1 JD Hunt JE Abraham JAS Khan University of Calgary Workshop of the Land Use Transport (LUT) Modelling Group London, UK 2 July 2005 Calgary Testbed for

1

JD HuntJE AbrahamJAS Khan

University of Calgary

Workshop of the Land Use Transport (LUT) Modelling GroupLondon, UK 2 July 2005

Calgary Testbed for Microsimulation Calgary Testbed for Microsimulation of Business Establishmentsof Business Establishments

Part 1Part 1Introduction and Synthetic Population GenerationIntroduction and Synthetic Population Generation

2

Introduction

ContextContext

• PhD Research at University of Calgary• student: JAS Khan• supervisor: JD Hunt• support: JE Abraham, as PDF

• ILUTE project funded by SSHRC:• focus on behavioural mechanisms influencing travel• more basic, less practical focus

• less concern about short-term practical relevance• more concern about longer-term possibilities for practice

• less focus on calibration, more on behaviour and potentials

3

Introduction

TermsTerms• Business Establishment (BE)

portion of a firm conducting activities at a single physical location

• Firmography:demography for BEs, covering:

- births- deaths- locations/relocation/migration

4

Approach

BE Representation - ObjectsBE Representation - Objects• BE properties:

• consumption behavior (production function), what is used per unit of output:

• commodities – three transportable categories• labour – a single category, produced by households• floorspace – two types, commercial and residential

• production behavior – fixed number of units of output in one of three commodities

• age (number of years BE active)

• geographical location - cell where located

5

Testbed Application

Hypothetical CaseHypothetical Case• test geography

• 9x9=81 zones, each with 100 cells• each cell 1 acre

• mix of sizes and types of establishments• start with 1,400 BEs

• mix of initial floorspace types and quantities in each cell• 3 floorspace types, range of allowable FARs

• static external economic conditions• runs for 100 years

6

Approach

Simulation MethodsSimulation Methods

• Monte Carlo Technique:• selecting changes in states from sampling distributions based on

probabilistic choice models and observed distributions

• Each BE considered once per year

• Floorspace development changes considered once per year

• Prices updated each 6 months

7

Approach

Behavior RepresentedBehavior Represented• Includes behavior of and interactions among:

• BEs• land development

• builds on, and extends, what’s been shown to work in

PECAS:PECAS:

Production, Exchange & Consumption Allocation System

8





Part 2Part 2FirmographyFirmography

9

Approach

BE Dynamic BehaviorBE Dynamic Behavior• New set of BEs added each year in each zone

•randomly assigned attributes•no contributions to input or output – a testrun of an ‘idea’

• Each year BE makes decision to• ‘stay’, in same floorspace area in same cell• ‘relocate’ to some other floorspace area in another cell• ‘leave’, the model area (emigrating or folding up) :

• Using a nested logit model• at top level of PECAS hierarchy of models• with composite utilities from PECAS hierarchy feeding up

• Prices change at each exchange zone every 6 months

2211 .)/( eqvrzzPricePricePrice btt

10





Part 3Part 3Location and Interaction TreatmentsLocation and Interaction Treatments

11

Reasons for using PECAS A model of the spatial economy, including labour,

goods, services, floorspace and capital Establish locations of activities (population and

employment) and spatial interactions Produce the “land use” inputs to transportation demand

forecasting models for practical transportation planning Based on Input-Output theory, but integrated with

discrete choice theory and disaggregated markets and including an explicit representation of transport costs

Quasi-equilibrium (aka quasi-dynamic), with floorspace supply modelled dynamically, and equilibrium solutions influenced by past conditions to introduce lags

Builds on theory and experience with MEPLAN and TRANUS planning models, but more behavioural and consistent with economic theory and discrete choice theory.

12

Production - Exchange - ConsumptionProduction - Exchange - Consumption

selling allocationselling allocationprocessprocess

totalconsumption

totalproduction

totalproduction

totalproduction

buying allocationbuying allocationprocessprocess

commodityflows

exchangezone

exchangezone

exchangezone

13

PPECASECAS Schematic Schematic

Logit allocation of activity to zones

Aggregation of commodity Zutilities, choice of production function

Allocation to exchange zonesbased on price at exchange zones and transportdisutility to/from exchange zones

Microsimulate establishment behaviour at this level in this research

In these two levels, replace equilibrium price representation

with dynamic price update representation based on excess

demand

14

Land development (LD)Land development (LD)

In practical work with PECAS, and in this research, a disaggregate microsimulation of Land Development (floorspace supply) is used: adjust the quantity of space over time in

response to changes in price represents the behaviour of developers and

landowners in land improvements decisions Land is represented here as small "grid cells"

Grid cells representing age, density, development types, & types of development allowed

