavioral data for trip rates friction factors auto occupancy rates
bull Freight considerations (3 step) ndash 3 step pluscommodity flows economic data tonnages by mode truck counts
bull Full multi-modal structure (4 step) ndash 3 step plustransit networks and paths logit mode choice model
bull Beyond 4 Step ndash synthetic household generation operational characteristics etc
Source Georgia DOTCambridge Systematics Inc
Statewide Model Data Sources A-Zbull American Travel Survey (ATS)
bull Bureau of Economic Analysis (BEA)
bull Census Journey-to-Work (JTW) data
bull Census Public Use Microsample (PUMS)
bull Census Transportation Planning Package (CTPP)
bull Claritas
bull Commodity Flow Survey (CFS)
Source US Census Bureau
Statewide Model Data Sources (Contrsquod)bull Dun amp Bradstreet (DampB) employment data
bull Employment Security data (ES-202)
bull Freight Analysis Framework (FAF)
bull Highway Performance Monitoring System (HPMS)
bull InfoUSA employment data
bull Metropolitan Planning Organizations (MPO)
bull NCHRP 365 Manual
bull National Highway Planning Network (NHPN)
Source FAF2 Provisional Commodity Origin-Destination Estimates Revised Methodology Report BattelleFHWA
Statewide Model Data Sources (Contrsquod)
bull National Household Travel Survey (NHTS)
bull Origin-destination surveys
bull Regional Economic Models Inc (REMI)
bull State Departments of Transportation
bull TRANSEARCH data
bull Vehicle Use and Identification Survey (VIUS)
bull Woods amp Poole employment forecasts
Source Bureau of Transportation Statistics NHTS
Data Sources on Long-Distance Travel
bull Census Journey-to-Work databull Long distance component
American Travel Survey (77 95) National Household Travel Survey (01) Ohio statewide surveys (02-04) Michigan statewide survey (04-05) California (10-12)
bull Surveys focused on long distance Oregon (96-98 07-08)
bull Others BTS Omnibus surveys Tourism surveys European practice NCHRP Report 735
NHTS and Long-Distance Travelbull Evolution of US national surveys
NPTS (69 77 83 90 95) ATS (77 95)
bull 2001 NHTS ndash Total of 60282 long distance person trips Data collected March 01 to May 02 22204 (37 percent) before 911 38078 (63 percent) afterwards
bull Long distance trip is longer than 50 miles one-way (vs 100 miles in 1995 ATS)
bull 28-day retrospective survey (including travel day)
bull Not your typical travelers Low frequency Distinct segment of the population
Weight Calculation for Long-Distance Travelers
bull Use to ensure valid population-level estimates Accounts for non-response under-coverage multiple phones etc
bull Multi-stage process Match national margins Adjust for data splitting (force sumexpfac = 277208169) Raking (match independent controls) race ethnicity race by
month ethnicity by month sex by age census region MSA status month by day of week
Trim outlier expansion factors max expfac = 4 times mean Replication (correct for sampling error)
bull At least two passes through the data
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Freight Model Types | |||||||||||||||||||
State | Type | Peer Exc | |||||||||||||||||
OR | 4 | 1 | Count of Type | Count of Type | Peer Exc | ||||||||||||||
WI | 3 | 1 | Type | Total | Type | 0 | 1 | Grand Total | |||||||||||
MO | 1 | 1 | 1 | 2 | 1 | 2 | 2 | 4 | |||||||||||
LA | 3 | 1 | 2 | 2 | 2 | 2 | 2 | 4 | |||||||||||
IN | 3 | 1 | 3 | 6 | 3 | 2 | 6 | 8 | |||||||||||
OH | 4 | 1 | 4 | 2 | 4 | 2 | 2 | ||||||||||||
KY | 3 | 1 | Grand Total | 12 | 3 | 3 | |||||||||||||
FL | 3 | 1 | Grand Total | 9 | 12 | 21 | |||||||||||||
VA | 3 | 1 | None | 2 | Not | Peer Exchange | |||||||||||||
NJ | 2 | 1 | Traditional | 2 | None | 2 | 2 | ||||||||||||
MA | 2 | 1 | Commodity | 6 | Traditional | 2 | 2 | ||||||||||||
NH | 1 | 1 | Econometric | 2 | Commodity | 2 | 6 | ||||||||||||
CA | 0 | Econometric | 2 | ||||||||||||||||
TX | 3 | 0 | Unknown | 3 | |||||||||||||||
MI | 3 | 0 | |||||||||||||||||
GA | 0 | ||||||||||||||||||
SC | 1 | 0 | |||||||||||||||||
DE | 2 | 0 | |||||||||||||||||
CT | 0 | ||||||||||||||||||
VT | 2 | 0 | |||||||||||||||||
ME | 1 | 0 | |||||||||||||||||
MN | 0 |
None | None | ||
Traditional | Traditional | ||
Commodity | Commodity | ||
Econometric | Econometric | ||
Unknown | Unknown |
Alternative Statewide Model Structures (Contrsquod)bull Full multi-modal structure (4 step)
Trip generation Trip distribution Mode choice Highway assignment
bull More common with mega-regions Focused on fixed guideway systems Intercity rail is important component
bull Pros Potential to model new intercity rail
including high speed rail Test greater range of potential
transportation solutions
bull Cons Requires coding of transit networks and
mode choice estimation Modeling of air travel might be needed
especially if intercity rail is a completely new mode of travel
Source California HSR AuthorityCambridge Systematics Inc
Alternative Statewide Model Structures (Contrsquod)
Beyond 4 Stepbull Potential components
Implement destination choice Tour-basedtrip chains Activity-based travel model Activity micro-simulation Traffic micro-simulation Integration with land use models Integration with economic models
bull Pros Ability to test a more complete set of
transportation strategies and solutions
bull Cons Extensive data requirements Cost to maintain Complexity
Source Oregon Dept of TransportationPBUniversity of Calgary
Deciding on Best Statewide Model Approach
bull What are the motivations for building a statewide travel demand modelbull What are the emerging transportation and policy issues to addressbull How significant is the impact of through travel on your statebull Do you have many multi-state MPOsbull How reliable (and current) are your survey databull Does your state have major international ports or border crossings
StudyArea
UrbanModel
AreaStudyArea
UrbanModelArea 1
Study AreaWithin OneModel Area
UrbanModelArea 2
Study AreasWithin Two Model Areas
StudyArea
UrbanModelArea
StudyArea
Statewide Model
Study AreaOutside Urban
Model Areas
Statewide ModelStatewide Model
Source Florida Department of TransportationCambridge Systematics Inc
Figure 317 Subarea Study Area Examples
Study
Area
Urban
Model
Area
Study
Area
Urban
Model
Area 1
Study Area
Within One
Model Area
Urban
Model
Area 2
Study Areas
Within Two
Model Areas
Study
Area
Urban
Model
Area
Study
Area
Statewide Model
Study Area
Outside Urban
Model Areas
Statewide Model
Statewide Model
Statewide Model Data Development Challenges
bull Primary data deficiencies Long distance travelers Rural resident trip-making
bull Challenges in compiling required data Age of data Confidentiality Coordination with data providers Proper level of geography Sufficient sample size
bull Four categories of data for discussion Networks and zone system Demographic and employment data Freight and other economic data NHTS and other behavioral data
Source Florida Department of TransportationCambridge Systematics Inc
Data Challengeshellipwhat is ldquolong distancerdquo
bull gt50 miles (American Traveler Survey)
bull gt100 miles (National Household Travel Survey)
bull IE trips (several states)
bull Trips made by residents for business personal business
bull Recreationaltourist trips
Topic ndash Similarities amp Differences with Urban Models
bull Zone systems
bull Networks
bull Travel behavior assumptions
bull Trip purposesmarket segments
bull Modal issues
bull Truck forecasting
Source Delaware Dept of TransportationWRampA Inc
Statewide vs Urban Model Zone Systems
bull Usually larger zones in statewide models than MPOs
bull Block groups are good starting point for statewide model TAZs
bull Block groups might need splitting along major transportation corridors
bull In small states MPO TAZs could equal SWM zones
bull In most states MPO TAZ equivalency tables will likely be neededdesirable
bull FAF disaggregation approach used for statewide model TAZs
Source FHWA Office of Freight Management and Operations
Source Florida Dept of TransportationCambridge Systematics Inc
FAF zones
Block groups = TAZs
Too small for statewide zones
Too big for statewide zones
Statewide vs Urban Model Networksbull Statewide model networks are sparse in
comparison with MPO models
bull At a minimum statewide networks should include Limited access highways Arterials Major collectors
bull For proper circulation patterns lower class facilities might be added
bull Outside state a very sparse network is needed for major truck corridors
bull Consistent level of detail for networks and zones
Source Tennessee Dept of TransportationCambridge Systematics Inc
MPO network
statewide network
Statewide vs Urban Model Travel Behaviorbull Understanding urban travel behavior
MPO models are usually developed with local household travel survey data
MPOs without recent HH survey data can borrow parameters from other MPO models or guidance documents such as NCHRP 716 Transferable Urban Travel Parameters
bull Rural and long-distance travel behavior A limited number of states have conducted
statewide household travel surveys American Travel Survey was designed to target
long-distance trips but was discontinued long ago
National Household Travel Survey Add-ons can be structured for a sufficient sample of rural trip-makers
NCHRP 735 Rural and Long-Distance Travel Parameters recently published for use in statewide models
Map of 2009 Add-on Surveys Source FHWA
Statewide vs Urban Model Trip Purposes
bull Typical MPO trip purposes are usually collapsed into HBW HBOHBNW and NHB for statewide models
bull Long-distance purposes Business Tourist Personal business
bull Trucks Commodity group is often
surrogate for trip purpose
bull External trips External generally means trip from
outside state Might exclude ldquohalordquo area
surrounding state boundary
Trip PurposeFL
2000 IN LA MA MS RI
Home-based Work 1914 1995 1900 1841 1606 2000
Home-based Nonwork 5547 5208 4770 5206 4917 5700
Home-based Other 2792 5208 4770 1938 4917 5700
Home-based Shop 1576 ndash ndash 1173 ndash ndash
Home-based SocialRecreation
1179 ndash ndash 1395 ndash ndash
Home-base RecreationShop ndash ndash ndash ndash ndash ndash
Home-based School ndash ndash ndash 700 ndash ndash
Nonhome-based 2539 2686 3100 2468 3289 2300
Nonhome-based Work ndash ndash ndash 1027 ndash ndash
Nonhome-based Nonwork ndash ndash ndash 1441 ndash ndash
Other ndash ndash ndash ndash ndash
Truck ndash ndash 170 ndash ndash ndash
Long-Distance 000 111 060 417 022 000
Long-Distance Business ndash ndash 020 284 011 ndash
Long-Distance Personal Business
ndash ndash ndash ndash ndash ndash
Long-Distance Tourist ndash ndash 010 027 000 ndash
Long-Distance Work ndash ndash ndash 105 ndash ndash
Long-Distance NonworkOther
ndash ndash 030 ndash 011 ndash
Statewide vs Urban Model Trip Purposes (Contrsquod)
Number of person trips
Mode Pleasure Business Personal business Other Total Percent
Personal vehicle 12185 5708 3120 789 21802 908Commercial air 1012 746 151 24 1933 8Train 74 150 14 0 238 1Other 4 2 5 12 23 01Ship 19 0 0 0 19 01
Total 13294 6606 3290 825 24015 100Percent 554 275 137 34 100
Percent by trip purpose
Mode Pleasure Business Personal business Other
Personal vehicle 917 864 948 956Commercial air 76 113 46 29Train 06 23 04 0Other 0 0 02 15Ship 01 0 0 0
Trip purpose
Trip purpose
Statewide vs Urban Model Modal Issuesbull Most statewide models are limited to highway trips
bull Modeling of local bus routes is likely overkill
bull Mode choice is usually just a conversion from person trips to vehicle trips via auto occupancy rates
bull Transit components of statewide models are more common in regions with history of fixed guideway transit
bull Transit model is needed to test options such as high speedintercity rail
ATS 95 NHTS 01 MDOT 03-04 ODOT 02-03Personal vehicle 81 89 89 88Commercial air 16 8 10 8Bus 2 2 04 19Train 05 01 0 005Other 05 09 06 205
Survey and yearPrimary mode of transport
Statewide vs Urban Model Modal Issues (Contrsquod)
3727 24 22
32
29 36
10
21
28 22
10
22
12
1529
11
44 43
0
10
20
30
40
50
60
70
80
90
100
Personalvehicle
Bus Train Marine Commercialair
Perc
ent o
f per
son
trips
lt300 300-499 500-599 1000-1999 gt2000
68
85 85 82 85
30
12 11 15 13
0
10
20
30
40
50
60
70
80
90
100
Busi-ness
Plea-sure
pBusi-ness
Vaca-tion
Week-end
Perc
ent o
f per
son
trips
Personal vehicle Commercial air All other
Mode by Trip Purpose Mode by Distance (miles)
Statewide vs Urban Model Modal Issues (Contrsquod)
bull Most statewide assignment models separate trucks from passenger cars Preloading process commonly used for trucks Takes into account more limited route options for truck travel
Source Ohio Dept of Transportation
Statewide vs Urban Model Truck Forecasting
bull Urban modelsndash Freight movements are partially
ldquoexternalrdquo to model (truncated at external zone)
ndash More general categories of truckcommercial trips used
ndash Strong reliance on Quick Response Freight Manual
bull Statewide modelsndash Rooted in commodity flows ndash Often use econometric
approaches to modelingndash Generally model entire length of
freight travel
Freight Model Type by Peer Exchange Participants and
NonParticipants
0123456789
None
Traditio
nal
Commod
ity
Econo
metric
Unkno
wn
NotPeer Exchange
Source TRB Electronic Circular E-C075 Statewide Travel Demand Modeling ndash A Peer Exchange
Topic ndash Data Needs and Considerations
bull Impact of model complexity on data needs
bull Overview of data sources
bull Long-Distance travel gt different focus than MPOregional models
Sources NHTS Weight calculation Pitfalls Key statistics European surveys Lessons learned
Impact of Model Complexity on Data Needsbull Trip table estimation (2 step) ndash highway network
zone system trip table
bull Highway model structure (3 step) ndash 2 step plusbehavioral data for trip rates friction factors auto occupancy rates
bull Freight considerations (3 step) ndash 3 step pluscommodity flows economic data tonnages by mode truck counts
bull Full multi-modal structure (4 step) ndash 3 step plustransit networks and paths logit mode choice model
bull Beyond 4 Step ndash synthetic household generation operational characteristics etc
Source Georgia DOTCambridge Systematics Inc
Statewide Model Data Sources A-Zbull American Travel Survey (ATS)
bull Bureau of Economic Analysis (BEA)
bull Census Journey-to-Work (JTW) data
bull Census Public Use Microsample (PUMS)
bull Census Transportation Planning Package (CTPP)
bull Claritas
bull Commodity Flow Survey (CFS)
Source US Census Bureau
Statewide Model Data Sources (Contrsquod)bull Dun amp Bradstreet (DampB) employment data
bull Employment Security data (ES-202)
bull Freight Analysis Framework (FAF)
bull Highway Performance Monitoring System (HPMS)
bull InfoUSA employment data
bull Metropolitan Planning Organizations (MPO)
bull NCHRP 365 Manual
bull National Highway Planning Network (NHPN)
Source FAF2 Provisional Commodity Origin-Destination Estimates Revised Methodology Report BattelleFHWA
Statewide Model Data Sources (Contrsquod)
bull National Household Travel Survey (NHTS)
bull Origin-destination surveys
bull Regional Economic Models Inc (REMI)
bull State Departments of Transportation
bull TRANSEARCH data
bull Vehicle Use and Identification Survey (VIUS)
bull Woods amp Poole employment forecasts
Source Bureau of Transportation Statistics NHTS
Data Sources on Long-Distance Travel
bull Census Journey-to-Work databull Long distance component
American Travel Survey (77 95) National Household Travel Survey (01) Ohio statewide surveys (02-04) Michigan statewide survey (04-05) California (10-12)
bull Surveys focused on long distance Oregon (96-98 07-08)
bull Others BTS Omnibus surveys Tourism surveys European practice NCHRP Report 735
NHTS and Long-Distance Travelbull Evolution of US national surveys
NPTS (69 77 83 90 95) ATS (77 95)
bull 2001 NHTS ndash Total of 60282 long distance person trips Data collected March 01 to May 02 22204 (37 percent) before 911 38078 (63 percent) afterwards
bull Long distance trip is longer than 50 miles one-way (vs 100 miles in 1995 ATS)
bull 28-day retrospective survey (including travel day)
bull Not your typical travelers Low frequency Distinct segment of the population
Weight Calculation for Long-Distance Travelers
bull Use to ensure valid population-level estimates Accounts for non-response under-coverage multiple phones etc
bull Multi-stage process Match national margins Adjust for data splitting (force sumexpfac = 277208169) Raking (match independent controls) race ethnicity race by
month ethnicity by month sex by age census region MSA status month by day of week
Trim outlier expansion factors max expfac = 4 times mean Replication (correct for sampling error)
bull At least two passes through the data
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Freight Model Types | |||||||||||||||||||
State | Type | Peer Exc | |||||||||||||||||
OR | 4 | 1 | Count of Type | Count of Type | Peer Exc | ||||||||||||||
WI | 3 | 1 | Type | Total | Type | 0 | 1 | Grand Total | |||||||||||
MO | 1 | 1 | 1 | 2 | 1 | 2 | 2 | 4 | |||||||||||
LA | 3 | 1 | 2 | 2 | 2 | 2 | 2 | 4 | |||||||||||
IN | 3 | 1 | 3 | 6 | 3 | 2 | 6 | 8 | |||||||||||
OH | 4 | 1 | 4 | 2 | 4 | 2 | 2 | ||||||||||||
KY | 3 | 1 | Grand Total | 12 | 3 | 3 | |||||||||||||
FL | 3 | 1 | Grand Total | 9 | 12 | 21 | |||||||||||||
VA | 3 | 1 | None | 2 | Not | Peer Exchange | |||||||||||||
NJ | 2 | 1 | Traditional | 2 | None | 2 | 2 | ||||||||||||
MA | 2 | 1 | Commodity | 6 | Traditional | 2 | 2 | ||||||||||||
NH | 1 | 1 | Econometric | 2 | Commodity | 2 | 6 | ||||||||||||
CA | 0 | Econometric | 2 | ||||||||||||||||
TX | 3 | 0 | Unknown | 3 | |||||||||||||||
MI | 3 | 0 | |||||||||||||||||
GA | 0 | ||||||||||||||||||
SC | 1 | 0 | |||||||||||||||||
DE | 2 | 0 | |||||||||||||||||
CT | 0 | ||||||||||||||||||
VT | 2 | 0 | |||||||||||||||||
ME | 1 | 0 | |||||||||||||||||
MN | 0 |
None | None | ||
Traditional | Traditional | ||
Commodity | Commodity | ||
Econometric | Econometric | ||
Unknown | Unknown |
Alternative Statewide Model Structures (Contrsquod)
Beyond 4 Stepbull Potential components
Implement destination choice Tour-basedtrip chains Activity-based travel model Activity micro-simulation Traffic micro-simulation Integration with land use models Integration with economic models
bull Pros Ability to test a more complete set of
transportation strategies and solutions
bull Cons Extensive data requirements Cost to maintain Complexity
Source Oregon Dept of TransportationPBUniversity of Calgary
Deciding on Best Statewide Model Approach
bull What are the motivations for building a statewide travel demand modelbull What are the emerging transportation and policy issues to addressbull How significant is the impact of through travel on your statebull Do you have many multi-state MPOsbull How reliable (and current) are your survey databull Does your state have major international ports or border crossings
StudyArea
UrbanModel
AreaStudyArea
UrbanModelArea 1
Study AreaWithin OneModel Area
UrbanModelArea 2
Study AreasWithin Two Model Areas
StudyArea
UrbanModelArea
StudyArea
Statewide Model
Study AreaOutside Urban
Model Areas
Statewide ModelStatewide Model
Source Florida Department of TransportationCambridge Systematics Inc
Figure 317 Subarea Study Area Examples
Study
Area
Urban
Model
Area
Study
Area
Urban
Model
Area 1
Study Area
Within One
Model Area
Urban
Model
Area 2
Study Areas
Within Two
Model Areas
Study
Area
Urban
Model
Area
Study
Area
Statewide Model
Study Area
Outside Urban
Model Areas
Statewide Model
Statewide Model
Statewide Model Data Development Challenges
bull Primary data deficiencies Long distance travelers Rural resident trip-making
bull Challenges in compiling required data Age of data Confidentiality Coordination with data providers Proper level of geography Sufficient sample size
bull Four categories of data for discussion Networks and zone system Demographic and employment data Freight and other economic data NHTS and other behavioral data
Source Florida Department of TransportationCambridge Systematics Inc
Data Challengeshellipwhat is ldquolong distancerdquo
bull gt50 miles (American Traveler Survey)
bull gt100 miles (National Household Travel Survey)
bull IE trips (several states)
bull Trips made by residents for business personal business
bull Recreationaltourist trips
Topic ndash Similarities amp Differences with Urban Models
bull Zone systems
bull Networks
bull Travel behavior assumptions
bull Trip purposesmarket segments
bull Modal issues
bull Truck forecasting
Source Delaware Dept of TransportationWRampA Inc
Statewide vs Urban Model Zone Systems
bull Usually larger zones in statewide models than MPOs
bull Block groups are good starting point for statewide model TAZs
bull Block groups might need splitting along major transportation corridors
bull In small states MPO TAZs could equal SWM zones
bull In most states MPO TAZ equivalency tables will likely be neededdesirable
bull FAF disaggregation approach used for statewide model TAZs
Source FHWA Office of Freight Management and Operations
Source Florida Dept of TransportationCambridge Systematics Inc
FAF zones
Block groups = TAZs
Too small for statewide zones
Too big for statewide zones
Statewide vs Urban Model Networksbull Statewide model networks are sparse in
comparison with MPO models
bull At a minimum statewide networks should include Limited access highways Arterials Major collectors
bull For proper circulation patterns lower class facilities might be added
bull Outside state a very sparse network is needed for major truck corridors
bull Consistent level of detail for networks and zones
Source Tennessee Dept of TransportationCambridge Systematics Inc
MPO network
statewide network
Statewide vs Urban Model Travel Behaviorbull Understanding urban travel behavior
MPO models are usually developed with local household travel survey data
MPOs without recent HH survey data can borrow parameters from other MPO models or guidance documents such as NCHRP 716 Transferable Urban Travel Parameters
bull Rural and long-distance travel behavior A limited number of states have conducted
statewide household travel surveys American Travel Survey was designed to target
long-distance trips but was discontinued long ago
National Household Travel Survey Add-ons can be structured for a sufficient sample of rural trip-makers
NCHRP 735 Rural and Long-Distance Travel Parameters recently published for use in statewide models
Map of 2009 Add-on Surveys Source FHWA
Statewide vs Urban Model Trip Purposes
bull Typical MPO trip purposes are usually collapsed into HBW HBOHBNW and NHB for statewide models
bull Long-distance purposes Business Tourist Personal business
bull Trucks Commodity group is often
surrogate for trip purpose
bull External trips External generally means trip from
outside state Might exclude ldquohalordquo area
surrounding state boundary
Trip PurposeFL
2000 IN LA MA MS RI
Home-based Work 1914 1995 1900 1841 1606 2000
Home-based Nonwork 5547 5208 4770 5206 4917 5700
Home-based Other 2792 5208 4770 1938 4917 5700
Home-based Shop 1576 ndash ndash 1173 ndash ndash
Home-based SocialRecreation
1179 ndash ndash 1395 ndash ndash
Home-base RecreationShop ndash ndash ndash ndash ndash ndash
Home-based School ndash ndash ndash 700 ndash ndash
Nonhome-based 2539 2686 3100 2468 3289 2300
Nonhome-based Work ndash ndash ndash 1027 ndash ndash
Nonhome-based Nonwork ndash ndash ndash 1441 ndash ndash
Other ndash ndash ndash ndash ndash
Truck ndash ndash 170 ndash ndash ndash
Long-Distance 000 111 060 417 022 000
Long-Distance Business ndash ndash 020 284 011 ndash
Long-Distance Personal Business
ndash ndash ndash ndash ndash ndash
Long-Distance Tourist ndash ndash 010 027 000 ndash
Long-Distance Work ndash ndash ndash 105 ndash ndash
Long-Distance NonworkOther
ndash ndash 030 ndash 011 ndash
Statewide vs Urban Model Trip Purposes (Contrsquod)
Number of person trips
Mode Pleasure Business Personal business Other Total Percent
Personal vehicle 12185 5708 3120 789 21802 908Commercial air 1012 746 151 24 1933 8Train 74 150 14 0 238 1Other 4 2 5 12 23 01Ship 19 0 0 0 19 01
Total 13294 6606 3290 825 24015 100Percent 554 275 137 34 100
Percent by trip purpose
Mode Pleasure Business Personal business Other
Personal vehicle 917 864 948 956Commercial air 76 113 46 29Train 06 23 04 0Other 0 0 02 15Ship 01 0 0 0
Trip purpose
Trip purpose
Statewide vs Urban Model Modal Issuesbull Most statewide models are limited to highway trips
bull Modeling of local bus routes is likely overkill
