TRANSCAD MODELING AT NCTCOG: WHAT WE NOW HAVE

TRANSCAD MODELING AT NCTCOG:

WHAT WE NOW HAVE

ForUT-Austin/NCTCOG Meeting

January 28, 2004

AGENDA

Background – Modeling Environment

Trip Generation And Distribution

Transit Skims

Mode Choice

Transit And Traffic Assignment

Future Issues

MODELING ENVIRONMENT

NCTCOG-Developed FORTRAN Programs (Mainframe)MOBILE5A Emissions AnalysisLatest Mobility Plan UpdateFor Legacy Applications

TRANPLAN (PC)Subarea Traffic Modeling (Legacy Applications)

TransCAD (PC-Windows)MOBILE6 Emissions AnalysisFuture 2030 Mobility PlanAll New Travel Modeling Activities

SOFTWARE SELECTIONDECISION

We Are Tired Of Maintaining/Updating OurOwn FORTRAN Programs

It’s Difficult, Being Our Own Software Vendor!And It Costs A Lot Of Money To Rent Mainframe Time

Competitive Travel Model Software Procurement ProcessRFP Sent To Travel Model Software Vendors (1998)

Selection CommitteeNCTCOG, Texas DOT, DART, And City of Fort Worth Staff

TransCAD Selected From The Five Submissions

TRANSCAD MODEL SIZE

4874 Zones Retained For ALL Modeling StepsFrom Trip Generation To Traffic/Transit AssignmentNumber Of Zone-To-Zone Pairs = 23.8 Million

Year 2025: 27,000 Roadway Links+ 9,600 Zone Connectors

Over 36,000 Coded Links22,000 Network Nodes

2025 Transit410 Coded One-Way Bus Lines And 36 Rail Lines14,500 Bus Stops And 171 Rail Stations

TRAVEL DEMANDFORECASTING PROCESS

UrbanActivity

TripFrequency

DestinationChoice

ModeChoice

RoadwayRouteChoice

TransitRouteChoice

FOUR-STEP TRANSCAD MODELING PROCESS

TRIP GENERATION

TRIP DISTRIBUTION

MODE CHOICE

ROADWAY ASSIGNMENT

TRAVEL TIMECONVERGENCE

TRANSIT ASSIGNMENT

NO

YES

ZONE LAYERDEMOGRAPHIC INFORMATION

ROADWAY NETWORK

TRANSIT NETWORK

LOO

P

ROADWAY SKIMS

TRANSIT SKIMS

INPUT

PROCESS

DECISION

PROBLEMS WITH INTUITIVE“FULL TRAVEL PATTERN”

APPROACH

The Great Problem Of DataModeling Algorithm ShortcomingsCalibration DifficultiesModel Maintenance IssuesThe Greater Problem Of PredictionComputational Power Problem

TRANSITION TO THE REAL WORLD

Actual Scope ÚModel Scope– All Trips gMotorized Trips– All Purposes g HBW, HNW, NHB, And Truck Purposes– All Occupations g Basic, Retail, Service– All Households g Income Size Categories– All Streets g Non-Local Streets

Individual Data Ú Aggregate DataNeed For Adequate Samples

– In Household Survey gRAA level– In Loading The Trips gTSZ level

Chained Models ÚSeparate Models

CREATION OF 4,874-ZONE STRUCTURE

Start With Year 2000 Census Block LayerGDT RectificationSome Block Splits (e.g., DFW And Love Field Airport)76,336 Blocks Aggregated to 6,399 zonesAdd 61 External Station “Tiny Circle” Zones = 6,460 Total Zones

Aggregate The 6,460-Zone Layer= 4,813 Internal Zones + 61 External Station Zones= 4,874 TAP zones

MODEL RUN TIMES(FOR 3.2 GHz PC)

