A Study of Alternative Land Use Forecasting Models

A Study of Alternative Land Use Forecasting Models

Soon Chung and Fang ZhaoLehman Center for Transportation

ResearchFlorida International University

Miami, Florida

11th TRB National Transportation Planning Applications Conference

Daytona Beach, FloridaMay 8, 2007

Acknowledgements

• This project was funded by the Florida Department of Transportation Systems Planning Office

• Terrence Corkery, AICP Project Manager

• Mr. Michael Neihart, Volusia County MPO

Outline

• Introduction• Objectives• UrbanSim• Study Area• Design of Simulation• Validation• Test Scenarios• Simulation Results• Findings

Introduction

• Transportation models need good land use forecasts

• Many land use models have been developed– Simple to complex – Integrated or not integrated– With/without economic theory

basis– Support community visioning– GIS platform

Objectives

• Understanding the state-of-the-art and state-of-the-practice of land use models

• Determining the data requirements

• Identifying application issues• Investigating need for data

processing and interfacing FSUTMS

• Identifying future research and implementation issues

What Is UrbanSim?• A land use microsimulation model• Developed by the University of

Washington• Provide new land use forecasting and

analysis capabilities– Based on economic theories – Model the interactions of markets and

policies, including dynamic disequilibrium– Design to interface activity-based models

• Open Source software – source code free to use, modify, and redistribute (available at www.urbansim.org)

UrbanSim Users• US Users

– Seattle– Eugene-Springfield– Houston– Honolulu– Salt Lake City– Phoenix– Detroit

• Europe Users– Amsterdam– Paris– Zurich

• Middle East Users – Tel Aviv• Potential Users - Downloaded from 80

Different Countries

UrbanSim Model Structure

Travel DemandModel System

ScenarioAssumptions

Economic andDemographic

Transition2

Location Choice4

Land Price6

Data Store

Travel ModelOutputs

Control Totals

MacroeconomicModel

User SpecifiedEvents

Accessibility1

Mobility3

Real EstateDevelopment

5SQL DatabaseData Export

GISVisualization

ExternalModels

UserInputs

Model Coordinator

Sub-Models• accessibility-model• household-transition-model• employment-transition-model• household-relocation-choice-model• employment-relocation-choice-model• household-location-choice-model• employment-non-home-based-

location-choice-model• employment-home-based-location-

choice-model• scaling-procedure-for-jobs-model• land-price-model• developer-model

Data Required• Grid Cells• Parcel Data• Property Tax Data• Employment Data (Info/USA)• Environmental Layers

– Water– Wetlands– Floodplains– Parks and open space– National forests– Steep slopes (DEM)– Stream buffers (riparian areas)

• Planning and Political Layers– Traffic Analysis Zones (TAZs)– Cities– Urban growth boundaries– Military– Major public lands– Tribal lands

• Streets

Land Price Model• Linear Regression Model• Dependent Variable – natural

logarithm of the total land value within a grid cell

• Independent Variables– Site characteristics

• Development type• Land use plan• Environmental constraints

– Regional accessibility• Access to population and employment

– Urban design-scale• Land use mix and density• Proximity to highways and arterials

Household Location Model

• Discrete Choice Model• Variables

– Housing Characteristics• Prices (cost to income ratio)• Development types (density, land use mix)• Housing age

– Regional accessibility• Job accessibility by auto-ownership group• Travel time to CBD and airport

– Urban design-scale (local accessibility)

• Neighborhood land use mix and density• Neighborhood employment

Employment Location Model• Discrete Choice Model

– Employment Home-Based Location Model – Employment Non-Home-Based Location

Model

• Variables– Real Estate Characteristics

• Price• Development type (land use mix, density)

– Regional accessibility• Access to population• Travel time to CBD, airport

– Urban design-scale• Proximity to highway, arterials• Local agglomeration economies within and

between sectors: center formation

Developer Model• Discrete Choice Model• Variables

– Site characteristics• Existing development characteristics• Land use plan• Environmental constraints