bull Mode choice is usually just a conversion from person trips to vehicle trips via auto occupancy rates
bull Transit components of statewide models are more common in regions with history of fixed guideway transit
bull Transit model is needed to test options such as high speedintercity rail
ATS 95 NHTS 01 MDOT 03-04 ODOT 02-03Personal vehicle 81 89 89 88Commercial air 16 8 10 8Bus 2 2 04 19Train 05 01 0 005Other 05 09 06 205
Survey and yearPrimary mode of transport
Statewide vs Urban Model Modal Issues (Contrsquod)
3727 24 22
32
29 36
10
21
28 22
10
22
12
1529
11
44 43
0
10
20
30
40
50
60
70
80
90
100
Personalvehicle
Bus Train Marine Commercialair
Perc
ent o
f per
son
trips
lt300 300-499 500-599 1000-1999 gt2000
68
85 85 82 85
30
12 11 15 13
0
10
20
30
40
50
60
70
80
90
100
Busi-ness
Plea-sure
pBusi-ness
Vaca-tion
Week-end
Perc
ent o
f per
son
trips
Personal vehicle Commercial air All other
Mode by Trip Purpose Mode by Distance (miles)
Statewide vs Urban Model Modal Issues (Contrsquod)
bull Most statewide assignment models separate trucks from passenger cars Preloading process commonly used for trucks Takes into account more limited route options for truck travel
Source Ohio Dept of Transportation
Statewide vs Urban Model Truck Forecasting
bull Urban modelsndash Freight movements are partially
ldquoexternalrdquo to model (truncated at external zone)
ndash More general categories of truckcommercial trips used
ndash Strong reliance on Quick Response Freight Manual
bull Statewide modelsndash Rooted in commodity flows ndash Often use econometric
approaches to modelingndash Generally model entire length of
freight travel
Freight Model Type by Peer Exchange Participants and
NonParticipants
0123456789
None
Traditio
nal
Commod
ity
Econo
metric
Unkno
wn
NotPeer Exchange
Source TRB Electronic Circular E-C075 Statewide Travel Demand Modeling ndash A Peer Exchange
Topic ndash Data Needs and Considerations
bull Impact of model complexity on data needs
bull Overview of data sources
bull Long-Distance travel gt different focus than MPOregional models
Sources NHTS Weight calculation Pitfalls Key statistics European surveys Lessons learned
Impact of Model Complexity on Data Needsbull Trip table estimation (2 step) ndash highway network
zone system trip table
bull Highway model structure (3 step) ndash 2 step plusbehavioral data for trip rates friction factors auto occupancy rates
bull Freight considerations (3 step) ndash 3 step pluscommodity flows economic data tonnages by mode truck counts
bull Full multi-modal structure (4 step) ndash 3 step plustransit networks and paths logit mode choice model
bull Beyond 4 Step ndash synthetic household generation operational characteristics etc
Source Georgia DOTCambridge Systematics Inc
Statewide Model Data Sources A-Zbull American Travel Survey (ATS)
bull Bureau of Economic Analysis (BEA)
bull Census Journey-to-Work (JTW) data
bull Census Public Use Microsample (PUMS)
bull Census Transportation Planning Package (CTPP)
bull Claritas
bull Commodity Flow Survey (CFS)
Source US Census Bureau
Statewide Model Data Sources (Contrsquod)bull Dun amp Bradstreet (DampB) employment data
bull Employment Security data (ES-202)
bull Freight Analysis Framework (FAF)
bull Highway Performance Monitoring System (HPMS)
bull InfoUSA employment data
bull Metropolitan Planning Organizations (MPO)
bull NCHRP 365 Manual
bull National Highway Planning Network (NHPN)
Source FAF2 Provisional Commodity Origin-Destination Estimates Revised Methodology Report BattelleFHWA
Statewide Model Data Sources (Contrsquod)
bull National Household Travel Survey (NHTS)
bull Origin-destination surveys
bull Regional Economic Models Inc (REMI)
bull State Departments of Transportation
bull TRANSEARCH data
bull Vehicle Use and Identification Survey (VIUS)
bull Woods amp Poole employment forecasts
Source Bureau of Transportation Statistics NHTS
Data Sources on Long-Distance Travel
bull Census Journey-to-Work databull Long distance component
American Travel Survey (77 95) National Household Travel Survey (01) Ohio statewide surveys (02-04) Michigan statewide survey (04-05) California (10-12)
bull Surveys focused on long distance Oregon (96-98 07-08)
bull Others BTS Omnibus surveys Tourism surveys European practice NCHRP Report 735
NHTS and Long-Distance Travelbull Evolution of US national surveys
NPTS (69 77 83 90 95) ATS (77 95)
bull 2001 NHTS ndash Total of 60282 long distance person trips Data collected March 01 to May 02 22204 (37 percent) before 911 38078 (63 percent) afterwards
bull Long distance trip is longer than 50 miles one-way (vs 100 miles in 1995 ATS)
bull 28-day retrospective survey (including travel day)
bull Not your typical travelers Low frequency Distinct segment of the population
Weight Calculation for Long-Distance Travelers
bull Use to ensure valid population-level estimates Accounts for non-response under-coverage multiple phones etc
bull Multi-stage process Match national margins Adjust for data splitting (force sumexpfac = 277208169) Raking (match independent controls) race ethnicity race by
month ethnicity by month sex by age census region MSA status month by day of week
Trim outlier expansion factors max expfac = 4 times mean Replication (correct for sampling error)
bull At least two passes through the data
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Freight Model Types | |||||||||||||||||||
State | Type | Peer Exc | |||||||||||||||||
OR | 4 | 1 | Count of Type | Count of Type | Peer Exc | ||||||||||||||
WI | 3 | 1 | Type | Total | Type | 0 | 1 | Grand Total | |||||||||||
MO | 1 | 1 | 1 | 2 | 1 | 2 | 2 | 4 | |||||||||||
LA | 3 | 1 | 2 | 2 | 2 | 2 | 2 | 4 | |||||||||||
IN | 3 | 1 | 3 | 6 | 3 | 2 | 6 | 8 | |||||||||||
OH | 4 | 1 | 4 | 2 | 4 | 2 | 2 | ||||||||||||
KY | 3 | 1 | Grand Total | 12 | 3 | 3 | |||||||||||||
FL | 3 | 1 | Grand Total | 9 | 12 | 21 | |||||||||||||
VA | 3 | 1 | None | 2 | Not | Peer Exchange | |||||||||||||
NJ | 2 | 1 | Traditional | 2 | None | 2 | 2 | ||||||||||||
MA | 2 | 1 | Commodity | 6 | Traditional | 2 | 2 | ||||||||||||
NH | 1 | 1 | Econometric | 2 | Commodity | 2 | 6 | ||||||||||||
CA | 0 | Econometric | 2 | ||||||||||||||||
TX | 3 | 0 | Unknown | 3 | |||||||||||||||
MI | 3 | 0 | |||||||||||||||||
GA | 0 | ||||||||||||||||||
SC | 1 | 0 | |||||||||||||||||
DE | 2 | 0 | |||||||||||||||||
CT | 0 | ||||||||||||||||||
VT | 2 | 0 | |||||||||||||||||
ME | 1 | 0 | |||||||||||||||||
MN | 0 |
None | None | ||
Traditional | Traditional | ||
Commodity | Commodity | ||
Econometric | Econometric | ||
Unknown | Unknown |
Deciding on Best Statewide Model Approach
bull What are the motivations for building a statewide travel demand modelbull What are the emerging transportation and policy issues to addressbull How significant is the impact of through travel on your statebull Do you have many multi-state MPOsbull How reliable (and current) are your survey databull Does your state have major international ports or border crossings
StudyArea
UrbanModel
AreaStudyArea
UrbanModelArea 1
Study AreaWithin OneModel Area
UrbanModelArea 2
Study AreasWithin Two Model Areas
StudyArea
UrbanModelArea
StudyArea
Statewide Model
Study AreaOutside Urban
Model Areas
Statewide ModelStatewide Model
Source Florida Department of TransportationCambridge Systematics Inc
Figure 317 Subarea Study Area Examples
Study
Area
Urban
Model
Area
Study
Area
Urban
Model
Area 1
Study Area
Within One
Model Area
Urban
Model
Area 2
Study Areas
Within Two
Model Areas
Study
Area
Urban
Model
Area
Study
Area
Statewide Model
Study Area
Outside Urban
Model Areas
Statewide Model
Statewide Model
Statewide Model Data Development Challenges
bull Primary data deficiencies Long distance travelers Rural resident trip-making
bull Challenges in compiling required data Age of data Confidentiality Coordination with data providers Proper level of geography Sufficient sample size
bull Four categories of data for discussion Networks and zone system Demographic and employment data Freight and other economic data NHTS and other behavioral data
Source Florida Department of TransportationCambridge Systematics Inc
Data Challengeshellipwhat is ldquolong distancerdquo
bull gt50 miles (American Traveler Survey)
bull gt100 miles (National Household Travel Survey)
bull IE trips (several states)
bull Trips made by residents for business personal business
bull Recreationaltourist trips
Topic ndash Similarities amp Differences with Urban Models
bull Zone systems
bull Networks
bull Travel behavior assumptions
bull Trip purposesmarket segments
bull Modal issues
bull Truck forecasting
Source Delaware Dept of TransportationWRampA Inc
Statewide vs Urban Model Zone Systems
bull Usually larger zones in statewide models than MPOs
bull Block groups are good starting point for statewide model TAZs
bull Block groups might need splitting along major transportation corridors
bull In small states MPO TAZs could equal SWM zones
bull In most states MPO TAZ equivalency tables will likely be neededdesirable
bull FAF disaggregation approach used for statewide model TAZs
Source FHWA Office of Freight Management and Operations
Source Florida Dept of TransportationCambridge Systematics Inc
FAF zones
Block groups = TAZs
Too small for statewide zones
Too big for statewide zones
Statewide vs Urban Model Networksbull Statewide model networks are sparse in
comparison with MPO models
bull At a minimum statewide networks should include Limited access highways Arterials Major collectors
bull For proper circulation patterns lower class facilities might be added
bull Outside state a very sparse network is needed for major truck corridors
bull Consistent level of detail for networks and zones
Source Tennessee Dept of TransportationCambridge Systematics Inc
MPO network
statewide network
Statewide vs Urban Model Travel Behaviorbull Understanding urban travel behavior
MPO models are usually developed with local household travel survey data
MPOs without recent HH survey data can borrow parameters from other MPO models or guidance documents such as NCHRP 716 Transferable Urban Travel Parameters
bull Rural and long-distance travel behavior A limited number of states have conducted
statewide household travel surveys American Travel Survey was designed to target
long-distance trips but was discontinued long ago
National Household Travel Survey Add-ons can be structured for a sufficient sample of rural trip-makers
NCHRP 735 Rural and Long-Distance Travel Parameters recently published for use in statewide models
Map of 2009 Add-on Surveys Source FHWA
Statewide vs Urban Model Trip Purposes
bull Typical MPO trip purposes are usually collapsed into HBW HBOHBNW and NHB for statewide models
bull Long-distance purposes Business Tourist Personal business
bull Trucks Commodity group is often
surrogate for trip purpose
bull External trips External generally means trip from
outside state Might exclude ldquohalordquo area
surrounding state boundary
Trip PurposeFL
2000 IN LA MA MS RI
Home-based Work 1914 1995 1900 1841 1606 2000
Home-based Nonwork 5547 5208 4770 5206 4917 5700
Home-based Other 2792 5208 4770 1938 4917 5700
Home-based Shop 1576 ndash ndash 1173 ndash ndash
Home-based SocialRecreation
1179 ndash ndash 1395 ndash ndash
Home-base RecreationShop ndash ndash ndash ndash ndash ndash
Home-based School ndash ndash ndash 700 ndash ndash
Nonhome-based 2539 2686 3100 2468 3289 2300
Nonhome-based Work ndash ndash ndash 1027 ndash ndash
Nonhome-based Nonwork ndash ndash ndash 1441 ndash ndash
Other ndash ndash ndash ndash ndash
Truck ndash ndash 170 ndash ndash ndash
Long-Distance 000 111 060 417 022 000
Long-Distance Business ndash ndash 020 284 011 ndash
Long-Distance Personal Business
ndash ndash ndash ndash ndash ndash
Long-Distance Tourist ndash ndash 010 027 000 ndash
Long-Distance Work ndash ndash ndash 105 ndash ndash
Long-Distance NonworkOther
ndash ndash 030 ndash 011 ndash
Statewide vs Urban Model Trip Purposes (Contrsquod)
Number of person trips
Mode Pleasure Business Personal business Other Total Percent
Personal vehicle 12185 5708 3120 789 21802 908Commercial air 1012 746 151 24 1933 8Train 74 150 14 0 238 1Other 4 2 5 12 23 01Ship 19 0 0 0 19 01
Total 13294 6606 3290 825 24015 100Percent 554 275 137 34 100
Percent by trip purpose
Mode Pleasure Business Personal business Other
Personal vehicle 917 864 948 956Commercial air 76 113 46 29Train 06 23 04 0Other 0 0 02 15Ship 01 0 0 0
Trip purpose
Trip purpose
Statewide vs Urban Model Modal Issuesbull Most statewide models are limited to highway trips
bull Modeling of local bus routes is likely overkill
bull Mode choice is usually just a conversion from person trips to vehicle trips via auto occupancy rates
bull Transit components of statewide models are more common in regions with history of fixed guideway transit
bull Transit model is needed to test options such as high speedintercity rail
ATS 95 NHTS 01 MDOT 03-04 ODOT 02-03Personal vehicle 81 89 89 88Commercial air 16 8 10 8Bus 2 2 04 19Train 05 01 0 005Other 05 09 06 205
Survey and yearPrimary mode of transport
Statewide vs Urban Model Modal Issues (Contrsquod)
3727 24 22
32
29 36
10
21
28 22
10
22
12
1529
11
44 43
0
10
20
30
40
50
60
70
80
90
100
Personalvehicle
Bus Train Marine Commercialair
Perc
ent o
f per
son
trips
lt300 300-499 500-599 1000-1999 gt2000
68
85 85 82 85
30
12 11 15 13
0
10
20
30
40
50
60
70
80
90
100
Busi-ness
Plea-sure
pBusi-ness
Vaca-tion
Week-end
Perc
ent o
f per
son
trips
Personal vehicle Commercial air All other
Mode by Trip Purpose Mode by Distance (miles)
Statewide vs Urban Model Modal Issues (Contrsquod)
bull Most statewide assignment models separate trucks from passenger cars Preloading process commonly used for trucks Takes into account more limited route options for truck travel
Source Ohio Dept of Transportation
Statewide vs Urban Model Truck Forecasting
bull Urban modelsndash Freight movements are partially
ldquoexternalrdquo to model (truncated at external zone)
ndash More general categories of truckcommercial trips used
ndash Strong reliance on Quick Response Freight Manual
bull Statewide modelsndash Rooted in commodity flows ndash Often use econometric
approaches to modelingndash Generally model entire length of
freight travel
Freight Model Type by Peer Exchange Participants and
NonParticipants
0123456789
None
Traditio
nal
Commod
ity
Econo
metric
Unkno
wn
NotPeer Exchange
Source TRB Electronic Circular E-C075 Statewide Travel Demand Modeling ndash A Peer Exchange
Topic ndash Data Needs and Considerations
bull Impact of model complexity on data needs
bull Overview of data sources
bull Long-Distance travel gt different focus than MPOregional models
Sources NHTS Weight calculation Pitfalls Key statistics European surveys Lessons learned
Impact of Model Complexity on Data Needsbull Trip table estimation (2 step) ndash highway network
zone system trip table
bull Highway model structure (3 step) ndash 2 step plusbehavioral data for trip rates friction factors auto occupancy rates
bull Freight considerations (3 step) ndash 3 step pluscommodity flows economic data tonnages by mode truck counts
bull Full multi-modal structure (4 step) ndash 3 step plustransit networks and paths logit mode choice model
bull Beyond 4 Step ndash synthetic household generation operational characteristics etc
Source Georgia DOTCambridge Systematics Inc
Statewide Model Data Sources A-Zbull American Travel Survey (ATS)
bull Bureau of Economic Analysis (BEA)
bull Census Journey-to-Work (JTW) data
bull Census Public Use Microsample (PUMS)
bull Census Transportation Planning Package (CTPP)
bull Claritas
bull Commodity Flow Survey (CFS)
Source US Census Bureau
Statewide Model Data Sources (Contrsquod)bull Dun amp Bradstreet (DampB) employment data
bull Employment Security data (ES-202)
bull Freight Analysis Framework (FAF)
bull Highway Performance Monitoring System (HPMS)
bull InfoUSA employment data
bull Metropolitan Planning Organizations (MPO)
bull NCHRP 365 Manual
bull National Highway Planning Network (NHPN)
Source FAF2 Provisional Commodity Origin-Destination Estimates Revised Methodology Report BattelleFHWA
Statewide Model Data Sources (Contrsquod)
bull National Household Travel Survey (NHTS)
bull Origin-destination surveys
bull Regional Economic Models Inc (REMI)
bull State Departments of Transportation
bull TRANSEARCH data
bull Vehicle Use and Identification Survey (VIUS)
bull Woods amp Poole employment forecasts
Source Bureau of Transportation Statistics NHTS
Data Sources on Long-Distance Travel
bull Census Journey-to-Work databull Long distance component
American Travel Survey (77 95) National Household Travel Survey (01) Ohio statewide surveys (02-04) Michigan statewide survey (04-05) California (10-12)
bull Surveys focused on long distance Oregon (96-98 07-08)
bull Others BTS Omnibus surveys Tourism surveys European practice NCHRP Report 735
NHTS and Long-Distance Travelbull Evolution of US national surveys
NPTS (69 77 83 90 95) ATS (77 95)
bull 2001 NHTS ndash Total of 60282 long distance person trips Data collected March 01 to May 02 22204 (37 percent) before 911 38078 (63 percent) afterwards
bull Long distance trip is longer than 50 miles one-way (vs 100 miles in 1995 ATS)
bull 28-day retrospective survey (including travel day)
bull Not your typical travelers Low frequency Distinct segment of the population
Weight Calculation for Long-Distance Travelers
bull Use to ensure valid population-level estimates Accounts for non-response under-coverage multiple phones etc
bull Multi-stage process Match national margins Adjust for data splitting (force sumexpfac = 277208169) Raking (match independent controls) race ethnicity race by
month ethnicity by month sex by age census region MSA status month by day of week
Trim outlier expansion factors max expfac = 4 times mean Replication (correct for sampling error)
bull At least two passes through the data
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Freight Model Types | |||||||||||||||||||
State | Type | Peer Exc | |||||||||||||||||
OR | 4 | 1 | Count of Type | Count of Type | Peer Exc | ||||||||||||||
WI | 3 | 1 | Type | Total | Type | 0 | 1 | Grand Total | |||||||||||
MO | 1 | 1 | 1 | 2 | 1 | 2 | 2 | 4 | |||||||||||
LA | 3 | 1 | 2 | 2 | 2 | 2 | 2 | 4 | |||||||||||
IN | 3 | 1 | 3 | 6 | 3 | 2 | 6 | 8 | |||||||||||
OH | 4 | 1 | 4 | 2 | 4 | 2 | 2 | ||||||||||||
KY | 3 | 1 | Grand Total | 12 | 3 | 3 | |||||||||||||
FL | 3 | 1 | Grand Total | 9 | 12 | 21 | |||||||||||||
VA | 3 | 1 | None | 2 | Not | Peer Exchange | |||||||||||||
NJ | 2 | 1 | Traditional | 2 | None | 2 | 2 | ||||||||||||
MA | 2 | 1 | Commodity | 6 | Traditional | 2 | 2 | ||||||||||||
NH | 1 | 1 | Econometric | 2 | Commodity | 2 | 6 | ||||||||||||
CA | 0 | Econometric | 2 | ||||||||||||||||
TX | 3 | 0 | Unknown | 3 | |||||||||||||||
MI | 3 | 0 | |||||||||||||||||
GA | 0 | ||||||||||||||||||
SC | 1 | 0 | |||||||||||||||||
DE | 2 | 0 | |||||||||||||||||
CT | 0 | ||||||||||||||||||
VT | 2 | 0 | |||||||||||||||||
ME | 1 | 0 | |||||||||||||||||
MN | 0 |
None | None | ||
Traditional | Traditional | ||
Commodity | Commodity | ||
Econometric | Econometric | ||
Unknown | Unknown |
Statewide Model Data Development Challenges
bull Primary data deficiencies Long distance travelers Rural resident trip-making
bull Challenges in compiling required data Age of data Confidentiality Coordination with data providers Proper level of geography Sufficient sample size
bull Four categories of data for discussion Networks and zone system Demographic and employment data Freight and other economic data NHTS and other behavioral data
Source Florida Department of TransportationCambridge Systematics Inc
Data Challengeshellipwhat is ldquolong distancerdquo
bull gt50 miles (American Traveler Survey)
bull gt100 miles (National Household Travel Survey)
bull IE trips (several states)
bull Trips made by residents for business personal business
bull Recreationaltourist trips
Topic ndash Similarities amp Differences with Urban Models
bull Zone systems
bull Networks
bull Travel behavior assumptions
bull Trip purposesmarket segments
bull Modal issues
bull Truck forecasting
Source Delaware Dept of TransportationWRampA Inc
Statewide vs Urban Model Zone Systems
bull Usually larger zones in statewide models than MPOs
bull Block groups are good starting point for statewide model TAZs
bull Block groups might need splitting along major transportation corridors
bull In small states MPO TAZs could equal SWM zones
bull In most states MPO TAZ equivalency tables will likely be neededdesirable
bull FAF disaggregation approach used for statewide model TAZs
Source FHWA Office of Freight Management and Operations
Source Florida Dept of TransportationCambridge Systematics Inc
FAF zones
Block groups = TAZs
Too small for statewide zones
Too big for statewide zones
Statewide vs Urban Model Networksbull Statewide model networks are sparse in
comparison with MPO models
bull At a minimum statewide networks should include Limited access highways Arterials Major collectors
bull For proper circulation patterns lower class facilities might be added
bull Outside state a very sparse network is needed for major truck corridors
bull Consistent level of detail for networks and zones
Source Tennessee Dept of TransportationCambridge Systematics Inc
MPO network
statewide network
Statewide vs Urban Model Travel Behaviorbull Understanding urban travel behavior
MPO models are usually developed with local household travel survey data
MPOs without recent HH survey data can borrow parameters from other MPO models or guidance documents such as NCHRP 716 Transferable Urban Travel Parameters
bull Rural and long-distance travel behavior A limited number of states have conducted
statewide household travel surveys American Travel Survey was designed to target
long-distance trips but was discontinued long ago
National Household Travel Survey Add-ons can be structured for a sufficient sample of rural trip-makers
NCHRP 735 Rural and Long-Distance Travel Parameters recently published for use in statewide models
Map of 2009 Add-on Surveys Source FHWA
Statewide vs Urban Model Trip Purposes
bull Typical MPO trip purposes are usually collapsed into HBW HBOHBNW and NHB for statewide models
bull Long-distance purposes Business Tourist Personal business
bull Trucks Commodity group is often
surrogate for trip purpose
bull External trips External generally means trip from
outside state Might exclude ldquohalordquo area
surrounding state boundary
Trip PurposeFL
2000 IN LA MA MS RI
Home-based Work 1914 1995 1900 1841 1606 2000
Home-based Nonwork 5547 5208 4770 5206 4917 5700
Home-based Other 2792 5208 4770 1938 4917 5700
Home-based Shop 1576 ndash ndash 1173 ndash ndash
Home-based SocialRecreation
1179 ndash ndash 1395 ndash ndash
Home-base RecreationShop ndash ndash ndash ndash ndash ndash
Home-based School ndash ndash ndash 700 ndash ndash
Nonhome-based 2539 2686 3100 2468 3289 2300
Nonhome-based Work ndash ndash ndash 1027 ndash ndash
Nonhome-based Nonwork ndash ndash ndash 1441 ndash ndash
Other ndash ndash ndash ndash ndash
Truck ndash ndash 170 ndash ndash ndash
Long-Distance 000 111 060 417 022 000
Long-Distance Business ndash ndash 020 284 011 ndash
Long-Distance Personal Business
ndash ndash ndash ndash ndash ndash
Long-Distance Tourist ndash ndash 010 027 000 ndash
Long-Distance Work ndash ndash ndash 105 ndash ndash
Long-Distance NonworkOther
ndash ndash 030 ndash 011 ndash
Statewide vs Urban Model Trip Purposes (Contrsquod)
Number of person trips
Mode Pleasure Business Personal business Other Total Percent
Personal vehicle 12185 5708 3120 789 21802 908Commercial air 1012 746 151 24 1933 8Train 74 150 14 0 238 1Other 4 2 5 12 23 01Ship 19 0 0 0 19 01
Total 13294 6606 3290 825 24015 100Percent 554 275 137 34 100
Percent by trip purpose
Mode Pleasure Business Personal business Other
Personal vehicle 917 864 948 956Commercial air 76 113 46 29Train 06 23 04 0Other 0 0 02 15Ship 01 0 0 0
Trip purpose
Trip purpose
Statewide vs Urban Model Modal Issuesbull Most statewide models are limited to highway trips
bull Modeling of local bus routes is likely overkill
bull Mode choice is usually just a conversion from person trips to vehicle trips via auto occupancy rates
bull Transit components of statewide models are more common in regions with history of fixed guideway transit
bull Transit model is needed to test options such as high speedintercity rail
ATS 95 NHTS 01 MDOT 03-04 ODOT 02-03Personal vehicle 81 89 89 88Commercial air 16 8 10 8Bus 2 2 04 19Train 05 01 0 005Other 05 09 06 205
