From One to Many Evolving the models and products to meet the needs at PSRC 2014 COG/MPO Mini-Conference on Socioeconomic Modeling July 18, 2014
Dec 29, 2015
From One to Many
Evolving the models and products to meet the needs at PSRC
2014 COG/MPO Mini-Conference on Socioeconomic Modeling
July 18, 2014
2
Outline• In the beginning…
• Life was simple
• New Demands from Modeling• Growth Management Planning• Parcel focus• Activity Based Models
• Expanded Products• Modeling emerging policy directions
• Recasting the Model & Products• Swiss Army Knife
Central Puget Sound Region
Area: 6,300 mi²16,300 km²
(16% urban)
As of 2011: Population 3,715,650
Jobs 1,853,900
Largest City Seattle – 612,000
Smallest City Index - 180
4 Counties
82 Municipalities
4
In the Beginning• Top-down, two-model structure
• Regional Forecasts• Zonal allocation• “Small Area Forecasts” – mix of
• Modeled output• Reviewer comments• Interpretations of policies• Assumptions of future plans and projects
• New Demands from Modeling• Growth Management Act in 1990
• Focus on land use plans and policies• Explicit representation of comp plans and Urban Growth
boundaries needed• Support AB Modeling
55
PSRC Modeling Suite – circa 2010-2013
Travel Forecasts – PSRC Travel Demand Models
Benefit-Cost Analysis Tool
Transport System - GeoDatabase
Air Quality Analysis – EPA
MOVES
Land Use Model - UrbanSim
Land Development Models
Household Location Models
Employment Location Models
Workplace Location Models
Regional Economic Forecasts – ECO Model
US Forecast (Exogenous Input)
Regional Forecast Model
Modeling emerging policy
6
Previous GMA Cycle Current GMA Cycle
CurrentComp Plans
CPPs Targets 2022/2025
OFM Projections
VISION 2040
Updated CPPs
Comp Plansdue 2015/16
Targets 2030-2035
OFM Projections
Land Use Forecast
Local Targets Representation
VISION 2020
Original concept = compare Forecast to emerging policy, a “Gap Analysis”
Forecast Products Package Two new future land use datasets:
1. Land Use Forecast- New land use forecast developed using PSRC’s UrbanSim
model
2. Local Targets Representation- Companion future land use dataset based on local 2030-2035
growth targets developed to align with VISION 2040’s Regional Growth Strategy
Additional resources:
3. VISION 2040 Gap Analysis- Comparison of Land Use Forecast and V2040 Regional Growth
Strategy
4. Planning Guidance & Technical Assistance
7
8
Three Months Later – Renaming….
9
… there’s still a void
Product for travel modeling, other planning uses
• Consistent with VISION 2040 (Regional Growth Strategy)
• Updated to reflect the 2012 Regional Economic Forecast
• Years: Horizon of 2040 with interim years available
• Consistency with locally adopted Growth Targets
10
A Third Product Land Use Baseline Land Use Targets Land Use Vision
What It Represents:
The region’s predicted development pattern based on current pre-VISION 2040 local
comprehensive plans and development regulations (circa
2012)
A future land use and development scenario based
on county/local growth targets developed to align with the
VISION 2040 Regional Growth Strategy
Future land use scenario consistent with the VISION
2040 Regional Growth Strategy, updated by the 2012 Regional Economic Forecast, and informed by local targets
Model: UrbanSim Allocation Method Allocation MethodRegional Forecast Assumption:
2012 Regional Economic Forecast
2005 Puget Sound Economic Forecast/ 2006 Small
Area Forecast
2012 Regional Economic Forecast
Data Variables:
- Total population- Group quarter population (by institutional/non-institutional)- Household population- Households (by income quartile)- Employment (by major sectors)
Geography: FAZ, CT, city/uninc’d urban/rural
Base Year: 2000 2010 2010Interim Years: Decadal through 2040 None 5-year intervals
Horizon Years: 2040 2025, 2030, 2031 & 2035 2040
11
2014 Forward: Recasting the Platform
• UrbanSim as the Swiss Army Knife (DRAFT)
Simulation Mode
Allocation Mode
Baseline Projections
Scenario Analysis
Sub-Regional Control Totals
Output Refinements
12
Redefining the Products
• Product concept (DRAFT – still to be vetted)
Simulation Mode
Allocation Mode
Baseline Projections
Scenario Analysis
Sub-Regional Control Totals
Output Refinements
Control run – aka “do nothing” alternativeNo post-processing, “it is what it is”
Inform thru comparisons to Control runOutput achieved thru policy & plan levers
Post-process to predefined totalsSynthesize for travel model
Generate policy-based distributionsReplace decision-rule allocation models
1313
Stress Test Concept:
Goal: Be in position this fall to demonstrate UrbanSim’s ability to support scenario analysis work
Reasoning: Land Use Baseline was the result of one set of assumptions (basically, the status quo)
It did not leverage the strengths of building a complex model with many “levers” that could be used to test policy outcomes.
We would like to test those levers to see how the model behaves – to build confidence in and support for the tool.
1414
Stress Test Approach:
Borrowing from ABAG’s approach in the Bay Area:
• Define & design Policy Levers
• Groupings of levers ultimately used to define tests
• Multi-round testing:• Initial Levers on extreme settings• Analysis of what worked, what needs work• Inform parallel model improvement work
15
Learned Lessons and Future Direction• Staffing
• Economic expertise• Inner-agency consortium (policy, modeling, outreach)
• Establish credibility prior to scenario analysis• Outreach, internal & external• Stress testing and model improvement• Expectations
• Review process• Re-assess = inputs instead of outputs? Break w tradition• Time & labor intensive – learning curve, Q&A,
implementation of changes
• Alternative inputs consensus• Straightforward to create scenario inputs, but are parcel-
level assumptions about future land use plans acceptable?
16
The review process – Draft Release• March 2012
• Workshop & documentation to explain basic UrbanSim workings & how to comment
• Focus on input correction & model improvementso Web mapping tool for reviewers
17
The review process – Revised Draft• December 2012
• Focus on both inputs & outputs• Begin implementing output Refinements - two ‘types’
• Year 2010 Refinement – attempt to correct for validation error• Forecast Refinement – adjust output in response to comments• Both processes – Zone adjustments, maintained regional
forecast totals.
• Use of Confidence Intervals for adjustment guidance
Forecast Products Package / Land Use Modeling Staff
Billy CharltonCarol NaitoRebeccah MaskinMark SimonsonHana SevcikovaPeter CaballeroMichael Jensen