Puget Sound Region - Pima ASSOCIATION of Government

Download Report

Transcript Puget Sound Region - Pima ASSOCIATION of Government

Modeling and Data at the Puget Sound Regional Council:

(For a Few Dollars More…) COG/MPO Mini-Conference SANDAG Friday, July 29th, 2005 Kevin Murphy [email protected]

Transportation leadership you can trust.

Jeff Frkonja [email protected]

Mark Simonson [email protected]

Who We Are Membership

King, Kitsap, Pierce and Snohomish Counties

• • • • • •

70 cities 4 Ports Tribes State agencies 7 Transit agencies Associate members Over 3.4 million residents An estimated 1.9 million jobs

Challenges of Growth In 1950:

• •

1,200,000 People 500,000 Jobs In 2000:

3,300,000 People

1,900,000 Jobs By 2040:

• •

5,000,000 People 3,000,000 Jobs

What We Do Key Responsibilities

Long range growth, economic and transportation planning

• •

Transportation funding Economic development coordination

• •

Regional data Forum for regional issues

Decision-Making

Organization FY 2006-07 Budget:

• •

$6.6 Million DSA ($20.2 Million Agency)

• •

17.3 DSA FTE (51.0 FTE Agency)

Business Practices to Support Systems

Data Systems And Analysis Products Current and Historical Data

Census tabulations

• •

Covered Employment Annual Pop & HH Estimates Forecasts (regional & sub-regional) Modeling (travel demand, air quality) GIS (analysis & mapping) Transportation Data Collection

• •

Surveys Counts Transportation Finance Data & Forecasts

Some Questions We Get Asked Impacts on the regional economy from:

Traffic congestion

Transportation revenue increases (taxes, fees, tolls, etc.) Return on particular transportation investments Aging population impacts What types of questions do you get asked?

Regional Economic & Demographic Forecasting Transportation leadership you can trust.

Regional (STEP) & Small Area Forecasts Two-Step, Top-Down Process

STEP (Synchronized Translator of Econometric Projections

4 County Region Regional Forecasts (Pop, Emp, HH) •

EMPAL (Employment Allocation Model)

219 Forecast Analysis Zones •

DRAM (Disaggregate Residential Allocation Model)

Individual Counties

PSRC Model Organization Regional Forecast Model -STEP-

-PSEF-

Land Use Sketch Planning Tool

-Index-

Land Use Model -DRAM/EMPAL-

-UrbanSim-

Transportation Tax Base / Revenue Model Travel Demand Model -EMME/2 current-

-EMME/2 improved-

Air Quality Model (Emmissions) -Mobile 6-

How the Models Work - STEP Economic base theory

Pre-1983, sectors were either export (basic) or local (non basic)

Revised to recognize aspect of both in each sector Exogenous US forecasts as input

Historically purchased from vendor Econometric model equations forecast 116 endogenous variables Boeing, Microsoft variables projected independently

How the Models Work - STEP Blocks OUTPUT

Core forecast block

Productivity Spending EMPLOYMENT

Productivity & output = employment

Wage Rates & CPI INCOME

Ind. employment, national wage rates Reg CPI

Demand for Labor Force POPULATION

Lagged link to employment growth

Switching from STEP to New Model (PSEF -?) RFP in 2004: Replacing STEP (NAICS data time series disruptions)

• • • • •

Meet our MPO, RTPO, Interlocal Agreement Obligations NAICS-friendly Support both old and new land use models Long-range forecast ability out 30 years Transparency, ease of use and maintenance for staff

How the Models Work - PSEF No Output Block Mixed Regression and ARIMA Model NAICS Sectoring Plan Quarterly Trend and Forecast Data Annual Forecasts at County-Level

Will be used as a waypoint for Small Area Forecasts E-views replaces Fortran

NAICS Sectoring Plan - PSEF

Other Variables - PSEF

Input Data - PSEF Long-range US forecasts (Global Insight) Regional trend data (1970-current)

Census, BEA, Washington State ESD (BLS) Just Wage & Salary Employment

Total Employment will need to be a post-processing task

Lessons Learned: Regional Forecasts Watching for secondary variable output / consistency

Ave HH Size

Recent Trends vs Long Range Trends US Exogenous Forecasts

Productivity, GDP Growth Member Jurisdiction Involvement

Questions of Others Linking regional forecasts with:

• •

traffic congestion / travel model forecasts transportation revenue policy (taxes, fees, tolls, etc.) Recognizing aging population

Lower Ave HH Size, different trip generation rates?

