BROWSE

SEARCH

ORDER DOCUMENTS

NEWSLETTER

DRAFT: A Review of Land Use Forecasting Methodologies for Metropolitan Planning Organizations

Rik Keller
Transportation Planner
Austin Transportation Study
P.O. Box 1088-Annex
Austin, TX 78767-1088
(512) 499-2275

Introduction

This report will provide an overview of land-use forecasting methodologies available for use by Metropolitan Planning Organizations (MPOs). The term 'land use forecasting' as used in this report refers to population and employment projections for small geographic units within a region, typically Traffic Analysis Zones (TAZs) or the equivalent. The report focuses on methodologies for allocation of projections by TAZ but also includes relevant information regarding regional projections and data collection methodology. It relies heavily on the analysis in a number of different sources. These have been credited where applicable.

The report is organized as follows:

1. Background on Forecasting Methodologies
   1. Overview
   2. Types of Land Use Forecasting Models
   3. Descriptions of Specific Land-Use Forecasting Models
   4. Conclusions on Land Use Forecasting Modeling
   5. Considerations in Choosing a Land Use Forecasting Model
2. Survey of MPOs on Land Use Forecasting Methodologies

Background on Forecasting Methodologies

Overview

Greig Harvey and Elizabeth Deakin in A Manual of Regional Transportation Modeling Practice for Air Quality Analysis⁽¹⁾ provide an overview of different types of land use forecasting methodologies. This section was derived from that analysis.

Harvey and Deakin outline three main approaches to producing population and employment allocations by zone. These are outlined below:

1. Negotiated Estimates

Description

• Draws upon local plans and projections to allocate by zone.
• May use a simple ratio approach: total growth is assumed to locate in various jurisdictions as a proportion of their estimated growth (e.g., everyone gets 67% of their projections).
• May use negotiations (e.g., regional agency staff facilitate meetings in which each local jurisdiction's anticipated growth rate is compared to historic performance and market potential, and compromises are made). Some areas develop both population and employment estimates in this manner and bring them into agreement in a negotiated process over the planned balance of housing and jobs.

Comments

• Negotiated estimates can be acceptable if the negotiations are disciplined by data and forecast of growth and development trends, including expected transportation investments and levels of service and local policies and desires for growth and development.
• Acceptable negotiated approaches typically will be based on employment data and employment trends, population data and trends, information on land availability and price, land use occupancy and rent data, information about zoning and other land use regulations, local and regional economic development plans and market assessments, data on building permits and construction starts, measures of current and anticipated transportation investments and level of service, local policies and desires for growth, and so on.

2. Scenario Approaches

Description

Construction of 2 or more land use and development alternatives. There are two main approaches to this scenario development:

1. Visionary plans -- Focus on how the region 'should' look in the future. Usually this plan is compared to a trends-extended plan to provide a framework for discussing transportation and land use planning.
2. Policy options -- Looks at choices actually available to decision-makers to develop feasible directions for land use and development. This approach emphasizes the specification and testing of land use and transportation options which are financially feasible and reflect market realities.

Comments

• Population and employment allocations which are firmly grounded upon data analysis, with or without models, are preferable to approaches based primarily on the aspirations of localities or individuals, without considering market or political realities. Scenario based allocations can be acceptable if they are supported by evidence and defensible forecasts.
• Approaches which use simple fair-share allocations or allocation based solely on local aspirations are not recommended except perhaps for scenario testing to investigate what would happen if cities' plans came true. Allocations developed as part of a scenario testing exercise similarly are acceptable to the extent that they are based on data and data-based forecasts.

3. Formal Mathematical Models

Descriptions

• There have been a number of forecasting tools developed to predict land use for transportation planning. In general terms, these models allocate jobs and housing to subareas within a region as functions of accessibility, land availability, population and employment by category, household income, and other factors.
• The main allocation modeling approach now in use in the U.S. is the DRAM/EMPAL component of the ITLUP software package which was developed as an addition to the standard four-step transportation model.

Comments

• Currently in use in a minority of regions because of political and policy sensitivities involved in land use allocation.
• There is a skepticism about land use models because the models are relatively limited in their policy sensitivity and because, to date, the forecasts they have produced have not been highly accurate.
• Land-use allocation models have only had moderate predictive capabilities in many applications, but they are viewed as preferable to approaches which simply ignore transportation/land use interactions.
• Such models are often complicated and expensive to run and also make several simplifying assumptions that don't accord with theory. Current allocation models contain several limitations which may restrict their utility:
  1. The behavioral context of many land use allocation models is low -- for the most part, they do not attempt to represent the location decision-making of either firms or households. Instead, they generally rely on extrapolation of past trends subject to a set of constraints. While discrete choice models of households' location hold promise, they have not been implemented in practice and equivalent models of business choice location have not yet emerged.
  2. Most allocation models describe transportation access in terms of highways only, yet in many cases the impact of transit on land use is a key policy issue.
  3. Land availability by zone is generally described in terms of amount, current use, and general zoning category. This does not take into account other significant factors in location decision, including availability of sewer, water, and other infrastructure and services, tax rates, crime rates and other factors.
  4. The models do not account for, or only weakly account for, market responses such as shifts in land values and rents and their impact on the intensity of uses in existing and new development.
  5. Time leads and lags which occur in transportation/land use interactions are poorly represented.

Types of Land Use Forecasting Models

The David Simmons Consultancy (DSC) in Available Methods for Land Use/Transport Interaction Modelling,⁽²⁾ and Frank Southworth in A Technical Review of Urban Land Use Transportation Models as Tools for Evaluating Vehicle Travel Reduction Strategies,⁽³⁾ provide frameworks for categorizing formal mathematical models used for land use forecasting. The two publications organize formal modeling methods into roughly similar categories (though with different names). This section is derived from those publications and synthesizes the categorization and descriptions. Discussion of each type of model includes a general description , example models, and strengths and weaknesses of the particular approach. More in-depth descriptions of specific models are presented in a following section.

Land use forecasting models are organized into the following categories:

1. Predictive Models -- purpose is to predict the behavior of urban systems by responding to 'design' inputs
   1. Static Models -- represent a single point in time
   2. Dynamic Models -- runs for a series of time periods
      1. Entropy Based Lowry Models -- based on statistical methods and analogies rather than on explicit economic or urban theories.
      2. Spatial Economics Base Input-Output Models -- based on integrating a range of separate economic models concerning the structure and functioning of the city at specific points in time.
      3. Activity Based Microanalytical Simulation Models -- based on analysis and modeling of the processes of urban change; upon more explicit representation of particular processes affecting different activities.
2. Optimizing Models -- purpose is to find a 'design' which optimizes a particular function

A. Predictive Models

1. Static Models

Description

• Represent a single point in time, with an objective of adding a land use dimension to a standard 'horizon-year' transport model.

Example Models

• TRANSTEP and TRACKS -- Australian packages developed mainly for traffic and transportation modeling, with some accessibility-based land use models.
• DSCMOD -- From David Simmons Consultancy. Land use add-on capable of working with a variety of transportation models. Requires and independent network model.
• IMREL -- Created as part of a series of transportation studies in Sweden. Depends on a separate transportation model.

Strengths

• With little effort can add an informed indication of land use effects to a transportation study without modifying the transportation modeling methodology.

Weaknesses

• Weaknesses follow from the characteristics of conventional transportation modeling. The 'horizon year' exists in isolation and there is no consideration of the time-scale over which the effects become apparent. This is problematic for land use modeling since many processes take time to have effect and many aspects of land use have complex dynamics.

2. Dynamic Models

Run for a series of time periods, with transportation changes taking one or more periods to have an impact on land use.

a. Entropy Based Lowry Models

Description

• Based originally on statistical methods and analogies rather than on explicit economic or urban theories.
• Most operational urban land use models today can trace their beginning to Lowry's (1964) A Model of Metropolis for the City of Pittsburgh. The original Lowry model incorporates the spatial distribution of population, employment, retailing (the entire service, or "non-basic" sector), and land use within a compact iterative procedure requiring only nine equations and three inequalities.
• In essence, the approach consists of linking together two spatial interaction models. One of these models allocates workers to a predefined set of land use zones on the basis of exogenously supplied basic employment levels. These workers and their families demand services, and these demands are met by means of a second spatial interaction model which allocates this service supply, in the form of 'non-basic' employment, across the same spatial zoning system.

Example Models

• DRAM/EMPAL -- Disaggregate Residential Allocation Models/Employment Allocation Model
• LILT -- Leeds Integrated Land Use/Transport Model

Strengths

• DRAM/EMPAL -- Some scope to introduce constraints or other influences, particularly to account for local knowledge. Calibration of model is relatively easy. Required data sources are generally available.
• LILT -- More elaborate and versatile than DRAM/EMPAL. Explicitly considers processes affecting the location of activities over time (similar to activity-based models). Represents the relationships between transport and the spatial distribution of populations, employment, jobs, shopping, housing, and land utilization.

Weaknesses

• DRAM/EMPAL
  • Modeling process is statistical rather than based on any explicit economic theory.
  • Little or no scope to introduce planning policies other than land zoning.
  • The absence of any mechanism for simulating the land market clearing process underlying multi-year infrastructure change. Clearly, more comprehensive simulation of the land market requires additional data that is generally difficult to collect, notably data on the pricing of land, housing, and other forms of development.
  • It is probably the weakest of the dynamic models.
• LILT
  • Land and property are included in the model but are not treated as a market, and prices are not considered. Implementation and calibration is very time-consuming due to greater complexity and data needs.

b. Spatial Economics Based Input-Output Models

Description

• Based on analysis and modeling of the processes of urban changes -- on more explicit representation than most models of the particular processes affecting the different activities represented.
• An intersectoral approach to economic analysis, this approach provides a general framework from which to being to integrate the manufacturing and other basic industrial activities, which are treated as exogenous inputs to the urban development process by Lowry-based models.
• Treats both land and transportation systems as market processes with endogenously determined costs. This involves the application of neoclassical economic theory to urban land use patterns, notably residential land use, which is allocated across space on the basis of a land market clearing process.
• Land use model integrates three distinct economic models:
  1. An input-output or social-accounting model,
  2. A random utility model of location choice, and
  3. A rent-density function based on the Alonso model.
• The three economic models are solved simultaneously in such a way that the land use model estimates:
  1. The quantity and location of activities and the use and rent of land, both subject to planning (zoning constraints).
  2. Interactions between different activities in different locations, such as the flow of labor from home to workplaces. These interactions can be directly converted into the demand for transportation, replacing the generation and distribution steps of the four-stage model.

Example Models

• MEPLAN -- from Marcial Echenique and Partners; used widely in Europe.
• TRANUS -- from Modelistica; used widely in Central and South America.

