Sensitivity Testing with the Oregon Statewide Integrated Model (SWIM2)

1
Sensitivity Testing with the Oregon Statewide
Integrated Model (SWIM2)
Session 709 Integrated Land Use and Transport
Models in Practice

• 88th Transportation Research Board Annual
  MeetingJanuary 2009, Washington DC
• Presenters Brian Gregor/ODOT Tara Weidner/PB

2
Outline

• Background
• Oregon SWIM2
• Calibration
• Prep for Model Application
• Results
• Scenario 1 Increased Highway Capacity
• Scenario 23 Increased Driving Costs (4 and
  10-fold)
• Concluding remarks

3
Acknowledgements

• Sponsorship
• Oregon Department of Transportation (ODOT)
• Agency Lead
• ODOT Transportation Planning Analysis Unit
  (TPAU).
• Funding
• Federal Highway Administration.
• Consultant Team
• Parsons Brinckerhoff lead
• HBA Specto Inc.
• EcoNorthwest
• Special thanks
• Alex Bettinardi of ODOT

4
Why Integrated Models in Oregon?

• New state and federal mandates in early 90's
  required entire system view
• ODOT modelers wanted to remain relevant in Oregon
  regulatory and policy environment

1995 OMIP Oregon Modeling Improvement Program
1996 TLUMIP Transportation Land Use Model
Integration Project
5
The Oregon Path Application-driven research
SWIM1 Model

• TRANUS/Oregon (2000)

• WV Forum (2001)
• East/Central OR Fwy (2001)
• Bridge Options Study (2002)
• Newberg Bypass (2004)
• OR Transportation Plan (2005)

SWIM2 Model

• OR Amtrak Ridership (2008/2009)
• OR Climate Change (2009)
• OR Freight Study (2009/2010)
• MPO Co. Connection (2009/2010)

• Assembled (2004-2005)
• Calibrated (2008)

6
Oregon Statewide Integrated Model (SWIM2)
Economy
Employment by Industry
Construction Totals
Production Totals
Spatial Models
Space Inventory
Production Interaction
Land Development
Synthetic Population
HH Labor
Labor Flows
Commodity Flows
Person Transport
External Transport
Goods Transport
Transport Models
OD Trip Tables
Assignment
7
SWIM2 Calibration

• Challenges
• Multi-dimensional targets, inconsistencies,
  different years, varying confidence levels
• e.g. Commodity flow data, floor space price data,
  floor space quantity data
• Unavailable inputs must be synthesized, adds
  errors
• e.g. Floor space quantity data
• As a result, more emphasis placed on behavioral
  response than ability to tightly match all targets

8
Iterative SWIM2 Calibration Process

• Stage 1 Estimate module parameters where data
  exists
• Stage 2 Calibrate individual sub-models in
  isolation
• Stage 3 Calibrate full model, baseyear over
  time
• including Sensitivity Tests of likely policy
  scenarios
• 20-years, in 3-year increments
• unrealistically large change, to identify
  direction of change
• proof-of-concept scenarios test behavioral
  response to likely policy levers

9
Preparations for Model Application

• Prepare ODOT staff to perform policy analysis
• Knowledge Transfer
• Ongoing Agency Staff Training
• Evaluate Model Variability (micro-simulation)
• User Interface Tools
• Model Runner System GUI (scenarios, start runs)
• Output Processing tools (automate, compress)
• File Management tools (model files, output
  processing)
• Computer Cluster
• Purchase Hardware
• Backups/archiving processes

10
SWIM2 Geographic Coverage

• Oregon Activity
• 3.7M pop 1.8M jobs
• 6 MPOs

2,950 alphazones 519 betazones
11
Scenario 1 Increased Hwy Capacity
12
Increased Capacity 2024 Household Change (vs.
Reference)
Portland Salem Corvallis
Bend Eugene Rogue Valley
13
Increased Capacity Households
EUG
Significant Growth at South terminus with
improved Portland connection
Total Households
PDX
Some Growth mid-valley
CORV
SAL
RV
Less Growth outside valley
BEND
14
Increased Capacity 2024 Floorspace Change
15
Increased Capacity 2024 Transport Results

• Truck Trip lengths increase, but times decline
  (7)
• Auto Tour lengths, times, and mode do not change

16
Scenario 23 Increased Driving Costs
17
Increased Costs (10-fold) More Detail
Portland Salem Corvallis
Eugene
18
Increased Costs MPO Density Curves
Proportion of MPO Land Area
Persons Per Acre

• Regional Centers densities gain 20
• Other areas 0-10 density gains

19
Increased Costs (10-fold) Employment
SAL
Centralization to 3 Regional Centers
RV
Total Employment
Eugene gets Salem overflow
BEND
EUG
PDX
CORV
Non MPO areas lose growth
NON-MPO
Portland urban inertia overshadowed by Salems
central location
20
Increased Costs (4-fold) 2024 Floorspace
Results

• Floorspace created in new regional centers.
• High prices reflect difficulty in meeting
  residential needs.
• Nonresidential floorspace changes muted.

21
Increased Costs 2024 Transport Results

• Truck Trip lengths and times decline (5-10)
• Auto Tour lengths and times decline (30-40),
  more sensitive
• Significant person mode shift

22
2024 Person Trips by Selected Modes
Not all MPO offer Transit Held current service
levels
23
Concluding Remarks

• Success!!
• Extreme scenarios accommodated
• Good direction magnitude in all components
  (activity, land development, transport)
• Understandable geographical patterns
• Internally consistent, price signals working
• Transition to Agency Policy Application
• well underway

24
Some challenges remain

• Assumption of fixed economy/halo effects
• File management
• Summarizing Outputs/Improving Visualization
• Runtimes/convergence
• Additional Scenario Testing

Note Updated TRB paper on ODOT
website http//www.oregon.gov/ODOT/TD/TPAU/referen
ces.shtmlStatewide_Model