Survey and yearPrimary mode of transport
Statewide vs Urban Model Modal Issues (Contrsquod)
3727 24 22
32
29 36
10
21
28 22
10
22
12
1529
11
44 43
0
10
20
30
40
50
60
70
80
90
100
Personalvehicle
Bus Train Marine Commercialair
Perc
ent o
f per
son
trips
lt300 300-499 500-599 1000-1999 gt2000
68
85 85 82 85
30
12 11 15 13
0
10
20
30
40
50
60
70
80
90
100
Busi-ness
Plea-sure
pBusi-ness
Vaca-tion
Week-end
Perc
ent o
f per
son
trips
Personal vehicle Commercial air All other
Mode by Trip Purpose Mode by Distance (miles)
Statewide vs Urban Model Modal Issues (Contrsquod)
bull Most statewide assignment models separate trucks from passenger cars Preloading process commonly used for trucks Takes into account more limited route options for truck travel
Source Ohio Dept of Transportation
Statewide vs Urban Model Truck Forecasting
bull Urban modelsndash Freight movements are partially
ldquoexternalrdquo to model (truncated at external zone)
ndash More general categories of truckcommercial trips used
ndash Strong reliance on Quick Response Freight Manual
bull Statewide modelsndash Rooted in commodity flows ndash Often use econometric
approaches to modelingndash Generally model entire length of
freight travel
Freight Model Type by Peer Exchange Participants and
NonParticipants
0123456789
None
Traditio
nal
Commod
ity
Econo
metric
Unkno
wn
NotPeer Exchange
Source TRB Electronic Circular E-C075 Statewide Travel Demand Modeling ndash A Peer Exchange
Topic ndash Data Needs and Considerations
bull Impact of model complexity on data needs
bull Overview of data sources
bull Long-Distance travel gt different focus than MPOregional models
Sources NHTS Weight calculation Pitfalls Key statistics European surveys Lessons learned
Impact of Model Complexity on Data Needsbull Trip table estimation (2 step) ndash highway network
zone system trip table
bull Highway model structure (3 step) ndash 2 step plusbehavioral data for trip rates friction factors auto occupancy rates
bull Freight considerations (3 step) ndash 3 step pluscommodity flows economic data tonnages by mode truck counts
bull Full multi-modal structure (4 step) ndash 3 step plustransit networks and paths logit mode choice model
bull Beyond 4 Step ndash synthetic household generation operational characteristics etc
Source Georgia DOTCambridge Systematics Inc
Statewide Model Data Sources A-Zbull American Travel Survey (ATS)
bull Bureau of Economic Analysis (BEA)
bull Census Journey-to-Work (JTW) data
bull Census Public Use Microsample (PUMS)
bull Census Transportation Planning Package (CTPP)
bull Claritas
bull Commodity Flow Survey (CFS)
Source US Census Bureau
Statewide Model Data Sources (Contrsquod)bull Dun amp Bradstreet (DampB) employment data
bull Employment Security data (ES-202)
bull Freight Analysis Framework (FAF)
bull Highway Performance Monitoring System (HPMS)
bull InfoUSA employment data
bull Metropolitan Planning Organizations (MPO)
bull NCHRP 365 Manual
bull National Highway Planning Network (NHPN)
Source FAF2 Provisional Commodity Origin-Destination Estimates Revised Methodology Report BattelleFHWA
Statewide Model Data Sources (Contrsquod)
bull National Household Travel Survey (NHTS)
bull Origin-destination surveys
bull Regional Economic Models Inc (REMI)
bull State Departments of Transportation
bull TRANSEARCH data
bull Vehicle Use and Identification Survey (VIUS)
bull Woods amp Poole employment forecasts
Source Bureau of Transportation Statistics NHTS
Data Sources on Long-Distance Travel
bull Census Journey-to-Work databull Long distance component
American Travel Survey (77 95) National Household Travel Survey (01) Ohio statewide surveys (02-04) Michigan statewide survey (04-05) California (10-12)
bull Surveys focused on long distance Oregon (96-98 07-08)
bull Others BTS Omnibus surveys Tourism surveys European practice NCHRP Report 735
NHTS and Long-Distance Travelbull Evolution of US national surveys
NPTS (69 77 83 90 95) ATS (77 95)
bull 2001 NHTS ndash Total of 60282 long distance person trips Data collected March 01 to May 02 22204 (37 percent) before 911 38078 (63 percent) afterwards
bull Long distance trip is longer than 50 miles one-way (vs 100 miles in 1995 ATS)
bull 28-day retrospective survey (including travel day)
bull Not your typical travelers Low frequency Distinct segment of the population
Weight Calculation for Long-Distance Travelers
bull Use to ensure valid population-level estimates Accounts for non-response under-coverage multiple phones etc
bull Multi-stage process Match national margins Adjust for data splitting (force sumexpfac = 277208169) Raking (match independent controls) race ethnicity race by
month ethnicity by month sex by age census region MSA status month by day of week
Trim outlier expansion factors max expfac = 4 times mean Replication (correct for sampling error)
bull At least two passes through the data
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Freight Model Types | |||||||||||||||||||
State | Type | Peer Exc | |||||||||||||||||
OR | 4 | 1 | Count of Type | Count of Type | Peer Exc | ||||||||||||||
WI | 3 | 1 | Type | Total | Type | 0 | 1 | Grand Total | |||||||||||
MO | 1 | 1 | 1 | 2 | 1 | 2 | 2 | 4 | |||||||||||
LA | 3 | 1 | 2 | 2 | 2 | 2 | 2 | 4 | |||||||||||
IN | 3 | 1 | 3 | 6 | 3 | 2 | 6 | 8 | |||||||||||
OH | 4 | 1 | 4 | 2 | 4 | 2 | 2 | ||||||||||||
KY | 3 | 1 | Grand Total | 12 | 3 | 3 | |||||||||||||
FL | 3 | 1 | Grand Total | 9 | 12 | 21 | |||||||||||||
VA | 3 | 1 | None | 2 | Not | Peer Exchange | |||||||||||||
NJ | 2 | 1 | Traditional | 2 | None | 2 | 2 | ||||||||||||
MA | 2 | 1 | Commodity | 6 | Traditional | 2 | 2 | ||||||||||||
NH | 1 | 1 | Econometric | 2 | Commodity | 2 | 6 | ||||||||||||
CA | 0 | Econometric | 2 | ||||||||||||||||
TX | 3 | 0 | Unknown | 3 | |||||||||||||||
MI | 3 | 0 | |||||||||||||||||
GA | 0 | ||||||||||||||||||
SC | 1 | 0 | |||||||||||||||||
DE | 2 | 0 | |||||||||||||||||
CT | 0 | ||||||||||||||||||
VT | 2 | 0 | |||||||||||||||||
ME | 1 | 0 | |||||||||||||||||
MN | 0 |
None | None | ||
Traditional | Traditional | ||
Commodity | Commodity | ||
Econometric | Econometric | ||
Unknown | Unknown |
Data Challengeshellipwhat is ldquolong distancerdquo
bull gt50 miles (American Traveler Survey)
bull gt100 miles (National Household Travel Survey)
bull IE trips (several states)
bull Trips made by residents for business personal business
bull Recreationaltourist trips
Topic ndash Similarities amp Differences with Urban Models
bull Zone systems
bull Networks
bull Travel behavior assumptions
bull Trip purposesmarket segments
bull Modal issues
bull Truck forecasting
Source Delaware Dept of TransportationWRampA Inc
Statewide vs Urban Model Zone Systems
bull Usually larger zones in statewide models than MPOs
bull Block groups are good starting point for statewide model TAZs
bull Block groups might need splitting along major transportation corridors
bull In small states MPO TAZs could equal SWM zones
bull In most states MPO TAZ equivalency tables will likely be neededdesirable
bull FAF disaggregation approach used for statewide model TAZs
Source FHWA Office of Freight Management and Operations
Source Florida Dept of TransportationCambridge Systematics Inc
FAF zones
Block groups = TAZs
Too small for statewide zones
Too big for statewide zones
Statewide vs Urban Model Networksbull Statewide model networks are sparse in
comparison with MPO models
bull At a minimum statewide networks should include Limited access highways Arterials Major collectors
bull For proper circulation patterns lower class facilities might be added
bull Outside state a very sparse network is needed for major truck corridors
bull Consistent level of detail for networks and zones
Source Tennessee Dept of TransportationCambridge Systematics Inc
MPO network
statewide network
Statewide vs Urban Model Travel Behaviorbull Understanding urban travel behavior
MPO models are usually developed with local household travel survey data
MPOs without recent HH survey data can borrow parameters from other MPO models or guidance documents such as NCHRP 716 Transferable Urban Travel Parameters
bull Rural and long-distance travel behavior A limited number of states have conducted
statewide household travel surveys American Travel Survey was designed to target
long-distance trips but was discontinued long ago
National Household Travel Survey Add-ons can be structured for a sufficient sample of rural trip-makers
NCHRP 735 Rural and Long-Distance Travel Parameters recently published for use in statewide models
Map of 2009 Add-on Surveys Source FHWA
Statewide vs Urban Model Trip Purposes
bull Typical MPO trip purposes are usually collapsed into HBW HBOHBNW and NHB for statewide models
bull Long-distance purposes Business Tourist Personal business
bull Trucks Commodity group is often
surrogate for trip purpose
bull External trips External generally means trip from
outside state Might exclude ldquohalordquo area
surrounding state boundary
Trip PurposeFL
2000 IN LA MA MS RI
Home-based Work 1914 1995 1900 1841 1606 2000
Home-based Nonwork 5547 5208 4770 5206 4917 5700
Home-based Other 2792 5208 4770 1938 4917 5700
Home-based Shop 1576 ndash ndash 1173 ndash ndash
Home-based SocialRecreation
1179 ndash ndash 1395 ndash ndash
Home-base RecreationShop ndash ndash ndash ndash ndash ndash
Home-based School ndash ndash ndash 700 ndash ndash
Nonhome-based 2539 2686 3100 2468 3289 2300
Nonhome-based Work ndash ndash ndash 1027 ndash ndash
Nonhome-based Nonwork ndash ndash ndash 1441 ndash ndash
Other ndash ndash ndash ndash ndash
Truck ndash ndash 170 ndash ndash ndash
Long-Distance 000 111 060 417 022 000
Long-Distance Business ndash ndash 020 284 011 ndash
Long-Distance Personal Business
ndash ndash ndash ndash ndash ndash
Long-Distance Tourist ndash ndash 010 027 000 ndash
Long-Distance Work ndash ndash ndash 105 ndash ndash
Long-Distance NonworkOther
ndash ndash 030 ndash 011 ndash
Statewide vs Urban Model Trip Purposes (Contrsquod)
Number of person trips
Mode Pleasure Business Personal business Other Total Percent
Personal vehicle 12185 5708 3120 789 21802 908Commercial air 1012 746 151 24 1933 8Train 74 150 14 0 238 1Other 4 2 5 12 23 01Ship 19 0 0 0 19 01
Total 13294 6606 3290 825 24015 100Percent 554 275 137 34 100
Percent by trip purpose
Mode Pleasure Business Personal business Other
Personal vehicle 917 864 948 956Commercial air 76 113 46 29Train 06 23 04 0Other 0 0 02 15Ship 01 0 0 0
Trip purpose
Trip purpose
Statewide vs Urban Model Modal Issuesbull Most statewide models are limited to highway trips
bull Modeling of local bus routes is likely overkill
bull Mode choice is usually just a conversion from person trips to vehicle trips via auto occupancy rates
bull Transit components of statewide models are more common in regions with history of fixed guideway transit
bull Transit model is needed to test options such as high speedintercity rail
ATS 95 NHTS 01 MDOT 03-04 ODOT 02-03Personal vehicle 81 89 89 88Commercial air 16 8 10 8Bus 2 2 04 19Train 05 01 0 005Other 05 09 06 205
Survey and yearPrimary mode of transport
Statewide vs Urban Model Modal Issues (Contrsquod)
3727 24 22
32
29 36
10
21
28 22
10
22
12
1529
11
44 43
0
10
20
30
40
50
60
70
80
90
100
Personalvehicle
Bus Train Marine Commercialair
Perc
ent o
f per
son
trips
lt300 300-499 500-599 1000-1999 gt2000
68
85 85 82 85
30
12 11 15 13
0
10
20
30
40
50
60
70
80
90
100
Busi-ness
Plea-sure
pBusi-ness
Vaca-tion
Week-end
Perc
ent o
f per
son
trips
Personal vehicle Commercial air All other
Mode by Trip Purpose Mode by Distance (miles)
Statewide vs Urban Model Modal Issues (Contrsquod)
bull Most statewide assignment models separate trucks from passenger cars Preloading process commonly used for trucks Takes into account more limited route options for truck travel
Source Ohio Dept of Transportation
Statewide vs Urban Model Truck Forecasting
bull Urban modelsndash Freight movements are partially
ldquoexternalrdquo to model (truncated at external zone)
ndash More general categories of truckcommercial trips used
ndash Strong reliance on Quick Response Freight Manual
bull Statewide modelsndash Rooted in commodity flows ndash Often use econometric
approaches to modelingndash Generally model entire length of
freight travel
Freight Model Type by Peer Exchange Participants and
NonParticipants
0123456789
None
Traditio
nal
Commod
ity
Econo
metric
Unkno
wn
NotPeer Exchange
Source TRB Electronic Circular E-C075 Statewide Travel Demand Modeling ndash A Peer Exchange
Topic ndash Data Needs and Considerations
bull Impact of model complexity on data needs
bull Overview of data sources
bull Long-Distance travel gt different focus than MPOregional models
Sources NHTS Weight calculation Pitfalls Key statistics European surveys Lessons learned
Impact of Model Complexity on Data Needsbull Trip table estimation (2 step) ndash highway network
zone system trip table
bull Highway model structure (3 step) ndash 2 step plusbehavioral data for trip rates friction factors auto occupancy rates
bull Freight considerations (3 step) ndash 3 step pluscommodity flows economic data tonnages by mode truck counts
bull Full multi-modal structure (4 step) ndash 3 step plustransit networks and paths logit mode choice model
bull Beyond 4 Step ndash synthetic household generation operational characteristics etc
Source Georgia DOTCambridge Systematics Inc
Statewide Model Data Sources A-Zbull American Travel Survey (ATS)
bull Bureau of Economic Analysis (BEA)
bull Census Journey-to-Work (JTW) data
bull Census Public Use Microsample (PUMS)
bull Census Transportation Planning Package (CTPP)
bull Claritas
bull Commodity Flow Survey (CFS)
Source US Census Bureau
Statewide Model Data Sources (Contrsquod)bull Dun amp Bradstreet (DampB) employment data
bull Employment Security data (ES-202)
bull Freight Analysis Framework (FAF)
bull Highway Performance Monitoring System (HPMS)
bull InfoUSA employment data
bull Metropolitan Planning Organizations (MPO)
bull NCHRP 365 Manual
bull National Highway Planning Network (NHPN)
Source FAF2 Provisional Commodity Origin-Destination Estimates Revised Methodology Report BattelleFHWA
Statewide Model Data Sources (Contrsquod)
bull National Household Travel Survey (NHTS)
bull Origin-destination surveys
bull Regional Economic Models Inc (REMI)
bull State Departments of Transportation
bull TRANSEARCH data
bull Vehicle Use and Identification Survey (VIUS)
bull Woods amp Poole employment forecasts
Source Bureau of Transportation Statistics NHTS
Data Sources on Long-Distance Travel
bull Census Journey-to-Work databull Long distance component
American Travel Survey (77 95) National Household Travel Survey (01) Ohio statewide surveys (02-04) Michigan statewide survey (04-05) California (10-12)
bull Surveys focused on long distance Oregon (96-98 07-08)
bull Others BTS Omnibus surveys Tourism surveys European practice NCHRP Report 735
NHTS and Long-Distance Travelbull Evolution of US national surveys
NPTS (69 77 83 90 95) ATS (77 95)
bull 2001 NHTS ndash Total of 60282 long distance person trips Data collected March 01 to May 02 22204 (37 percent) before 911 38078 (63 percent) afterwards
bull Long distance trip is longer than 50 miles one-way (vs 100 miles in 1995 ATS)
bull 28-day retrospective survey (including travel day)
bull Not your typical travelers Low frequency Distinct segment of the population
Weight Calculation for Long-Distance Travelers
bull Use to ensure valid population-level estimates Accounts for non-response under-coverage multiple phones etc
bull Multi-stage process Match national margins Adjust for data splitting (force sumexpfac = 277208169) Raking (match independent controls) race ethnicity race by
month ethnicity by month sex by age census region MSA status month by day of week
Trim outlier expansion factors max expfac = 4 times mean Replication (correct for sampling error)
bull At least two passes through the data
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Freight Model Types | |||||||||||||||||||
State | Type | Peer Exc | |||||||||||||||||
OR | 4 | 1 | Count of Type | Count of Type | Peer Exc | ||||||||||||||
WI | 3 | 1 | Type | Total | Type | 0 | 1 | Grand Total | |||||||||||
MO | 1 | 1 | 1 | 2 | 1 | 2 | 2 | 4 | |||||||||||
LA | 3 | 1 | 2 | 2 | 2 | 2 | 2 | 4 | |||||||||||
IN | 3 | 1 | 3 | 6 | 3 | 2 | 6 | 8 | |||||||||||
OH | 4 | 1 | 4 | 2 | 4 | 2 | 2 | ||||||||||||
KY | 3 | 1 | Grand Total | 12 | 3 | 3 | |||||||||||||
FL | 3 | 1 | Grand Total | 9 | 12 | 21 | |||||||||||||
VA | 3 | 1 | None | 2 | Not | Peer Exchange | |||||||||||||
NJ | 2 | 1 | Traditional | 2 | None | 2 | 2 | ||||||||||||
MA | 2 | 1 | Commodity | 6 | Traditional | 2 | 2 | ||||||||||||
NH | 1 | 1 | Econometric | 2 | Commodity | 2 | 6 | ||||||||||||
CA | 0 | Econometric | 2 | ||||||||||||||||
TX | 3 | 0 | Unknown | 3 | |||||||||||||||
MI | 3 | 0 | |||||||||||||||||
GA | 0 | ||||||||||||||||||
SC | 1 | 0 | |||||||||||||||||
DE | 2 | 0 | |||||||||||||||||
CT | 0 | ||||||||||||||||||
VT | 2 | 0 | |||||||||||||||||
ME | 1 | 0 | |||||||||||||||||
MN | 0 |
None | None | ||
Traditional | Traditional | ||
Commodity | Commodity | ||
Econometric | Econometric | ||
Unknown | Unknown |
Topic ndash Similarities amp Differences with Urban Models
bull Zone systems
bull Networks
bull Travel behavior assumptions
bull Trip purposesmarket segments
bull Modal issues
bull Truck forecasting
Source Delaware Dept of TransportationWRampA Inc
Statewide vs Urban Model Zone Systems
bull Usually larger zones in statewide models than MPOs
bull Block groups are good starting point for statewide model TAZs
bull Block groups might need splitting along major transportation corridors
bull In small states MPO TAZs could equal SWM zones
bull In most states MPO TAZ equivalency tables will likely be neededdesirable
bull FAF disaggregation approach used for statewide model TAZs
Source FHWA Office of Freight Management and Operations
Source Florida Dept of TransportationCambridge Systematics Inc
FAF zones
Block groups = TAZs
Too small for statewide zones
Too big for statewide zones
Statewide vs Urban Model Networksbull Statewide model networks are sparse in
comparison with MPO models
bull At a minimum statewide networks should include Limited access highways Arterials Major collectors
bull For proper circulation patterns lower class facilities might be added
bull Outside state a very sparse network is needed for major truck corridors
bull Consistent level of detail for networks and zones
Source Tennessee Dept of TransportationCambridge Systematics Inc
MPO network
statewide network
Statewide vs Urban Model Travel Behaviorbull Understanding urban travel behavior
MPO models are usually developed with local household travel survey data
MPOs without recent HH survey data can borrow parameters from other MPO models or guidance documents such as NCHRP 716 Transferable Urban Travel Parameters
bull Rural and long-distance travel behavior A limited number of states have conducted
statewide household travel surveys American Travel Survey was designed to target
long-distance trips but was discontinued long ago
National Household Travel Survey Add-ons can be structured for a sufficient sample of rural trip-makers
NCHRP 735 Rural and Long-Distance Travel Parameters recently published for use in statewide models
Map of 2009 Add-on Surveys Source FHWA
Statewide vs Urban Model Trip Purposes
bull Typical MPO trip purposes are usually collapsed into HBW HBOHBNW and NHB for statewide models
bull Long-distance purposes Business Tourist Personal business
bull Trucks Commodity group is often
surrogate for trip purpose
bull External trips External generally means trip from
outside state Might exclude ldquohalordquo area
surrounding state boundary
Trip PurposeFL
2000 IN LA MA MS RI
Home-based Work 1914 1995 1900 1841 1606 2000
Home-based Nonwork 5547 5208 4770 5206 4917 5700
Home-based Other 2792 5208 4770 1938 4917 5700
Home-based Shop 1576 ndash ndash 1173 ndash ndash
Home-based SocialRecreation
1179 ndash ndash 1395 ndash ndash
Home-base RecreationShop ndash ndash ndash ndash ndash ndash
Home-based School ndash ndash ndash 700 ndash ndash
Nonhome-based 2539 2686 3100 2468 3289 2300
Nonhome-based Work ndash ndash ndash 1027 ndash ndash
Nonhome-based Nonwork ndash ndash ndash 1441 ndash ndash
Other ndash ndash ndash ndash ndash
Truck ndash ndash 170 ndash ndash ndash
Long-Distance 000 111 060 417 022 000
Long-Distance Business ndash ndash 020 284 011 ndash
Long-Distance Personal Business
ndash ndash ndash ndash ndash ndash
Long-Distance Tourist ndash ndash 010 027 000 ndash
Long-Distance Work ndash ndash ndash 105 ndash ndash
Long-Distance NonworkOther
ndash ndash 030 ndash 011 ndash
Statewide vs Urban Model Trip Purposes (Contrsquod)
Number of person trips
Mode Pleasure Business Personal business Other Total Percent
Personal vehicle 12185 5708 3120 789 21802 908Commercial air 1012 746 151 24 1933 8Train 74 150 14 0 238 1Other 4 2 5 12 23 01Ship 19 0 0 0 19 01
Total 13294 6606 3290 825 24015 100Percent 554 275 137 34 100
Percent by trip purpose
Mode Pleasure Business Personal business Other
Personal vehicle 917 864 948 956Commercial air 76 113 46 29Train 06 23 04 0Other 0 0 02 15Ship 01 0 0 0
Trip purpose
Trip purpose
Statewide vs Urban Model Modal Issuesbull Most statewide models are limited to highway trips
bull Modeling of local bus routes is likely overkill
bull Mode choice is usually just a conversion from person trips to vehicle trips via auto occupancy rates
bull Transit components of statewide models are more common in regions with history of fixed guideway transit
bull Transit model is needed to test options such as high speedintercity rail
ATS 95 NHTS 01 MDOT 03-04 ODOT 02-03Personal vehicle 81 89 89 88Commercial air 16 8 10 8Bus 2 2 04 19Train 05 01 0 005Other 05 09 06 205
Survey and yearPrimary mode of transport
Statewide vs Urban Model Modal Issues (Contrsquod)
3727 24 22
32
29 36
10
21
28 22
10
22
12
1529
11
44 43
0
10
20
30
40
50
60
70
80
90
100
Personalvehicle
Bus Train Marine Commercialair
Perc
ent o
f per
son
trips
lt300 300-499 500-599 1000-1999 gt2000
68
85 85 82 85
30
12 11 15 13
0
10
20
30
40
50
60
70
80
90
100
Busi-ness
Plea-sure
pBusi-ness
Vaca-tion
Week-end
Perc
ent o
f per
son
trips
Personal vehicle Commercial air All other
Mode by Trip Purpose Mode by Distance (miles)
Statewide vs Urban Model Modal Issues (Contrsquod)
bull Most statewide assignment models separate trucks from passenger cars Preloading process commonly used for trucks Takes into account more limited route options for truck travel
Source Ohio Dept of Transportation
Statewide vs Urban Model Truck Forecasting
bull Urban modelsndash Freight movements are partially
ldquoexternalrdquo to model (truncated at external zone)
ndash More general categories of truckcommercial trips used
ndash Strong reliance on Quick Response Freight Manual
bull Statewide modelsndash Rooted in commodity flows ndash Often use econometric
approaches to modelingndash Generally model entire length of
freight travel
Freight Model Type by Peer Exchange Participants and
NonParticipants
0123456789
None
Traditio
nal
Commod
ity
Econo
metric
Unkno
wn
NotPeer Exchange
Source TRB Electronic Circular E-C075 Statewide Travel Demand Modeling ndash A Peer Exchange
Topic ndash Data Needs and Considerations
bull Impact of model complexity on data needs
bull Overview of data sources
bull Long-Distance travel gt different focus than MPOregional models
Sources NHTS Weight calculation Pitfalls Key statistics European surveys Lessons learned
Impact of Model Complexity on Data Needsbull Trip table estimation (2 step) ndash highway network
zone system trip table
bull Highway model structure (3 step) ndash 2 step plusbehavioral data for trip rates friction factors auto occupancy rates
bull Freight considerations (3 step) ndash 3 step pluscommodity flows economic data tonnages by mode truck counts
bull Full multi-modal structure (4 step) ndash 3 step plustransit networks and paths logit mode choice model
bull Beyond 4 Step ndash synthetic household generation operational characteristics etc
Source Georgia DOTCambridge Systematics Inc
Statewide Model Data Sources A-Zbull American Travel Survey (ATS)
bull Bureau of Economic Analysis (BEA)
bull Census Journey-to-Work (JTW) data
bull Census Public Use Microsample (PUMS)
bull Census Transportation Planning Package (CTPP)
bull Claritas
bull Commodity Flow Survey (CFS)
Source US Census Bureau
Statewide Model Data Sources (Contrsquod)bull Dun amp Bradstreet (DampB) employment data