Strengths

• Greatest strength of these models is that of all commercial land use model packages, they come closest to modeling interrelated variables in cities.
• Allow analysis of different kinds of policies -- investment, regulation, and pricing -- in both land use and transportation. Rents and utilities can be calculated throughout the land use and transportation calculations, allowing the analysis to extend into cost-benefit evaluation of plans and policies.
• The highly synthetic nature of the models allows most of the description of even the base situation to be estimated within the model, reducing the reliance on observed data.
• Because most of the model is cross-sectional, and only a few variables are explicitly forecast into terms of changes over time, it is possible to implement the model with very little data other than that for the base year.

Weaknesses

• The calibration process is difficult and time-consuming and usually limits it to trying to fit a predetermined specification as well as possible to the data set. Neither the process nor the software encourage testing alternative model formulations. The simultaneity of the three main model components means that, to a large extent, they have to be calibrated jointly, by running the model in a formal search for fit to multiple observed variables. Because each component's inputs are another's outputs, the distinction between dependent and independent variables does not apply and conventional statistical tools cannot readily be used.
• Validation of base year results is problematic. Calibration is difficult since everything in the model is linked. For example, the most readily observed data, link traffic flows, depend on the route assignment which depends on the modal choice model, which depends on the distribution modeling, which depends on the input-output process, etc.
• There is no way to integrate data for years subsequent to the base year. Model may be accurate for the base year, but inaccurate in reproducing the later data set.
• Intrinsically weak in representing the actual processes of urban change because of highly synthetic approach.

c. Activity Based Microanalytical Simulation Models

Description

• Uses a microsimulation approach to model the interrelated choices made by individuals. Looks at detailed information for individuals and households.
• Microanalytic simulation, or 'microsimulation' for short, revers to the method of generating random numbers from within predescribed probability distributions, which numbers are then assigned to a specific response or response value. For travel model purposes, such a response may be associated with a particular traveler attribute or with a specific travel choice. The idea is to generate a series of traveler attributes and/or travel choices in this manner, to build up a detailed representation of specific trips or multi-trip travel activity patterns. Summing over all of these individually simulated travel patterns provides aggregate values for planning studies.
• Based on integrating a range of separately developed economic models concerning the processes of change in the structure and functioning of the city over time.
• The models covered in this section use a sampling technique to deal with the difficulty of dealing with the range of possibilities of change over a given time period. The models use a 'Monte Carlo' simulation to model what happens to a sample of households. Sample results are aggregated to larger scales for final results.

Example Models

• IRPUD -- Developed at the Institute for Spatial Planning at the University of Dortmund, Germany. It is a detailed and powerful model of Dortmund, Germany, and although it has had excellent results, it would be a huge undertaking to implement it in another city.
• MASTER -- (Micro-Analytical Simulation of Transport, Employment and Residence Model) Developed at University College, London, to address a range of problems that were too complex to be represented at the aggregate level of the LILT model.
• TRANSIMS -- A modeling effort funded by the U.S. Federal Highway Administration as an experiment in simulating a complete area-wide set of individual traveler-based urban activity patterns.

Strengths

• The principle utility of the microsimulation approach is that it lets us incorporate a number of dimensions of both individuals and their choice processes which would otherwise require an excessive amount of disaggregation in model-based accounts. Allows level of detail that is impractical in aggregate approaches.
• It is relatively easy to understand and to implement. To create a piece of software to simulate a particular process using 'Monte Carlo' simulation, all that is required is a suitable random number generating routine, a suitable probability distribution (or the raw data itself, perhaps in histogram form), a routine for allocating values between 0.0 and 1.0 to randomly select choices on the basis of this distribution (data), and a routine for collecting the results of the sampling exercise.
• It allows not only the explicit tracking of simulated individuals' (travelers, households, companies) status over time, but also a detailed tracking of the simulated changes in the use of individual land lots. Where suitable time series data exists, evan at a quite aggregate level of resolution, this provides a useful means of checking the reasonableness of the model processes underlying the simulated outcomes. Scrutiny of such microsimulated temporal paths has also been found to provide useful insight into the implications of using alternative model forms as well as alternative parameter values to replicate a particular multistage process. Certainly, microsimulation offers a good deal of flexibility in experimenting with different event sequencing, which is not present in traditional land use or transportation planning models. Its application to the nature and timing of landowner and land developer decision-making processes offers an interesting possibility.

Weaknesses

• There are difficulties because of the randomness and sampling inherent in the model. Because choice processes are random at the individual level, the total results have a degree of randomness. Must use multiple runs to average out results.
• A significant challenge involves the acquisition of suitable data, the determination of a suitably representative sample size for analysis purposes (for which well established methods exist in most cases), and the ability to place such sampled responses within an appropriate modeling framework. Procedures must also be developed to capture the cumulative effects of common activities such as traffic congestion and spatial agglomeration of commercial or industrial activity. This raises some interesting and challenging questions for model design, issues not yet clearly elaborated within the literature.
• Microsimulation is not panacea for data hungry simulation models. Using the technique in conjunction with suitable travel activity scheduling models and sample enumeration techniques offers some attractive possibilities. However, it is necessary to establish carefully the accuracy of the mechanisms being simulated and, in particular, the applicability of generic procedures to different traveler groups.

B. Optimizing Models

Description

• A normative approach; reflecting a long held interest within the planning profession for best possible solutions.
• Emphasis on prediction of future outcomes, or indeed, the replication of current or past ones, is replaced here by efforts to define, or to 'design' more efficient urban futures.
• The resulting urban location-allocation models usually take the form of convex mathematical programs subject to a set of linear planning constraints.
• Intended as a tool which can find a "design" which optimizes a particular function.

Example Models

• TOPAZ - Technique for Optimum Placement of Activities into Zones - originally developed in Australia and used widely there as well as in a few U.S. applications.
• POLIS - currently used in the San Francisco Bay Area.

Strengths

• Can make use of generally simpler mathematical forms that are readily tied to theories of system efficiency, cost minimization, or net gain.
• Avoids the need to account for a wide range of empirically observed idiosyncrasies.
• Uses mathematical programming frameworks to state the urban land use-transportation problem as a single, if rather complex, mathematical formulation.

Weaknesses

• Informative for research and long-term planning, but difficult to link to practical planning problems of particular cities.

Descriptions of Specific Land Use Forecasting Models

This section provides a more in-depth look at some of the major models in use, including one of each of the three kinds of dynamic predictive models (an entropy based Lowry model, a spatial economics based input-output model, and an activity based microanalytical simulation model), and one optimizing model. In addition, there is a description of a recently developed hybrid model which includes an input-output based model and a GIS allocation model. This section is primarily descriptive in nature and is intended to provide insight into the functioning of the different types of models. Strengths and weaknesses of the various model types have been listed in the previous section.

Entropy Based Lowry Model - DRAM/EMPAL (Disaggregate Residential Allocation Model/Employment Allocation Model)

• The employment allocation sub-model, EMPAL, allocates employment to zones using exogenous forecasts of total employment by type (basic and non-basic), together with zone-specific employment levels and growth trends and zone-specific measures of accessibility to the work force.
• The residential allocation sub-model, DRAM, then forecasts the future location of households given this distribution of employment and the attractiveness (including accessibility) of the zones. Future land consumption also is forecasted using base year information and exogenous forecasts.
• DRAM and EMPAL are based on singly-constrained spatial interaction model formulations incorporating multivariate attractiveness functions:
  • Potential employment in zone j occupying workers from zone I is calculated on the basis of the previous population of zone I, the previous attractiveness of zone j (as indicated by jobs and land area), and the current disutility of travel between the zones.
  • Future employment at zone j then is calculated as a function of previous employment and the total potential in j from all worker-generating zones.
  • Similarly the number of households living in zone I is treated as a function of the attractiveness of zone I (in terms of developable land) and the accessibility of zone I to employment zones.
  • An additional factor in the attractiveness equation has the effect of ensuring that households of similar types will tend to cluster together.
• Data requirements on a zonal level include:
  • employment for each district employment sector (usually 3-5),
  • population by category (usually 4 income groups),
  • land allocation by activity,
  • transportation descriptors (links, speeds, capacities), and
  • data on employment by sector for the period subsequent to the base year are used for calibration;
• Forecasts require:
  • estimates of regional employment by sector,
  • regional population by category,
  • trips per person by trip purpose, and
  • future networks (links, speeds, capacities).
• Forecasts are done on a zonal level and in five-year increments, with output from one forecast year becoming the input for the next.
• Land use outputs include:
  • employment by sector,
  • population by income group and working/non working category,
  • land allocation by activity, and
  • vacant land.
• Travel outputs include:
  • trips by origin/destination and purpose,
  • trips by purpose,
  • income group and mode,
  • average travel time by origin/destination and mode,
  • energy consumption by trip purpose, social/car ownership group, and mode, and
  • air pollution estimates are an additional option.

Spatial Economics Based Input-Output Model - MEPLAN

MEPLAN is a land use allocation model used in a number of overseas applications but not yet implemented in the U.S. The various versions of MEPLAN incorporate an approach based on input/output models. The following is a summary of the MEPLAN model:

• Land and transportation are treated as two parallel and interacting markets. Behavior in each system is modeled as a response to price or price-like signals (including travel disutility).
• A key relationship is the effect on locational accessibility of travel cost and time changes, which find their way back, in a temporally lagged manner, into a set of activity-location models.
• In terms of the simulated dynamic, land use is influenced by the pattern of use in the prior period and by previous period transport accessibilities. Transport is influenced by previous infrastructure and present activity patterns arising from land use.
• Location of economic activities and the interactions between them are predicted by an input/output model.
• Input/output model represent movements of labor, plus the demand from industry and households for services, floor space and land.
• Households are differentiated by socio-economic group, having different qualifications for labor and differing consumption demands.
• Supply and demand (for both activities and space) respond to 'prices' which may be expressed in monetary terms or disutilities.
• The demands for transport are calculated directly from the interactions predicted by the spatial economic system defined within the land use model. The need for a trip distribution modeling step is obviated by the direct translation of what are termed trade flows, or "trades" from the land use model into suitable modal volumes.
• An interzonal mix of 'trade flows' between production and demand zones is produced, subject to specified constraints on production in particular zones, and a distribution matrix is developed based on elasticities of travel conditions and prices.
• Interzonal flows are then converted to trips by purpose and socio-economic group. An elaborate interface between the land use and transportation models translates these trade flows (labor, materials, services) into mode specific trip matrices. Trips are assigned to modes by logit models and, subsequently, onto the highway network using a probabilistic, multipath assignment routine that takes into account costs and congested travel.

The MEPLAN model requires a large amount of data inputs and its specific applications are time consuming. While data requirements for a fully implemented model are potentially rather daunting, it is claimed that the generality of this highly synthetic modeling framework allows it to be tailored to handle relatively modest data inputs no more than less comprehensive systems which do not contain any land rent, production costs, or other pricing variables. However, there are often difficulties involved in selecting the model's many parameters, typically involving extensive iterations and retrials, not always in a purely automated fashion.

