UCIME: The Urban Change Integrated Modeling Interface
Keith C. ClarkeDepartment of Geography
UC Santa Barbara
Congratulations Mike Goodchild!
UCSB Geography’s second National Academy member
Funding
NSF Urban Research InitiativeLos Alamos National Laboratories (UCOP, NSF)USGS (Menlo Park)Santa Barbara Economic Community Project
www.geog.ucsb.edu/~kclarke/ucime
UCIME Objectives1. Develop a multi-scale infrastructure for
modeling the dynamics of US urban areas 2. Modeling and analysis of the growth
patterns of selected metropolitan regions 3. Predict land use intensity and population
distribution by coupling models of physical and human processes
4. Implement and disseminate an integrated GIS-based modeling environment for research and policy
The goal Informed, participatory local
decision making Simultaneous consideration of
regional and local issues Multi-scale consistency Focus on issues based
education
The foundation: Data Applications in many US cities Applications outside the US
(Australia, Portugal, Mexico, Brazil) Needed dense test case: Santa
Barbara Historical time series data Exclusions Land use Metropolitan data
The South Coast
Aerial Photos
Years: 1929, 1943, 1954, 1967, 1986, 1997
1929
Assessors’ Digital Parcel Map
Roads input
1929
1999
2005
GIS/EM: The integration challenge
Park & Wagner TGIS 1997
•Isolated•Loose•Tight•Integrated
Model integration
Share data from GIS Have common input/output layers Link inputs to outputs Have a single user interface (UCIME) Hide the models from the user Interact via scenarios (integrate via
planning/decision-making process)
Urban models in UCIME Population density structure SLEUTH (Urban form and land use) SCOPE By very loose coupling
Hydrology Air quality Wildfire hazard
SLEUTH archeology Clarke cellular automaton urban growth model
(UGM) Multiple applications (e.g. San Francisco,
Washington/Baltimore) Project Gigalopolis Applications: Chicago, New York, Portland,
Philadelphia, MAIA, Albuquerque, Detroit, Mexico City, Lisbon, Santa Barbara
1998/9 funding made model portable and web-based (USGS: EROS Data Center, EPA Collaboration)
1999-02 work extended and integrated model with other efforts (LANL and USGS collaboration, NSF Urban Research Initiative, SBECP)
EPA has provided significant input (MPI)
Gigalopolis: Project Goal Use historical data for urban areas to
understand present day urbanization Simulate using a Cellular Automaton
Model (SLEUTH) Run the model into the future Simulate alternative futures Compare across scale and cities Apply to Urban Dynamics cities
Cellular Automata Gridded world Cells have finite states Rules define state transitions Time is incremental Cells are autonomous, act as agents Self-replicating machines: Von
Neumann Classic example is Conway’s LIFE
Urban Cellular Automata Cells are pixels States are land uses Time is “units”, e.g. years Rules determine growth and change Different models have different rule sets Many models now developed, few
tested Requiem for large scale models (Lee)
Model tight couples land use change So far works at Anderson Level 1 and 2 Calibration for MAIA and Lower 48 States Needs two LULC layers Based on the concept of deltatrons Generates synthetic LU change based on
transition matrix and enforced spatial/temporal autocorrelation
Applies CA in change space
Why SLEUTH?
Slope
Land Cover
Excluded
Urban
Transportatio
n
Hillshade
1900 1925 1950 1975 2000
Project Web Site Set of background materials, e.g.