Land Use Forecasting: DRAM & EMPAL Transportation leadership you can trust.

How the Models Work – DRAM and EMPAL

Base Year Employment Base Year Pop & HH Base Year Land Use EMPAL Current Yr Employment DRAM Initial Travel Impedances

From PSRC Travel Demand Model

Current Yr Pop & HH Current Yr Land Use

DRAM/EMPAL Land Use Forecast Data Total Population

• •

Household population Group Quarters population Total Households

Percent Multi-Family, Single Family

Income quartiles Total Jobs By Sector

• •

Manufacturing WTCU (Wholesale, Transportation, Communications, Utilities)

• •

Retail FIRES (Finance, Insurance, Real Estate, Services)

Government and Education

Current Land Use Forecast Geography 219 Forecast Analysis Zones (FAZs) Built from 2000 Census Tracts

Building Consensus for Models & Forecasts No longer adopt forecasts Boards approval needed for RFPs and contracts Include non-PSRC staff on RFP, interview teams for consultants TACs for model and forecast work Extensive review & outreach through Regional Technical Forum monthly meetings UrbanSim example

Multiple workshops to cover issues involved in implementing new model

Land Use Forecasting: Moving to UrbanSim Transportation leadership you can trust.

Survey Results from 2001 Study – Important Aspects of Land Use Model 1.

2.

3.

4.

5.

6.

Analyze Effects of Land Use on Transportation Analyze Multimodal Assignments Promote Common Use of Data Manage Data Needs Analyze All Modes of Travel Analyze Effects of Land Use Policies 7.

8.

Support Visualization Techniques Analyze Effects of Transportation Pricing Policies 9.

Analyze Effects of Growth Management Policies 10.

Analyze Effects of Transportation on Land Use

Land Use Model Changes Changing Demands: GMA and more complex analysis questions:

• •

More “what if” questions Model policies and land use impacts – Better interaction between transportation and land use

More flexible reporting geography Our DRAM/EMPAL Limitations:

• •

Zonal geography No implicit land use plan inputs Direction from PSRC Boards during Destination 2030 Update = Improve land use modeling ability RFQ issued in 2002

Entered into interagency agreement and annual contracts with UW Center for Urban Simulation and Policy Analysis (CUSPA – Dr. Paul Waddell) = The UrbanSim Model

UrbanSim Overview

http://www.urbansim.org/

Modeling “Actors” instead of zones Notable Advantages

Potential new output (built SQFT, land value)

Direct modeling of land use plans, development constraints such as wetlands, floodplains, etc.

Geographic flexibility Very Data Hungry

Assessor’s files, Census, Employment Data (Key Input), Land Use plans, Environmental constraints

Modeled Unit = 150 Meter Grid cell (5.5 Acres)

Roughly 790,000 in region (versus 219 FAZs)

UrbanSim Schematic

Changes in Land Use Forecasts: Employment Existing EMPAL Detail: Total Jobs By Sector

• •

Manufacturing WTCU (Wholesale, Transportation, Communications, Utilities)

Retail

FIRES (Finance, Insurance, Real Estate, Services)

Government and Education UrbanSim Detail: One Record per Job ID Sector

1 Resource 2 Construction 3 Manufacturing - Aviation 4 Manufacturing - Other 5 Transportation 6 Communications and Utilities 7 Wholesale Trade 8 Eating and Drinking Places 9 Other Retail Trade 10 Finance, Insurance, and Real Estate 11 Producer Services 12 Consumer Services 13 Health Services 14 Federal Government, Civilian 15 Federal Government, Military 16 Education, K-12 17 Education, Higher 18 State, Local Government

PSRC Category

Res Con Res Con Manuf Manuf WTCU WTCU WTCU Retail Retail FIRES FIRES FIRES FIRES Gov Gov Educ Educ Gov

Changes in Land Use Forecasts: Residential Existing DRAM Detail: Total Population