Activity Based Microanalytical Simulation Model - MASTER (Micro-Analytical Simulation of Transport, Employment and Residence

The MASTER model, developed in the University College, London, is an integrated land use transportation model based on microsimulation, using Monte Carlo methods to simulate the decision processes that a set of individuals and their households go through over time. The central purpose of the model is to examine the influence of transportation on land use.

Households are considered one at a time, but are grouped together at certain points in the simulation to allow use of aggregate values. The model treats the inter-related choices made by individuals in some detail. It includes explicit representation of life-changes including birth, death, marriage, divorce, obtaining/changing/losing job, etc.

The following indicates some of the scope of the simulation process:

• Households are considered one at a time, but are grouped together at certain points in the simulation to allow use of aggregate values.
• The first processes considered are demographic ones, including aging, giving birth, dying, divorce, and marriage. Population change is modeled explicitly within the model. Marriage and divorce lead to the creation of new households.
• Both public and private housing markets are recognized, and dwelling occupancies are tracked from one period to the next.
• Choice of residence zone is based on a weighted function of generalized travel to work costs for the head of household.
• Both the supply of jobs and dwellings are exogenous, zone-specific inputs to the model.
• Choice of dwelling type is based on household size and composition. Other household members' job selections are also considered, and changes in economic status for simulated individuals may include redundancy and retirement.
• The transportation processes modeled are becoming an auto license holder, car ownership, car availability, and choice of mode to work: each variously functions of age, sex, household income, household composition, and mode-specific costs of travel. To change mode of travel to work, either a change in job or home location or a change in vehicle ownership or availability must occur or a significant change in travel costs must be introduced.
• Logit forms are used to select the mode of travel. If family members work along the same travel corridor, carpooling is also possible. Only the work trip is discussed, shopping and other trip purposes are not included in the process described. Assignment of traffic to specific routes is also not included in the model, therefore, congestion is not modeled explicitly.

Optimizing Model - POLIS (Projective Optimization Land use Information System)

POLIS is a combined land use-transportation model built around a single mathematical programming formulation. It has been applied in the San Francisco Bay region by the Association of Bay Area Governments (ABAG). The model can be stated as a single mathematical program which seeks to maximize jointly the locational surplus associated with multimodal travel to work, retail, and local service sector travel, and, significantly and jointly, the agglomeration benefits accruing to basic-sector employers.

A more complete description of the POLIS model and the general projection process conducted by ABAG in the San Francisco region is available in the section surveying MPO land use forecasting methodology.

Other Models

Robert A. Johnston of the Division of Environmental Studies of the University of California at Davis and Tomas de le Barra of Modelistica (producers of the TRANUS model) recently released a research paper which describes a study using a suite of modeling packages to provide a comprehensive evaluation of regional plans.⁽⁴⁾

The concept behind the study was to use a comprehensive market-based urban model (in this case, the TRANUS model) in conjunction with a GIS-based land allocation model (in this case, the California Urban Futures Model (CUFM)) in order to produce detailed land use projections in GIS which could then be used to derive a variety of environmental impact assessment models. The practical objective is to evaluate regional 'sketch plans' for impacts on economic welfare emissions, energy use, important habitats, prime agricultural lands, wetlands and so forth. The eventual goal is that the models could be used to design regional plans that satisfy certain criteria.

TRANUS is a dynamic spatial allocation model, based in random utility theory and bid-rent theory, similar to the MEPLAN model. Johnston and de la Barra chose the TRANUS model over MEPLAN because TRANUS runs in Windows, handles interregional as well as regional travel and goods movement, has multidimensional multipath stochastic route choice, and because of greater ease of calibration.

The study was carried out for the four-county Sacramento Area Council of Governments (SACOG) planning region. SACOG's district structure was adopted in the study in order to simplify data comparisons with their travel model outputs. A simplified version of SACOG's transportation network model was used.

Transportation Data

The transportation calibration data inputs included: road counts; public transport route boardings; value of walk, wait, and ride time by mode; average parking costs by zone, free-flow speeds by link type; transit fares; operating costs by transit operator and auto user; fuel consumption; average occupancy for autos by trip purpose and for transit vehicles; car availability by trip purpose by household income class; number or trips by zone pair; proportion of trips in morning peak purpose; and cordon volumes.

Land Use Data

The land use calibration data inputs included: number of households by income group by zone; average number of persons per household by income class; average acres per dwelling by income class by zone; average acres per employee by activity type by zone; land sales prices per acre by type of land use and density; land use designation in local plans, and zone; number of employees by employment category by residence zone and workplace zone by income class; average income per capita by income class; household expenditures for land, travel, retail, and other; and flows of schoolers by residence zone and school zone by income class.

Scenario Data

The future scenario input data included (by 5-year period): network changes; changes in transit headways and fares; roadway tolls; parking charges; allowable growth in each land use by zone; building density caps by land use by zone; and projections of total regional employment.

Scenarios

Four different long-range growth scenarios for the year 2015 were developed:

1. Trend Scenario - including a few small projects such as a modest LRT extension, a few miles of new HOV lanes, and some arterial widening.
2. HOV Lanes - about 200 lane-miles of new freeway HOV lanes on all major freeways in the urbanized area.
3. Beltways + HOV - HOV as above plus an outer 3 outer beltway segments.
4. Light Rail Transit (LRT) and Pricing - Major LRT extensions with improved headways for buses and LRT and more bus feeders and park and ride lots as needed, plus average all-day parking charges of $5 in the CBD and $2 elsewhere.

The major improvements were scheduled in 2005, so that there would be two rounds of land use effects (at 5 year model iteration increments) for most facilities.

The GIS Model

The land use projections (acres by land use type by zone) for 2015 from TRANUS are fed into a rule based land use allocation model using GIS as its data structure. Linking the two models allowed the study to perform market-based policy experiments with TRANUS and to use CUFM to produce detailed land use maps (GIS coverages) for environmental impact assessment.

The GIS model used in the study is the California Urban Futures Model (CUFM), a large-area housing simulation model with capabilities for analyzing transportation improvements. by John Landis at UC-Berkeley. It is a nonlinear programming model that allocates residential land development to polygons which are ranked according to profitability for the developer. Profitability is calculated as a function of accessibility to roads and services, slope, local government fees, land prices, and several other variables. Developable Land Units (DLUs) are created by overlaying a variety of GIS coverages, such as city boundaries. wetlands, slope, land use type, and roads, which produces polygons in the GIS.

Industrial land uses are allocated first by CUFM to polygons which are industrially designated in local land use plans. Then commercial and residential uses at declining densities are allocated, with Residential Very-Low Density allocated last. All land uses are allocated to DLUs in order of the DLU accessibility rankings, except for Residential Very-Low Density, which is randomly allocated to DLUs to simulate amenity-seeking locational behavior.

Vacant land rate statistics are only taken into consideration in the allocation process for residential low and very-low uses, which cannot generally 'outbid' existing land uses but which consume over 80% of the land in all scenarios. For residential low and very-low uses, the vacant land rate is determined at the minor traffic analysis zone level. If over 75% of the minor TAZ is vacant, up to 90% of the vacant land can be allocated. If 50-75% is vacant, up to 75/O of the vacant land can be allocated. If 10-25% is vacant, up to 50% can be allocated, and if less than 10% is vacant only up to 10% of the vacant land may be developed. This system leaves at least 7% of vacant land in any zone and discourages development of land where the vacancy rate is low. Even in built-up areas, many U.S. regions contain considerable amounts of vacant lands.

The GIS outputs of the land use allocation produces a somewhat artificial-looking land use pattern. However, the objective of the process is a realistic layout of land uses in terms of types of locations. The authors of the study claim that the study has provided a reasonable allocation of land uses in this regard.

Future steps that will be taken in this project to improve the modeling process include recalibrating the TRANUS model, adjusting model parameters, improving land price datasets, and comparison of model projections to expert panel predictions. In addition, the project will attempt to design scenarios to reduce environmental degradation, use projections in combination with travel models to generate regional economic welfare measures for different scenarios, add additional environmental impact models, and perform full social welfare evaluations.

For further information on this project, refer to Johnston's and de la Barra's paper; and the Modelistica home page for information on TRANUS (http://www.modelistica.com/).

Conclusions on Land Use Forecasting Modeling

David Simmons makes the point that "the subject of urban modeling has flourished in a mainly academic context, during two decades in which the possibilities of such modeling have very largely been ignored in planning practice."⁽⁵⁾ In a review conducted by Cambridge Systematics and Hague Consulting⁽⁶⁾ in 1991, only a few of the top 18 metropolitan areas were using integrated models in their planning processes.

There are several reasons for the lack of implementation of models in common planning practice. Frank Southworth provides an overview of these issues.⁽⁷⁾ The following section is derived from Southworth:

In their relatively brief history, land use-transportation models have been subjected to a good deal of criticism (see the Winter, 1994 edition of the Journal of the American Planning Association for a retrospective). Past criticisms have tended to revolve around:

1. practical issues of data availability and quality, as well as computational requirements and ease of use,
2. the role such models are to play in the planning process, and
3. conceptual issues of model realism and hence usefulness.

While recent computational advances have done much to remove concerns over both computer costs and computer run times, the other issues remain.

Role of Land Use/Transportation Models

The role of large scale urban land use-transportation simulation models remains a cause for debate. Should they be considered as tactical or as strategic planning tools? If used as tactical planning tools, their most common application would probably be to evaluate travel policies along specific urban corridors, with an eye to an environmentally influenced benefit-cost ratio being realized within a suitable time period. Even so, such an evaluation period might cover as long as 15 or 20 years depending on the project being proposed (i.e., up to the expected lifetime of a typical urban highway pavement, if the addition of new infrastructure is involved).

However, a danger with using models solely to analyze individual travel reduction projects is the potential for disjointed, piecemeal planning. Ideally, we need to find a way to embed such project evaluations within more strategically developed, area-wide transportation plans. If these plans are to make the sort of contributions to petroleum savings and CO₂ reductions which have come from more efficient engine and fuel technologies, area-wide impacts will almost certainly be required. We also need to think in terms of longer planning horizons. Even a planning horizon of 30 years may not be long enough to capture the true impacts of a plan which contains significant transportation infrastructure investments. Such plans may go onto influence urban form, and therefore urban travel activity, for many years into the future.

Some experts argue that in searching for such a strategic role we may be trying to get too much detail into our models. As we add more detail and functionality to what are already rather ambitious models, we lose flexibility in their application and increase expensive data requirements. Modeling procedures might be better viewed as an aid to comprehensiveness of understanding, rather than comprehensiveness in forecasting. This correlates with the use of integrated models as tools for evaluating the robustness and resilience, rather than the details, of alternative urban and regional plans. In the end, perhaps, accurate prediction matters less than flexible normative planning, based on an intelligent assessment of the most likely directions of certain trends.