– Urban design-scale• Proximity to highway and arterials• Proximity to existing development• Neighborhood land use mix and property

values• Recent development in neighborhood

– Regional• Access to population and employment• Travel time to CBD, airport

Possible Scenarios

• Macroeconomic Assumptions– Household and employment control

totals

• Development constraints– Can select any combination of

• Political and planning overlays• Environmental overlays• Land use plan designation

– Determine which development types cannot occur

• Transportation infrastructure• User-specified events

Study Area Selection Criteria

• Have recent household survey data

• Up-to-date GIS data, including parcel-level property data

• Being relatively self-contained

Volusia County• 1,263 square

miles• Population –

443,343 in 2000• Surrounded by

Flagler, Marion, Lake, Seminole, and Brevard counties (most rural)

Volusia County Planning Regions

Simulation Process

UrbanSim TranPlan

HouseholdsJobs

ZDATA1ZDATA2

Composite Utility by Auto OwnershipHighway Travel Times

Accessibility_InputTravel_DataZones

1 2

3

4

TP*

UrbanSim

TP

UrbanSim UrbanSim UrbanSimUrbanSim UrbanSim UrbanSim

TP TP TP TP TPTP

2000 2020201720132010200720052002

*TP: FSUTMS/TranPlan

Validation

Model Output was compared with

1. Model results adopted in the LRTP

2. 2005 InfoUSA Employment

Data

Cumulative Percentage of TAZs vs. Differences in Zonal Households and Population

0

20

40

60

80

100

<= -1000 <= -500 <= -200 <= -100 <= 0 <= 100 <= 200 <= 500 <= 1000 > 1000

Differences

Cum

ulat

ive

Per

cent

age

2000 Total Households

2000 Total Population





Cumulative Percentage of Employment Differences between UrbanSim and the 2005 InfoUSA Data

0

10

20

30

40

50

60

70

80

90

100

<= -1000 <= -500 <= -200 <= -100 <= 0 <= 100 <= 200 <= 500 <= 1000 > 1000

Absolute Difference

Cum

ulat

ed P

erce

nt

Industrial Emp.

Commercial Emp.

Service Emp.

Total Emp.

Spatial Distribution of Differences between UrbanSim and 2005 InfoUSA Data

Five Scenarios

• 3 alternatives from Volusia County 2020 LRTP– Alternative #2

– Alternative #3

– Final Plan

• 3 projections (low, mid, high) of population from Bureau of Economic and Business Research