bull Employment Security data (ES-202)
bull Freight Analysis Framework (FAF)
bull Highway Performance Monitoring System (HPMS)
bull InfoUSA employment data
bull Metropolitan Planning Organizations (MPO)
bull NCHRP 365 Manual
bull National Highway Planning Network (NHPN)
Source FAF2 Provisional Commodity Origin-Destination Estimates Revised Methodology Report BattelleFHWA
Statewide Model Data Sources (Contrsquod)
bull National Household Travel Survey (NHTS)
bull Origin-destination surveys
bull Regional Economic Models Inc (REMI)
bull State Departments of Transportation
bull TRANSEARCH data
bull Vehicle Use and Identification Survey (VIUS)
bull Woods amp Poole employment forecasts
Source Bureau of Transportation Statistics NHTS
Data Sources on Long-Distance Travel
bull Census Journey-to-Work databull Long distance component
American Travel Survey (77 95) National Household Travel Survey (01) Ohio statewide surveys (02-04) Michigan statewide survey (04-05) California (10-12)
bull Surveys focused on long distance Oregon (96-98 07-08)
bull Others BTS Omnibus surveys Tourism surveys European practice NCHRP Report 735
NHTS and Long-Distance Travelbull Evolution of US national surveys
NPTS (69 77 83 90 95) ATS (77 95)
bull 2001 NHTS ndash Total of 60282 long distance person trips Data collected March 01 to May 02 22204 (37 percent) before 911 38078 (63 percent) afterwards
bull Long distance trip is longer than 50 miles one-way (vs 100 miles in 1995 ATS)
bull 28-day retrospective survey (including travel day)
bull Not your typical travelers Low frequency Distinct segment of the population
Weight Calculation for Long-Distance Travelers
bull Use to ensure valid population-level estimates Accounts for non-response under-coverage multiple phones etc
bull Multi-stage process Match national margins Adjust for data splitting (force sumexpfac = 277208169) Raking (match independent controls) race ethnicity race by
month ethnicity by month sex by age census region MSA status month by day of week
Trim outlier expansion factors max expfac = 4 times mean Replication (correct for sampling error)
bull At least two passes through the data
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Freight Model Types | |||||||||||||||||||
State | Type | Peer Exc | |||||||||||||||||
OR | 4 | 1 | Count of Type | Count of Type | Peer Exc | ||||||||||||||
WI | 3 | 1 | Type | Total | Type | 0 | 1 | Grand Total | |||||||||||
MO | 1 | 1 | 1 | 2 | 1 | 2 | 2 | 4 | |||||||||||
LA | 3 | 1 | 2 | 2 | 2 | 2 | 2 | 4 | |||||||||||
IN | 3 | 1 | 3 | 6 | 3 | 2 | 6 | 8 | |||||||||||
OH | 4 | 1 | 4 | 2 | 4 | 2 | 2 | ||||||||||||
KY | 3 | 1 | Grand Total | 12 | 3 | 3 | |||||||||||||
FL | 3 | 1 | Grand Total | 9 | 12 | 21 | |||||||||||||
VA | 3 | 1 | None | 2 | Not | Peer Exchange | |||||||||||||
NJ | 2 | 1 | Traditional | 2 | None | 2 | 2 | ||||||||||||
MA | 2 | 1 | Commodity | 6 | Traditional | 2 | 2 | ||||||||||||
NH | 1 | 1 | Econometric | 2 | Commodity | 2 | 6 | ||||||||||||
CA | 0 | Econometric | 2 | ||||||||||||||||
TX | 3 | 0 | Unknown | 3 | |||||||||||||||
MI | 3 | 0 | |||||||||||||||||
GA | 0 | ||||||||||||||||||
SC | 1 | 0 | |||||||||||||||||
DE | 2 | 0 | |||||||||||||||||
CT | 0 | ||||||||||||||||||
VT | 2 | 0 | |||||||||||||||||
ME | 1 | 0 | |||||||||||||||||
MN | 0 |
None | None | ||
Traditional | Traditional | ||
Commodity | Commodity | ||
Econometric | Econometric | ||
Unknown | Unknown |
Statewide vs Urban Model Zone Systems
bull Usually larger zones in statewide models than MPOs
bull Block groups are good starting point for statewide model TAZs
bull Block groups might need splitting along major transportation corridors
bull In small states MPO TAZs could equal SWM zones
bull In most states MPO TAZ equivalency tables will likely be neededdesirable
bull FAF disaggregation approach used for statewide model TAZs
Source FHWA Office of Freight Management and Operations
Source Florida Dept of TransportationCambridge Systematics Inc
FAF zones
Block groups = TAZs
Too small for statewide zones
Too big for statewide zones
Statewide vs Urban Model Networksbull Statewide model networks are sparse in
comparison with MPO models
bull At a minimum statewide networks should include Limited access highways Arterials Major collectors
bull For proper circulation patterns lower class facilities might be added
bull Outside state a very sparse network is needed for major truck corridors
bull Consistent level of detail for networks and zones
Source Tennessee Dept of TransportationCambridge Systematics Inc
MPO network
statewide network
Statewide vs Urban Model Travel Behaviorbull Understanding urban travel behavior
MPO models are usually developed with local household travel survey data
MPOs without recent HH survey data can borrow parameters from other MPO models or guidance documents such as NCHRP 716 Transferable Urban Travel Parameters
bull Rural and long-distance travel behavior A limited number of states have conducted
statewide household travel surveys American Travel Survey was designed to target
long-distance trips but was discontinued long ago
National Household Travel Survey Add-ons can be structured for a sufficient sample of rural trip-makers
NCHRP 735 Rural and Long-Distance Travel Parameters recently published for use in statewide models
Map of 2009 Add-on Surveys Source FHWA
Statewide vs Urban Model Trip Purposes
bull Typical MPO trip purposes are usually collapsed into HBW HBOHBNW and NHB for statewide models
bull Long-distance purposes Business Tourist Personal business
bull Trucks Commodity group is often
surrogate for trip purpose
bull External trips External generally means trip from
outside state Might exclude ldquohalordquo area
surrounding state boundary
Trip PurposeFL
2000 IN LA MA MS RI
Home-based Work 1914 1995 1900 1841 1606 2000
Home-based Nonwork 5547 5208 4770 5206 4917 5700
Home-based Other 2792 5208 4770 1938 4917 5700
Home-based Shop 1576 ndash ndash 1173 ndash ndash
Home-based SocialRecreation
1179 ndash ndash 1395 ndash ndash
Home-base RecreationShop ndash ndash ndash ndash ndash ndash
Home-based School ndash ndash ndash 700 ndash ndash
Nonhome-based 2539 2686 3100 2468 3289 2300
Nonhome-based Work ndash ndash ndash 1027 ndash ndash
Nonhome-based Nonwork ndash ndash ndash 1441 ndash ndash
Other ndash ndash ndash ndash ndash
Truck ndash ndash 170 ndash ndash ndash
Long-Distance 000 111 060 417 022 000
Long-Distance Business ndash ndash 020 284 011 ndash
Long-Distance Personal Business
ndash ndash ndash ndash ndash ndash
Long-Distance Tourist ndash ndash 010 027 000 ndash
Long-Distance Work ndash ndash ndash 105 ndash ndash
Long-Distance NonworkOther
ndash ndash 030 ndash 011 ndash
Statewide vs Urban Model Trip Purposes (Contrsquod)
Number of person trips
Mode Pleasure Business Personal business Other Total Percent
Personal vehicle 12185 5708 3120 789 21802 908Commercial air 1012 746 151 24 1933 8Train 74 150 14 0 238 1Other 4 2 5 12 23 01Ship 19 0 0 0 19 01
Total 13294 6606 3290 825 24015 100Percent 554 275 137 34 100
Percent by trip purpose
Mode Pleasure Business Personal business Other
Personal vehicle 917 864 948 956Commercial air 76 113 46 29Train 06 23 04 0Other 0 0 02 15Ship 01 0 0 0
Trip purpose
Trip purpose
Statewide vs Urban Model Modal Issuesbull Most statewide models are limited to highway trips
bull Modeling of local bus routes is likely overkill
bull Mode choice is usually just a conversion from person trips to vehicle trips via auto occupancy rates
bull Transit components of statewide models are more common in regions with history of fixed guideway transit
bull Transit model is needed to test options such as high speedintercity rail
ATS 95 NHTS 01 MDOT 03-04 ODOT 02-03Personal vehicle 81 89 89 88Commercial air 16 8 10 8Bus 2 2 04 19Train 05 01 0 005Other 05 09 06 205
Survey and yearPrimary mode of transport
Statewide vs Urban Model Modal Issues (Contrsquod)
3727 24 22
32
29 36
10
21
28 22
10
22
12
1529
11
44 43
0
10
20
30
40
50
60
70
80
90
100
Personalvehicle
Bus Train Marine Commercialair
Perc
ent o
f per
son
trips
lt300 300-499 500-599 1000-1999 gt2000
68
85 85 82 85
30
12 11 15 13
0
10
20
30
40
50
60
70
80
90
100
Busi-ness
Plea-sure
pBusi-ness
Vaca-tion
Week-end
Perc
ent o
f per
son
trips
Personal vehicle Commercial air All other
Mode by Trip Purpose Mode by Distance (miles)
Statewide vs Urban Model Modal Issues (Contrsquod)
bull Most statewide assignment models separate trucks from passenger cars Preloading process commonly used for trucks Takes into account more limited route options for truck travel
Source Ohio Dept of Transportation
Statewide vs Urban Model Truck Forecasting
bull Urban modelsndash Freight movements are partially
ldquoexternalrdquo to model (truncated at external zone)
ndash More general categories of truckcommercial trips used
ndash Strong reliance on Quick Response Freight Manual
bull Statewide modelsndash Rooted in commodity flows ndash Often use econometric
approaches to modelingndash Generally model entire length of
freight travel
Freight Model Type by Peer Exchange Participants and
NonParticipants
0123456789
None
Traditio
nal
Commod
ity
Econo
metric
Unkno
wn
NotPeer Exchange
Source TRB Electronic Circular E-C075 Statewide Travel Demand Modeling ndash A Peer Exchange
Topic ndash Data Needs and Considerations
bull Impact of model complexity on data needs
bull Overview of data sources
bull Long-Distance travel gt different focus than MPOregional models
Sources NHTS Weight calculation Pitfalls Key statistics European surveys Lessons learned
Impact of Model Complexity on Data Needsbull Trip table estimation (2 step) ndash highway network
zone system trip table
bull Highway model structure (3 step) ndash 2 step plusbehavioral data for trip rates friction factors auto occupancy rates
bull Freight considerations (3 step) ndash 3 step pluscommodity flows economic data tonnages by mode truck counts
bull Full multi-modal structure (4 step) ndash 3 step plustransit networks and paths logit mode choice model
bull Beyond 4 Step ndash synthetic household generation operational characteristics etc
Source Georgia DOTCambridge Systematics Inc
Statewide Model Data Sources A-Zbull American Travel Survey (ATS)
bull Bureau of Economic Analysis (BEA)
bull Census Journey-to-Work (JTW) data
bull Census Public Use Microsample (PUMS)
bull Census Transportation Planning Package (CTPP)
bull Claritas
bull Commodity Flow Survey (CFS)
Source US Census Bureau
Statewide Model Data Sources (Contrsquod)bull Dun amp Bradstreet (DampB) employment data
bull Employment Security data (ES-202)
bull Freight Analysis Framework (FAF)
bull Highway Performance Monitoring System (HPMS)
bull InfoUSA employment data
bull Metropolitan Planning Organizations (MPO)
bull NCHRP 365 Manual
bull National Highway Planning Network (NHPN)
Source FAF2 Provisional Commodity Origin-Destination Estimates Revised Methodology Report BattelleFHWA
Statewide Model Data Sources (Contrsquod)
bull National Household Travel Survey (NHTS)
bull Origin-destination surveys
bull Regional Economic Models Inc (REMI)
bull State Departments of Transportation
bull TRANSEARCH data
bull Vehicle Use and Identification Survey (VIUS)
bull Woods amp Poole employment forecasts
Source Bureau of Transportation Statistics NHTS
Data Sources on Long-Distance Travel
bull Census Journey-to-Work databull Long distance component
American Travel Survey (77 95) National Household Travel Survey (01) Ohio statewide surveys (02-04) Michigan statewide survey (04-05) California (10-12)
bull Surveys focused on long distance Oregon (96-98 07-08)
bull Others BTS Omnibus surveys Tourism surveys European practice NCHRP Report 735
NHTS and Long-Distance Travelbull Evolution of US national surveys
NPTS (69 77 83 90 95) ATS (77 95)
bull 2001 NHTS ndash Total of 60282 long distance person trips Data collected March 01 to May 02 22204 (37 percent) before 911 38078 (63 percent) afterwards
bull Long distance trip is longer than 50 miles one-way (vs 100 miles in 1995 ATS)
bull 28-day retrospective survey (including travel day)
bull Not your typical travelers Low frequency Distinct segment of the population
Weight Calculation for Long-Distance Travelers
bull Use to ensure valid population-level estimates Accounts for non-response under-coverage multiple phones etc
bull Multi-stage process Match national margins Adjust for data splitting (force sumexpfac = 277208169) Raking (match independent controls) race ethnicity race by
month ethnicity by month sex by age census region MSA status month by day of week
Trim outlier expansion factors max expfac = 4 times mean Replication (correct for sampling error)
bull At least two passes through the data
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Freight Model Types | |||||||||||||||||||
State | Type | Peer Exc | |||||||||||||||||
OR | 4 | 1 | Count of Type | Count of Type | Peer Exc | ||||||||||||||
WI | 3 | 1 | Type | Total | Type | 0 | 1 | Grand Total | |||||||||||
MO | 1 | 1 | 1 | 2 | 1 | 2 | 2 | 4 | |||||||||||
LA | 3 | 1 | 2 | 2 | 2 | 2 | 2 | 4 | |||||||||||
IN | 3 | 1 | 3 | 6 | 3 | 2 | 6 | 8 | |||||||||||
OH | 4 | 1 | 4 | 2 | 4 | 2 | 2 | ||||||||||||
KY | 3 | 1 | Grand Total | 12 | 3 | 3 | |||||||||||||
FL | 3 | 1 | Grand Total | 9 | 12 | 21 | |||||||||||||
VA | 3 | 1 | None | 2 | Not | Peer Exchange | |||||||||||||
NJ | 2 | 1 | Traditional | 2 | None | 2 | 2 | ||||||||||||
MA | 2 | 1 | Commodity | 6 | Traditional | 2 | 2 | ||||||||||||
NH | 1 | 1 | Econometric | 2 | Commodity | 2 | 6 | ||||||||||||
CA | 0 | Econometric | 2 | ||||||||||||||||
TX | 3 | 0 | Unknown | 3 | |||||||||||||||
MI | 3 | 0 | |||||||||||||||||
GA | 0 | ||||||||||||||||||
SC | 1 | 0 | |||||||||||||||||
DE | 2 | 0 | |||||||||||||||||
CT | 0 | ||||||||||||||||||
VT | 2 | 0 | |||||||||||||||||
ME | 1 | 0 | |||||||||||||||||
MN | 0 |
None | None | ||
Traditional | Traditional | ||
Commodity | Commodity | ||
Econometric | Econometric | ||
Unknown | Unknown |
Statewide vs Urban Model Networksbull Statewide model networks are sparse in
comparison with MPO models
bull At a minimum statewide networks should include Limited access highways Arterials Major collectors
bull For proper circulation patterns lower class facilities might be added
bull Outside state a very sparse network is needed for major truck corridors
bull Consistent level of detail for networks and zones
Source Tennessee Dept of TransportationCambridge Systematics Inc
MPO network
statewide network
Statewide vs Urban Model Travel Behaviorbull Understanding urban travel behavior
MPO models are usually developed with local household travel survey data
MPOs without recent HH survey data can borrow parameters from other MPO models or guidance documents such as NCHRP 716 Transferable Urban Travel Parameters
bull Rural and long-distance travel behavior A limited number of states have conducted
statewide household travel surveys American Travel Survey was designed to target
long-distance trips but was discontinued long ago
National Household Travel Survey Add-ons can be structured for a sufficient sample of rural trip-makers
NCHRP 735 Rural and Long-Distance Travel Parameters recently published for use in statewide models
Map of 2009 Add-on Surveys Source FHWA
Statewide vs Urban Model Trip Purposes
bull Typical MPO trip purposes are usually collapsed into HBW HBOHBNW and NHB for statewide models
bull Long-distance purposes Business Tourist Personal business
bull Trucks Commodity group is often
surrogate for trip purpose
bull External trips External generally means trip from
outside state Might exclude ldquohalordquo area
surrounding state boundary
Trip PurposeFL
2000 IN LA MA MS RI
Home-based Work 1914 1995 1900 1841 1606 2000
Home-based Nonwork 5547 5208 4770 5206 4917 5700
Home-based Other 2792 5208 4770 1938 4917 5700
Home-based Shop 1576 ndash ndash 1173 ndash ndash
Home-based SocialRecreation
1179 ndash ndash 1395 ndash ndash
Home-base RecreationShop ndash ndash ndash ndash ndash ndash
Home-based School ndash ndash ndash 700 ndash ndash
Nonhome-based 2539 2686 3100 2468 3289 2300
Nonhome-based Work ndash ndash ndash 1027 ndash ndash
Nonhome-based Nonwork ndash ndash ndash 1441 ndash ndash
Other ndash ndash ndash ndash ndash
Truck ndash ndash 170 ndash ndash ndash
Long-Distance 000 111 060 417 022 000
Long-Distance Business ndash ndash 020 284 011 ndash
Long-Distance Personal Business
ndash ndash ndash ndash ndash ndash
Long-Distance Tourist ndash ndash 010 027 000 ndash
Long-Distance Work ndash ndash ndash 105 ndash ndash
Long-Distance NonworkOther
ndash ndash 030 ndash 011 ndash
Statewide vs Urban Model Trip Purposes (Contrsquod)
Number of person trips
Mode Pleasure Business Personal business Other Total Percent
Personal vehicle 12185 5708 3120 789 21802 908Commercial air 1012 746 151 24 1933 8Train 74 150 14 0 238 1Other 4 2 5 12 23 01Ship 19 0 0 0 19 01
Total 13294 6606 3290 825 24015 100Percent 554 275 137 34 100
Percent by trip purpose
Mode Pleasure Business Personal business Other
Personal vehicle 917 864 948 956Commercial air 76 113 46 29Train 06 23 04 0Other 0 0 02 15Ship 01 0 0 0
Trip purpose
Trip purpose
Statewide vs Urban Model Modal Issuesbull Most statewide models are limited to highway trips
bull Modeling of local bus routes is likely overkill
bull Mode choice is usually just a conversion from person trips to vehicle trips via auto occupancy rates
bull Transit components of statewide models are more common in regions with history of fixed guideway transit
bull Transit model is needed to test options such as high speedintercity rail
ATS 95 NHTS 01 MDOT 03-04 ODOT 02-03Personal vehicle 81 89 89 88Commercial air 16 8 10 8Bus 2 2 04 19Train 05 01 0 005Other 05 09 06 205
Survey and yearPrimary mode of transport
Statewide vs Urban Model Modal Issues (Contrsquod)
3727 24 22
32
29 36
10
21
28 22
10
22
12
1529
11
44 43
0
10
20
30
40
50
60
70
80
90
100
Personalvehicle
Bus Train Marine Commercialair
Perc
ent o
f per
son
trips
lt300 300-499 500-599 1000-1999 gt2000
68
85 85 82 85
30
12 11 15 13
0
10
20
30
40
50
60
70
80
90
100
Busi-ness
Plea-sure
pBusi-ness
Vaca-tion
Week-end
Perc
ent o
f per
son
trips
Personal vehicle Commercial air All other
Mode by Trip Purpose Mode by Distance (miles)
Statewide vs Urban Model Modal Issues (Contrsquod)
bull Most statewide assignment models separate trucks from passenger cars Preloading process commonly used for trucks Takes into account more limited route options for truck travel
Source Ohio Dept of Transportation
Statewide vs Urban Model Truck Forecasting
bull Urban modelsndash Freight movements are partially
ldquoexternalrdquo to model (truncated at external zone)
ndash More general categories of truckcommercial trips used
ndash Strong reliance on Quick Response Freight Manual
bull Statewide modelsndash Rooted in commodity flows ndash Often use econometric
approaches to modelingndash Generally model entire length of
freight travel
Freight Model Type by Peer Exchange Participants and
NonParticipants
0123456789
None
Traditio
nal
Commod
ity
Econo
metric
Unkno
wn
NotPeer Exchange
Source TRB Electronic Circular E-C075 Statewide Travel Demand Modeling ndash A Peer Exchange
Topic ndash Data Needs and Considerations
bull Impact of model complexity on data needs
bull Overview of data sources
bull Long-Distance travel gt different focus than MPOregional models
Sources NHTS Weight calculation Pitfalls Key statistics European surveys Lessons learned
Impact of Model Complexity on Data Needsbull Trip table estimation (2 step) ndash highway network
zone system trip table
bull Highway model structure (3 step) ndash 2 step plusbehavioral data for trip rates friction factors auto occupancy rates
bull Freight considerations (3 step) ndash 3 step pluscommodity flows economic data tonnages by mode truck counts
bull Full multi-modal structure (4 step) ndash 3 step plustransit networks and paths logit mode choice model
bull Beyond 4 Step ndash synthetic household generation operational characteristics etc
Source Georgia DOTCambridge Systematics Inc
Statewide Model Data Sources A-Zbull American Travel Survey (ATS)
bull Bureau of Economic Analysis (BEA)
bull Census Journey-to-Work (JTW) data
bull Census Public Use Microsample (PUMS)
bull Census Transportation Planning Package (CTPP)
bull Claritas
bull Commodity Flow Survey (CFS)
Source US Census Bureau
Statewide Model Data Sources (Contrsquod)bull Dun amp Bradstreet (DampB) employment data
bull Employment Security data (ES-202)
bull Freight Analysis Framework (FAF)
bull Highway Performance Monitoring System (HPMS)
bull InfoUSA employment data
bull Metropolitan Planning Organizations (MPO)
bull NCHRP 365 Manual
bull National Highway Planning Network (NHPN)
Source FAF2 Provisional Commodity Origin-Destination Estimates Revised Methodology Report BattelleFHWA
Statewide Model Data Sources (Contrsquod)
bull National Household Travel Survey (NHTS)
bull Origin-destination surveys
bull Regional Economic Models Inc (REMI)
bull State Departments of Transportation
bull TRANSEARCH data
bull Vehicle Use and Identification Survey (VIUS)
bull Woods amp Poole employment forecasts
Source Bureau of Transportation Statistics NHTS
Data Sources on Long-Distance Travel
bull Census Journey-to-Work databull Long distance component
American Travel Survey (77 95) National Household Travel Survey (01) Ohio statewide surveys (02-04) Michigan statewide survey (04-05) California (10-12)
bull Surveys focused on long distance Oregon (96-98 07-08)
bull Others BTS Omnibus surveys Tourism surveys European practice NCHRP Report 735
NHTS and Long-Distance Travelbull Evolution of US national surveys
NPTS (69 77 83 90 95) ATS (77 95)
bull 2001 NHTS ndash Total of 60282 long distance person trips Data collected March 01 to May 02 22204 (37 percent) before 911 38078 (63 percent) afterwards
bull Long distance trip is longer than 50 miles one-way (vs 100 miles in 1995 ATS)
bull 28-day retrospective survey (including travel day)
bull Not your typical travelers Low frequency Distinct segment of the population
Weight Calculation for Long-Distance Travelers
bull Use to ensure valid population-level estimates Accounts for non-response under-coverage multiple phones etc
bull Multi-stage process Match national margins Adjust for data splitting (force sumexpfac = 277208169) Raking (match independent controls) race ethnicity race by
month ethnicity by month sex by age census region MSA status month by day of week
Trim outlier expansion factors max expfac = 4 times mean Replication (correct for sampling error)
bull At least two passes through the data
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Freight Model Types | |||||||||||||||||||
State | Type | Peer Exc | |||||||||||||||||
OR | 4 | 1 | Count of Type | Count of Type | Peer Exc | ||||||||||||||
WI | 3 | 1 | Type | Total | Type | 0 | 1 | Grand Total | |||||||||||
MO | 1 | 1 | 1 | 2 | 1 | 2 | 2 | 4 | |||||||||||
LA | 3 | 1 | 2 | 2 | 2 | 2 | 2 | 4 | |||||||||||
IN | 3 | 1 | 3 | 6 | 3 | 2 | 6 | 8 | |||||||||||
OH | 4 | 1 | 4 | 2 | 4 | 2 | 2 | ||||||||||||
KY | 3 | 1 | Grand Total | 12 | 3 | 3 | |||||||||||||
FL | 3 | 1 | Grand Total | 9 | 12 | 21 | |||||||||||||
VA | 3 | 1 | None | 2 | Not | Peer Exchange | |||||||||||||
NJ | 2 | 1 | Traditional | 2 | None | 2 | 2 | ||||||||||||
MA | 2 | 1 | Commodity | 6 | Traditional | 2 | 2 | ||||||||||||
NH | 1 | 1 | Econometric | 2 | Commodity | 2 | 6 | ||||||||||||
CA | 0 | Econometric | 2 | ||||||||||||||||
TX | 3 | 0 | Unknown | 3 | |||||||||||||||
MI | 3 | 0 | |||||||||||||||||
GA | 0 | ||||||||||||||||||
SC | 1 | 0 | |||||||||||||||||
DE | 2 | 0 | |||||||||||||||||
CT | 0 | ||||||||||||||||||
VT | 2 | 0 | |||||||||||||||||
ME | 1 | 0 | |||||||||||||||||
MN | 0 |
None | None | ||
Traditional | Traditional | ||
Commodity | Commodity | ||
Econometric | Econometric | ||
Unknown | Unknown |
Statewide vs Urban Model Travel Behaviorbull Understanding urban travel behavior
MPO models are usually developed with local household travel survey data
MPOs without recent HH survey data can borrow parameters from other MPO models or guidance documents such as NCHRP 716 Transferable Urban Travel Parameters
bull Rural and long-distance travel behavior A limited number of states have conducted
statewide household travel surveys American Travel Survey was designed to target
long-distance trips but was discontinued long ago
National Household Travel Survey Add-ons can be structured for a sufficient sample of rural trip-makers