To carry out such planning, mathematical, computer-based models would seem to be our only realistic alternative if we wish to apply, and properly test the results of applying, a formally developed logic behind our planning decisions. Without reasonably comprehensive models, we cannot hope to simulate the often nonintuitive effects of combining a wide range of policy options within any single plan. This, however, raises the issue of how we gain confidence in our model-based results. Such a question moves us on to issues of model validation.

Model Validation

Validation means carrying out checks to establish how well a model did in forecasting a future situation by comparing the model's results with observed data. However, remarkably few validation exercises are reported in the modeling literature. Travel data availability constitutes the major constraint on validation exercises to date, especially data covering time intervals long enough to capture some of the important changes in urban infrastructure and land use. Clearly, greater emphasis on validating the models is required, including the establishment of procedures to track the major data sources necessary to calibrate them. This constitutes the most significant obstacle to model validation and, by implication, further useful model development. More comprehensive models mean more demanding data requirements.

Given current data limitations, how are we to assess the value of such models in a strategic context? When an analysis task involves forecasting over a long period of time with substantial deviation from historical experience to be expected, an assessment of a simulation model is best focused on 'realism in process.' This contrasts with more direct assessment of a model's predictive capability, involving the above discussed comparison of model results against a known, and empirically observed, reality; a validation process termed 'realism in performance.' At the present time any discussions of current model weaknesses and associated research needs are necessarily focused heavily on such realism in process. However, more realism in process suggests that we also use more behaviorally based (i.e., more realistic) models. That is, it suggests that we focus more attention on how travelers behave and, for the purposes of policy impact assessments, how such behavior changes over time once policies are implemented which act upon it. This in turn suggests that more attention be given to the collection and use of longitudinal data sets. In particular, multiwave traveler panel surveys - collecting information from the same group of travelers at discrete time intervals - are an important data collection option. A first step is to determine which are the major variables of interest to such longitudinal analyses and (since cost of data collection remains the major constraint) which data we can effectively relegate to less regular data collection activities. To do so, we need to better understand the causes of current variability in travel demand.

Integration of Modeling with the Planning Process

A strong argument can be made that as far as land use-transportation modeling efforts to date are concerned, tool making is more advanced than theory. Research has drawn on a great variety of mathematical, statistical, and computational methods in its search for empirical validation and subsequent practical applicability. Yet the application of many of these techniques is much less widespread within the planning profession than might be expected. Few practicing regional or metropolitan planners calibrate their own multinomial logit models or experiment with alternative land availability or density constraints as part of a nonlinear mathematical programming exercise. Nor is the issue simply one of technical training. In order to encourage practicing planners to make greater use of the models which do exist, the models need to be made easier to use.

If planners from more than one jurisdictional level (local, metropolitan, statewide, or regional) could be brought together by use of a common, easy to use modeling, possibly game-playing software, then improved models could possibly be transformed into consensus building tools, rather than the seemingly arcane components of a planning process in which only one or possibly two experts within any metropolitan planning agency have anything to do with them directly.

The emergence of reasonably priced and generally accessible geographic information system (GIS) software is the latest step in this development of decision-support tools. By linking a relational database management tool to software programs for manipulating spatial primitives (points, lines, polygons), adding the land use and transportation modeling subroutines themselves, and building around all of these an easy to use, map-based interface, we have the principal components of a spatial decision support system (an SDSS). Ongoing developments in the SDSS arena promise more effective manipulation of both spatial and nonspatial data elements, in the short term through the more efficient selection of which computations to carry out via database manipulations and which to continue to model through the more context-specific algorithms. The field of urban transportation modeling is only now beginning to make use of such GIS tools.

However, current commercial GIS packages are still some way from being the spatial decision-support tools we need. Experience with such software in the field of urban and regional transportation modeling has been quite limited to date. Education in how to construct, adapt, and use such software tools is now required within the transportation planning profession.

Current Efforts at Model Development

The traditional four-step transportation planning model has been the focus of a good deal of criticism for many years. Much of the criticism within the modeling literature argues that we need to place both household and company-based travel decisions within more behaviorally realistic decision-making frameworks. Treatment of travel as a good composed of separately modeled attributes of frequency, mode, destination, and route choices is being challenged. Within the United States, the need for metropolitan planning organizations to address the vehicle travel reduction requirements of the 1990 Clean Air Act Amendments (CAAA) is now leading to a new round of model development, known as the Transportation Model Improvement Program (TMIP).

The Travel Model Improvement Program (TMIP) is an initiative sponsored by the Federal Highway Administration, the Federal Transit Administration, the Office of the Secretary of Transportation, and the Environmental Protection Agency.⁽⁸⁾ The purpose of the Program is to conduct research and development to improve travel demand and supply forecasting models. The models developed in this program will determine the effects of transportation improvements on congestion, air quality, and land development. The Clean Air Act Amendments of 1990 (CAAA) provided major motivation for travel model improvements. That act mandates details and accuracy not currently available from the travel models. Requirements of the Intermodal Surface Transportation Efficiency Act of 1991 (ISTEA) have also prompted travel model improvements.

The objectives of the Program are:

1. To increase the policy sensitivity of existing travel forecasting procedures and their ability to respond to emerging issues including environmental concerns, growth management, and changes in personal and household activity patterns, along with the traditional transportation issues,
2. To redesign the travel forecasting process to reflect today's traveler behavior, to respond to greater information needs placed on the forecasting process, and to take advantage of changes in data collection technology, and
3. To make travel forecasting model results more useful for decision makers.

The TMIP is organized in five activity tracks. Track E of the project is involved in assembling information relating to land use forecasting methodologies.

TMIP sponsored a conference on land use modeling in February 1995.⁽⁹⁾ It was structured as a series of parallel workshops, with one day spent on reviewing the current state of the practice, and a second day dedicated to identifying directions for long-term model development.

The conference proceedings report includes the following papers:

"Current and Future Land Use Models" by Michael Wegener, Ph.D., Institute of Spatial Planning, University of Dortmund, Germany. Provides a review of the current state of the art of operational land use transportation models using criteria such as comprehensiveness, overall structure, theoretical foundations, modeling techniques, dynamics, data requirements, calibration and validation, and operationality and applicability. In addition the paper provides speculation regarding future model development.

"Draft: Data Requirements for Land Use Modeling: First Thoughts and a Preliminary Assessment" by Michael Batty, Carmeller J. Cote, David Howes, Pat Pelligrini, and Xiaohua Zheng, State University of New York. Provides a preliminary assessment of various data sources available for land use modeling and forecasting. Discusses the current problems in existing data sources and the compilation of data for modeling purposes.

"Understanding the Decision Makers: Policy Requirements for Land Use Modeling", Robert T. Dunphy, Urban Land Institute. Provides a review of the role of forecasts and market information in real estate development decisions.

The report also includes summarized results from the 6 workshops held in the conference. It is beyond the scope of this report to summarize the workshop reports. For a full report of conference proceedings contact:

Joe Kammerman
TMIP Report Request
c/o Federal Highway Administration
400 7th Street, SW
HEPM-30, Room 3232
Washington DC 20590
Phone: 202.366.4054
Fax: 202.366.7660
email: joseph.kammerman@fhwa.dot.gov

Considerations in Choosing a Land Use Forecasting Model

This section is derived from a paper entitled Land Use Models in Transportation Planning: A review of Past Developments and Current Best Practice by Britton Harris.⁽¹⁰⁾ It provides an overview of some of the major considerations that should be taken into account in selecting a land use modeling package.

Conceptual Considerations

• Must be realistic; must reproduce the behavior of system users and independent developers in the real world under varying conditions.
• Must accept as inputs not only the real influences on behavior such as household income and auto ownership, but also possible policy inputs such as parking policies, fuel prices and various influences on land use.
• The results of modeling must provide acceptable detail with respect to the policy objectives of the planning process:
  1. For transportation planning this means that they must reflect not only cost and convenience, but also emissions and other environmental impacts.
  2. In land use, this means that they must reflect the impact of plans on development, densities, and user satisfaction in various land and labor markets.
  3. Both must be concerned with overall objectives like amenities, environmental impacts, equity, desegregation, and the conservation of public and private resources.
• These requirements lead to the necessity of using increasingly disaggregated models which, however, increase computational costs and turnaround time and adding to the difficulties of securing and using adequate data. A compromise must be made between perfectibility and practicality of use.

Practical Considerations

• Relevant considerations include the user friendliness of the model, the ease with which new planning scenarios can be entered for testing, the readability of the results, required computational equipment, and the turn-around time for experiments and plan testing.
• The most important consideration revolves around data requirements. Many important needs for data can be met by the U.S. Census, but there are some problems. These are listed in the following:
• Transportation planning areas should, at some level of aggregation, be coterminous with Census tract boundaries.
• Land use information as such cannot be secured from the Census. Vacant land must be measured in every zone for the purposes of determining the future potential for expansion. The measurement of land devoted to different purposes is necessary to determine densities and land requirements for future development. Distinctions should be made between industrial, commercial, service, retail trade, residential and other uses.
• Employment data by zone is needed for many reasons. Different trip types correspond in part with different types of employment, and employment may be the best available indicator of different levels of certain activities such as retail trade. For some metropolitan areas, the census has tabulated employment by tract and by income, but this is not adequate for studying employment by industry type.
• Methods for calibrating models differ; cross-section calibration is simpler and requires less data. Calibration using multiple time periods can force a choice between the better detail of more recent data and the availability of more than one set of observations.
• Land use models can be run for larger zones with less transportation detail than transportation models. This requires the preparation of aggregated or spider networks, with some attendant difficulties in establishing capacities. Land use models can generate trip-tables for the journey to work and for home and work-based shopping, but for input into a detailed transportation model, these would have to be disaggregated.

Final Recommendations

• A model should be selected which is moderately disaggregated and whose underlying concepts are as realistic and as economically based as possible.
• Transportation conditions and available choices as to housing, industrial sites, access to amenities and to the labor force, should enter intimately into all locational decisions which are modeled in the system.
• An accurate delineation of choices implies that the model will distinguish among different types of housing and other developed space, or different types of land for development.
• The model should be doubly constrained, and with meaningful constraints at both origins and destinations. Wherever possible, the equilibrium which is sought should be a form of "market clearing."
• The degree of disaggregation should cover two to four types of households, probably separated by income level. Many types of housing (not all represented in a given residential area), and at least three types of employment, including manufacturing, retail trade, and other services, some of which should be broken into subclasses.
• Data requirements and methods of calibration should be well-specified by the vendor, with the cooperation of the users.
• Running times and equipment requirements are very important, and special consideration must be given to trade-offs between speed and accuracy.