Household population

Group Quarters population Total Households

Percent Multi-Family, Single Family

Income quartiles UrbanSim Detail: One Record for each Household

Changes in Land Use Forecasts: Land Use Data

• • • •

NEW INPUTS: Implicit to Model compared to DRAM/EMPAL Assessor’s Files Land Use Designations Environmental Areas Land and Building Assessed Value

New Land Use Categories: PLUs and DevType IDs Planned Land Use (PLU) = Comprehensive Plan designations in UrbanSim Development Type IDs = “Built” attributes of each grid cell, based on

• • •

Housing Units Non-Residential Square Feet Environmental Overlays

UrbanSim Data: Plan Types (Comprehensive Land Use Plans) Model Comp Plan Designations Implicitly

• • •

Four-County Aggregate Classifications Part of Model Specification (Can’t add on the fly) One of two parts of the “Constraint” Process

UrbanSim: Development Type IDs (Built Land Use) Or, Overall Land Use Mix of each Grid cell

Measures of units/square feet of built environment

Part of Model Specification (Can’t add on the fly)

One of two parts of the “Constraint” Process

Data Acquisition and Pre Processing: Current LU (Development Type)

Data Acquisition and Pre Processing: Planned LU

Changing the PLU Categories Triple Balancing Act

• • •

Detail in comp plans Job categories Development Type IDs Assign each (660) comp plan code to PLU

Started with 20+, wound up with 19 final PLU codes

More detail in Residential, Commercial, Industrial, Mixed Use, and Government/Tribal/Military

New PLUs

Sample Maps of New PLUs

Comp Plan vs Zoning Example Mixed Use in Comp Plan

2-5 du/ac, Office, Comm Bus Multiple Zoning Classes R4 R5

Comp Plan Descriptions & Consistency Light Yellow = Single Family High Density Residential…

Was in 12+ DU / Acre 6 DU /Acre 3-5 DU /Acre

Centroid vs ‘Majority Rules’ Approach

New PLU Acreage Summaries

DevType IDs

Example: Development Constraints Table Example: RES-Light (1 4 DU/Acre)

PLU + DevTypeIDs = Development Constraints Table

Lessons Learned: Land Use Models Involve local staff in data assembly issues and forecast results review Plan for the update and maintenance

• •

Staff retention CUSPA automated a lot of data processing applications Underestimated time spent on data cleaning

Allow time for 2-3 loops, data assembly, model testing Hard to gauge the “correct altitude” to fly at for dat cleaning

• •

IE Employment data to parcels Other uses of base year data

Reviewer concerns vs impacts on the model

Questions for Others Plancast vs Forecast

Balancing plans & comments against model results How strict or loose to model comp plans?

Regarding Employment Data Transportation leadership you can trust.

Different Employment Databases Geocoded Points 1 Covered employment Total employment Factors to ESD Totals Factors from STEP database Specific adjustments 2 Covered employment 3 Total employment 4 “Modeling” employment

Assemble Employment Data ES202 business inventory from Employment Securities Division Government and Educational Survey, PSRC Assign employment sectors (based on STEP model sectors) Manual verification of major employer geocoding to parcel

Parcels, Streets, and Manual Matches Arc-Info Arcview Interns

Assign Employment to Parcels Provides cross-checking of employment and parcel data (should be consistent) Automated procedures for assignment of businesses to parcels

• •

Operates on one census block at a time Uses multiple decision rules

− −

Address of business falls between 2 parcels Availability of nonresidential SQFT

− − −

Tax-exempt properties Sector to Land Use probability distribution by FAZ group Check for mis-geocoding to wrong block

Field verification of algorithm on small sample of blocks

Impute Missing Data on Parcels Automated imputation procedures for:

• • • •

Land Use code Year Built Housing Units Sqft Based on spatial query of nearby parcels with similar characteristics Uses SQL queries and Perl scripts

Interagency Agreement: Restrictions on Data Use Confidentiality – Require reviewers and users of database to sign agreement

• • •

Geocoding accuracy Travel demand modeling GMA analysis Suppression – Publication rules to prevent individual employers from being identified

• • •

One employer accounts for 80% or more of total employment There are less than 3 employers If showing totals, suppression of one value means one other must be suppressed

Appendix A

Step-By-Step UrbanSim Data Assembly Methodology Transportation leadership you can trust.