15

• Floorspace changes each year

Approach

LD Dynamic BehaviorLD Dynamic Behavior

No Change

Change options

Add more of the same

Redevelop as same

Redevelop as different

Developer choice

16

Taz # of GridCells Area/cell Dev Type Min Dev Max Dev Min Age Max Age Z scheme203 6 43560 Vacant 21666.66667 21833.3333 10 15 R1204 85 43560 Vacant 21666.66667 21833.3333 10 15 R1204 15 43560 Vacant 0 0 0 0 R1205 50 43560 Vacant 21666.66667 21833.3333 10 15 R1205 50 43560 Vacant 0 0 0 0 R1206 100 43560 Vacant 0 0 0 0 R1207 100 43560 Vacant 0 0 0 0 C1ORR1208 100 43560 Vacant 0 0 0 0 R1209 100 43560 Vacant 0 0 0 0 C1ORR1210 100 43560 Vacant 0 0 0 0 R1301 100 43560 Vacant 0 0 0 0 R1302 90 43560 Vacant 0 0 0 0 R1302 10 43560 Vacant 21666.66667 21833.3333 10 15 R1303 15 43560 Vacant 0 0 0 0 R1303 85 43560 Residential 21666.66667 21833.3333 10 15 R1303 100 43560 Vacant 0 0 0 0 R1304 92 43560 Residential 21666.66667 21833.3333 10 15 R1304 8 43560 Vacant 0 0 0 0 R1305 95 43560 Vacant 0 0 0 0 R1305 5 43560 Residential 21666.66667 21833.3333 10 15 R1306 55 43560 Vacant 0 0 0 0 Vacant306 45 43560 Vacant 0 0 0 0 C3ORR3307 90 43560 Vacant 0 0 0 0 C1ORR1307 10 43560 Vacant 0 0 0 0 C3ORR3308 65 43560 Vacant 0 0 0 0 R1

Testbed Application

Initial FloorspaceInitial Floorspace

17

Zoning Scheme Allowed Development Type Maximum FARR1 Vacant 0R1 Residential 0.5R2 Vacant 0R2 Residential 1R3 Vacant 0R3 Residential 15C1 Vacant 0C1 Commercial 1C2 Vacant 0C2 Commercial 2C3 Vacant 0C3 Commercial 15C2ORR2 Vacant 0C2ORR2 Commercial 2C2ORR2 Residential 5C1ORC2 Commercial 1.5C1ORC2 Vacant 0C3ORR3 Vacant 0C3ORR3 Commercial 15C3ORR3 Residential 15C1ORR1 Vacant 0C1ORR1 Commercial 1C1ORR1 Residential 1Vacant Vacant 0

Testbed Application

ZoningZoning

18

Testbed Application

CalibrationCalibration• adjust parameters until:

• model stable• results not unreasonable

• work iteratively, parameter by parameter

• note emphasis on understanding and testing rather than practical results

19

Testbed Application- Calibration move or stay or leave’ dispersion parameter values. move or stay or leave’ dispersion parameter values. Unreasonable value (too-high)Unreasonable value (too-high)

20

Testbed Application- Calibration

21

Testbed ApplicationCalibration – Reasonable valueCalibration – Reasonable value

22

Testbed ApplicationCalibration – Reasonable value (Cont.)Calibration – Reasonable value (Cont.)

23





Part 4Part 4Scenarios and Future WorkScenarios and Future Work

24

Testbed ApplicationAlternative Policy Test 1Alternative Policy Test 1

Reference Case30% floorspace rent increase

zones 202, 203, 204

25


Reference Case30% floorspace rent subsidy

zones 202, 203, 204

26


Reference Case50% development cost subsidy

zones 202, 203, 204

27


Reference Case30% price increase

various zones

28


Reference Casedouble travel

costs everywhere

29


Reference Casenew freeway; reducedtravel times and costs

30


Reference Case increase allowable developmentall kinds zones 202, 203, 204

31


Reference Casechanged allowable developmentfor commercial, various zones

more

less

32

ConclusionsConclusions• Model and approach work:

• reasonable aggregate behavior from disaggregate treatment• policy responses as expected

• Increased understanding of system:• link from individual BEs to aggregate patterns• new perceptions of

• Von Thunen• Alonso• Hotelling – Ice Cream Salesmen on a Beach• Christaller Central Place Theory

• Seems to offer interesting potential way ahead in practical modelling, but ...

33

ConclusionsConclusions• Run times an issue:

• 10 hours for 100 years of hypothetical scenario• using 1GHz computer with 800 MB of RAM

• Range of BE sizes and types:• no growing and declining, just appearing and ‘leaving’• range of technologies, but simplified

• Reality much more complex, with many more commodity types, interactions and markets

• Substantial data issues also side-stepped here

• Gap to full-blown practical application• greater than with mirco-simulation of households• a ways into future . . . . more questions arising

Documents

1 JD Hunt JE Abraham JAS Khan University of Calgary Workshop of the Land Use Transport (LUT) Modelling Group London, UK 2 July 2005 Calgary Testbed for