Survey of MPOs About Land Use Forecasting Methods

This section is intended to provide a review of existing Metropolitan Planning Organization (MPO) land use forecasting practice.

A survey of the MPOs of the 35 largest U.S. metropolitan areas conducted in 1994 and 1995 by the Institute of Urban and Regional Development at the University of California at Berkeley included information on land use forecasting methodology. This study included the following findings about the 35 MPOs:

• "Eleven reported using DRAM/EMPAL (Disaggregate Residential Allocation Model/ Employment Allocation Model) or a variant, often in conjunction with less formal methods such as local review.
• Twelve reported using some other type of model, usually developed in-house. These were usually based on some combination of regional demographic and employment forecasts, allocated to a zonal level, in conjunction with local land use plans.
• Nine reported basing land use projections primarily on informal assessments of development trends and local conditions.
• Three reported doing no land use forecasting although one of these was in the process of developing a model." ⁽¹¹⁾

An informal survey of MPOs was conducted by the Austin Transportation Study, the MPO for the Austin, Texas region, in late 1996 and early 1997. The survey initially started with MPOs for regions with roughly the same population size as Austin's, but expanded to include the activities of several other MPOs which have noteworthy projection methodologies. 13 MPOs were included in the survey.

Questions concentrated on the methodology used to forecast population and employment by Traffic Analysis Zone (TAZ) or equivalent, though some information is presented on regional forecasting methodology and data collection procedures.

The following chart summarizes the results of the survey. More detailed descriptions follow.

MPO Region	Land Use Forecasting Methodology for TAZs
Albuquerque, New Mexico	In-house statistical regression methodology based on historical data. Constrained/modified by land use plans and expert knowledge.
Charlotte, North Carolina	'Delphi'-type technique - successive rounds of expert negotiation.
Columbus, Ohio	Expert knowledge based on land use plans and development activity. Local review.
Denver/Boulder/Longmont, Colorado	In-house statistical model based on multi-variate attractiveness index and local comprehensive plans. Model is designed to study alternate urban form scenarios for the preparation of a long-range integrated regional development and transportation plan.
Las Vegas, Nevada	In-house statistical model based on multivariate attractiveness index and future land use plans. Also factors in the location of major development projects.
Louisville, Kentucky/Indiana	In-house statistical regression methodology based on historical data. Local review.
Madison, Wisconsin	Reiterative statistical process based on historical trend analysis combined with land use plans and constraints.
Minneapolis-St. Paul, Minnesota	Performed by localities using regional projections developed by the MPO. Methodology varies among localities.
Orlando/Kissimmee. Florida	Implementing a new system using DRAM/EMPAL for allocation at the sector level and a small-area model for allocation at the TAZ level.
Phoenix, Arizona	DRAM/EMPAL model for county and subregional projection. GIS-based (ARC/INFO GRID) small area model (SAM) based on attractiveness index is used for allocation by 1 acre grid cells. This projection is aggregated into a TAZ projection.
Portland-Vancouver, Oregon, Washington	Uses a mix of technical analysis (statistical regression based on historic trend data) and a series of expert review panels in an iterative process to allocate the regional forecast to subareas. Sub-area projections are allocated to 1 /16 acre grid cells based on comprehensive plan data in a GIS-based system (ARC/INFO GRID). The small-area projections are aggregated into TAZ projections and submitted to another series of expert review panels for final review.
Providence, Rhode Island	Uses proportional statistical allocation. Modified by local review.
San Francisco, California	Uses POLIS optimization-type model for regional projection. Subarea Projection Model (SAM) is used for small-area forecast.

Albuquerque, New Mexico

1990 Population - Albuquerque NM, MSA - 589,131

MPO: Middle Rio Grande Council of Governments
Dave Abrams
317 Commercial St,. NE, Suite 300
Albuquerque, NM 87102-3429
(505) 257-1750
fax: (505) 247-1753 (call to confirm receipt)

Population Projection Methodology

Uses an in-house developed calibrated regression technique based on historical land use data. The dependent variable is the population growth index for each TAZ.

Independent variables:

1. TAZ proportion of vacant developable residential land. Usually this is calculated as the acres in a TAZ divided by the acres in a subcounty area of 20-80 TAZs.
2. Housing price quartile (on a 1-4 scale) - in the Albuquerque market, higher priced areas tend to attract development at higher rates.
3. Proportion of residential land developed. This has a negative correlation with growth, since areas that are 80 to 90% built-out generally only have less desirable lots available.
4. Inverse of travel time from the TAZ to the CBD (infill variable). The inverse is used to accentuate differentiation between areas that are closer to the CBD.

Employment Projection Methodology

1. Allocate all of the employment that is 'on-line' - i.e. known business development that is planned for the next year or two.
2. Rank TAZs (1 -10 scale) within the subarea by attractiveness to basic employment and by attractiveness to population serving development. This ranking is done using a consensus technique among planning staff.
3. Determine land capacity - acreage and density. Look at land use plans.
4. Use algorithm which allocates the employment based on the attractiveness rate. The algorithm reassigns employment to other zones when a TAZ has 'filled' its employment capacity.

Charlotte, North Carolina

1990 Population-Charlotte-Gastonia-Rock Hill, NC-SCMSA (NC part) - 1,030,596

MPO: Charlotte Department of Transportation
Joe McClelland
600 E. 4th St.
Charlotte, NC 28202
(704) 336-3908
fax: (704) 336-4400

County and sub-county (11 subareas) were forecast by outside consultant; allocation to TAZs was performed by local staff using a Delphi technique (several rounds of expert negotiation). For neighboring Union County area, state level forecasts are combined with forecasts for adjacent areas of Mecklenburg County.⁽¹²⁾

Although they have examined options for doing multi-county regional modeling,⁽¹³⁾ they don't have the resources and the political will does not exist to do this at this time.

Columbus, Ohio

1990 Population - Columbus, OH MSA - 1,377,419

MPO: Middle Ohio Regional Planning Commission
Nancy Reger
285 E. Main
Columbus, OH 43215
(614) 228-2663
fax: (614) 228-1904

Population

County population projections for 2020 provided by the Ohio Department of Development are used as a control figure. Distribution process by TAZ (over 1,000 zones in the region) is based on subdivision, permitting and building activity. The local comprehensive land use plans, sewer and water connections, and school district boundaries are also looked at. These forecasts are balanced against the control figures. The projections are then brought to the cities for review. These are revised as necessary.

Employment

A land use employment base data set is updated every five years. This is done with a workforce survey of all the new businesses in the area. Also uses a Polk Directory (business directory by address) to compare possible land use or business changes along the major employment corridors.

The distribution of employment by industry is obtained from the Ohio Department of Labor Statistics. An estimated employment figure is derived from population forecasts which factor in commuters from other regions, residents commuting to other regions and unemployment rates.

Employment is distributed into the TAZs using the Ohio Dept. of Labor Statistics forecasts as a guide to distribute new employment into retail, office, and industrial sectors; by examining the cities' comprehensive plans and permitting activity; and through information about future development plans (often found in the newspaper).

Denver/Boulder/Longmont, Colorado

1990 Population - Denver-Boulder, CO CMSA - 1,848,319

MPO: Denver Regional Council of Governments
Tim Sheesley
2480 West 26th Ave., Suite 200-B
Denver, CO 80211-5580
(303) 480-6765
main: (303) 455-1000

As part of a long-range planning process, the Metro Vision 2020 project, DRCOG is modeling alternative allocations of households and employment by Traffic Analysis Zone (TAZ). The Metro Vision project is designed to study alternative urban forms leading to the preparation of a new long-range, integrated regional development and transportation plan for the Denver region. Four Metro Vision alternative urban form scenarios were chosen for consideration. The four alternatives were meant to define very different visions of the future development of the region. The four alternatives were:

• A dispersed pattern similar to current trends,
• A compact form which encourages redevelopment rather than continued development on the fringe,
• A corridor form that has urban development following radial transportation facilities, and
• A satellite form that assumes most development will occur in outlying communities, both existing and now.

In developing the model for the project, DRCOG had several factors which were of key importance:

• Needed a method that could adequately model alternative urban development patterns which are significantly different from previous historical trends.
• The method needed to be systematically and consistently applied across the region for each of the alternative urban from scenarios.
• The method needed a sound theoretical basis for determining the relationships in the model and it needed to be clearly documented as well as easily modified between iterations to reflect differing policy options

Both 'top-down' and 'bottom-up' approaches were used to combine professional knowledge and field information in the 'visioning' process.

Sub-Regional Control Values

The initial step in the modeling process sets sub-regional control values for households and employment which clearly differentiate the four alternatives. This is because the transportation, air quality, and water quality models are relatively insensitive so that relatively large differences between alternatives are needed to test their impacts on the evaluation criteria.

Five sub-regional analysis areas were defined to characterize portions of the region with similar demographic characteristics. The population, household, and control totals for these areas were derived from a series of invited-expert workshops. The figures developed in these workshops were reviewed and modified by technical advisory committees and staff.

Zonal Attractiveness Index

A Zonal Attractiveness Index was developed to distribute growth within the sub-regional analysis areas based on the characteristics of individual zones and their resulting attractiveness to development under each alternative.

To employ the Zonal Attractiveness Index at the TAZ level, a large spreadsheet model was created that included a number of data variables for each TAZ. The variables were weighted and evaluated by a number of 'if-then' statements to rank order the TAZs for the suitability for development and then the TAZs were then 'filled-in' in this order by the allocation algorithm. By changing the weightings and the 'if then' statements for each alternative, the process allows for the consistent, predictable and relatively rapid development of data sets.

Land Use Data

Land use data by TAZ was obtained from the region from a variety of sources. Local jurisdictions were asked to update land use coverage for ten land use types. Where no local information was provided, DRCOG staff used aerial photos and zoning maps to estimate land use coverages.

The land use information was used to allocate population and employment using a variety of strategies based on the regional alternative that was being modeled.

Attractiveness Index Variables

DRCOG staff identified 17 index variables that could explain variations in attractiveness of land for development and the potential location of population and employment growth given differing policy options. Variables were selected from those having theoretical or intuitive explanatory strength in forecasting the location of growth as well as to consider the interaction and linkages between transportation, urban form and environmental quality. These were derived from local knowledge, planning judgment and current literature. The GIS system was used to generate percentages of land use attributes such as open space and water bodies by TAZ.

Index Weightings

Once the index variables were identified and index values calculated, the final step in developing the Zonal Attractiveness Index was the development of weights to these values in order to establish the relative effect of the indices for each alternative. Ideally, this weighting process would be the result of research and empirical evidence, however due to time constraints and the lack of historical examples of the development patterns, this could not be done. Instead, a structured process of expert judgment was used.