UrbanSim Data Integration Process Input Data Data Store

Parcel file Business Establishment File Census PUMS, STF3 Data Integration Process Households

HouseholdID

Persons Workers Children Age of Head Income GridId Jobs

JobID

Sector GridId GIS Overlays: Environmental UGB City County Traffic Zone Grid Cell

GridId

Total Housing Units Vacant Housing Units Total Nonres Sqft Vacant Nonres Sqft Development Type Land Value Residential Imp Value Nonres Imp Value Environ Overlays UGB City County Traffic Zone

UrbanSim Data Preparation

Coverage: King, Kitsap, Pierce, Snohomish Base Year: 2000

• • • •

Input databases:

Parcels from each county (2001)

Employment data from ES202 and survey of Government and Educational Establishments Census data from PUMS, SF3 Transportation model outputs Environmental GIS layers Planning and political GIS layers

Major Steps in Data Preparation 9.

10.

11.

12.

1.

2.

3.

4.

5.

6.

7.

8.

Determine study area boundary Generate grid over study area Assemble and standardize parcel data Impute missing data on parcels Assemble employment data Assign employment to parcels Convert Parcel data to grid Convert other GIS layers to grid Assign Development Types Synthesize household database Diagnose data quality and make refinements Document data and process

1. Determine study area boundary Initial application will be to 4-County Central Puget Sound

King, Kitsap, Pierce, Snohomish Potential later extension to other counties

Island, Mason, Skagit, Thurston

2. Generate Grid Over Study Area Uses grid cell size of 150 x 150 meters Areas in water or outside project boundary coded as NODATA

150 Meter Grid Cells

3. Assemble and Standardize Parcels Parcel database assembly for all 4 counties

• •

Conversion of county land use codes to regional standard Consolidation of key fields:

Lot size

− − − − − − − −

Land use Housing units Sqft building space Year built Zoning Land use plan Assessed land value Assessed improvement value Microsoft Access Version MySQL with Replication

Parcel Data

Parcel Counts:

King County:

• • • •

Kitsap County: Pierce County: Snohomish County: Region Total: 542,446 100,336 260,230 211,677 1,114,689

Generalized Land Uses - Parcel Agriculture Civic and Quasi-Public Commercial Fisheries Forest, harvestable Forest, protected Government Group Quarters Hospital, Convalescent Center Industrial Military Mining Mobile Home Park

Generalized Land Uses - Parcel Multi-Family Residential Office Park and Open Space Parking Recreation Right-of-Way School Single Family Residential Transportation, Communication, Utilities Tribal Vacant Warehousing Water

4. Impute Missing Data on Parcels Automated imputation procedures for:

• • • •

Land Use code Year Built Housing Units Sqft Based on spatial query of nearby parcels with similar characteristics Uses SQL queries and Perl scripts

5. Assemble Employment Data ES202 business inventory from Employment Securities Division Government and Educational Survey, PSRC Assign employment sectors (based on STEP model sectors) Manual verification of major employer geocoding to parcel

6. Assign Employment to Parcels Provides cross-checking of employment and parcel data (should be consistent) Automated procedures for assignment of businesses to parcels

• •

Operates on one census block at a time Uses multiple decision rules

− −

Address of business falls between 2 parcels Availability of nonresidential SQFT

− − −

Tax-exempt properties Sector to Land Use probability distribution by FAZ group Check for mis-geocoding to wrong block

Field verification of algorithm on small sample of blocks

7. Convert Parcel Data to Grid GIS overlay of parcels on gridcells Allocate parcel quantities to gridcells in proportion to land area in each cell Aggregate data in grid cells Convert employment from parcel geocoding to grid cell

8. Convert Other GIS Layers to Grid Environmental Layers

• •

Completed:

− −

Water Wetlands

− − −

Floodplains Parks and Open Space National Forests Pending – need feedback on definitions to use for:

Steep slopes

Stream buffers (riparian areas)

Convert Other GIS Layers to Grid Planning/Political Layers

Completed:

− − − − − −

Cities Counties Urban Growth Boundaries Military Major Public Lands Tribal Lands Note: Current data sources may be replaced if better data are available All grid-based data stored as attributes on gridcells table

