170 likes | 313 Views
UrbanSim at PSRC. UrbanSim Users Group April 29 th -May 1 st 2009. The last 16 months. Worked the edges of the models – adjustments and refinements, not major additions 1 st implementation of UrbanSim contains many ‘Good enough for now’ decisions ‘Wish list’ for next time around.
E N D
UrbanSim at PSRC UrbanSim Users Group April 29th-May 1st 2009
The last 16 months • Worked the edges of the models – adjustments and refinements, not major additions • 1st implementation of UrbanSim contains many ‘Good enough for now’ decisions • ‘Wish list’ for next time around
The Challenges we encountered • Getting familiar with the models and software • Developer Model • Infill and redevelopment • Translating comp plans to constraints • Matching constraints and templates • Location Choice Models • Adds and edits to the specifications during validation • Balancing vacancy rates and HLCM results • Job capacity – turning buildings into cubicles • Future Year Land Use inputs and scenario changes • How best to make changes to constraints table - GIS or DB queries • Unplaced HHs, Jobs, and Development Capacity • We ran out of room…relaxed constraints to allow for sensitivity tests
Validation Work • 2000-2006 validation provided framework for most of the adjustments & discoveries
Vacant Redevelopable Partially Vacant / Underutilized Parcels available to the Developer Model • Underutilized (Infill) • Function of current density versus zoned or comp plan density – started with 2.5 ratio threshold: • A 30,000 SQFT parcel has a single housing unit, a 1200 SQFT single-story ranch house. The existing density value is 1.45 DU/Acre – 43,560 SQFT/30,000 SQFT. • Current comp plan designation calls for a maximum of 4 DU/Acre – 4 / 1.45 = 2.75, exceeds 2.5 and therefore meets our criteria for under-developed. • Lot size of 4/DU per acre is 43,560/4 = 10,890 SQFT. This is applied to the housing unit’s building record in UrbanSim, leaving 30,000-10,890 = 19,110 SQFT as vacant land area . • Redevelopable • Valuation ratio – currently using improvement value (building value) over total value (land + building) • Currently using 0.33 as trigger for a redevelopable property 7
Developer Model CalculationsExample 1 • Vacant Parcel # 1104255 - Arlington • 10,019 SQFT • 6 to 24 DU/Acre MF/Condo Only • Builds Condo building in 2030 with 4 units
Developer Model CalculationsExample 2 • Vacant Parcel # 1047486 - Uninc. Snohomish County • 19,602 SQFT • 4 to 6 DU/Acre - SF and MF/Condo allowed • Builds 2 SF units in 2023
Real Estate Price ModelResearch and Design Questions • What other variables should we be attempting to model? • Incorporate vacancy rates and allowable development (land use plan designation? • What travel measures to use? • Generalized Costs – most future year assumptions involve tolls • Continuous versus threshold measures • Separate models for land value, improvement value? • Different factors for each?
Developer ModelsResearch and Design Questions for Project Sampling / Selection portion • How vacancy rates influence the market • Currently used as a ‘on/off’ switch – change to an adjustment to the expected return on investment? • Add ability to model cost incentives / disincentives on development by type, location • IE policy to encourage affordable housing in urban centers • Are the ROI values realistic? • Negative profits • Correct spread between higher density and lower density proposals on the same parcel?
Developer ModelsResearch and Design Questions for Project Proposal portion • How to model comp plans? Strict versus loose • Should constraints degrade over time? • How to stock the development_templates table? • Historical patterns, future designs, mix? • Ensure templates match up with constraints • How best to flag parcels for infill and redevelopment • Varies among jurisdictions now for Buildable Lands program • Improvement value should be high enough to prevent parcel from re-entering the redevelopment cycle for period of time (the ‘recycled parcels’ issue) • Allow developer model to aggregate / disaggregate parcels • Bending reality in template lot size and maximum FAR for high-density areas
Examining Sensitivity to Transportation Plan / Accessibility Changes
King County Pop Change 2000 to 2040: Each Alt compared to baseline
Next 6-9 Months • Consolidation Phase • Install and test GUI, built-in synthesizer • Write some of this stuff down • Model Refinements • Tune infill assumptions • How model handles redevelopment (no ban) • Thorough 'price model' review and adjustments • Better relationship between cost of housing and HH income in HCLM aka 'affordable housing' • Rural Vested Lots / allow DU's on undersized parcels • Explore segmented or nested HLCM structure • Scale model improvements in ELCM • Cost incentive levers (instead of only DU, FAR constraints) • Updated Inputs • Future Year land use, Major planned developments, 2006 ‘Light’ base year • Use for forecast product • Model our Vision 2040 plan • Continue to support Transportation 2040 analysis • Grow comfort level with member jurisdictions
Lessons learned so far • Pull on those threads…where there is smoke, there is fire…<choose your cliché> • Value of the validation cycle in uncovering needed model refinements • Troubleshooting takes time, error messages aren’t yet particularly helpful for non-programmers • Think multiple models & inputs • Example: What can be built on a parcel dependent on both the comp plan, and what templates UrbanSim can build given the parcel characteristics and comp plan translation • Code is fragile – update with care • Advantages of the parcel model • How development actually occurs – represent land economics • Modeling redevelopment, infill critical for representing GMA-related growth policies