Several stages of expert review and discussion were used to develop and modify the weightings for the different growth scenarios. Finally, the weightings for the indices by each alternative were multiplied by the normalized index values for each TAZ and summed to produce the Zonal Attractiveness Index for each TAZ. The TAZs were then rank ordered on the basis of their ZIAs for each alternative and the resulting order formed the basis for the allocation algorithms used in the spreadsheet.

Metro Vision Distribution Process

Once attractiveness indices are determined for each TAZ, the amount of new growth the zone can absorb is calculated. The distribution process involves introducing the various model inputs to a series of equations that determine how many households or jobs will be added to each TAZ according to the zone's analysis area.

The region was divided into five concentric ring Analysis Areas based on similar characteristics that are noticeable as growth takes place through time. The five analysis areas are: Central Business District (CBD), Inner, Fringe, Satellite, and Rural. The TAZs are sorted by highest weighted importance in each Analysis Area.

The distribution algorithms take into account the following factors: developable land percentage residential and commercial development potential, percentage of developed land in the Analysis Area, residential and employment density, percentage of zone that is in the projected 2020 urban area, and number of projected new households and employment.

Employment and households are distributed to the TAZs starting with the most attractive zones, until the control total for each Analysis Area by alternative is met.

The final step in the allocation modeling process is to have local governments review and verify data inputs and density assumptions to assess the reasonableness of densities associated with existing conditions and known development projects.

Las Vegas, Nevada

1990 Population - Las Vegas, NV MSA - 741,459

MPO: Clark County Regional Transportation Commission
Dennis Mewshaw (mewshaw@co.clark.nv.us)
301 East Clark Ave., Suite 300
Las Vegas, NV 89101
(702) 455-5777
main: (702) 455-4481
fax: (702) 455-5959

Estimates regional control totals using a committee made up of representatives from the county, cities in the region, and quasi-governmental entities (public utilities, etc.). Estimates are approved by the RTC commission

Regional control totals are sent to an outside consultant for TAZ assignment (see below).

Looking to more fully integrate the process. Want to be able to evaluate how transportation investments will affect timing of land absorption, development densities, etc. Looking to integrate projections with GIS system to evaluate transportation and development scenarios - in order to look at 'what-if' options.

Applied Economics

Also has done economic projection work for Maricopa Association of Governments
Rick Brammer
5957 East Windrose
Scottsdale, AZ 85254
(602) 922-9397
fax: (602) 922-2670

Uses a spreadsheet-based small area allocation model. Uses data supplied by the county on existing land use and future land use (compilation of area comprehensive plans). Allocates population and employment using a multivariate attractiveness index. Attractiveness index variables include: population, basic employment, service employment, income, accessibility; and proximity to freeways, airports, and rail lines. Some difficulties include the conversion of industry by SIC code to land use type.

The small area allocation model is adjusted for specific Las Vegas conditions (land use pattern is different than other areas). Also maintains a database on large-scale development projects which affect location decisions.

Louisville, Kentucky/Indiana

1990 Population - Louisville, KY-IN MSA (KY part) - 770,591

MPO: Kentuckiana Regional Planning and Development Agency
Lori Kelsey
11520 Commonwealth Drive
Louisville, KY 40299
(502) 266-6084
fax: (502) 266-5047

Five county region with 756 zones. Uses county-level forecast data for population and households from the Kentucky State Data Center, Indiana State Data Center and Louisville Area Forecasting Project. Uses county-level forecast information for employment by place-of-work from the Louisville Area Forecasting Project. Starts with year 2020 county-level controls and works down to TAZs.

Household Projection (Non-group quarters population projection was done through a similar process of review)

Bullitt Clark and Floyd counties: The relationship of zonal change to county level change from 1980 to 1990 was held constant and applied to the year 2020 county household control totals. This yielded an initial 2020 TAZ forecast of households. In some cases, this method caused sharp decreases or negative results. When this occurred, either a 1980 or 1990 TAZ linear trend was applied or the 1990 household number was held constant through the year 2020.

Comments and data from local and published sources were used to adjust the initial trend on a zone-by zone basis. The entire dataset was adjusted again to closely match the year 2020 household county control totals. The results were submitted to representative agencies in the three counties for review. All changes received were incorporated into the projections. The data was adjusted one final time to make it consistent with the approved county control totals.

Oldham County: TAZ level shares of county household growth were estimated (based on building permit activity) by the Oldham County Planning and Zoning Commission staff. The projected zonal shares of growth were applied to the expected county household growth from 1990 to 2020. The result was added to the 1990 TAZ data to yield 2020 zonal households.

The dataset was reviewed by the Oldham County Planning and Zoning Commission. Their changes were incorporated into the projections. The data was adjusted one final time to make it consistent with the approved county control totals.

Jefferson County: The most urbanized county in the region - Jefferson County (Louisville) - is doing its own forecasting as part of its 2020 comprehensive plan (see below). Jefferson County accounts for around 511 of the 756 TSZs in the region. Projected zone level household data was provided by the Louisville-Jefferson County Planning Commission staff (see description below).

Employment Projection

Bullitt Clark. Floyd and Oldham counties: Year 1980 TAZ-level employment by place-of-work data was not available for Bullitt and Oldham counties. It was available only for a few of the zones in Clark and Floyd counties; therefore, the 1980 TAZ level employment data was not used for this forecast. Because 1990 TAZ level employment data was available, the initial forecast allocated year 2020 jobs to the same places and in the same proportions as they had existed in 1990. The relationship of zonal employment to county employment was held constant and applied to the year 2020 county employment control total. Retail and non-retail categories were held to their 1990 proportions as well.

The initial forecast was examined on a zone-by-zone basis and adjusted accordingly, using locally provided data about the locations of future jobs. A subsequent adjustment brought these figures in line with the 2020 county level control totals for total, retail and non-retail employment.

These initial forecasts were reviewed by agencies in Bullitt, Oldham, Clark and Floyd counties. Their changes were incorporated into the forecasts, and the entire dataset was adjusted one last time to make it consistent with the year 2020 county level controls for total, retail and non-retail employment.

Jefferson County: Louisville-Jefferson County Planning Commission staff provided zonal forecasts of total, retail and non retail employment by place-of-work (see below for description).

Louisville-Jefferson County Planning Commission
Ed Mellet
531 Court Place, Suite 900
Louisville, KY 40402-3396
(502) 574-6230
fax: (502) 574-8129

Uses a 'reiterative' process. Start with regional projection, move down to market areas projection to census tract projection and then to TAZ projection. Then uses Delphi review as part of reiterative process to eliminate the extremes. The general process is described below.

Population and Housing

1. Started with county-level projection.
2. Divided county into 13 market areas. Each market area consists of groupings of the county's 1 78 census tracts. The tracts were grouped for common topography, housing mix, density, and socioeconomic character.
3. Initial projection of net new housing units for 1995, 2000, 2010 and 2020 was made based for each market area. This allocation was made at the census tract level and then accumulated to the market area level for review. This initial projection was a straight line projection from the 1980 to 1990 trend and ignored land availability constraints. The projection used the 1990 mix of SF and MF housing units as well as the current development density.
4. Initial projection was constrained at the tract level to account for known policy in market areas. For example, projected declines in two market areas were capped to reflect housing commitments by local agencies. Also, projected no new housing units for several census tracts in one market area because of airport expansion.
5. The initial projection yielded a shortage of available land by the year 2000 for 2 market areas. Reallocated housing units to other market areas and increased projected housing densities. Reallocation was made using the consensus judgment of a panel of experts from area utilities, the Chamber of Commerce and the Planning Commission.
6. Projections were then mailed out to a panel of local experts for their review and recommendations 'Delphi review.' Adjustments were made to the projections.
7. Population projections were made using the forecasts of housing units. The number of persons per housing unit was calculated at the census tract level using 1990 data. The downward trend in persons per household in the county was assumed to apply equally across the county. Vacancy rates were assumed as constant, and estimates of the number of households were made from estimates of housing units. Population estimates were produced by multiplying estimated number of households by number of persons per household.

Employment

1. Started with census tract data on job location by industry.
2. Projected number of jobs by industry in each census tract using 1) each tract's share of 1990 total county jobs in each industry; and 2) the growth rate forecast for the county total in that industry.
3. In order to account for recent market developments not accounted for in 1990 data, mailed out the 'constant share' job projections to a panel of local experts to refine job projections by industry and census tract. Adjustments were made to constant share projections.

Traffic Zone Forecasts

1. Based on census tract control numbers for both population and employment and allocated to the traffic zone subparts based on available vacant land and existing housing and employment mix characteristics.
2. Dwelling unit density, employee density, land use and vacant land figures were developed by constraint category for each traffic zone. This data was adjusted for anomalies due to differences in the timing of land use. population or employment measurement of classification irregularities that were more apparent at the traffic zone level.
3. Land use data was generated in 1992 while the census data was from 1990.
4. In larger undeveloped census tracts that were composed of several traffic zones some adjustments were made to place growth in unconstrained areas that bordered existing development rather than distributing it evenly across the tract.

Madison, Wisconsin

1990 Population - Madison WI, MSA - 367,085

MPO: Dane County Regional Planning Commission
Bob McDonald, Director of Transportation Planning
Transportation Planning Division
217 South Hamilton Street, Suite 403
Madison, WI 53703-3238
(608) 266-4518
fax: (608) 266-9117

Uses the county forecast from the Demographic Service Center of the Wisconsin State Department of Administration as the 'control total.' This forecast is disaggregated for urban and rural areas of the county. The MPO doesn't forecast for rural areas (outside of Urban Service Areas (USAs)) since these are supposed to remain undeveloped. Forecasts are suballocated down to the jurisdiction area (USAs of the cities and towns) by looking at both trends and plans. A 'discounted linear regression' is used to forecast population, dwelling units, and employment trends - this method looks at previous Census figures ('70, '80, and '90) as well as recent estimates, with more weight given to recent figures. The forecasting method also takes into account developable land. The total individual USA forecasts are then compared to the urban allocation total and the individual figures are reconciled on a proportional basis so that the totals match.

To assign population. dwelling unit, and employment projections by Traffic Analysis Zone (TAZ), a land use allocation model was used to look at the amount of developable land left in the TAZ and the history of development (i.e., how much land has been developed and when). Also the TAZ is looked at in relation to its 'Planning Analysis Zone', a larger aggregation which corresponds (in urbanized areas) to the census tract level. As the TAZs are assigned population and employment, several iterations are done across the whole system, until there is a balance between what the trends predict and the land use plans allow for.

The MPO projections are done at the TAZ level every 10 years, but a trend report is released every year showing new office building and residential building permit data.