Scenarios

ScenariosTransportation

AlternativeProjection

1 Final Medium

2 Alternative 2 Medium

3 Alternative 3 Medium

4 Final Low

5 Final High

Comparison of Volumes in 2020LRTP - Scenario 1

Spatial Distribution of Households from Scenario 12020 – Base Year

Spatial Distribution of Employment from Scenario 12020 – Base Year

Comparison of Households between the 2020 LRTP and Scenario1

Planning Region

1997 2000 2010 2020

LRTP Scenario 1 LRTP Scenario 1 LRTP Scenario 1

Northeast 84,022 80,899 106,863 96,383 119,592 97,987

Southeast 23,846 22,116 31,880 30,424 35,437 32,410

Central 5,689 6,754 18,312 8,087 29,739 8,505

Northwest 2,288 2,176 2,855 2,741 3,255 2,842

Central-west 19,178 19,145 24,590 25,992 27,705 28,937

Southwest 36,300 40,492 40,950 47,198 44,648 49,367

Comparison of Employment between the 2020 LRTP and Scenario1

Planning Region

1997 2000 2010 2020

LRTP Scenario 1 LRTP Scenario 1 LRTP Scenario 1

Northeast 100,648 101,315 130,583 106,643 130,583 118,548

Southeast 16,220 15,103 28,919 19,039 28,919 23,803

Central 2,626 6,778 9,761 8,369 9,761 10,565

Northwest 3,789 3,602 3,809 2,927 3,809 3,055

Central-west 20,434 25,440 30,587 36,443 30,587 46,432

Southwest 17,340 20,591 24,984 26,049 24,984 31,969

Comparison of Alternative 2 and Final Plan

Comparison of Volumes in 2020Scenario 2 – Scenario 1

Spatial Distribution of Households from Scenario 22020 – Base Year

Spatial Distribution of Employment from Scenario 2 2020 – Base Year

Comparison of Households between Scenario1 and Scenario 2

PlanningRegion

2010 2020

Scenario 1 Scenario 2 Scenario 1 Scenario 2

Northeast 96,383 96,515 97,987 98,145

Southeast 30,424 30,486 32,410 32,372

Central 8,087 8,077 8,505 8,492

Northwest 2,741 2,748 2,842 2,848

Central-west 25,992 25,993 28,937 28,766

Southwest 47,198 47,214 49,367 49,391

Comparison of Employment between Scenario1 and Scenario 2

PlanningRegion

2010 2020


Northeast 106,643 106,420 118,548 118,318

Southeast 19,039 19,647 23,803 24,561

Central 8,369 8,133 10,565 10,096

Northwest 2,927 2,958 3,055 2,985

Central-west 36,443 36,427 46,432 46,676

Southwest 26,049 25,999 31,969 31,781

Comparison of Alternative 3 and Final Plan


Spatial Distribution of Households from Scenario 3 2020 – Base Year

Spatial Distribution of Employment from Scenario 32020 – Base Year


PlanningRegion

2010 2020


Northeast 96,383 96,397 97,987 97,899

Southeast 30,424 30,470 32,410 32,210

Central 8,087 8,051 8,505 8,448

Northwest 2,741 2,740 2,842 2,847

Central-west 25,992 26,003 28,937 28,916

Southwest 47,198 47,215 49,367 49,263


PlanningRegion

2010 2020


Northeast 106,643 107,154 118,548 118,100

Southeast 19,039 19,599 23,803 24,726

Central 8,369 8,594 10,565 11,172

Northwest 2,927 2,943 3,055 2,974

Central-west 36,443 36,576 46,432 46,243

Southwest 26,049 25,781 31,969 31,255

Comparison of Volumes in 2020 Scenario 4 – Scenario 1


PlanningRegion

2010 2020


Northeast 96,383 78,728 97,987 63,487

Southeast 30,424 27,121 32,410 25,767

Central 8,087 8,053 8,505 8,135

Northwest 2,741 2,704 2,842 2,593

Central-west 25,992 25,311 28,937 26,405

Southwest 47,198 46,596 49,367 46,032


PlanningRegion

2010 2020


Northeast 106,643 91,757 118,548 82,841

Southeast 19,039 15,076 23,803 13,157

Central 8,369 6,402 10,565 7,373

Northwest 2,927 2,567 3,055 2,036

Central-west 36,443 28,679 46,432 27,235

Southwest 26,049 20,471 31,969 18,612



PlanningRegion

2010 2020


Northeast 96,383 96,325 97,987 97,766

Southeast 30,424 30,511 32,410 31,854

Central 8,087 8,002 8,505 8,310

Northwest 2,741 2,759 2,842 2,835

Central-west 25,992 25,687 28,937 28,183

Southwest 47,198 46,827 49,367 48,572


PlanningRegion

2010 2020


Northeast 106,643 115,927 118,548 136,304

Southeast 19,039 21,412 23,803 26,950

Central 8,369 8,969 10,565 12,147

Northwest 2,927 3,229 3,055 3,621

Central-west 36,443 38,156 46,432 50,922

Southwest 26,049 28,565 31,969 36,708

Findings• High data requirements

– Impute missing data– Join property TAX data with parcel layer– Join employment data with parcel layer

• Data mining and synthetic data cleaning tools, currently being designed, will facilitate working with messy data

• Model estimation requires knowledge of discrete choice models

• Consultations with local government agencies are desirable in developing model specifications and estimating model parameters

Documents

A Study of Alternative Land Use Forecasting Models