NCHRP 735 Rural and Long-Distance Travel Parameters recently published for use in statewide models
Map of 2009 Add-on Surveys Source FHWA
Statewide vs Urban Model Trip Purposes
bull Typical MPO trip purposes are usually collapsed into HBW HBOHBNW and NHB for statewide models
bull Long-distance purposes Business Tourist Personal business
bull Trucks Commodity group is often
surrogate for trip purpose
bull External trips External generally means trip from
outside state Might exclude ldquohalordquo area
surrounding state boundary
Trip PurposeFL
2000 IN LA MA MS RI
Home-based Work 1914 1995 1900 1841 1606 2000
Home-based Nonwork 5547 5208 4770 5206 4917 5700
Home-based Other 2792 5208 4770 1938 4917 5700
Home-based Shop 1576 ndash ndash 1173 ndash ndash
Home-based SocialRecreation
1179 ndash ndash 1395 ndash ndash
Home-base RecreationShop ndash ndash ndash ndash ndash ndash
Home-based School ndash ndash ndash 700 ndash ndash
Nonhome-based 2539 2686 3100 2468 3289 2300
Nonhome-based Work ndash ndash ndash 1027 ndash ndash
Nonhome-based Nonwork ndash ndash ndash 1441 ndash ndash
Other ndash ndash ndash ndash ndash
Truck ndash ndash 170 ndash ndash ndash
Long-Distance 000 111 060 417 022 000
Long-Distance Business ndash ndash 020 284 011 ndash
Long-Distance Personal Business
ndash ndash ndash ndash ndash ndash
Long-Distance Tourist ndash ndash 010 027 000 ndash
Long-Distance Work ndash ndash ndash 105 ndash ndash
Long-Distance NonworkOther
ndash ndash 030 ndash 011 ndash
Statewide vs Urban Model Trip Purposes (Contrsquod)
Number of person trips
Mode Pleasure Business Personal business Other Total Percent
Personal vehicle 12185 5708 3120 789 21802 908Commercial air 1012 746 151 24 1933 8Train 74 150 14 0 238 1Other 4 2 5 12 23 01Ship 19 0 0 0 19 01
Total 13294 6606 3290 825 24015 100Percent 554 275 137 34 100
Percent by trip purpose
Mode Pleasure Business Personal business Other
Personal vehicle 917 864 948 956Commercial air 76 113 46 29Train 06 23 04 0Other 0 0 02 15Ship 01 0 0 0
Trip purpose
Trip purpose
Statewide vs Urban Model Modal Issuesbull Most statewide models are limited to highway trips
bull Modeling of local bus routes is likely overkill
bull Mode choice is usually just a conversion from person trips to vehicle trips via auto occupancy rates
bull Transit components of statewide models are more common in regions with history of fixed guideway transit
bull Transit model is needed to test options such as high speedintercity rail
ATS 95 NHTS 01 MDOT 03-04 ODOT 02-03Personal vehicle 81 89 89 88Commercial air 16 8 10 8Bus 2 2 04 19Train 05 01 0 005Other 05 09 06 205
Survey and yearPrimary mode of transport
Statewide vs Urban Model Modal Issues (Contrsquod)
3727 24 22
32
29 36
10
21
28 22
10
22
12
1529
11
44 43
0
10
20
30
40
50
60
70
80
90
100
Personalvehicle
Bus Train Marine Commercialair
Perc
ent o
f per
son
trips
lt300 300-499 500-599 1000-1999 gt2000
68
85 85 82 85
30
12 11 15 13
0
10
20
30
40
50
60
70
80
90
100
Busi-ness
Plea-sure
pBusi-ness
Vaca-tion
Week-end
Perc
ent o
f per
son
trips
Personal vehicle Commercial air All other
Mode by Trip Purpose Mode by Distance (miles)
Statewide vs Urban Model Modal Issues (Contrsquod)
bull Most statewide assignment models separate trucks from passenger cars Preloading process commonly used for trucks Takes into account more limited route options for truck travel
Source Ohio Dept of Transportation
Statewide vs Urban Model Truck Forecasting
bull Urban modelsndash Freight movements are partially
ldquoexternalrdquo to model (truncated at external zone)
ndash More general categories of truckcommercial trips used
ndash Strong reliance on Quick Response Freight Manual
bull Statewide modelsndash Rooted in commodity flows ndash Often use econometric
approaches to modelingndash Generally model entire length of
freight travel
Freight Model Type by Peer Exchange Participants and
NonParticipants
0123456789
None
Traditio
nal
Commod
ity
Econo
metric
Unkno
wn
NotPeer Exchange
Source TRB Electronic Circular E-C075 Statewide Travel Demand Modeling ndash A Peer Exchange
Topic ndash Data Needs and Considerations
bull Impact of model complexity on data needs
bull Overview of data sources
bull Long-Distance travel gt different focus than MPOregional models
Sources NHTS Weight calculation Pitfalls Key statistics European surveys Lessons learned
Impact of Model Complexity on Data Needsbull Trip table estimation (2 step) ndash highway network
zone system trip table
bull Highway model structure (3 step) ndash 2 step plusbehavioral data for trip rates friction factors auto occupancy rates
bull Freight considerations (3 step) ndash 3 step pluscommodity flows economic data tonnages by mode truck counts
bull Full multi-modal structure (4 step) ndash 3 step plustransit networks and paths logit mode choice model
bull Beyond 4 Step ndash synthetic household generation operational characteristics etc
Source Georgia DOTCambridge Systematics Inc
Statewide Model Data Sources A-Zbull American Travel Survey (ATS)
bull Bureau of Economic Analysis (BEA)
bull Census Journey-to-Work (JTW) data
bull Census Public Use Microsample (PUMS)
bull Census Transportation Planning Package (CTPP)
bull Claritas
bull Commodity Flow Survey (CFS)
Source US Census Bureau
Statewide Model Data Sources (Contrsquod)bull Dun amp Bradstreet (DampB) employment data
bull Employment Security data (ES-202)
bull Freight Analysis Framework (FAF)
bull Highway Performance Monitoring System (HPMS)
bull InfoUSA employment data
bull Metropolitan Planning Organizations (MPO)
bull NCHRP 365 Manual
bull National Highway Planning Network (NHPN)
Source FAF2 Provisional Commodity Origin-Destination Estimates Revised Methodology Report BattelleFHWA
Statewide Model Data Sources (Contrsquod)
bull National Household Travel Survey (NHTS)
bull Origin-destination surveys
bull Regional Economic Models Inc (REMI)
bull State Departments of Transportation
bull TRANSEARCH data
bull Vehicle Use and Identification Survey (VIUS)
bull Woods amp Poole employment forecasts
Source Bureau of Transportation Statistics NHTS
Data Sources on Long-Distance Travel
bull Census Journey-to-Work databull Long distance component
American Travel Survey (77 95) National Household Travel Survey (01) Ohio statewide surveys (02-04) Michigan statewide survey (04-05) California (10-12)
bull Surveys focused on long distance Oregon (96-98 07-08)
bull Others BTS Omnibus surveys Tourism surveys European practice NCHRP Report 735
NHTS and Long-Distance Travelbull Evolution of US national surveys
NPTS (69 77 83 90 95) ATS (77 95)
bull 2001 NHTS ndash Total of 60282 long distance person trips Data collected March 01 to May 02 22204 (37 percent) before 911 38078 (63 percent) afterwards
bull Long distance trip is longer than 50 miles one-way (vs 100 miles in 1995 ATS)
bull 28-day retrospective survey (including travel day)
bull Not your typical travelers Low frequency Distinct segment of the population
Weight Calculation for Long-Distance Travelers
bull Use to ensure valid population-level estimates Accounts for non-response under-coverage multiple phones etc
bull Multi-stage process Match national margins Adjust for data splitting (force sumexpfac = 277208169) Raking (match independent controls) race ethnicity race by
month ethnicity by month sex by age census region MSA status month by day of week
Trim outlier expansion factors max expfac = 4 times mean Replication (correct for sampling error)
bull At least two passes through the data
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Freight Model Types | |||||||||||||||||||
State | Type | Peer Exc | |||||||||||||||||
OR | 4 | 1 | Count of Type | Count of Type | Peer Exc | ||||||||||||||
WI | 3 | 1 | Type | Total | Type | 0 | 1 | Grand Total | |||||||||||
MO | 1 | 1 | 1 | 2 | 1 | 2 | 2 | 4 | |||||||||||
LA | 3 | 1 | 2 | 2 | 2 | 2 | 2 | 4 | |||||||||||
IN | 3 | 1 | 3 | 6 | 3 | 2 | 6 | 8 | |||||||||||
OH | 4 | 1 | 4 | 2 | 4 | 2 | 2 | ||||||||||||
KY | 3 | 1 | Grand Total | 12 | 3 | 3 | |||||||||||||
FL | 3 | 1 | Grand Total | 9 | 12 | 21 | |||||||||||||
VA | 3 | 1 | None | 2 | Not | Peer Exchange | |||||||||||||
NJ | 2 | 1 | Traditional | 2 | None | 2 | 2 | ||||||||||||
MA | 2 | 1 | Commodity | 6 | Traditional | 2 | 2 | ||||||||||||
NH | 1 | 1 | Econometric | 2 | Commodity | 2 | 6 | ||||||||||||
CA | 0 | Econometric | 2 | ||||||||||||||||
TX | 3 | 0 | Unknown | 3 | |||||||||||||||
MI | 3 | 0 | |||||||||||||||||
GA | 0 | ||||||||||||||||||
SC | 1 | 0 | |||||||||||||||||
DE | 2 | 0 | |||||||||||||||||
CT | 0 | ||||||||||||||||||
VT | 2 | 0 | |||||||||||||||||
ME | 1 | 0 | |||||||||||||||||
MN | 0 |
None | None | ||
Traditional | Traditional | ||
Commodity | Commodity | ||
Econometric | Econometric | ||
Unknown | Unknown |
Statewide vs Urban Model Trip Purposes
bull Typical MPO trip purposes are usually collapsed into HBW HBOHBNW and NHB for statewide models
bull Long-distance purposes Business Tourist Personal business
bull Trucks Commodity group is often
surrogate for trip purpose
bull External trips External generally means trip from
outside state Might exclude ldquohalordquo area
surrounding state boundary
Trip PurposeFL
2000 IN LA MA MS RI
Home-based Work 1914 1995 1900 1841 1606 2000
Home-based Nonwork 5547 5208 4770 5206 4917 5700
Home-based Other 2792 5208 4770 1938 4917 5700
Home-based Shop 1576 ndash ndash 1173 ndash ndash
Home-based SocialRecreation
1179 ndash ndash 1395 ndash ndash
Home-base RecreationShop ndash ndash ndash ndash ndash ndash
Home-based School ndash ndash ndash 700 ndash ndash
Nonhome-based 2539 2686 3100 2468 3289 2300
Nonhome-based Work ndash ndash ndash 1027 ndash ndash
Nonhome-based Nonwork ndash ndash ndash 1441 ndash ndash
Other ndash ndash ndash ndash ndash
Truck ndash ndash 170 ndash ndash ndash
Long-Distance 000 111 060 417 022 000
Long-Distance Business ndash ndash 020 284 011 ndash
Long-Distance Personal Business
ndash ndash ndash ndash ndash ndash
Long-Distance Tourist ndash ndash 010 027 000 ndash
Long-Distance Work ndash ndash ndash 105 ndash ndash
Long-Distance NonworkOther
ndash ndash 030 ndash 011 ndash
Statewide vs Urban Model Trip Purposes (Contrsquod)
Number of person trips
Mode Pleasure Business Personal business Other Total Percent
Personal vehicle 12185 5708 3120 789 21802 908Commercial air 1012 746 151 24 1933 8Train 74 150 14 0 238 1Other 4 2 5 12 23 01Ship 19 0 0 0 19 01
Total 13294 6606 3290 825 24015 100Percent 554 275 137 34 100
Percent by trip purpose
Mode Pleasure Business Personal business Other
Personal vehicle 917 864 948 956Commercial air 76 113 46 29Train 06 23 04 0Other 0 0 02 15Ship 01 0 0 0
Trip purpose
Trip purpose
Statewide vs Urban Model Modal Issuesbull Most statewide models are limited to highway trips
bull Modeling of local bus routes is likely overkill
bull Mode choice is usually just a conversion from person trips to vehicle trips via auto occupancy rates
bull Transit components of statewide models are more common in regions with history of fixed guideway transit
bull Transit model is needed to test options such as high speedintercity rail
ATS 95 NHTS 01 MDOT 03-04 ODOT 02-03Personal vehicle 81 89 89 88Commercial air 16 8 10 8Bus 2 2 04 19Train 05 01 0 005Other 05 09 06 205
Survey and yearPrimary mode of transport
Statewide vs Urban Model Modal Issues (Contrsquod)
3727 24 22
32
29 36
10
21
28 22
10
22
12
1529
11
44 43
0
10
20
30
40
50
60
70
80
90
100
Personalvehicle
Bus Train Marine Commercialair
Perc
ent o
f per
son
trips
lt300 300-499 500-599 1000-1999 gt2000
68
85 85 82 85
30
12 11 15 13
0
10
20
30
40
50
60
70
80
90
100
Busi-ness
Plea-sure
pBusi-ness
Vaca-tion
Week-end
Perc
ent o
f per
son
trips
Personal vehicle Commercial air All other
Mode by Trip Purpose Mode by Distance (miles)
Statewide vs Urban Model Modal Issues (Contrsquod)
bull Most statewide assignment models separate trucks from passenger cars Preloading process commonly used for trucks Takes into account more limited route options for truck travel
Source Ohio Dept of Transportation
Statewide vs Urban Model Truck Forecasting
bull Urban modelsndash Freight movements are partially
ldquoexternalrdquo to model (truncated at external zone)
ndash More general categories of truckcommercial trips used
ndash Strong reliance on Quick Response Freight Manual
bull Statewide modelsndash Rooted in commodity flows ndash Often use econometric
approaches to modelingndash Generally model entire length of
freight travel
Freight Model Type by Peer Exchange Participants and
NonParticipants
0123456789
None
Traditio
nal
Commod
ity
Econo
metric
Unkno
wn
NotPeer Exchange
Source TRB Electronic Circular E-C075 Statewide Travel Demand Modeling ndash A Peer Exchange
Topic ndash Data Needs and Considerations
bull Impact of model complexity on data needs
bull Overview of data sources
bull Long-Distance travel gt different focus than MPOregional models
Sources NHTS Weight calculation Pitfalls Key statistics European surveys Lessons learned
Impact of Model Complexity on Data Needsbull Trip table estimation (2 step) ndash highway network
zone system trip table
bull Highway model structure (3 step) ndash 2 step plusbehavioral data for trip rates friction factors auto occupancy rates
bull Freight considerations (3 step) ndash 3 step pluscommodity flows economic data tonnages by mode truck counts
bull Full multi-modal structure (4 step) ndash 3 step plustransit networks and paths logit mode choice model
bull Beyond 4 Step ndash synthetic household generation operational characteristics etc
Source Georgia DOTCambridge Systematics Inc
Statewide Model Data Sources A-Zbull American Travel Survey (ATS)
bull Bureau of Economic Analysis (BEA)
bull Census Journey-to-Work (JTW) data
bull Census Public Use Microsample (PUMS)
bull Census Transportation Planning Package (CTPP)
bull Claritas
bull Commodity Flow Survey (CFS)
Source US Census Bureau
Statewide Model Data Sources (Contrsquod)bull Dun amp Bradstreet (DampB) employment data
bull Employment Security data (ES-202)
bull Freight Analysis Framework (FAF)
bull Highway Performance Monitoring System (HPMS)
bull InfoUSA employment data
bull Metropolitan Planning Organizations (MPO)
bull NCHRP 365 Manual
bull National Highway Planning Network (NHPN)
Source FAF2 Provisional Commodity Origin-Destination Estimates Revised Methodology Report BattelleFHWA
Statewide Model Data Sources (Contrsquod)
bull National Household Travel Survey (NHTS)
bull Origin-destination surveys
bull Regional Economic Models Inc (REMI)
bull State Departments of Transportation
bull TRANSEARCH data
bull Vehicle Use and Identification Survey (VIUS)
bull Woods amp Poole employment forecasts
Source Bureau of Transportation Statistics NHTS
Data Sources on Long-Distance Travel
bull Census Journey-to-Work databull Long distance component
American Travel Survey (77 95) National Household Travel Survey (01) Ohio statewide surveys (02-04) Michigan statewide survey (04-05) California (10-12)
bull Surveys focused on long distance Oregon (96-98 07-08)
bull Others BTS Omnibus surveys Tourism surveys European practice NCHRP Report 735
NHTS and Long-Distance Travelbull Evolution of US national surveys
NPTS (69 77 83 90 95) ATS (77 95)
bull 2001 NHTS ndash Total of 60282 long distance person trips Data collected March 01 to May 02 22204 (37 percent) before 911 38078 (63 percent) afterwards
bull Long distance trip is longer than 50 miles one-way (vs 100 miles in 1995 ATS)
bull 28-day retrospective survey (including travel day)
bull Not your typical travelers Low frequency Distinct segment of the population
Weight Calculation for Long-Distance Travelers
bull Use to ensure valid population-level estimates Accounts for non-response under-coverage multiple phones etc
bull Multi-stage process Match national margins Adjust for data splitting (force sumexpfac = 277208169) Raking (match independent controls) race ethnicity race by
month ethnicity by month sex by age census region MSA status month by day of week
Trim outlier expansion factors max expfac = 4 times mean Replication (correct for sampling error)
bull At least two passes through the data
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Freight Model Types | |||||||||||||||||||
State | Type | Peer Exc | |||||||||||||||||
OR | 4 | 1 | Count of Type | Count of Type | Peer Exc | ||||||||||||||
WI | 3 | 1 | Type | Total | Type | 0 | 1 | Grand Total | |||||||||||
MO | 1 | 1 | 1 | 2 | 1 | 2 | 2 | 4 | |||||||||||
LA | 3 | 1 | 2 | 2 | 2 | 2 | 2 | 4 | |||||||||||
IN | 3 | 1 | 3 | 6 | 3 | 2 | 6 | 8 | |||||||||||
OH | 4 | 1 | 4 | 2 | 4 | 2 | 2 | ||||||||||||
KY | 3 | 1 | Grand Total | 12 | 3 | 3 | |||||||||||||
FL | 3 | 1 | Grand Total | 9 | 12 | 21 | |||||||||||||
VA | 3 | 1 | None | 2 | Not | Peer Exchange | |||||||||||||
NJ | 2 | 1 | Traditional | 2 | None | 2 | 2 | ||||||||||||
MA | 2 | 1 | Commodity | 6 | Traditional | 2 | 2 | ||||||||||||
NH | 1 | 1 | Econometric | 2 | Commodity | 2 | 6 | ||||||||||||
CA | 0 | Econometric | 2 | ||||||||||||||||
TX | 3 | 0 | Unknown | 3 | |||||||||||||||
MI | 3 | 0 | |||||||||||||||||
GA | 0 | ||||||||||||||||||
SC | 1 | 0 | |||||||||||||||||
DE | 2 | 0 | |||||||||||||||||
CT | 0 | ||||||||||||||||||
VT | 2 | 0 | |||||||||||||||||
ME | 1 | 0 | |||||||||||||||||
MN | 0 |
None | None | ||
Traditional | Traditional | ||
Commodity | Commodity | ||
Econometric | Econometric | ||
Unknown | Unknown |
Statewide vs Urban Model Trip Purposes (Contrsquod)
Number of person trips
Mode Pleasure Business Personal business Other Total Percent
Personal vehicle 12185 5708 3120 789 21802 908Commercial air 1012 746 151 24 1933 8Train 74 150 14 0 238 1Other 4 2 5 12 23 01Ship 19 0 0 0 19 01
Total 13294 6606 3290 825 24015 100Percent 554 275 137 34 100
Percent by trip purpose
Mode Pleasure Business Personal business Other
Personal vehicle 917 864 948 956Commercial air 76 113 46 29Train 06 23 04 0Other 0 0 02 15Ship 01 0 0 0
Trip purpose
Trip purpose
Statewide vs Urban Model Modal Issuesbull Most statewide models are limited to highway trips
bull Modeling of local bus routes is likely overkill
bull Mode choice is usually just a conversion from person trips to vehicle trips via auto occupancy rates
bull Transit components of statewide models are more common in regions with history of fixed guideway transit
bull Transit model is needed to test options such as high speedintercity rail
ATS 95 NHTS 01 MDOT 03-04 ODOT 02-03Personal vehicle 81 89 89 88Commercial air 16 8 10 8Bus 2 2 04 19Train 05 01 0 005Other 05 09 06 205
Survey and yearPrimary mode of transport
Statewide vs Urban Model Modal Issues (Contrsquod)
3727 24 22
32
29 36
10
21
28 22
10
22
12
1529
11
44 43
0
10
20
30
40
50
60
70
80
90
100
Personalvehicle
Bus Train Marine Commercialair
Perc
ent o
f per
son
trips
lt300 300-499 500-599 1000-1999 gt2000
68
85 85 82 85
30
12 11 15 13
0
10
20
30
40
50
60
70
80
90
100
Busi-ness
Plea-sure
pBusi-ness
Vaca-tion
Week-end
Perc
ent o
f per
son
trips
Personal vehicle Commercial air All other
Mode by Trip Purpose Mode by Distance (miles)
Statewide vs Urban Model Modal Issues (Contrsquod)
bull Most statewide assignment models separate trucks from passenger cars Preloading process commonly used for trucks Takes into account more limited route options for truck travel
Source Ohio Dept of Transportation
Statewide vs Urban Model Truck Forecasting
bull Urban modelsndash Freight movements are partially
ldquoexternalrdquo to model (truncated at external zone)
ndash More general categories of truckcommercial trips used
ndash Strong reliance on Quick Response Freight Manual
bull Statewide modelsndash Rooted in commodity flows ndash Often use econometric
approaches to modelingndash Generally model entire length of
freight travel
Freight Model Type by Peer Exchange Participants and
NonParticipants
0123456789
None
Traditio
nal
Commod
ity
Econo
metric
Unkno
wn
NotPeer Exchange
Source TRB Electronic Circular E-C075 Statewide Travel Demand Modeling ndash A Peer Exchange
Topic ndash Data Needs and Considerations
bull Impact of model complexity on data needs
bull Overview of data sources
bull Long-Distance travel gt different focus than MPOregional models
Sources NHTS Weight calculation Pitfalls Key statistics European surveys Lessons learned
Impact of Model Complexity on Data Needsbull Trip table estimation (2 step) ndash highway network
zone system trip table
bull Highway model structure (3 step) ndash 2 step plusbehavioral data for trip rates friction factors auto occupancy rates
bull Freight considerations (3 step) ndash 3 step pluscommodity flows economic data tonnages by mode truck counts
bull Full multi-modal structure (4 step) ndash 3 step plustransit networks and paths logit mode choice model
bull Beyond 4 Step ndash synthetic household generation operational characteristics etc
Source Georgia DOTCambridge Systematics Inc
Statewide Model Data Sources A-Zbull American Travel Survey (ATS)
bull Bureau of Economic Analysis (BEA)
bull Census Journey-to-Work (JTW) data
bull Census Public Use Microsample (PUMS)
bull Census Transportation Planning Package (CTPP)
bull Claritas
bull Commodity Flow Survey (CFS)
Source US Census Bureau
Statewide Model Data Sources (Contrsquod)bull Dun amp Bradstreet (DampB) employment data
bull Employment Security data (ES-202)
bull Freight Analysis Framework (FAF)
bull Highway Performance Monitoring System (HPMS)
bull InfoUSA employment data
bull Metropolitan Planning Organizations (MPO)
bull NCHRP 365 Manual
bull National Highway Planning Network (NHPN)
Source FAF2 Provisional Commodity Origin-Destination Estimates Revised Methodology Report BattelleFHWA
Statewide Model Data Sources (Contrsquod)
bull National Household Travel Survey (NHTS)
bull Origin-destination surveys
bull Regional Economic Models Inc (REMI)
bull State Departments of Transportation
bull TRANSEARCH data
bull Vehicle Use and Identification Survey (VIUS)
bull Woods amp Poole employment forecasts
Source Bureau of Transportation Statistics NHTS
Data Sources on Long-Distance Travel
bull Census Journey-to-Work databull Long distance component
American Travel Survey (77 95) National Household Travel Survey (01) Ohio statewide surveys (02-04) Michigan statewide survey (04-05) California (10-12)
bull Surveys focused on long distance Oregon (96-98 07-08)
bull Others BTS Omnibus surveys Tourism surveys European practice NCHRP Report 735
NHTS and Long-Distance Travelbull Evolution of US national surveys
NPTS (69 77 83 90 95) ATS (77 95)
bull 2001 NHTS ndash Total of 60282 long distance person trips Data collected March 01 to May 02 22204 (37 percent) before 911 38078 (63 percent) afterwards
bull Long distance trip is longer than 50 miles one-way (vs 100 miles in 1995 ATS)
bull 28-day retrospective survey (including travel day)
bull Not your typical travelers Low frequency Distinct segment of the population
Weight Calculation for Long-Distance Travelers
bull Use to ensure valid population-level estimates Accounts for non-response under-coverage multiple phones etc
bull Multi-stage process Match national margins Adjust for data splitting (force sumexpfac = 277208169) Raking (match independent controls) race ethnicity race by
month ethnicity by month sex by age census region MSA status month by day of week
Trim outlier expansion factors max expfac = 4 times mean Replication (correct for sampling error)
bull At least two passes through the data
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Freight Model Types | |||||||||||||||||||
State | Type | Peer Exc | |||||||||||||||||
OR | 4 | 1 | Count of Type | Count of Type | Peer Exc | ||||||||||||||
WI | 3 | 1 | Type | Total | Type | 0 | 1 | Grand Total | |||||||||||
MO | 1 | 1 | 1 | 2 | 1 | 2 | 2 | 4 | |||||||||||
LA | 3 | 1 | 2 | 2 | 2 | 2 | 2 | 4 | |||||||||||
IN | 3 | 1 | 3 | 6 | 3 | 2 | 6 | 8 | |||||||||||
OH | 4 | 1 | 4 | 2 | 4 | 2 | 2 | ||||||||||||
KY | 3 | 1 | Grand Total | 12 | 3 | 3 | |||||||||||||
FL | 3 | 1 | Grand Total | 9 | 12 | 21 | |||||||||||||
VA | 3 | 1 | None | 2 | Not | Peer Exchange | |||||||||||||
NJ | 2 | 1 | Traditional | 2 | None | 2 | 2 | ||||||||||||
MA | 2 | 1 | Commodity | 6 | Traditional | 2 | 2 | ||||||||||||
NH | 1 | 1 | Econometric | 2 | Commodity | 2 | 6 | ||||||||||||
CA | 0 | Econometric | 2 | ||||||||||||||||
TX | 3 | 0 | Unknown | 3 | |||||||||||||||
MI | 3 | 0 | |||||||||||||||||
GA | 0 | ||||||||||||||||||
SC | 1 | 0 | |||||||||||||||||
DE | 2 | 0 | |||||||||||||||||
CT | 0 | ||||||||||||||||||
VT | 2 | 0 | |||||||||||||||||
ME | 1 | 0 | |||||||||||||||||
MN | 0 |
None | None | ||
Traditional | Traditional | ||