The Regional Planning Commission came up with 5 alternate land use scenarios ranging from not assigning any growth to outlying areas, to allowing clustered outlying growth, to a continuation of existing scattered outlying growth. It demonstrated the consequences of these various scenarios with the traffic modeling system. At certain levels of employment and population concentration in the central city, for example, there was an unacceptable increase in congestion downtown, even accounting for increased commuter rail/ light rail dedicated busway/ pedestrian/ bicycle traffic. The staff recommendation was for development that did occur on the urban fringe to be 'nodal' type development, with higher densities than were usually being developed and with a 50-50 countywide mix of SF and MF units. While recent development was averaging 3-4 dwelling units (D.U.s / acre) for SF and 10-12 D.U.s/ acre for MF, the plan recommended 5 D.U.s/acre for SF and 16 D.U.s/acre for MF.

There was also a staff recommendation to limit growth in rural areas to 12% of the total growth in the area, and also to recommend that the rural growth that does occur should be clustered along already existing villages or subdivisions.

While there was a strong initial reaction against the density recommendations, staff was able to show that the real estate market, both nationally and locally, was moving toward these development densities for various economic and demographic reasons. Staff was also able to demonstrate the benefits of increasing densities in more affordable housing, fewer bedroom communities, less traffic congestion, closer services, etc.

Staff held a workshop in which the major developers in the region were invited to participate in a growth allocation exercise. Given that the area was to receive a set amount of growth, they were asked where should this growth be directed. In the end, the results from the developers' group were very similar to the staff recommendations.

Minneapolis-St. Paul, Minnesota

1990 Population-Minneapolis-St. Paul, MN-WIMSA (MN part)-2,413,873

MPO: Metropolitan Council, Twin Cities Area
Michael Munson
Mears Park Centre
230 West Fifth Street
St. Paul, MN 55101
(612) 291-6331
fax: (612) 291-6442

The Metro Council does regional demographic and employment forecasts for the whole area and allocations to sub areas within the region. The allocation assigns population and employment forecasts to the 7 counties and 189 cities and townships in the region. Development is restricted outside of the Metropolitan Urban Services Area. This allocation is done by looking at land use trends (using directional and expanding ring models), local growth (or anti-growth policies) and available land supply.

The allocation focuses on household and employment numbers. Household numbers relate to land consumption and are more accurate to forecast at a small scale. The Council used to do elaborate forecasting (household income, etc.), but now just forecasts households, population, employment and retail employment for traffic modeling.

The Metro Council makes projections for all the 141 cities and 48 townships in the area and then has the localities make assignments of these projections for the traffic assignment zones in their jurisdiction (over 1,200 zones in the region). The Metro Council feels that the local governments are more familiar with the development patterns that are likely to result from local planning and zoning, and that this method is more accurate. Previously a set of equations was used that related growth to available land. Example: if 90% of a zone was developed, the equation would fill up 90% of the zone with a land use pattern (employment and housing numbers) similar to the existing pattern and 10% similar to the rest of the city as a whole. If a zone was 90% empty, the equation would fill up 90% of the zone with a land use pattern similar to the city the zone was in (taking into account the city-level density).

Note: The City of Plymouth in the Minneapolis-St. Paul area was also contacted about their projection methodology in order to give an example of what a fast-growing city in the region is doing.

City of Plymouth
Ann Hurlburt
3400 Plymouth Blvd.
Plymouth, MN 55447
(612) 509-5401
fax: (612) 509-5407

Conducts 'holding-capacity' analysis to compare regional projections done by the Metro Council with actual development potential. This analysis is based on a GIS-based land use inventory and looks at development trends and the characteristics of land available within the TAZs. The projections done by the Metro Council can be modified based on local input. The local population and employment allocations by TAZ are distributed based on this local analysis.

Orlando/Kissimmee, Florida

1990 Population - Orlando FL, MSA - 1,072,748

MPO: Orlando Urban Area Metropolitan Planning Organization
Transportation Planning Division
Dennis Hooker
1011 Wymore Road, Suite 105
Winter Park, FL 32789
(407) 623-1075, ex. 310
fax: (407) 623-1084

Currently in the middle of changing their system. Previously used census figures as a base and updated the county numbers using data from the University of Florida Statistical Abstract as a cap. Looked at the comprehensive plans of the counties in the region and calculated percent of developable land for traffic serial zones. Populate TAZs based on developable land percentages.

Now moving to a new system. Contracting with Dun and Bradstreet for employer and employee data by SIC code and street address. Using GIS system to geocode this data. Implementing a DRAM/EMPAL model to allocate population and employment at the sector level. Will use a small-area allocation model to allocate DRAM/EMPAL sector allocations to TAZs. This small-area model will be based on developable land percentages in each TAZ.

Phoenix, Arizona

1990 Population - Phoenix, AZ MSA - 2,122,101

MPO: Maricopa Association of Governments
Cathy Arthur
2901 West Durango Street
Phoenix, AZ 85009
(602) 506-4117
main: (602) 254-6308

Uses a two-tiered system (for over 1,500 TAZs). A DRAM/EMPAL model is used for the county and subregional projections. A small area model (SAM) is used to allocate projections from the subregions to the TAZs.

Started a GIS Analysis and Data Enhancement Study in August, 1994 performed by Barton-Aschman Associates, Inc. in association with AMPG, Inc.⁽¹⁴⁾ The study had three objectives: 1) improvement in the transportation network databases to support the recently overhauled travel forecasting models; 2) improvement in the DRAM/EMPAL land use forecasting model to ensure that land use forecasts remain consistent with the measures of accessibility on which they are based.; and 3) develop a GIS-based Subarea Allocation Model (SAM) for disaggregating the regional forecasting data derived from the DRAM/EMPAL model.

In addition to the need to disaggregate the regional DRAM/EMPAL-based projections into TAZ geographies, other objectives of the SAM model were to provide socio-economic forecast information for a wide range of geographies, such as water district or municipal boundaries. and to provide data to implement 'focused travel models.' Because of these demands, MAG data is starting to be maintained in a 'Minimal Analysis Zone (MAZ) system, which is highly disaggregated, but which can be easily aggregated into any number of geographies.

The implementation of the MAZ system has been accomplished using the GRID capabilities of the ARC/INFO GIS system. GRID is a raster representation of geography. In this case a 220-foot grid (resulting in cells a little over 1 acre in size) was used for the MAZ structure of the region.

The SAM model, which was implemented in an ARC/INFO AML (ARC Macro Language) application, is designed to disaggregate forecast information from DRAM/EMPAL by simulating land use decisions made by developers. The key elements are as follows:

1. A base year land use coverage was created and populated with information about households and employment, including information about 'special' population groups (motel rooms, mobile home park locations, nursing homes, etc.) which play a major role in trip generation.
2. The SAM model allocates growth. Growth estimates for households and employment come from DRAM/EMPAL for Regional Analysis Zones (RAZs) in comparison with the base year data set. Growth estimates for other land uses (e.g. special population groups) come from countywide control totals.
3. Lands which are eligible for development are identified by the model. The criteria by which land is eligible for development include (1) land is represented as vacant and developable in the existing land use cover and (2) land is appropriately designated for development in the general plan cover.
4. Lands which are eligible for development are then evaluated for growth potential through a scoring system which reflects locational preference criteria. A wide variety of criteria can play a role in judging the growth potential for land, including (1) proximity to the urban area, (2) proximity to major arterial highways, and (3) proximity to other developed land. Special characteristics, for example the propensity for motels to be built almost exclusively around freeway interchanges, can be reflected in the scoring system.
5. Growth is then allocated to the highest ranking land until it is completely absorbed. Growth is absorbed at density levels associated with the general plan designations, which can be altered as needed to completely absorb growth for any time frame.
6. A new forecast land use cover, combining both the existing land use cover as well as lands identified to absorb growth is then generated. This is used as the 'base year' for the next iteration.

Additional contacts:
Planning Technologies, Inc.
Mike Corlett
Albuquerque, NM
(505) 872-4808

Was a key member of the MAG project for Barton-Aschman and is now with Planning Technologies, Inc. Is also implementing a GIS-based system for Albuquerque.

Portland-Vancouver, Oregon-Washington

1990 Population - Portland-Vancouver, OR-WA CMSA (OR part) - 1,239,842

MPO: Metropolitan Service District (Metro)
Sonny Conder
600 Grand Ave., NE
Portland, OR 97232
(503) 797-1700
fax: (503) 797-1794

A regional economic and demographic forecast was prepared using a Metro-developed econometric model using national growth assumptions. The regional economic model covers all 5 counties in the Portland-Vancouver CMSA and uses national growth assumptions as its primary inputs. This model combines assumptions about U.S and global conditions, regional developments, and demographic assumptions to determine future regional employment and population growth in a consistent theoretical framework.

The forecast methodology for allocating growth to smaller areas includes the following four elements:

1. Start with a regional forecast of population, households and employment to use as control totals to constrain the allocation to smaller units of geography.
2. Produce a 'technically-based' spatial allocation of growth by using time series data on population and employment to represent market demand.
3. Use land availability and comprehensive plan designations to measure the supply/capacity of each subarea in order to constrain the technical allocations.
4. Use expert panels - Growth Allocation Workshops - to review and revise the technical allocations.

The regional forecast and allocation process has been explicitly connected with the region-wide planning policy, the Region 2040 urban growth plan. Policy assumptions for the forecast include:

1. The region will grow into a denser and somewhat more compact form than has been the trend over the last 50 years. Densities will increase from approximately four Dwelling Units (DU) per acre to about five DU per acre.
2. The Urban Growth Boundary (UGB) is assumed to expand in order to maintain a 20 year land supply for residential purposes.
3. The density and pattern of growth will be affected by the level and type of transportation investment.
4. Metro and local government will actively encourage infill and redevelopment as well as increased densities.
5. Local governments outside of Metro will manage their growth in a similar fashion.

Metro used a mix of technical analysis and expert review in an iterative process to allocate the regional forecast to subareas. The process included the following main elements:

1. The region was divided into six major market areas. An assumption was made that trends evident in the subareas would not be affected by any particular Region 2040 growth policy other than land availability.
2. Trend growth projections were made using regression equations and historical data from 1970 to 1994.
3. Housing and job growth projections were compared to the development capacity for each market area by looking at local comprehensive plans and the Region 2040 growth capacity assumptions.
4. The results were presented to a Growth Allocation Workshop. Participants reviewed data and adjusted estimates for market areas by shifting the excess growth to adjacent market areas or by agreeing to make sufficient regulatory changes to provide for the additional required capacity.
5. These revised market area projections were then used as subarea control totals. Forecasts were developed for each of the 20 planning districts using a methodology identical to that for forecasting growth for the six major market areas (regression analysis using historical trend data).
6. A second series of Growth Allocation Workshops was held to compare district projections to capacity limits. Forecasts were adjusted in the same manner that they were for the major market areas.
7. The adjusted projections for the 20 planning districts were disaggregated into 1/16 acre grid cells (52 ft. grid size) according to the designation and land status in the Region 2040 plan. The allocation is based on parcel level data, comprehensive plans and three categories of land use. These categories include developed land, re-developable land, and vacant land. The allocation is further constrained by environmental considerations and policy-based area density profiles. These density profiles relate specifically to urban centers, transportation corridors, and residential and commercial areas. The allocation is performed at the cell level (considering all these factors) utilizing GRID and ARC/INFO. Mapping projected growth to exact locations allows the local planning staff to make a precise assessment of the likelihood of such growth occurring at a particular location. This approach also tells local planning staff what regulatory and investment changes need to be made to achieve the Region 2040 design capacity in any particular site.
8. The spatially allocated household and employment projections were aggregated into 1260 Traffic Analysis Zones (TAZs). Individual jurisdictions then reviewed the household and employment allocations for their own TAZs as well as the 1/16 acre grid allocations.
9. The final round of Growth Allocation Workshops was held to review and revise the TAZ-level household and employment growth allocations.
10. The 1/16 acre grid allocation of households and employment was revised for consistency with the TAZ-level allocations.