Full “No Feedback” Model Run = 647 Minutes (10.8 Hours)Trip Generation = 1.0 minuteRoadway Skimming (4) = 11 minutesTrip Distribution = 11 minutesMarket Segmentation = 6 minutesTransit Prep And Skimming (4) = 77 minutesMode Choice (13) = 65 minutesMatrix Preparations (For Transit Assignment) = 10 minutesTransit Assignment (4) = 21 minutesMatrix Preparations (For Traffic Assignment) = 98 minutesTraffic Assignment (3) = 347 minutes (5.8 hours)

TRIP GENERATION

GISDK Macro LanguageSeven “Regular” Internal-Internal Trip Purposes4 HBW, 1 HNW, 1 NHB, And 1 Truck

InputsPopulation, Households, Income, And Basic/Retail/Service JobsSpecial Generators (Shopping Malls, Colleges, Hospitals, Airports)

HOUSEHOLD INCOME DISTRIBUTION

0

0.2

0.4

0.6

0.8

1

1.2

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5

Average Zone Income / Regional Median Income

Inc

om

e G

rou

p P

ort

ion

s i

n Z

on

e

Qrtl. 1

Qrtl. 2

Qrtl. 3

Qrtl. 4

Total = 1

0.11

0.13

0.19

0.57

HOUSEHOLD INCOMEDISTRIBUTION ADJUSTMENT

• Enforcing 25% Of Households To Be In Each Quartile Requires Applying Iterative Proportional Fitting (IPF) To The Raw Distributed Households

HOUSEHOLD SIZE DISTRIBUTION

0

0.2

0.4

0.6

0.8

1

1.2

1 1.5 2 2.5 3 3.5 4 4.5 5 5.5

Avg. Zone Household Size

Ho

us

eh

old

Siz

e p

ort

ion

s i

n Z

on

e

HH1

HH2

HH3

HH4

HH5

HH6+

JOINT INCOME ANDHOUSEHOLD SIZE DISTRIBUTION

• Using The Regional Distribution As A Seed, We Create A Joint Distribution For Each Zone

0.3330.2500.3100.3660.3280.243QT4

0.2920.2870.2860.2970.2840.268QT3

0.2160.2450.2380.1990.2200.247QT2

0.1590.2190.1660.1390.1690.201QT1

HH6+HH5HH4HH3HH2HH1

TRIP PRODUCTION STRATIFICATION

Home-Based Work (Trips Made With Car Or Transit)4 Income Groups And 6 Household Sizes

Home-Based Non-Work (Trips Made With Car Or Transit)4 Income Groups And 6 Household Sizes

Non-Home-Based (Trips Made With Car Or Transit)4 Income Groups And 6 Household Sizes

Truck (Trips By Non-Transit Vehicles With 6 Or More Tires)4 Employment Types (Including Household) And 5 Area Types

TRIP PRODUCTION RATETABLE FOR HBW

4.4583.2132.8662.8662.1921.2884

3.1683.1682.7712.7562.1921.2883

3.5242.9082.5832.4911.9161.2882

2.0032.0032.3542.0821.3470.8701

6+54321

Household SizeInc.Q.

AREA TYPE CALCULATIONS

Based On RAA-Level Calculations:b = 1.59 (Population / Employment Ratio In 1999)ad = [pop + b*(bas + ret + ser)] / landArea30 <ad ==> Area Type = 27.5 <ad<=30 ==> Area Type = 31.8 <ad<=7.5 ==> Area Type = 4ad<=1.8 ==> Area Type = 5

CBD Zones Always Considered Area Type 1

Changes Based On Being An Island And Being CloseTo Thresholds

EMPLOYMENT INCOME DISTRIBUTION

The Distribution Of Employment Across Income Quartiles At Work Is Needed, To Prevent Misallocation Of Trips During Trip Distribution

Income Distribution At Workplace Based On Household Income And Number Of Nearby Employees

IPF Used To Adjust Distributions To Regional Ratios

TRIP ATTRACTION STRATIFICATION

Home-Based Work4 Income Groups, 5 Area types, And 3 Employment Types

Home-Based Non-Work4 Employment Types (Including Household) And 5 Area Types

Non-home-based4 Employment Types (Including Household) And 5 Area Types

Truck4 Employment Types (Including Household) And 5 Area Types

TRIP BALANCING

HBW Productions Balanced To Attractions

HNW Attractions Balanced To Productions

NHB Attractions Balanced To Productions And Then For Each Zone Production Is Set To Equal Attraction