Commodity | Commodity | ||
Econometric | Econometric | ||
Unknown | Unknown |
Statewide vs Urban Model Modal Issuesbull Most statewide models are limited to highway trips
bull Modeling of local bus routes is likely overkill
bull Mode choice is usually just a conversion from person trips to vehicle trips via auto occupancy rates
bull Transit components of statewide models are more common in regions with history of fixed guideway transit
bull Transit model is needed to test options such as high speedintercity rail
ATS 95 NHTS 01 MDOT 03-04 ODOT 02-03Personal vehicle 81 89 89 88Commercial air 16 8 10 8Bus 2 2 04 19Train 05 01 0 005Other 05 09 06 205
Survey and yearPrimary mode of transport
Statewide vs Urban Model Modal Issues (Contrsquod)
3727 24 22
32
29 36
10
21
28 22
10
22
12
1529
11
44 43
0
10
20
30
40
50
60
70
80
90
100
Personalvehicle
Bus Train Marine Commercialair
Perc
ent o
f per
son
trips
lt300 300-499 500-599 1000-1999 gt2000
68
85 85 82 85
30
12 11 15 13
0
10
20
30
40
50
60
70
80
90
100
Busi-ness
Plea-sure
pBusi-ness
Vaca-tion
Week-end
Perc
ent o
f per
son
trips
Personal vehicle Commercial air All other
Mode by Trip Purpose Mode by Distance (miles)
Statewide vs Urban Model Modal Issues (Contrsquod)
bull Most statewide assignment models separate trucks from passenger cars Preloading process commonly used for trucks Takes into account more limited route options for truck travel
Source Ohio Dept of Transportation
Statewide vs Urban Model Truck Forecasting
bull Urban modelsndash Freight movements are partially
ldquoexternalrdquo to model (truncated at external zone)
ndash More general categories of truckcommercial trips used
ndash Strong reliance on Quick Response Freight Manual
bull Statewide modelsndash Rooted in commodity flows ndash Often use econometric
approaches to modelingndash Generally model entire length of
freight travel
Freight Model Type by Peer Exchange Participants and
NonParticipants
0123456789
None
Traditio
nal
Commod
ity
Econo
metric
Unkno
wn
NotPeer Exchange
Source TRB Electronic Circular E-C075 Statewide Travel Demand Modeling ndash A Peer Exchange
Topic ndash Data Needs and Considerations
bull Impact of model complexity on data needs
bull Overview of data sources
bull Long-Distance travel gt different focus than MPOregional models
Sources NHTS Weight calculation Pitfalls Key statistics European surveys Lessons learned
Impact of Model Complexity on Data Needsbull Trip table estimation (2 step) ndash highway network
zone system trip table
bull Highway model structure (3 step) ndash 2 step plusbehavioral data for trip rates friction factors auto occupancy rates
bull Freight considerations (3 step) ndash 3 step pluscommodity flows economic data tonnages by mode truck counts
bull Full multi-modal structure (4 step) ndash 3 step plustransit networks and paths logit mode choice model
bull Beyond 4 Step ndash synthetic household generation operational characteristics etc
Source Georgia DOTCambridge Systematics Inc
Statewide Model Data Sources A-Zbull American Travel Survey (ATS)
bull Bureau of Economic Analysis (BEA)
bull Census Journey-to-Work (JTW) data
bull Census Public Use Microsample (PUMS)
bull Census Transportation Planning Package (CTPP)
bull Claritas
bull Commodity Flow Survey (CFS)
Source US Census Bureau
Statewide Model Data Sources (Contrsquod)bull Dun amp Bradstreet (DampB) employment data
bull Employment Security data (ES-202)
bull Freight Analysis Framework (FAF)
bull Highway Performance Monitoring System (HPMS)
bull InfoUSA employment data
bull Metropolitan Planning Organizations (MPO)
bull NCHRP 365 Manual
bull National Highway Planning Network (NHPN)
Source FAF2 Provisional Commodity Origin-Destination Estimates Revised Methodology Report BattelleFHWA
Statewide Model Data Sources (Contrsquod)
bull National Household Travel Survey (NHTS)
bull Origin-destination surveys
bull Regional Economic Models Inc (REMI)
bull State Departments of Transportation
bull TRANSEARCH data
bull Vehicle Use and Identification Survey (VIUS)
bull Woods amp Poole employment forecasts
Source Bureau of Transportation Statistics NHTS
Data Sources on Long-Distance Travel
bull Census Journey-to-Work databull Long distance component
American Travel Survey (77 95) National Household Travel Survey (01) Ohio statewide surveys (02-04) Michigan statewide survey (04-05) California (10-12)
bull Surveys focused on long distance Oregon (96-98 07-08)
bull Others BTS Omnibus surveys Tourism surveys European practice NCHRP Report 735
NHTS and Long-Distance Travelbull Evolution of US national surveys
NPTS (69 77 83 90 95) ATS (77 95)
bull 2001 NHTS ndash Total of 60282 long distance person trips Data collected March 01 to May 02 22204 (37 percent) before 911 38078 (63 percent) afterwards
bull Long distance trip is longer than 50 miles one-way (vs 100 miles in 1995 ATS)
bull 28-day retrospective survey (including travel day)
bull Not your typical travelers Low frequency Distinct segment of the population
Weight Calculation for Long-Distance Travelers
bull Use to ensure valid population-level estimates Accounts for non-response under-coverage multiple phones etc
bull Multi-stage process Match national margins Adjust for data splitting (force sumexpfac = 277208169) Raking (match independent controls) race ethnicity race by
month ethnicity by month sex by age census region MSA status month by day of week
Trim outlier expansion factors max expfac = 4 times mean Replication (correct for sampling error)
bull At least two passes through the data
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Freight Model Types | |||||||||||||||||||
State | Type | Peer Exc | |||||||||||||||||
OR | 4 | 1 | Count of Type | Count of Type | Peer Exc | ||||||||||||||
WI | 3 | 1 | Type | Total | Type | 0 | 1 | Grand Total | |||||||||||
MO | 1 | 1 | 1 | 2 | 1 | 2 | 2 | 4 | |||||||||||
LA | 3 | 1 | 2 | 2 | 2 | 2 | 2 | 4 | |||||||||||
IN | 3 | 1 | 3 | 6 | 3 | 2 | 6 | 8 | |||||||||||
OH | 4 | 1 | 4 | 2 | 4 | 2 | 2 | ||||||||||||
KY | 3 | 1 | Grand Total | 12 | 3 | 3 | |||||||||||||
FL | 3 | 1 | Grand Total | 9 | 12 | 21 | |||||||||||||
VA | 3 | 1 | None | 2 | Not | Peer Exchange | |||||||||||||
NJ | 2 | 1 | Traditional | 2 | None | 2 | 2 | ||||||||||||
MA | 2 | 1 | Commodity | 6 | Traditional | 2 | 2 | ||||||||||||
NH | 1 | 1 | Econometric | 2 | Commodity | 2 | 6 | ||||||||||||
CA | 0 | Econometric | 2 | ||||||||||||||||
TX | 3 | 0 | Unknown | 3 | |||||||||||||||
MI | 3 | 0 | |||||||||||||||||
GA | 0 | ||||||||||||||||||
SC | 1 | 0 | |||||||||||||||||
DE | 2 | 0 | |||||||||||||||||
CT | 0 | ||||||||||||||||||
VT | 2 | 0 | |||||||||||||||||
ME | 1 | 0 | |||||||||||||||||
MN | 0 |
None | None | ||
Traditional | Traditional | ||
Commodity | Commodity | ||
Econometric | Econometric | ||
Unknown | Unknown |
Statewide vs Urban Model Modal Issues (Contrsquod)
3727 24 22
32
29 36
10
21
28 22
10
22
12
1529
11
44 43
0
10
20
30
40
50
60
70
80
90
100
Personalvehicle
Bus Train Marine Commercialair
Perc
ent o
f per
son
trips
lt300 300-499 500-599 1000-1999 gt2000
68
85 85 82 85
30
12 11 15 13
0
10
20
30
40
50
60
70
80
90
100
Busi-ness
Plea-sure
pBusi-ness
Vaca-tion
Week-end
Perc
ent o
f per
son
trips
Personal vehicle Commercial air All other
Mode by Trip Purpose Mode by Distance (miles)
Statewide vs Urban Model Modal Issues (Contrsquod)
bull Most statewide assignment models separate trucks from passenger cars Preloading process commonly used for trucks Takes into account more limited route options for truck travel
Source Ohio Dept of Transportation
Statewide vs Urban Model Truck Forecasting
bull Urban modelsndash Freight movements are partially
ldquoexternalrdquo to model (truncated at external zone)
ndash More general categories of truckcommercial trips used
ndash Strong reliance on Quick Response Freight Manual
bull Statewide modelsndash Rooted in commodity flows ndash Often use econometric
approaches to modelingndash Generally model entire length of
freight travel
Freight Model Type by Peer Exchange Participants and
NonParticipants
0123456789
None
Traditio
nal
Commod
ity
Econo
metric
Unkno
wn
NotPeer Exchange
Source TRB Electronic Circular E-C075 Statewide Travel Demand Modeling ndash A Peer Exchange
Topic ndash Data Needs and Considerations
bull Impact of model complexity on data needs
bull Overview of data sources
bull Long-Distance travel gt different focus than MPOregional models
Sources NHTS Weight calculation Pitfalls Key statistics European surveys Lessons learned
Impact of Model Complexity on Data Needsbull Trip table estimation (2 step) ndash highway network
zone system trip table
bull Highway model structure (3 step) ndash 2 step plusbehavioral data for trip rates friction factors auto occupancy rates
bull Freight considerations (3 step) ndash 3 step pluscommodity flows economic data tonnages by mode truck counts
bull Full multi-modal structure (4 step) ndash 3 step plustransit networks and paths logit mode choice model
bull Beyond 4 Step ndash synthetic household generation operational characteristics etc
Source Georgia DOTCambridge Systematics Inc
Statewide Model Data Sources A-Zbull American Travel Survey (ATS)
bull Bureau of Economic Analysis (BEA)
bull Census Journey-to-Work (JTW) data
bull Census Public Use Microsample (PUMS)
bull Census Transportation Planning Package (CTPP)
bull Claritas
bull Commodity Flow Survey (CFS)
Source US Census Bureau
Statewide Model Data Sources (Contrsquod)bull Dun amp Bradstreet (DampB) employment data
bull Employment Security data (ES-202)
bull Freight Analysis Framework (FAF)
bull Highway Performance Monitoring System (HPMS)
bull InfoUSA employment data
bull Metropolitan Planning Organizations (MPO)
bull NCHRP 365 Manual
bull National Highway Planning Network (NHPN)
Source FAF2 Provisional Commodity Origin-Destination Estimates Revised Methodology Report BattelleFHWA
Statewide Model Data Sources (Contrsquod)
bull National Household Travel Survey (NHTS)
bull Origin-destination surveys
bull Regional Economic Models Inc (REMI)
bull State Departments of Transportation
bull TRANSEARCH data
bull Vehicle Use and Identification Survey (VIUS)
bull Woods amp Poole employment forecasts
Source Bureau of Transportation Statistics NHTS
Data Sources on Long-Distance Travel
bull Census Journey-to-Work databull Long distance component
American Travel Survey (77 95) National Household Travel Survey (01) Ohio statewide surveys (02-04) Michigan statewide survey (04-05) California (10-12)
bull Surveys focused on long distance Oregon (96-98 07-08)
bull Others BTS Omnibus surveys Tourism surveys European practice NCHRP Report 735
NHTS and Long-Distance Travelbull Evolution of US national surveys
NPTS (69 77 83 90 95) ATS (77 95)
bull 2001 NHTS ndash Total of 60282 long distance person trips Data collected March 01 to May 02 22204 (37 percent) before 911 38078 (63 percent) afterwards
bull Long distance trip is longer than 50 miles one-way (vs 100 miles in 1995 ATS)
bull 28-day retrospective survey (including travel day)
bull Not your typical travelers Low frequency Distinct segment of the population
Weight Calculation for Long-Distance Travelers
bull Use to ensure valid population-level estimates Accounts for non-response under-coverage multiple phones etc
bull Multi-stage process Match national margins Adjust for data splitting (force sumexpfac = 277208169) Raking (match independent controls) race ethnicity race by
month ethnicity by month sex by age census region MSA status month by day of week
Trim outlier expansion factors max expfac = 4 times mean Replication (correct for sampling error)
bull At least two passes through the data
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Freight Model Types | |||||||||||||||||||
State | Type | Peer Exc | |||||||||||||||||
OR | 4 | 1 | Count of Type | Count of Type | Peer Exc | ||||||||||||||
WI | 3 | 1 | Type | Total | Type | 0 | 1 | Grand Total | |||||||||||
MO | 1 | 1 | 1 | 2 | 1 | 2 | 2 | 4 | |||||||||||
LA | 3 | 1 | 2 | 2 | 2 | 2 | 2 | 4 | |||||||||||
IN | 3 | 1 | 3 | 6 | 3 | 2 | 6 | 8 | |||||||||||
OH | 4 | 1 | 4 | 2 | 4 | 2 | 2 | ||||||||||||
KY | 3 | 1 | Grand Total | 12 | 3 | 3 | |||||||||||||
FL | 3 | 1 | Grand Total | 9 | 12 | 21 | |||||||||||||
VA | 3 | 1 | None | 2 | Not | Peer Exchange | |||||||||||||
NJ | 2 | 1 | Traditional | 2 | None | 2 | 2 | ||||||||||||
MA | 2 | 1 | Commodity | 6 | Traditional | 2 | 2 | ||||||||||||
NH | 1 | 1 | Econometric | 2 | Commodity | 2 | 6 | ||||||||||||
CA | 0 | Econometric | 2 | ||||||||||||||||
TX | 3 | 0 | Unknown | 3 | |||||||||||||||
MI | 3 | 0 | |||||||||||||||||
GA | 0 | ||||||||||||||||||
SC | 1 | 0 | |||||||||||||||||
DE | 2 | 0 | |||||||||||||||||
CT | 0 | ||||||||||||||||||
VT | 2 | 0 | |||||||||||||||||
ME | 1 | 0 | |||||||||||||||||
MN | 0 |
None | None | ||
Traditional | Traditional | ||
Commodity | Commodity | ||
Econometric | Econometric | ||
Unknown | Unknown |
Statewide vs Urban Model Modal Issues (Contrsquod)
bull Most statewide assignment models separate trucks from passenger cars Preloading process commonly used for trucks Takes into account more limited route options for truck travel
Source Ohio Dept of Transportation
Statewide vs Urban Model Truck Forecasting
bull Urban modelsndash Freight movements are partially
ldquoexternalrdquo to model (truncated at external zone)
ndash More general categories of truckcommercial trips used
ndash Strong reliance on Quick Response Freight Manual
bull Statewide modelsndash Rooted in commodity flows ndash Often use econometric
approaches to modelingndash Generally model entire length of
freight travel
Freight Model Type by Peer Exchange Participants and
NonParticipants
0123456789
None
Traditio
nal
Commod
ity
Econo
metric
Unkno
wn
NotPeer Exchange
Source TRB Electronic Circular E-C075 Statewide Travel Demand Modeling ndash A Peer Exchange
Topic ndash Data Needs and Considerations
bull Impact of model complexity on data needs
bull Overview of data sources
bull Long-Distance travel gt different focus than MPOregional models
Sources NHTS Weight calculation Pitfalls Key statistics European surveys Lessons learned
Impact of Model Complexity on Data Needsbull Trip table estimation (2 step) ndash highway network
zone system trip table
bull Highway model structure (3 step) ndash 2 step plusbehavioral data for trip rates friction factors auto occupancy rates
bull Freight considerations (3 step) ndash 3 step pluscommodity flows economic data tonnages by mode truck counts
bull Full multi-modal structure (4 step) ndash 3 step plustransit networks and paths logit mode choice model
bull Beyond 4 Step ndash synthetic household generation operational characteristics etc
Source Georgia DOTCambridge Systematics Inc
Statewide Model Data Sources A-Zbull American Travel Survey (ATS)
bull Bureau of Economic Analysis (BEA)
bull Census Journey-to-Work (JTW) data
bull Census Public Use Microsample (PUMS)
bull Census Transportation Planning Package (CTPP)
bull Claritas
bull Commodity Flow Survey (CFS)
Source US Census Bureau
Statewide Model Data Sources (Contrsquod)bull Dun amp Bradstreet (DampB) employment data
bull Employment Security data (ES-202)
bull Freight Analysis Framework (FAF)
bull Highway Performance Monitoring System (HPMS)
bull InfoUSA employment data
bull Metropolitan Planning Organizations (MPO)
bull NCHRP 365 Manual
bull National Highway Planning Network (NHPN)
Source FAF2 Provisional Commodity Origin-Destination Estimates Revised Methodology Report BattelleFHWA
Statewide Model Data Sources (Contrsquod)
bull National Household Travel Survey (NHTS)
bull Origin-destination surveys
bull Regional Economic Models Inc (REMI)
bull State Departments of Transportation
bull TRANSEARCH data
bull Vehicle Use and Identification Survey (VIUS)
bull Woods amp Poole employment forecasts
Source Bureau of Transportation Statistics NHTS
Data Sources on Long-Distance Travel
bull Census Journey-to-Work databull Long distance component
American Travel Survey (77 95) National Household Travel Survey (01) Ohio statewide surveys (02-04) Michigan statewide survey (04-05) California (10-12)
bull Surveys focused on long distance Oregon (96-98 07-08)
bull Others BTS Omnibus surveys Tourism surveys European practice NCHRP Report 735
NHTS and Long-Distance Travelbull Evolution of US national surveys
NPTS (69 77 83 90 95) ATS (77 95)
bull 2001 NHTS ndash Total of 60282 long distance person trips Data collected March 01 to May 02 22204 (37 percent) before 911 38078 (63 percent) afterwards
bull Long distance trip is longer than 50 miles one-way (vs 100 miles in 1995 ATS)
bull 28-day retrospective survey (including travel day)
bull Not your typical travelers Low frequency Distinct segment of the population
Weight Calculation for Long-Distance Travelers
bull Use to ensure valid population-level estimates Accounts for non-response under-coverage multiple phones etc
bull Multi-stage process Match national margins Adjust for data splitting (force sumexpfac = 277208169) Raking (match independent controls) race ethnicity race by
month ethnicity by month sex by age census region MSA status month by day of week
Trim outlier expansion factors max expfac = 4 times mean Replication (correct for sampling error)
bull At least two passes through the data
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Freight Model Types | |||||||||||||||||||
State | Type | Peer Exc | |||||||||||||||||
OR | 4 | 1 | Count of Type | Count of Type | Peer Exc | ||||||||||||||
WI | 3 | 1 | Type | Total | Type | 0 | 1 | Grand Total | |||||||||||
MO | 1 | 1 | 1 | 2 | 1 | 2 | 2 | 4 | |||||||||||
LA | 3 | 1 | 2 | 2 | 2 | 2 | 2 | 4 | |||||||||||
IN | 3 | 1 | 3 | 6 | 3 | 2 | 6 | 8 | |||||||||||
OH | 4 | 1 | 4 | 2 | 4 | 2 | 2 | ||||||||||||
KY | 3 | 1 | Grand Total | 12 | 3 | 3 | |||||||||||||
FL | 3 | 1 | Grand Total | 9 | 12 | 21 | |||||||||||||
VA | 3 | 1 | None | 2 | Not | Peer Exchange | |||||||||||||
NJ | 2 | 1 | Traditional | 2 | None | 2 | 2 | ||||||||||||
MA | 2 | 1 | Commodity | 6 | Traditional | 2 | 2 | ||||||||||||
NH | 1 | 1 | Econometric | 2 | Commodity | 2 | 6 | ||||||||||||
CA | 0 | Econometric | 2 | ||||||||||||||||
TX | 3 | 0 | Unknown | 3 | |||||||||||||||
MI | 3 | 0 | |||||||||||||||||
GA | 0 | ||||||||||||||||||
SC | 1 | 0 | |||||||||||||||||
DE | 2 | 0 | |||||||||||||||||
CT | 0 | ||||||||||||||||||
VT | 2 | 0 | |||||||||||||||||
ME | 1 | 0 | |||||||||||||||||
MN | 0 |
None | None | ||
Traditional | Traditional | ||
Commodity | Commodity | ||
Econometric | Econometric | ||
Unknown | Unknown |
Statewide vs Urban Model Truck Forecasting
bull Urban modelsndash Freight movements are partially
ldquoexternalrdquo to model (truncated at external zone)
ndash More general categories of truckcommercial trips used
ndash Strong reliance on Quick Response Freight Manual
bull Statewide modelsndash Rooted in commodity flows ndash Often use econometric
approaches to modelingndash Generally model entire length of
freight travel
Freight Model Type by Peer Exchange Participants and
NonParticipants
0123456789
None
Traditio
nal
Commod
ity
Econo
metric
Unkno
wn
NotPeer Exchange
Source TRB Electronic Circular E-C075 Statewide Travel Demand Modeling ndash A Peer Exchange
Topic ndash Data Needs and Considerations
bull Impact of model complexity on data needs
bull Overview of data sources
bull Long-Distance travel gt different focus than MPOregional models
Sources NHTS Weight calculation Pitfalls Key statistics European surveys Lessons learned
Impact of Model Complexity on Data Needsbull Trip table estimation (2 step) ndash highway network
zone system trip table
bull Highway model structure (3 step) ndash 2 step plusbehavioral data for trip rates friction factors auto occupancy rates
bull Freight considerations (3 step) ndash 3 step pluscommodity flows economic data tonnages by mode truck counts
bull Full multi-modal structure (4 step) ndash 3 step plustransit networks and paths logit mode choice model
bull Beyond 4 Step ndash synthetic household generation operational characteristics etc
Source Georgia DOTCambridge Systematics Inc
Statewide Model Data Sources A-Zbull American Travel Survey (ATS)
bull Bureau of Economic Analysis (BEA)
bull Census Journey-to-Work (JTW) data
bull Census Public Use Microsample (PUMS)
bull Census Transportation Planning Package (CTPP)
bull Claritas
bull Commodity Flow Survey (CFS)
Source US Census Bureau
Statewide Model Data Sources (Contrsquod)bull Dun amp Bradstreet (DampB) employment data
bull Employment Security data (ES-202)
bull Freight Analysis Framework (FAF)
bull Highway Performance Monitoring System (HPMS)
bull InfoUSA employment data
bull Metropolitan Planning Organizations (MPO)
bull NCHRP 365 Manual
bull National Highway Planning Network (NHPN)
Source FAF2 Provisional Commodity Origin-Destination Estimates Revised Methodology Report BattelleFHWA
Statewide Model Data Sources (Contrsquod)
bull National Household Travel Survey (NHTS)
bull Origin-destination surveys
bull Regional Economic Models Inc (REMI)
bull State Departments of Transportation
bull TRANSEARCH data
bull Vehicle Use and Identification Survey (VIUS)
bull Woods amp Poole employment forecasts
Source Bureau of Transportation Statistics NHTS
Data Sources on Long-Distance Travel
bull Census Journey-to-Work databull Long distance component
American Travel Survey (77 95) National Household Travel Survey (01) Ohio statewide surveys (02-04) Michigan statewide survey (04-05) California (10-12)
bull Surveys focused on long distance Oregon (96-98 07-08)
bull Others BTS Omnibus surveys Tourism surveys European practice NCHRP Report 735
NHTS and Long-Distance Travelbull Evolution of US national surveys
NPTS (69 77 83 90 95) ATS (77 95)
bull 2001 NHTS ndash Total of 60282 long distance person trips Data collected March 01 to May 02 22204 (37 percent) before 911 38078 (63 percent) afterwards
bull Long distance trip is longer than 50 miles one-way (vs 100 miles in 1995 ATS)
bull 28-day retrospective survey (including travel day)
bull Not your typical travelers Low frequency Distinct segment of the population
Weight Calculation for Long-Distance Travelers
bull Use to ensure valid population-level estimates Accounts for non-response under-coverage multiple phones etc
bull Multi-stage process Match national margins Adjust for data splitting (force sumexpfac = 277208169) Raking (match independent controls) race ethnicity race by
month ethnicity by month sex by age census region MSA status month by day of week
Trim outlier expansion factors max expfac = 4 times mean Replication (correct for sampling error)
bull At least two passes through the data
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Freight Model Types | |||||||||||||||||||
State | Type | Peer Exc | |||||||||||||||||
OR | 4 | 1 | Count of Type | Count of Type | Peer Exc | ||||||||||||||
WI | 3 | 1 | Type | Total | Type | 0 | 1 | Grand Total | |||||||||||
MO | 1 | 1 | 1 | 2 | 1 | 2 | 2 | 4 | |||||||||||
LA | 3 | 1 | 2 | 2 | 2 | 2 | 2 | 4 | |||||||||||
IN | 3 | 1 | 3 | 6 | 3 | 2 | 6 | 8 | |||||||||||
OH | 4 | 1 | 4 | 2 | 4 | 2 | 2 | ||||||||||||
KY | 3 | 1 | Grand Total | 12 | 3 | 3 | |||||||||||||
FL | 3 | 1 | Grand Total | 9 | 12 | 21 | |||||||||||||
VA | 3 | 1 | None | 2 | Not | Peer Exchange | |||||||||||||
NJ | 2 | 1 | Traditional | 2 | None | 2 | 2 | ||||||||||||
MA | 2 | 1 | Commodity | 6 | Traditional | 2 | 2 | ||||||||||||
NH | 1 | 1 | Econometric | 2 | Commodity | 2 | 6 | ||||||||||||
CA | 0 | Econometric | 2 | ||||||||||||||||
TX | 3 | 0 | Unknown | 3 | |||||||||||||||
MI | 3 | 0 | |||||||||||||||||
GA | 0 | ||||||||||||||||||
SC | 1 | 0 | |||||||||||||||||
DE | 2 | 0 | |||||||||||||||||
CT | 0 | ||||||||||||||||||
VT | 2 | 0 | |||||||||||||||||
ME | 1 | 0 | |||||||||||||||||
MN | 0 |
None | None | ||
Traditional | Traditional | ||
Commodity | Commodity | ||
Econometric | Econometric | ||
Unknown | Unknown |
Topic ndash Data Needs and Considerations
bull Impact of model complexity on data needs
bull Overview of data sources
bull Long-Distance travel gt different focus than MPOregional models
Sources NHTS Weight calculation Pitfalls Key statistics European surveys Lessons learned
Impact of Model Complexity on Data Needsbull Trip table estimation (2 step) ndash highway network
zone system trip table
bull Highway model structure (3 step) ndash 2 step plusbehavioral data for trip rates friction factors auto occupancy rates
bull Freight considerations (3 step) ndash 3 step pluscommodity flows economic data tonnages by mode truck counts