GIS Data Sources (Page 1) National Forests at 500k

Source: Washington State Department of Transportation Military Bases at 500k

Source: Washington State Department of Transportation Shoreline Management Act – Streams

Source: Washington State Department of Ecology Q3 Flood Data, King, Kitsap, Pierce, Snohomish

Source: Washington State Department of Ecology State Tribal Lands

Source: Washington State Department of Ecology National Wetlands Inventory

• •

Source: Puget Sound Regional Council Procedures: The wetlands have been identified using high altitude aerial photography and classified by the Cowardin Classification Scheme.

GIS Data Sources (Page 2) Park and Open Space

• •

Source: Puget Sound Regional Council Procedures: Regional Council staff collected the data from the four counties and their local jurisdictions. Major Public Lands

• •

Source: Puget Sound Regional Council Procedures: Spatial delineation was digitized by the Department of Natural Resources Division of Information Technology from 1:100,000 DNR Public Lands Quads and Bureau of Land Management 1:100,000 Public Lands Quads. Waterbodies

Source: Puget Sound Regional Council DEM30

Source: Puget Sound Regional Council Urban Growth Boundary

Source: Puget Sound Regional Council

9. Assign Development Types 25 Development Types Assigned Type 25 is Vacant Undevelopable

Composite of characteristics used to assign:

Percent of cell in water, wetland, floodplain, steep slope, public lands, etc.

− −

Need feedback on conditions to use Implication: undevelopable cells preserved in the model All cells not classified as Undevelopable are assigned a type using a lookup table based on the number of housing units, sqft of nonresidential space, and mix of uses

Development Types

Devtype Name 1 2 3 4 5 6 7 8 9 10 11 19 20 21 22 23 24 12 13 14 15 16 17 18 25 R1 R2 R3 R4 R5 R6 R7 R8 M1 M2 M3 M4 M5 M6 M7 M8 C1 C2 C3 I1 I2 I3 GV VacantDevelopable Undevelopable UnitsLow UnitsHigh SqftLow SqftHigh Primary Use 65000 9 30 30 30 30 99999 99999 99999 0 9 9 0 9 9 99999 0 0 14 21 30 75 1 4 9 0 0 0 0 0 0 0 10 15 22 31 1 2 5 76 1 10 10 10 10 31 31 31 0 0 0 0 0 0 0 0 0 999 Residential 999 Residential 999 Residential 2499 Residential 2499 Residential 2499 Residential 4999 Residential 0 1000 2500 5000 4999 Residential 4999 Mixed_R/C 4999 Mixed_R/C 24999 Mixed_R/C 25000 49999 Mixed_R/C 50000 9999999 Mixed_R/C 5000 24999 Mixed_R/C 25000 49999 Mixed_R/C 50000 9999999 Mixed_R/C 1000 25000 24999 Commercial 49999 Commercial 50000 9999999 Commercial 1000 24999 Industrial 25000 49999 Industrial 50000 9999999 Industrial 0 9999999 Government 0 0 VacantDevelopable 0 0 Undevelopable

10. Synthesize Household Database Need spatial distribution of households Beckman (1995) developed household synthesis methodology for TRANSIMS We extended Beckman’s approach:

Parcel-based housing counts

• •

Discount by vacancy rate to get target household count Assign household characteristics:

− −

Joint probability distribution from PUMS IPF scale to tract marginal distributions from SF3 Application of the synthesizer will need to wait for Census Bureau release of 5% PUMS

11. Diagnose data quality and make refinements Data Quality Indicators

• •

Automated database queries Before and after each major imputation or allocation procedure

Different geographic levels:

Parcel

− − − − − −

Grid cell (150 meter) Census block TAZ FAZ Group City County

Data Quality Indicators Example: Parcels Missing Year Built

• • • •

King Kitsap Pierce Snohomish 13% 31% 41% 19%

12. Document Data and Process Overview of Data Processing

Major steps, procedures, decisions Data Summaries Data Quality Indicators

Before and after processing Data Preparation Tools – User Guide

• • • • • •

Data imputation Household Synthesis Job Allocation Conversion to grid Assignment of Development Types Data Quality Indicator Queries