Truck Attractions Balanced To Productions

DEMOGRAPHICS FORSPECIAL GENERATORS

Current Special GeneratorsRegional Shopping Malls With Over 500,000 Square Feet (20)Colleges/Universities With Over 1,500 Enrolled Students (27)Hospitals With Over 300 Service Employees (42)DFW And Love Field Airport Terminals (Special Treatment)

Special Input Data Fields (e.g., For Shopping Malls)SGRETAIL = Number of Jobs In Zone That Relate To

The Shopping MallSGUNIT = Total Leasable Square Footage (Thousands Of Square

Feet)HBW, HNW, NHB, And Truck Trip Rates Per Thousand Square

Feet

EXTERNAL STATION TRIP TABLES

Internal-External And External-Internal (IE/EI) Weekday Passenger Vehicles (Total Trip Ends)

External-External (EE) Weekday Passenger Vehicles

IE/EI Weekday Trucks (Six Or More Tires)

EE Weekday Trucks (Six Or More Tires)

TRIP DISTRIBUTION

Gamma-Format Gravity Model (7 Purposes)Four HBW Groups (Income Quartiles) – AM PeakHNW (Non-Airport) -- OffPeakNHB (Non-Airport) -- OffPeakTrucks (Vehicles With Six Or More Tires) -- OffPeak

Base Year Trip Table Factoring (6 Purposes)HNW And NHB Airport TripsFour External-Related Auto/Truck Trips

ZONE TO ZONE SKIM TABLESFOR MODE CHOICE

Four AM Peak Skims (6:30a – 8:59a)Roadway – Without HOV Links Available (Drive Alone)Roadway – With HOV Links Available (Shared Ride 2 And 3+)Transit – Drive Access (PA Format)Transit – Walk Access (PA Format)

Four OffPeak SkimsRoadway Is 18-hour Offpeak

Without HOV Links Available (Drive Alone)With HOV Links Available (Shared Ride 2 And 3+)

Transit Is 6-hour Mid-Day Offpeak (9:00a – 2:59p)Drive Access (PA Format)Walk Access (PA Format)

MODE CHOICE INPUTSAuto Travel

Roadway Travel TimeRoadway Length (Operating Cost)Daily Parking Cost

Transit TravelIn-Vehicle Transit Travel Time (Includes Dwell)Walk (Or Drive) Access TimeWalk Transfer And Egress TimeInitial And Transfer Wait TimeTransit FareMarket Segment And Area Type Constants

MARKET SEGMENTATION

Objective: To Account For Differences In Commuter Behavior

Segmentation Basis (For HBW And HNW)– Household Income (3)

• Low• Medium• High

– Vehicle Availability For A Household (2)• Vehicles Less Than Persons• Vehicles Greater Than Or Equal To Persons

HBW MODE CHOICE:MAINFRAME VS TRANSCAD

Choice

Drive Alone Shared Ride 3+Shared Ride 2 Transit Drive Transit Walk

Choice

Drive Alone Shared Ride 3+Shared Ride 2

Transit Drive Transit WalkAuto

HNW MODE CHOICE:MAINFRAME VS TRANSCAD

Choice

Drive Alone Shared Ride 2+ Transit Drive Transit Walk

Choice

Shared Ride 3+Shared Ride 2 Transit Drive Transit Walk

Shared RideDrive Alone Transit

NHB MODE CHOICE:MAINFRAME VS TRANSCAD

Choice

Drive Alone Shared Ride 2+ Transit Drive Transit Walk

Choice

Drive Alone Shared Ride 3+Shared Ride 2 Transit Drive Transit Walk

POST MODE CHOICE

13 Market Segments5 Mode SharesTrucks (From Trip Distribution)

PA To OD

Time Of Day Factoring And Aggregation

Vehicle Trip Table Factoring

TRANSIT ASSIGNMENT

Four Multi-Path (Pathfinder) Production-Attraction Assignments

For All HBW Transit TripsPeak Transit-Initial Drive Access (Park-and-Ride)Peak Transit-Initial Walk Access (No Park-and-Ride)