bull Full multi-modal structure (4 step) ndash 3 step plustransit networks and paths logit mode choice model
bull Beyond 4 Step ndash synthetic household generation operational characteristics etc
Source Georgia DOTCambridge Systematics Inc
Statewide Model Data Sources A-Zbull American Travel Survey (ATS)
bull Bureau of Economic Analysis (BEA)
bull Census Journey-to-Work (JTW) data
bull Census Public Use Microsample (PUMS)
bull Census Transportation Planning Package (CTPP)
bull Claritas
bull Commodity Flow Survey (CFS)
Source US Census Bureau
Statewide Model Data Sources (Contrsquod)bull Dun amp Bradstreet (DampB) employment data
bull Employment Security data (ES-202)
bull Freight Analysis Framework (FAF)
bull Highway Performance Monitoring System (HPMS)
bull InfoUSA employment data
bull Metropolitan Planning Organizations (MPO)
bull NCHRP 365 Manual
bull National Highway Planning Network (NHPN)
Source FAF2 Provisional Commodity Origin-Destination Estimates Revised Methodology Report BattelleFHWA
Statewide Model Data Sources (Contrsquod)
bull National Household Travel Survey (NHTS)
bull Origin-destination surveys
bull Regional Economic Models Inc (REMI)
bull State Departments of Transportation
bull TRANSEARCH data
bull Vehicle Use and Identification Survey (VIUS)
bull Woods amp Poole employment forecasts
Source Bureau of Transportation Statistics NHTS
Data Sources on Long-Distance Travel
bull Census Journey-to-Work databull Long distance component
American Travel Survey (77 95) National Household Travel Survey (01) Ohio statewide surveys (02-04) Michigan statewide survey (04-05) California (10-12)
bull Surveys focused on long distance Oregon (96-98 07-08)
bull Others BTS Omnibus surveys Tourism surveys European practice NCHRP Report 735
NHTS and Long-Distance Travelbull Evolution of US national surveys
NPTS (69 77 83 90 95) ATS (77 95)
bull 2001 NHTS ndash Total of 60282 long distance person trips Data collected March 01 to May 02 22204 (37 percent) before 911 38078 (63 percent) afterwards
bull Long distance trip is longer than 50 miles one-way (vs 100 miles in 1995 ATS)
bull 28-day retrospective survey (including travel day)
bull Not your typical travelers Low frequency Distinct segment of the population
Weight Calculation for Long-Distance Travelers
bull Use to ensure valid population-level estimates Accounts for non-response under-coverage multiple phones etc
bull Multi-stage process Match national margins Adjust for data splitting (force sumexpfac = 277208169) Raking (match independent controls) race ethnicity race by
month ethnicity by month sex by age census region MSA status month by day of week
Trim outlier expansion factors max expfac = 4 times mean Replication (correct for sampling error)
bull At least two passes through the data
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Freight Model Types | |||||||||||||||||||
State | Type | Peer Exc | |||||||||||||||||
OR | 4 | 1 | Count of Type | Count of Type | Peer Exc | ||||||||||||||
WI | 3 | 1 | Type | Total | Type | 0 | 1 | Grand Total | |||||||||||
MO | 1 | 1 | 1 | 2 | 1 | 2 | 2 | 4 | |||||||||||
LA | 3 | 1 | 2 | 2 | 2 | 2 | 2 | 4 | |||||||||||
IN | 3 | 1 | 3 | 6 | 3 | 2 | 6 | 8 | |||||||||||
OH | 4 | 1 | 4 | 2 | 4 | 2 | 2 | ||||||||||||
KY | 3 | 1 | Grand Total | 12 | 3 | 3 | |||||||||||||
FL | 3 | 1 | Grand Total | 9 | 12 | 21 | |||||||||||||
VA | 3 | 1 | None | 2 | Not | Peer Exchange | |||||||||||||
NJ | 2 | 1 | Traditional | 2 | None | 2 | 2 | ||||||||||||
MA | 2 | 1 | Commodity | 6 | Traditional | 2 | 2 | ||||||||||||
NH | 1 | 1 | Econometric | 2 | Commodity | 2 | 6 | ||||||||||||
CA | 0 | Econometric | 2 | ||||||||||||||||
TX | 3 | 0 | Unknown | 3 | |||||||||||||||
MI | 3 | 0 | |||||||||||||||||
GA | 0 | ||||||||||||||||||
SC | 1 | 0 | |||||||||||||||||
DE | 2 | 0 | |||||||||||||||||
CT | 0 | ||||||||||||||||||
VT | 2 | 0 | |||||||||||||||||
ME | 1 | 0 | |||||||||||||||||
MN | 0 |
None | None | ||
Traditional | Traditional | ||
Commodity | Commodity | ||
Econometric | Econometric | ||
Unknown | Unknown |
Topic ndash Data Needs and Considerations
bull Impact of model complexity on data needs
bull Overview of data sources
bull Long-Distance travel gt different focus than MPOregional models
Sources NHTS Weight calculation Pitfalls Key statistics European surveys Lessons learned
Impact of Model Complexity on Data Needsbull Trip table estimation (2 step) ndash highway network
zone system trip table
bull Highway model structure (3 step) ndash 2 step plusbehavioral data for trip rates friction factors auto occupancy rates
bull Freight considerations (3 step) ndash 3 step pluscommodity flows economic data tonnages by mode truck counts
bull Full multi-modal structure (4 step) ndash 3 step plustransit networks and paths logit mode choice model
bull Beyond 4 Step ndash synthetic household generation operational characteristics etc
Source Georgia DOTCambridge Systematics Inc
Statewide Model Data Sources A-Zbull American Travel Survey (ATS)
bull Bureau of Economic Analysis (BEA)
bull Census Journey-to-Work (JTW) data
bull Census Public Use Microsample (PUMS)
bull Census Transportation Planning Package (CTPP)
bull Claritas
bull Commodity Flow Survey (CFS)
Source US Census Bureau
Statewide Model Data Sources (Contrsquod)bull Dun amp Bradstreet (DampB) employment data
bull Employment Security data (ES-202)
bull Freight Analysis Framework (FAF)
bull Highway Performance Monitoring System (HPMS)
bull InfoUSA employment data
bull Metropolitan Planning Organizations (MPO)
bull NCHRP 365 Manual
bull National Highway Planning Network (NHPN)
Source FAF2 Provisional Commodity Origin-Destination Estimates Revised Methodology Report BattelleFHWA
Statewide Model Data Sources (Contrsquod)
bull National Household Travel Survey (NHTS)
bull Origin-destination surveys
bull Regional Economic Models Inc (REMI)
bull State Departments of Transportation
bull TRANSEARCH data
bull Vehicle Use and Identification Survey (VIUS)
bull Woods amp Poole employment forecasts
Source Bureau of Transportation Statistics NHTS
Data Sources on Long-Distance Travel
bull Census Journey-to-Work databull Long distance component
American Travel Survey (77 95) National Household Travel Survey (01) Ohio statewide surveys (02-04) Michigan statewide survey (04-05) California (10-12)
bull Surveys focused on long distance Oregon (96-98 07-08)
bull Others BTS Omnibus surveys Tourism surveys European practice NCHRP Report 735
NHTS and Long-Distance Travelbull Evolution of US national surveys
NPTS (69 77 83 90 95) ATS (77 95)
bull 2001 NHTS ndash Total of 60282 long distance person trips Data collected March 01 to May 02 22204 (37 percent) before 911 38078 (63 percent) afterwards
bull Long distance trip is longer than 50 miles one-way (vs 100 miles in 1995 ATS)
bull 28-day retrospective survey (including travel day)
bull Not your typical travelers Low frequency Distinct segment of the population
Weight Calculation for Long-Distance Travelers
bull Use to ensure valid population-level estimates Accounts for non-response under-coverage multiple phones etc
bull Multi-stage process Match national margins Adjust for data splitting (force sumexpfac = 277208169) Raking (match independent controls) race ethnicity race by
month ethnicity by month sex by age census region MSA status month by day of week
Trim outlier expansion factors max expfac = 4 times mean Replication (correct for sampling error)
bull At least two passes through the data
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Freight Model Types | |||||||||||||||||||
State | Type | Peer Exc | |||||||||||||||||
OR | 4 | 1 | Count of Type | Count of Type | Peer Exc | ||||||||||||||
WI | 3 | 1 | Type | Total | Type | 0 | 1 | Grand Total | |||||||||||
MO | 1 | 1 | 1 | 2 | 1 | 2 | 2 | 4 | |||||||||||
LA | 3 | 1 | 2 | 2 | 2 | 2 | 2 | 4 | |||||||||||
IN | 3 | 1 | 3 | 6 | 3 | 2 | 6 | 8 | |||||||||||
OH | 4 | 1 | 4 | 2 | 4 | 2 | 2 | ||||||||||||
KY | 3 | 1 | Grand Total | 12 | 3 | 3 | |||||||||||||
FL | 3 | 1 | Grand Total | 9 | 12 | 21 | |||||||||||||
VA | 3 | 1 | None | 2 | Not | Peer Exchange | |||||||||||||
NJ | 2 | 1 | Traditional | 2 | None | 2 | 2 | ||||||||||||
MA | 2 | 1 | Commodity | 6 | Traditional | 2 | 2 | ||||||||||||
NH | 1 | 1 | Econometric | 2 | Commodity | 2 | 6 | ||||||||||||
CA | 0 | Econometric | 2 | ||||||||||||||||
TX | 3 | 0 | Unknown | 3 | |||||||||||||||
MI | 3 | 0 | |||||||||||||||||
GA | 0 | ||||||||||||||||||
SC | 1 | 0 | |||||||||||||||||
DE | 2 | 0 | |||||||||||||||||
CT | 0 | ||||||||||||||||||
VT | 2 | 0 | |||||||||||||||||
ME | 1 | 0 | |||||||||||||||||
MN | 0 |
Topic ndash Data Needs and Considerations
bull Impact of model complexity on data needs
bull Overview of data sources
bull Long-Distance travel gt different focus than MPOregional models
Sources NHTS Weight calculation Pitfalls Key statistics European surveys Lessons learned
Impact of Model Complexity on Data Needsbull Trip table estimation (2 step) ndash highway network
zone system trip table
bull Highway model structure (3 step) ndash 2 step plusbehavioral data for trip rates friction factors auto occupancy rates
bull Freight considerations (3 step) ndash 3 step pluscommodity flows economic data tonnages by mode truck counts
bull Full multi-modal structure (4 step) ndash 3 step plustransit networks and paths logit mode choice model
bull Beyond 4 Step ndash synthetic household generation operational characteristics etc
Source Georgia DOTCambridge Systematics Inc
Statewide Model Data Sources A-Zbull American Travel Survey (ATS)
bull Bureau of Economic Analysis (BEA)
bull Census Journey-to-Work (JTW) data
bull Census Public Use Microsample (PUMS)
bull Census Transportation Planning Package (CTPP)
bull Claritas
bull Commodity Flow Survey (CFS)
Source US Census Bureau
Statewide Model Data Sources (Contrsquod)bull Dun amp Bradstreet (DampB) employment data
bull Employment Security data (ES-202)
bull Freight Analysis Framework (FAF)
bull Highway Performance Monitoring System (HPMS)
bull InfoUSA employment data
bull Metropolitan Planning Organizations (MPO)
bull NCHRP 365 Manual
bull National Highway Planning Network (NHPN)
Source FAF2 Provisional Commodity Origin-Destination Estimates Revised Methodology Report BattelleFHWA
Statewide Model Data Sources (Contrsquod)
bull National Household Travel Survey (NHTS)
bull Origin-destination surveys
bull Regional Economic Models Inc (REMI)
bull State Departments of Transportation
bull TRANSEARCH data
bull Vehicle Use and Identification Survey (VIUS)
bull Woods amp Poole employment forecasts
Source Bureau of Transportation Statistics NHTS
Data Sources on Long-Distance Travel
bull Census Journey-to-Work databull Long distance component
American Travel Survey (77 95) National Household Travel Survey (01) Ohio statewide surveys (02-04) Michigan statewide survey (04-05) California (10-12)
bull Surveys focused on long distance Oregon (96-98 07-08)
bull Others BTS Omnibus surveys Tourism surveys European practice NCHRP Report 735
NHTS and Long-Distance Travelbull Evolution of US national surveys
NPTS (69 77 83 90 95) ATS (77 95)
bull 2001 NHTS ndash Total of 60282 long distance person trips Data collected March 01 to May 02 22204 (37 percent) before 911 38078 (63 percent) afterwards
bull Long distance trip is longer than 50 miles one-way (vs 100 miles in 1995 ATS)
bull 28-day retrospective survey (including travel day)
bull Not your typical travelers Low frequency Distinct segment of the population
Weight Calculation for Long-Distance Travelers
bull Use to ensure valid population-level estimates Accounts for non-response under-coverage multiple phones etc
bull Multi-stage process Match national margins Adjust for data splitting (force sumexpfac = 277208169) Raking (match independent controls) race ethnicity race by
month ethnicity by month sex by age census region MSA status month by day of week
Trim outlier expansion factors max expfac = 4 times mean Replication (correct for sampling error)
bull At least two passes through the data
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Topic ndash Data Needs and Considerations
bull Impact of model complexity on data needs
bull Overview of data sources
bull Long-Distance travel gt different focus than MPOregional models
Sources NHTS Weight calculation Pitfalls Key statistics European surveys Lessons learned
Impact of Model Complexity on Data Needsbull Trip table estimation (2 step) ndash highway network
zone system trip table
bull Highway model structure (3 step) ndash 2 step plusbehavioral data for trip rates friction factors auto occupancy rates
bull Freight considerations (3 step) ndash 3 step pluscommodity flows economic data tonnages by mode truck counts
bull Full multi-modal structure (4 step) ndash 3 step plustransit networks and paths logit mode choice model
bull Beyond 4 Step ndash synthetic household generation operational characteristics etc
Source Georgia DOTCambridge Systematics Inc
Statewide Model Data Sources A-Zbull American Travel Survey (ATS)
bull Bureau of Economic Analysis (BEA)
bull Census Journey-to-Work (JTW) data
bull Census Public Use Microsample (PUMS)
bull Census Transportation Planning Package (CTPP)
bull Claritas
bull Commodity Flow Survey (CFS)
Source US Census Bureau
Statewide Model Data Sources (Contrsquod)bull Dun amp Bradstreet (DampB) employment data
bull Employment Security data (ES-202)
bull Freight Analysis Framework (FAF)
bull Highway Performance Monitoring System (HPMS)
bull InfoUSA employment data
bull Metropolitan Planning Organizations (MPO)
bull NCHRP 365 Manual
bull National Highway Planning Network (NHPN)
Source FAF2 Provisional Commodity Origin-Destination Estimates Revised Methodology Report BattelleFHWA
Statewide Model Data Sources (Contrsquod)
bull National Household Travel Survey (NHTS)
bull Origin-destination surveys
bull Regional Economic Models Inc (REMI)
bull State Departments of Transportation
bull TRANSEARCH data
bull Vehicle Use and Identification Survey (VIUS)
bull Woods amp Poole employment forecasts
Source Bureau of Transportation Statistics NHTS
Data Sources on Long-Distance Travel
bull Census Journey-to-Work databull Long distance component
American Travel Survey (77 95) National Household Travel Survey (01) Ohio statewide surveys (02-04) Michigan statewide survey (04-05) California (10-12)
bull Surveys focused on long distance Oregon (96-98 07-08)
bull Others BTS Omnibus surveys Tourism surveys European practice NCHRP Report 735
NHTS and Long-Distance Travelbull Evolution of US national surveys
NPTS (69 77 83 90 95) ATS (77 95)
bull 2001 NHTS ndash Total of 60282 long distance person trips Data collected March 01 to May 02 22204 (37 percent) before 911 38078 (63 percent) afterwards
bull Long distance trip is longer than 50 miles one-way (vs 100 miles in 1995 ATS)
bull 28-day retrospective survey (including travel day)
bull Not your typical travelers Low frequency Distinct segment of the population
Weight Calculation for Long-Distance Travelers
bull Use to ensure valid population-level estimates Accounts for non-response under-coverage multiple phones etc
bull Multi-stage process Match national margins Adjust for data splitting (force sumexpfac = 277208169) Raking (match independent controls) race ethnicity race by
month ethnicity by month sex by age census region MSA status month by day of week
Trim outlier expansion factors max expfac = 4 times mean Replication (correct for sampling error)
bull At least two passes through the data
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Topic ndash Data Needs and Considerations
bull Impact of model complexity on data needs
bull Overview of data sources
bull Long-Distance travel gt different focus than MPOregional models
Sources NHTS Weight calculation Pitfalls Key statistics European surveys Lessons learned
Impact of Model Complexity on Data Needsbull Trip table estimation (2 step) ndash highway network
zone system trip table
bull Highway model structure (3 step) ndash 2 step plusbehavioral data for trip rates friction factors auto occupancy rates
bull Freight considerations (3 step) ndash 3 step pluscommodity flows economic data tonnages by mode truck counts
bull Full multi-modal structure (4 step) ndash 3 step plustransit networks and paths logit mode choice model
bull Beyond 4 Step ndash synthetic household generation operational characteristics etc
Source Georgia DOTCambridge Systematics Inc
Statewide Model Data Sources A-Zbull American Travel Survey (ATS)
bull Bureau of Economic Analysis (BEA)
bull Census Journey-to-Work (JTW) data
bull Census Public Use Microsample (PUMS)
bull Census Transportation Planning Package (CTPP)
bull Claritas
bull Commodity Flow Survey (CFS)
Source US Census Bureau
Statewide Model Data Sources (Contrsquod)bull Dun amp Bradstreet (DampB) employment data
bull Employment Security data (ES-202)
bull Freight Analysis Framework (FAF)
bull Highway Performance Monitoring System (HPMS)
bull InfoUSA employment data
bull Metropolitan Planning Organizations (MPO)
bull NCHRP 365 Manual
bull National Highway Planning Network (NHPN)
Source FAF2 Provisional Commodity Origin-Destination Estimates Revised Methodology Report BattelleFHWA
Statewide Model Data Sources (Contrsquod)
bull National Household Travel Survey (NHTS)
bull Origin-destination surveys
bull Regional Economic Models Inc (REMI)
bull State Departments of Transportation
bull TRANSEARCH data
bull Vehicle Use and Identification Survey (VIUS)
bull Woods amp Poole employment forecasts
Source Bureau of Transportation Statistics NHTS
Data Sources on Long-Distance Travel
bull Census Journey-to-Work databull Long distance component
American Travel Survey (77 95) National Household Travel Survey (01) Ohio statewide surveys (02-04) Michigan statewide survey (04-05) California (10-12)
bull Surveys focused on long distance Oregon (96-98 07-08)
bull Others BTS Omnibus surveys Tourism surveys European practice NCHRP Report 735
NHTS and Long-Distance Travelbull Evolution of US national surveys
NPTS (69 77 83 90 95) ATS (77 95)
bull 2001 NHTS ndash Total of 60282 long distance person trips Data collected March 01 to May 02 22204 (37 percent) before 911 38078 (63 percent) afterwards
bull Long distance trip is longer than 50 miles one-way (vs 100 miles in 1995 ATS)
bull 28-day retrospective survey (including travel day)
bull Not your typical travelers Low frequency Distinct segment of the population
Weight Calculation for Long-Distance Travelers
bull Use to ensure valid population-level estimates Accounts for non-response under-coverage multiple phones etc
bull Multi-stage process Match national margins Adjust for data splitting (force sumexpfac = 277208169) Raking (match independent controls) race ethnicity race by
month ethnicity by month sex by age census region MSA status month by day of week
Trim outlier expansion factors max expfac = 4 times mean Replication (correct for sampling error)
bull At least two passes through the data
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Impact of Model Complexity on Data Needsbull Trip table estimation (2 step) ndash highway network
zone system trip table
bull Highway model structure (3 step) ndash 2 step plusbehavioral data for trip rates friction factors auto occupancy rates
bull Freight considerations (3 step) ndash 3 step pluscommodity flows economic data tonnages by mode truck counts
bull Full multi-modal structure (4 step) ndash 3 step plustransit networks and paths logit mode choice model
bull Beyond 4 Step ndash synthetic household generation operational characteristics etc
Source Georgia DOTCambridge Systematics Inc
Statewide Model Data Sources A-Zbull American Travel Survey (ATS)
bull Bureau of Economic Analysis (BEA)
bull Census Journey-to-Work (JTW) data
bull Census Public Use Microsample (PUMS)
bull Census Transportation Planning Package (CTPP)
bull Claritas
bull Commodity Flow Survey (CFS)
Source US Census Bureau
Statewide Model Data Sources (Contrsquod)bull Dun amp Bradstreet (DampB) employment data
bull Employment Security data (ES-202)
bull Freight Analysis Framework (FAF)
bull Highway Performance Monitoring System (HPMS)
bull InfoUSA employment data
bull Metropolitan Planning Organizations (MPO)
bull NCHRP 365 Manual
bull National Highway Planning Network (NHPN)
Source FAF2 Provisional Commodity Origin-Destination Estimates Revised Methodology Report BattelleFHWA
Statewide Model Data Sources (Contrsquod)
bull National Household Travel Survey (NHTS)
bull Origin-destination surveys
bull Regional Economic Models Inc (REMI)
bull State Departments of Transportation
bull TRANSEARCH data
bull Vehicle Use and Identification Survey (VIUS)
bull Woods amp Poole employment forecasts
Source Bureau of Transportation Statistics NHTS
Data Sources on Long-Distance Travel
bull Census Journey-to-Work databull Long distance component
American Travel Survey (77 95) National Household Travel Survey (01) Ohio statewide surveys (02-04) Michigan statewide survey (04-05) California (10-12)
bull Surveys focused on long distance Oregon (96-98 07-08)
bull Others BTS Omnibus surveys Tourism surveys European practice NCHRP Report 735
NHTS and Long-Distance Travelbull Evolution of US national surveys
NPTS (69 77 83 90 95) ATS (77 95)
bull 2001 NHTS ndash Total of 60282 long distance person trips Data collected March 01 to May 02 22204 (37 percent) before 911 38078 (63 percent) afterwards
bull Long distance trip is longer than 50 miles one-way (vs 100 miles in 1995 ATS)
bull 28-day retrospective survey (including travel day)
bull Not your typical travelers Low frequency Distinct segment of the population
Weight Calculation for Long-Distance Travelers
bull Use to ensure valid population-level estimates Accounts for non-response under-coverage multiple phones etc
bull Multi-stage process Match national margins Adjust for data splitting (force sumexpfac = 277208169) Raking (match independent controls) race ethnicity race by
month ethnicity by month sex by age census region MSA status month by day of week
Trim outlier expansion factors max expfac = 4 times mean Replication (correct for sampling error)
bull At least two passes through the data
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Statewide Model Data Sources A-Zbull American Travel Survey (ATS)
bull Bureau of Economic Analysis (BEA)
bull Census Journey-to-Work (JTW) data
bull Census Public Use Microsample (PUMS)
bull Census Transportation Planning Package (CTPP)
bull Claritas
bull Commodity Flow Survey (CFS)
Source US Census Bureau
Statewide Model Data Sources (Contrsquod)bull Dun amp Bradstreet (DampB) employment data
bull Employment Security data (ES-202)
bull Freight Analysis Framework (FAF)
bull Highway Performance Monitoring System (HPMS)
bull InfoUSA employment data
bull Metropolitan Planning Organizations (MPO)
bull NCHRP 365 Manual
bull National Highway Planning Network (NHPN)
Source FAF2 Provisional Commodity Origin-Destination Estimates Revised Methodology Report BattelleFHWA
Statewide Model Data Sources (Contrsquod)
bull National Household Travel Survey (NHTS)
bull Origin-destination surveys
bull Regional Economic Models Inc (REMI)
bull State Departments of Transportation
bull TRANSEARCH data
bull Vehicle Use and Identification Survey (VIUS)
bull Woods amp Poole employment forecasts
Source Bureau of Transportation Statistics NHTS
Data Sources on Long-Distance Travel
bull Census Journey-to-Work databull Long distance component
American Travel Survey (77 95) National Household Travel Survey (01) Ohio statewide surveys (02-04) Michigan statewide survey (04-05) California (10-12)
bull Surveys focused on long distance Oregon (96-98 07-08)
bull Others BTS Omnibus surveys Tourism surveys European practice NCHRP Report 735
NHTS and Long-Distance Travelbull Evolution of US national surveys
NPTS (69 77 83 90 95) ATS (77 95)
bull 2001 NHTS ndash Total of 60282 long distance person trips Data collected March 01 to May 02 22204 (37 percent) before 911 38078 (63 percent) afterwards
bull Long distance trip is longer than 50 miles one-way (vs 100 miles in 1995 ATS)
bull 28-day retrospective survey (including travel day)
bull Not your typical travelers Low frequency Distinct segment of the population
Weight Calculation for Long-Distance Travelers
bull Use to ensure valid population-level estimates Accounts for non-response under-coverage multiple phones etc
bull Multi-stage process Match national margins Adjust for data splitting (force sumexpfac = 277208169) Raking (match independent controls) race ethnicity race by