publications Documentation as web pages in HTML Model discussion list Source Code for model in C Version 3.0 now on web for download Uses utilities and GD GIF libraries Parallel version requires MPI Set of sample calibration data demo_city http://www.ncgia.ucsb.edu/projects/gig/ncgia.htmlhttp://www.ncgia.ucsb.edu/projects/gig/ncgia.html
Project Web Site: Shareware C code and Documentation
Calibration Most essential element Ensures realism Ensures accountability and
repeatability Tests sensitivity Required for complex systems
models Conducted in Monte Carlo mode
The Method “Brute force calibration” Phased exploration of parameter space Start with coarse parameter steps and
coarsened spatial data Step to finer and finer data as calibration
proceeds Good rather than best solution 5 parameters 0-100 = 101^5
permutations
Calibrationpast
“present”
For n Monte Carlo iterations
For n coefficient sets
Predicting the presentfrom the past
The Problem Model calibration for a medium sized data set and minimal data layers requires about 1200 CPU hours on a typical workstation
CS calls problem tractability
Implementations to date DEC Alpha Silicon Graphics (Indy 10000 and O2) Silicon Graphics Origin 2000 cluster 32
processors: 2GB RAM SunBlade 1000 Rolla, MS MCMC Beowulf Linux Cluster Supercomputers (NESC EPA: NC)
Cray C-90 and T3D Cray T3E-1200
SLEUTH Outputs Statistics Logs Images Uncertainty maps Animations
Prediction (the future from the present)
Probability Images
Land Cover Uncertainty
Alternate Scenarios
Model simulation
2040 Scenarios
User requests Animations Probability-free forecasts Detail! “Report cards” More attributes
SCOPE
South Coast Outlook and Participation Experience
Economic Community Project
Board has representatives of non-profits, businesses, community activists, local government, UCSB, etc.
Mission:Mission: To act as a catalyst for creating a sustainable regional planning process for the South Coast which will support both a viable economy and preserve and enhance the quality of life over the next twenty years and beyond.
Background Funding from Irvine Foundation, S.B.
Foundation, local cities and SB County ECP and UCSB collaboration Developed 4 land use principles first
SCOPE archeology Originally called UGROW: Will Orr,
Prescott College with NASA funding Rewritten in Powersim for Santa
Barbara Ported to STELLA by Jeff Onsted Designed using stakeholder focus
groups, intensive collaboration and public outreach
• HH size
• Vacancy rates
Lower inc.
Middle
Business/Jobs
• Workers• Students• Retirees
Lower inc.
Middle
Upper
• Wage class jobs• Commuters
• Unemploy rate
Service/Retail
Office/Lt Mfg
• Density• Housing• Business• Ag/Open
Population
Housing
Land Use
Land availability
Workeravailability
Attractiveness to in-migration
Attractiveness to in-migration
Development
Qualityof Life
SCOPE focuses on the interrelationships among five sectors
Trafficcongestion
Climate and
SettingDevelopment
Land availability
Upper
Rebuilding
Build affordable
Growth control
Trans. policy
Urbanlimit
Tax/subsidy Commuting
Housing availability
Stella version of SCOPE
Stella Interface(WebSIM): zenith.geog.ucsb.edu
Scenario BasisScenario Basis
• RedRed Unrestrained Development
• GreenGreen Urban Growth Boundary honored
• YellowYellow No Commercial growth and unrestrained residential
Total Population
Population Density model
P = f(Land Use, dRoads, Age, dCenter)
Tested for Santa Barbara with 2000 census dataGranularity = LU polygon INT census tracts
Used to make forecasts from future maps
The UCIME Web Interface
Data/World Scenarios
Users/Decision MakersIndividual
Group
Convergence
Scenario as model/plan bridge
Environmentalist
0
10
20
30
40
50
1 4 7 10 13 16 19 22 25
Question
Sce
nar
io D
ista
nce
Unrestrained
Follow Current
Road Growth
Growth Boundary
EnvironmentFriendly
No Commercial
Neutral Responses
0
10
20
30
40
50
1 4 7 10 13 16 19 22
Questions
Sce
nar
io D
ista
nce Unrestrained
Follow Current
Road Growth
Growth Boundary
Environment Friendly
No Commercial
Good scenario sets Themes can be single or multiple How many? 7+-2 Relevance: Policy implication Comprehensive (drivers) Diverse Creative: Role for Visualization Transparent Coherent: properly formulated and plausible Consistent
Scenario Difference
Conclusion Sets of models when integrated are more
powerful than when used alone, or when one metamodel is formulated
Users want more than model results Users want credibility in modeling/ers Users don’t want control of models A single user interface for multiple
models must use a macro view Scenarios are key to bridging models and
views