For All HNW And NHB Transit TripsOffpeak Transit-Initial Drive Access (Park-and-Ride)Offpeak Transit-Initial Walk Access (No Park-and-Ride)

TRAFFIC ASSIGNMENT PREPARATION

Trip Table AggregationAM Peak Period (2.5 Hours)PM Peak Period (3.5 Hours)OffPeak Period (18 Hours)

K FactoringCompensate For Gravity Model LimitationsOD Estimation Procedure To Help With IdentificationAdjustments/Checks Based On Screen Line Results

TRAFFIC ASSIGNMENT

User Equilibrium Generalized Cost (Three 30-Iteration Assignments)

A.M. Peak (6:30a – 8:59a: 2.5 hours)P.M. Peak (3:00p – 6:29p: 3.5 hours)OffPeak (18 hours)

Four Vehicle Classes Loaded SimultaneouslyDrive AloneShared-Ride “Sees” HOV LanesShared-Ride “Doesn’t See” HOV LanesTrucks (Vehicles With 6 Or More Tires)

SPEED VS V/C RATIO

0

10

20

30

40

50

60

70

0.10 0.40 0.70 1.00 1.30 1.60 1.90

V/C Ratio

Sp

eed

(m

ph

)

Assignment Speed Post Process Speed

POST-PROCESSING OF LINK SPEEDS

Example: AM Peak Directional Link Speeds

Allocate (Based on ATR Factors) The 2.5-Hour AM Peak Assignment Volume Into Three Sub-Periods

6:30a – 6:59a7:00a – 7:59a8:00a – 9:00a

Calculate V/C ratios For Each Sub-PeriodNote: Capacity for 30-minute period is ½ the hourly capacity

Apply The “Post Process” Volume Delay Curves

TRAFFIC MODEL LIMITATIONS(First Page)

Ideally, the congested speeds directly from traffic assignment should be used in trip distribution—but we “post process” because the calibrated parameters do not give us realistic speeds

A low “C” parameter in the volume-delay equation (maximum minutes of delay per mile) has important implications for high-congestion scenarios

Related to above: “Peak Spreading” is not directly considered; in the future, we may consider peak hour and “shoulder of the peak” assignments

TRAFIC MODEL LIMITATIONS(Second Page)

The Offpeak assignment represents 18 hours of the day—perhaps a future breakdown into Mid-Day Offpeak (9:00 a.m. –2:59 p.m.) and Evening/Night Offpeak (6:30 p.m. to 6:29 a.m.)

We have no observed data to directly calibrate HOV-Toll usage; instead, we have to rely on our calibrated “shared-ride trip table HOV/NoHOV factoring” and “toll road value of time” modeling

The “HOV/NoHOV” factoring procedure is just one way of “making things work”

Special link impedances could also have been used

TRAFFIC MODEL LIMITATIONS(Third Page)

All passenger vehicles are assumed to have the same value of time

Trucks (By Our Choice) are given a Passenger Car Equivalency of 1.0

“Real” link capacities are more complex than our look-up tables

CAUTION on number of UE assignment iterations: complete “closure” is NOT practical!

AUTOMATED NETWORK CONVERSION

WHAT HAPPENS NEXTFOR NCTCOG

One Key To Good Modeling Is The Quality Of The Overall Information System

Improvements/Updates To Modeling ProceduresExpansion Of The Modeled AreaDestination Choice Instead Of Gravity Model Trip DistributionLOGSUM ConsistenciesNew Travel Surveys In 2006Activity-Based Implementation (After Full Testing!)

Traffic MicrosimulationTransCAD Subarea Analysis To Create “Windowed” OD TableDallas CBD Project Underway (Using VisSim)