month ethnicity by month sex by age census region MSA status month by day of week
Trim outlier expansion factors max expfac = 4 times mean Replication (correct for sampling error)
bull At least two passes through the data
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Statewide Model Data Sources (Contrsquod)bull Dun amp Bradstreet (DampB) employment data
bull Employment Security data (ES-202)
bull Freight Analysis Framework (FAF)
bull Highway Performance Monitoring System (HPMS)
bull InfoUSA employment data
bull Metropolitan Planning Organizations (MPO)
bull NCHRP 365 Manual
bull National Highway Planning Network (NHPN)
Source FAF2 Provisional Commodity Origin-Destination Estimates Revised Methodology Report BattelleFHWA
Statewide Model Data Sources (Contrsquod)
bull National Household Travel Survey (NHTS)
bull Origin-destination surveys
bull Regional Economic Models Inc (REMI)
bull State Departments of Transportation
bull TRANSEARCH data
bull Vehicle Use and Identification Survey (VIUS)
bull Woods amp Poole employment forecasts
Source Bureau of Transportation Statistics NHTS
Data Sources on Long-Distance Travel
bull Census Journey-to-Work databull Long distance component
American Travel Survey (77 95) National Household Travel Survey (01) Ohio statewide surveys (02-04) Michigan statewide survey (04-05) California (10-12)
bull Surveys focused on long distance Oregon (96-98 07-08)
bull Others BTS Omnibus surveys Tourism surveys European practice NCHRP Report 735
NHTS and Long-Distance Travelbull Evolution of US national surveys
NPTS (69 77 83 90 95) ATS (77 95)
bull 2001 NHTS ndash Total of 60282 long distance person trips Data collected March 01 to May 02 22204 (37 percent) before 911 38078 (63 percent) afterwards
bull Long distance trip is longer than 50 miles one-way (vs 100 miles in 1995 ATS)
bull 28-day retrospective survey (including travel day)
bull Not your typical travelers Low frequency Distinct segment of the population
Weight Calculation for Long-Distance Travelers
bull Use to ensure valid population-level estimates Accounts for non-response under-coverage multiple phones etc
bull Multi-stage process Match national margins Adjust for data splitting (force sumexpfac = 277208169) Raking (match independent controls) race ethnicity race by
month ethnicity by month sex by age census region MSA status month by day of week
Trim outlier expansion factors max expfac = 4 times mean Replication (correct for sampling error)
bull At least two passes through the data
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Statewide Model Data Sources (Contrsquod)
bull National Household Travel Survey (NHTS)
bull Origin-destination surveys
bull Regional Economic Models Inc (REMI)
bull State Departments of Transportation
bull TRANSEARCH data
bull Vehicle Use and Identification Survey (VIUS)
bull Woods amp Poole employment forecasts
Source Bureau of Transportation Statistics NHTS
Data Sources on Long-Distance Travel
bull Census Journey-to-Work databull Long distance component
American Travel Survey (77 95) National Household Travel Survey (01) Ohio statewide surveys (02-04) Michigan statewide survey (04-05) California (10-12)
bull Surveys focused on long distance Oregon (96-98 07-08)
bull Others BTS Omnibus surveys Tourism surveys European practice NCHRP Report 735
NHTS and Long-Distance Travelbull Evolution of US national surveys
NPTS (69 77 83 90 95) ATS (77 95)
bull 2001 NHTS ndash Total of 60282 long distance person trips Data collected March 01 to May 02 22204 (37 percent) before 911 38078 (63 percent) afterwards
bull Long distance trip is longer than 50 miles one-way (vs 100 miles in 1995 ATS)
bull 28-day retrospective survey (including travel day)
bull Not your typical travelers Low frequency Distinct segment of the population
Weight Calculation for Long-Distance Travelers
bull Use to ensure valid population-level estimates Accounts for non-response under-coverage multiple phones etc
bull Multi-stage process Match national margins Adjust for data splitting (force sumexpfac = 277208169) Raking (match independent controls) race ethnicity race by
month ethnicity by month sex by age census region MSA status month by day of week
Trim outlier expansion factors max expfac = 4 times mean Replication (correct for sampling error)
bull At least two passes through the data
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Data Sources on Long-Distance Travel
bull Census Journey-to-Work databull Long distance component
American Travel Survey (77 95) National Household Travel Survey (01) Ohio statewide surveys (02-04) Michigan statewide survey (04-05) California (10-12)
bull Surveys focused on long distance Oregon (96-98 07-08)
bull Others BTS Omnibus surveys Tourism surveys European practice NCHRP Report 735
NHTS and Long-Distance Travelbull Evolution of US national surveys
NPTS (69 77 83 90 95) ATS (77 95)
bull 2001 NHTS ndash Total of 60282 long distance person trips Data collected March 01 to May 02 22204 (37 percent) before 911 38078 (63 percent) afterwards
bull Long distance trip is longer than 50 miles one-way (vs 100 miles in 1995 ATS)
bull 28-day retrospective survey (including travel day)
bull Not your typical travelers Low frequency Distinct segment of the population
Weight Calculation for Long-Distance Travelers
bull Use to ensure valid population-level estimates Accounts for non-response under-coverage multiple phones etc
bull Multi-stage process Match national margins Adjust for data splitting (force sumexpfac = 277208169) Raking (match independent controls) race ethnicity race by
month ethnicity by month sex by age census region MSA status month by day of week
Trim outlier expansion factors max expfac = 4 times mean Replication (correct for sampling error)
bull At least two passes through the data
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
NHTS and Long-Distance Travelbull Evolution of US national surveys
NPTS (69 77 83 90 95) ATS (77 95)
bull 2001 NHTS ndash Total of 60282 long distance person trips Data collected March 01 to May 02 22204 (37 percent) before 911 38078 (63 percent) afterwards
bull Long distance trip is longer than 50 miles one-way (vs 100 miles in 1995 ATS)
bull 28-day retrospective survey (including travel day)
bull Not your typical travelers Low frequency Distinct segment of the population
Weight Calculation for Long-Distance Travelers
bull Use to ensure valid population-level estimates Accounts for non-response under-coverage multiple phones etc
bull Multi-stage process Match national margins Adjust for data splitting (force sumexpfac = 277208169) Raking (match independent controls) race ethnicity race by
month ethnicity by month sex by age census region MSA status month by day of week
Trim outlier expansion factors max expfac = 4 times mean Replication (correct for sampling error)
bull At least two passes through the data
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Weight Calculation for Long-Distance Travelers
bull Use to ensure valid population-level estimates Accounts for non-response under-coverage multiple phones etc
bull Multi-stage process Match national margins Adjust for data splitting (force sumexpfac = 277208169) Raking (match independent controls) race ethnicity race by
month ethnicity by month sex by age census region MSA status month by day of week
Trim outlier expansion factors max expfac = 4 times mean Replication (correct for sampling error)
bull At least two passes through the data
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Key Statistics for Long-Distance Travel
bull Incidence amp frequency (trip rates)
bull Mode of travel
bull Trip duration
bull Trip purpose
bull Party size
bull Placetype of stay
Raw WeightedDid not stay overnight 9901 1476180913 537 537Friend or relatives home 3300 528190020 192 729Hotel motel bed amp breakfast resort 3049 463335160 168 897Camper trailer tent or other recreational vehicle 637 87821810 32 929Owned cabin condominium vacation home timesha 554 78526719 29 958Rented cabin condominium or vacation home 432 61616966 22 98Overnight in automobile plane ship train etc 116 16599669 06 986Dormitory youth hostel 81 13850288 05 991Military housing 37 6980894 03 994Other 52 8830037 03 997Corporate owned housing 26 5723523 02 999Conference center for participants only 16 2188649 01 100YMCA 2 241128 0 100
Total 18203 2750085777 100
Per-centLodging at farthest destination
NHTS observations Cumu-lative
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Topic ndash Validation amp Reasonableness Checking Transferability
bull Likely range of values for SWMsbull NCHRP Report 735 Long-Distance
and Rural Transferable Parametersbull SWM reasonableness comparisons
Trip generation Aggregate trip rates Trip purpose percentages
Trip distribution Average trip lengths Percent intrazonal trips Analyze district-to-district trip movements
Mode choice Average auto occupancy rates Assess reasonableness of mode shares (for
multimodal models)
Trip assignment Root mean square error Volume-over-count ratios
Source Florida Dept of TransportationCambridge Systematics Inc
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Likely Range of Values for Statewide Models
bull Varies significantly Size of state Density of development
bull Network links per TAZbull Population per TAZ
bull Model checks against Census Persons per household Employees per household Autos per household
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
AL AZ DE FL 2000 IN KY LA MD MA MS MO NJ
Number of Zones 1081 1073 908 4000 4720 4753 1313 1739 3069 3305 2392 2813
Number of Links 4607 6412 10047 - 34500 77272 - 167150 - - - 44000Ratio of Links per Zone 426 598 1106 - 731 1626 - 9612 - - - 1564
Persons per Household or DU 226 226 228 219 24 231 242 247 242 245 229 254
EmploymentPopulation 043 043 049 044 049 045 042 05 05 042 048 047
AutosHousehold or DU 16 119 15 138 168 157 14 156 144 156 156 148
PopulationTAZ 4113 5611 863 3995 1288 850 3403 3045 2068 860 2339 2991
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
NCHRP Report 735 Long-Distance amp Rural Transferable Parameters for Statewide Models
bull Differences in rural and long-distance travelbull Statewide model statistics on rural and long-distance travelbull Transferability of rural and long-distance model parametersbull Consideration of other trip characteristicsbull Developed recommended
transferable model parameters 1995 ATS 2001
NHTS 2009 NHTS Statewide Surveys GPS Surveys
0102030405060
Percent Of Trips Percent Of Miles
Figure 1 Vehicle Trips and VMT by Trip Length
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Statewide Model Reasonableness ndash Trip Generationbull Aggregate trip rates
Provided for only a few statewide models Differs little from urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
As documented in 2008 Model Calibration and Validation Standards Final Report for Florida DOT (results from models throughout the US)
NCHRP Report 365 Travel Estimation Techniques for Urban Planning 1998 (tripsHH)
Statewide Model Results Guidance Documents
AZ CAFL
2000 MA UT VA WIFDOT Low
FDOT High
NCHRP Low
NCHRP High
Person TripsTAZ 27213 - 13026 7489 2134 16197 12788 15000
Person TripsPerson 485 195 326 362 308 362 447 33 4 - -
Person TripsHousehold (DU)
1095 541 821 94 894 95 1033 8 10 68 124
HBW Person TripsHousehold
132 - - 173 138 138 173 - - - -
Person TripsEmployee 1136 441 738 726 654 723 876 - - 129 14
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Statewide Model Reasonableness ndashTrip Generation
bull Trip purpose percentages
Less trip purposes than found in typical urban amp regional models
Percentages by purpose not that different from urban amp regional models as long distance trips are overshadowed by others
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AL AZFL
2000 LA MA MS UT
Home-based Work 1919 1333 1914 1900 1841 1606 1550Home-based Nonwork 4622 - 5547 4770 5206 4917 4510
Home-based Other - 5513 2792 4770 1938 4917 4510Home-based Shop - - 1576 - 1173 - -
Home-based SocialRecreation - - 1179 - 1395 - -
Home-based School - - - - 700 - -Nonhome-Based - 3002
Nonhome-based Work - - - - 1027 - -Nonhome-based
Nonwork - - - - 1441 - -Other 2180 - - - - - -Truck 779 - - 170 - - -
Long-Distance - 151Long-Distance
Business - - - 020 284 011 -Long-Distance Tourist - - - 010 027 000 -
Long-Distance Work - - - - 105 - 030
Long-Distance NonworkOther - - - 030 - 011 060
Internal-External Business - - - - - 044 -
Internal-External Tourist - - - - - 044 -
Internal-External Other - - - - - 077 -
External 075 - - - 067 - -
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Statewide Model Reasonableness
ndash Trip Distribution
bull Average trip lengths (minutes)
Vary by state Not that different
from MPO short distance purposes
However long-distance purposes exceed maximum trip lengths found in urban amp regional models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ FL 2000 LA MA RI TX UT VA
Home-based Work 261 1838 173 2305 2159 16 1779 146
Home-based OtherNonwork 202 1622 122 1457 1564 906 1298 108
Home-based Shop - 1677 - 1506 - - - -
Home-based SocialRecreation - 1555 - 1619 - - - -
Home-based School - - - 1393 - - - -
Nonhome-based 186 1306 93 - 1445 968 1237 2317
Nonhome-based Work - - - 1791 - - - -
Nonhome-based Other - - - 1562 - - - -
Other - - - - - 1083 - -
Long-Distance Business 174 13467 166 - - - - 12713
Long-Distance Personal Business - - - - - - - 12458
Long-Distance Tourist - - 172 - - - - 12667
Long-Distance Work - - - - - 20082 8954 -
Long-Distance NonworkOther - - 164 - - 19971 8173 -
Internal-External Business - - - - - - 5126 -
Internal-External Other - - - - - - 5549 -
Internal-External Short Distance - 4282 - - - 30 - -
Internal-External Long Distance - - - - - 12575 - -
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Statewide Model Reasonableness ndash Trip Distribution
bull Percent intrazonaltrips
Considerably higher than urban amp regional models
Result of larger TAZs found in statewide models
Trip PurposeFL
2000 MA PA UT WI
Home-Based Work 94 63 - 55 292
Home-Based Nonwork 174 - - - -
Home-Based Shop 156 165 - - 410
Home-Based SocialRecreation 373 153 - 16 374
Home-Based School - - - - 515
Home-Based Other 252 170 - 85 536
Nonhome-Based - - - 83 544
Nonhome-Based Work - 184 - - -
Nonhome-Based Other - 192 - - -
TOTAL 199 - 377 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Statewide Model Reasonableness ndash Mode Choice
bull Average auto occupancy rates
Similar to urban amp regional models for short-distance trips
Much higher for long-distance trips unique to statewide models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide Model Results
Trip Purpose AZ CA FL 2000 KY LA MS RI TX UT
Home-based Work 119 - 11 - 115 11 112 - -
Home-based Nonwork - 119 - - - - - - -
Home-based Other 18 154 17 - 178 165 156 - -
Home-based Shop - - 18 - - - - - -
Home-based SocialRecreation - 149 194 - - - - - -
Nonhome-based 167 - 171 - 179 156 156 - -
Work - - - - - - - 133 -
Non-Work - - - - - - - 206 -
Long-Distance Business 165 119 - 18 186 139 - - 182
Long-Distance Tourist - - - 331 344 255 - - 269
Long-Distance Work - - - 243 - - - - -
Long-Distance Rec - 173 - - - - - - -
Long-Distance NonworkOther - 131 - - 264 205 - - 269
Internal-External Business - - - - - 15 - - -
Internal-External Tourist - - - - - 255 - - -
Internal-External Other - - - - - 226 - - -
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error Not commonly reported for statewide models Areawide RMSE similar to urban amp regional models RMSE for low volume roadways of which there are plenty in statewide models generally fails to meet
urban model accuracy standards Volume groups vary among models
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
10000-25000 and 25000+ ranges used in Texas model reference
25000+ range used in Alabama model reference
50000+ range used in Tennessee model reference
Statewide Model Results Guidance Documents
Range AL AZ FL 2000 IN TN TX UT WIFDOT
AcceptableFDOT
Preferable Oregon DOT Michigan
1 - 5000 1418 1036 6094229-9906
272-218 90-290 57-147
4182-8061 100 45 11576 50-200
5000 -10000 807 569 434
3552-4297 159 70 53-76 32 45 35 4314 25
10000 -20000 655-824 367 327
3133-356
254-355 610
340-810 223 300-350 250-270 2873 20
20000 -30000 571 275 259 2934 202 400 45 1933 27 15 2584 20
30000 -40000 362 - 214 2193 182 400 36 1362 25 15 3025 15
40000 -50000 - - 174 1574 88 - 36 1391 25 15 3025 15
50000 -60000 - - 145 51 - - - 20 10 3025 10
gt 60000 - -109-149
1322-146 - - 41 - 19 10 192 10
Total 822 56 326 3942 - 90 49 3921 45 35
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Statewide Model Reasonableness ndash Trip Assignment
bull Percent root mean square error (contrsquod)
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
0
20
40
60
80
100
120
140
160
180
200
0 10000 20000 30000 40000 50000 60000 70000
Perc
ent R
MSE
Midpoint Volume Range
Statewide Model RMSE Distribution by State
AL AZ FL
IN MD MI
OH OR TN
TX UT WI
FDOT Acceptable FDOT Preferable Oregon DOT
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Statewide Model Reasonableness ndash Trip Assignment
bull Volume-over-count ratios
Similar to urban amp regional models
Infrequently summarized by urban vs rural area type
This limits ability to fully assess model accuracy
Facility Type ALFL
2000 MA MI TX WI
Interstate - - - 100 084 094
Freeway 101 098 101 - - 108
Expressway - - - - - 105
US Highway - - - - 103 -
State Highway - - - - 055 -
Arterials - - 097 - - -
Prinicpal Arterial 105 - - - - 099
Major Arterial - 101 - 099 - -
Minor Arterial 094 103 - 103 - 106
Collector 087 101 097 - - -
Major Collector - - - 088 - -
Minor Collector - - - 095 - -
Source NCHRP 8-36B Task 91 Validation and Sensitivity Considerations for Statewide Models Cambridge Systematics Inc
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Statewide-Urban Model InterfaceRun the statewide model
Run the urban model
Synthetic matrix estimation
Convergence
y
n
targets and constraints
estimators
data
link flows
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Data Sharing with Urban and Statewide Models
bull Pros of data sharing
Minimizes cost of data developmentreview
Achieves greater consistency with regional planning efforts
bull Cons of data sharing
Requires continuous coordination to maintain consistency with land use forecasts and status of planned projects
Methodologies such as network coding and socioeconomic forecasting will vary throughout statewide model
Source Florida Department of Transportation
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Statewide Model Independent of MPO DataProcessbull Pros
Model updates can be made more expeditiously
Common methodologies can be employed throughout the model Unified land use forecasting
model Common set of network
attributes Matrix estimation alternative
Set of forecasts independent from local politics
bull Cons
Could result in inconsistencies with regional planning
Results could be unrealistic
Source Florida Department of Transportation
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Additional Considerations on External Tripsbull In addition to using statewide models to forecast external trip control
totals for urban and regional modelshellip
Use statewide model as source for E-II-E and E-E trip tables Would require disaggregation of statewide trip table to represent internal urbanregional
model TAZs Would require aggregation of statewide trip table to represent external urbanregional
model TAZs Would require adjustments to account for statewide model assignment errors at external
crossings for urbanregional models
Source Tennessee Department of TransportationCambridge Systematics Inc
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Other Potential Uses in Urban and Regional Models
bull In addition to using statewide models as a source for external trips in urban and regional modelshellip
Time-of-day estimate (assuming statewide time-of-day capabilities)
Vehicle class definitions (most statewide models load truck trips as a separate trip purposepre-load)
External transit trips (assuming statewide multi-modal capabilities)
Source Tennessee Department of TransportationCambridge Systematics Inc
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Topic ndash Statewide Model Shortcomings amp Areas to Improve
bull Travel behavior data on visitors and long-distance trip makers
bull Trip generation rates and other model parameters for rural residents
bull Origin-destination data at state line external zones
bull Consistency between network and zone system
bull Daily and seasonal variability
bull Other shortcomingsSource USDOT Bureau of Transportation Statistics American Travel Survey
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Topology
Internal-external (EI)
Through (EE)
External-internal (EI)
Internal (I)
Visitor (non-resident long distance)Resident (long distance travel)
Resident
Strangers (non-resident long distance)
Mindset Boundaries rarr Behavioral
Modeled area
Visitors
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
American Long Distance Personal Travel Program
bull American Long-Distance Personal Travel Data amp Modeling Program identified
bull Goal is to support data on long distance personal travel Collection Synthesis Analysis
bull Coverage areas Within into and out of US
bull Carry out design and application of models
bull FHWA national origin-destination matrix by mode Study recently completed
bull FHWA Exploratory Advanced Research Program Design of new approach for
long-distance travel survey approachinstrument and model framework underway
Source Oak Ridge National Laboratory University of Maryland
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Rural Trip Rates and Other Model Parameters
bull Most household travel surveys focus on urban trip-making
bull Rural sample from National Household Travel Survey
bull NHTS Add-On surveys for specific states
eg Florida Add-on targeted a sample of rural residents
Source Florida Dept of TransportationCambridge Systematics Inc
Rural counties depicted in gold
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Origin-Destination Data at State Line External Zones
bull State line crossings analogous to MPO external zones
bull Likewise IEEE split is needed (through trips)
bull Unlike MPO models need trip patterns to major internal destinations
bull Also helpful to know trip purposes and vehicle classes (HOV truck etc)
bull Some potential for anonymous cellular data but long tracking distances could be challenging
bull GPS data for truck travel (ATRI) Source Georgia DOT Cambridge Systematics Inc
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Consistency Between Network and Zone System
bull Statewide model transportation networks are typically too dense (rarely too sparse)
bull Statewide zone systems to support dense networks = long run times
bull Resulting statewide model zones often too large for corridor validation
bull Nested zone structures might hold promise
Source Ohio Department of Transportation
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Daily and Seasonal Variability
bull Statewide models usually represent something akin to AADT (MPO models more likely to reflect peak conditions)
bull Peak seasons could vary across the state
bull Major roadways outside MPO areas might even experience peak traffic conditions during weekends
bull Normalizing required
Source New Jersey Institute of Technology
RT-HIS NPTA
Weekday Weekend Weekday Weekend
Sample Size (Number of
Households) 4541 275 321 128
Estimated Mean (Number of Trips per HH) 880 771 1035 853
Difference between Weekend and Weekday 04 14
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
Other Potential Shortcomings of Statewide Models
bull High cost of maintenance
Might require personnel with broader skills (GIS econometricfreight models decision support systems)
Ongoing data collection (beyond typical MPO needs)
bull Vague or poorly defined goals and objectives (unlike MPO models not statutorily required)
bull Often developed with single purpose in mind (eg conduct statewide plan)
Source Tennessee Department of Transportation
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
MiniQuiz 3bull Source(s) for Long-Distance Travel Survey Data
A Select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above
bull Best source(s) for Rural Travel Survey DataA 2009 NHTS and select statewide household surveys
B 2001 NHTS
C 1995 ATS
D All of the above