The forecast review and revision process lasted over eight months. The RLIS (Regional Land Information System) database and specific growth management plan (Region 2040) allowed policy and forecast data to be combined and evaluated at a very detailed and realistic level.⁽¹⁵⁾

The small area allocation system relies on policy-based land use assumptions contained in the Region 2040 plan and therefore can be looked at as simulating desired goals rather than strictly as a predictive model.⁽¹⁶⁾ However, it does provide a detailed look at expected results if existing land use plans are followed. It also provides a mechanism to test 'what-if' strategies about the effects of changing land use plans.

San Francisco, California

1990 Population - San Francisco-Oakland-San Jose, CA CMSA - 6,253,311

MPO: Metropolitan Transportation Commission (MTC)
Association of Bay Area Governments (ABAG) - does regional forecasting
Hing Wong
Metro Center
101 8th Street
Oakland, CA 94607-4756
(510) 464-7966
main: (510) 464-7900
fax: (510) 464-7970

Modeling System used for Projections 96⁽¹⁷⁾

This projection system, designed to predict growth and distribution of population, households, employment, income, and labor force characteristics, is structured around three components. These are: a) the regional economic and demographic forecasting system; b) the county employment, population, and income forecasting system; and c) the distribution of jobs and households as a function of available land, assumptions about density and travel demand within counties in the region. The distribution of jobs and households, as well as the total growth forecast, is heavily influenced by information gathered in the Local Policy Survey.

Regional Economic-Demographic System

The projection of regional employment, income, output, population, labor force, and occupational demand is performed by the Regional Economic-Demographic System (REDS). REDS is an analytical and econometric model which uses a non-survey input/output (I/O) model to drive the interaction in the system. A general overview of the model can be found in "The Design and Implementation of a Regional Economic-Demographic Simulation Model," by R. Brady and C. M. Yang in the Annals of Regional Science, November 1983.

The basic equations and input/output model are updated every two years. The most recent update occurred in 1995. The user of REDS may change up to sixteen variables to affect the model's projection behavior. The system is designed to be user friendly. REDS divides the economy into thirty-five industry sectors, and predicts the output, job demand, and capital requirements of each sector. The demand for jobs drives the labor force model which interacts with the migration model. The population model is a Cohort-Survival Model.

REDS has approximately thirty-three equations in the system. Some are statistical equations developed from time series data and hence constantly updated; others are analytical equations based upon observed behavior in the economy. The latter equations are either differential or difference equations.

County Employment Forecasting System

The projections of employment and income for each of the nine counties of the Bay Area were obtained from the County Employment Forecasting System (CEFS). CEFS is an econometric model that makes efficient use of the limited employment data available at the subregional level. H produces county forecasts consistent with the regional employment forecasts of REDS. A complete and thorough discussion of the model can be found in "Industrial and Spatial Interdependency in Modeling: An Empirical Forecasting Model for the Counties in the San Francisco Bay Region" by P. Prastacos and R. Brady in the Annals of Regional Science, July 1985.

CEFS recognizes thirty-two sectors, each sector representing a two-digit SIC code sector or a major industrial group. There is one equation for each sector and county. The equations were specified to account for the industrial and spatial interdependency of activities. Jobs in a particular sector are often dependent on job levels in other sectors in the same county and the region. Spatial interaction is determined by linking employment growth in competing counties and in the entire region with that of employment growth in the dependent counties. Local serving employment is more heavily dependent upon local population and income levels.

CEFS uses ordinary least squares technique to develop predictive equations with data from the County Business Patterns reports for years 1964 to 1992. The results of the regressions were very good and indicate that the relationships depicted in the equations are of empirical value and that they do reflect the economy of the counties. Both the R-squares for the equations and the t-values for the individual coefficients were acceptable. Additionally, a dynamic simulation of the estimated model over the period 1964 to 1992 showed that the employment levels forecasted by CEFS are close to actual historical data. After updating the statistical equations, ABAG produces a report that provides both the statistical information and the updated equations. The most recent update was released in August 1994. The report title and author are: "CEFS, A County Employment Forecasting System for the San Francisco Bay Region," by E. K. Caindec.

Population and Household Forecasts

ABAG uses trend analysis to determine long-term growth forecasts for each county's population and households. The latest time series uses data from 1975 to 1994. Linear, exponential, and geometric regression time series equations are used to predict future growth. The results of these trend equations are summed and averaged.

Trend data are constrained by local development policies which limit housing production, and hence household growth. In several counties, household and population growth in the forecast exceed the aggregate of local policies over the long term. Short-term growth, however, closely follows development policies.

Subcounty Allocation System

The allocation of population, housing, and employment at the sub-county (zonal) level was carried out using the Projective Optimization Land Use Information System (POLIS). POLIS, a land use and transportation model, has replaced the PLUM and BEMOD models which were used at ABAG prior to 1980 for land use and zonal population and employment projections. A discussion on the structure of POLIS can be found in the ABAG reports "A Description of POLIS: The Projective Optimization Land use Information System," by P. Prastacos & E. K. Caindec, 1995, and "The Basics of POLIS," by E. K. Caindec, 1991.

The allocation process in POLIS is based on several criteria, some reflecting the behavior of individuals and some describing physical and planning constraints. Residential choice is determined by the travel-to work and shipping behavior, the availability and attractiveness of housing, and the current levels of nearby employment. Retail activity is located in proximity to population centers to maximize sales revenue. The locational patterns of the other industries are influenced by the accessibility to labor supply, the proximity to other similar industries and local development policies.

POLIS is a structured mathematical programming, optimization-type problem. That is, the allocation of population and employment is optimized with respect to an objective function or goal while at the same time satisfying planning constraints. POLIS converges after several iterations on a solution that optimally allocates jobs and households, subject to the constraints. It results in housing, employment and trip flow patterns which are consistent with each other and the land use constraints.

The form of the objective function in POLIS is derived from the random utility theory and describes the behavior of individuals (employees) to select among a set of alternatives the one maximizing their utility.

The constraints of the model describe the housing and land supplies, the development policies of the different cities, and the employment/housing to be allocated among all the zones within a county.

The Bay Region is subdivided into 119 zones in the POLIS system. Job data are derived from the County Employment Forecasting System (CEFS). The thirty-two employment categories in CEFS are aggregated separately for each county into four sectors: 1) Manufacturing and Wholesale Trade; 2) Transportation, Communications, Utilities (TCU) and Finance, Insurance, and Real Estate (F.I.R.E.); 3) Retail Trade; and 4) Services. Countywide estimates of household demand, population, and employed resident growth are also provided. Finally, detailed land use information on potential growth by employment type is provided as input to the system.

Recent calibrations of POLIS indicate that the mathematical structure reasonably simulates historical behavior. ABAG has just completed the process of re-calibrating POLIS using the 1990 Journey to Work and data collected and provided by the Metropolitan Transportation Commission. This calibration was completed in 1994.

ABAG uses a second subcounty model, the Subarea Projections Model (SAM), to allocate the results obtained from the POLIS model to census tracts.⁽¹⁸⁾ The model is described in the ABAG report, "Subarea Projections Model (SAM): Allocating Employment and Population, Projecting Household Income, and Land Use Accounting," February 1993.

Review of Forecasts

All county and subregional forecasts are reviewed by local governments. This review process has several objectives. First, forecasting for 126 cities and unincorporated areas and nine counties is a complicated process. Although the models ABAG uses are state-of-the-art, models are imperfect replications of reality. Second, review by local governments helps ABAG to identify problems at the small area forecast level.

References

Available Methods for Land Use/Transport Interaction Modelling; David Simmons, David Simmons Consultancy (10 Jesus Lane, Cambridge CB5 8BA, England); March 1995.

Comprehensive Modeling for Long-Range Planning: Integrated Urban Models and GIS; Robert A. Johnston and Tomas de la Barra; presented at the Transportation Research Board Annual Meeting, January 1997.

GIS Analysis and Data Enhancement Study; prepared for the Maricopa Association of Governments (MAG) by Barton-Aschman and Associates, Inc.; January 1996.

Land Use Models in Transportation Planning: A Review of Past Developments and Current Best Practice; Britton Harris; January 1996.

Land Use and Travel Survey Data: A Survey of the Metropolitan Planning Organizations of the 35 Largest U.S. Metropolitan Areas; Chris Porter, Laura Melendy, and Elizabeth Deakin, Institute of Urban and Regional Development, University of California at Berkeley; February 1996.

Making the Land Use Transportation Air Quality Connection, Volume 1: Modeling Practices; prepared for 1000 Friends of Oregon (Portland, Oregon) by Cambridge Systematics, Inc. and Hague Consulting Group; 1991.

A Manual of Regional Transportation Modeling Practice for Air Quality Analysis (Version 1.0); prepared by Greig Harvey and Elizabeth Deakin for the National Association of Regional Councils; July 1993.

Metro Vision 2020 Population and Employment Modeling Process, Denver Regional Council of Governments (DRCOG), October 1994.

Projections 96; The Association of Bay Area Governments (ABAG); December 1995.

Review of Land Use Models and Recommended Model for Delaware Valley Regional Planning Commission (DVRPC); prepared for DVRPC by URS Consultants, Inc., September 1996.

A Technical Review of Urban Land Use Transportation Models as Tools for Evaluating Vehicle Travel Reduction Strategies; Frank Southworth, Center for Transportation Analysis, Energy Division, Oak Ridge National Laboratory; July 1995.

Travel Model Improvement Program Land Use Modeling Conference Proceedings, February 19-21, 1995; prepared for the U.S. Department of Transportation (document #: DOT-T-96-09) by Gordon A. Shunk, Patricia L. Bass, Cynthia A. Weatherby, and Lynette J. Engelke, Texas Transportation Institute; February 1995.

The 2015 Regional Forecast and Urban Development Patterns: Population, Households, Employment and Income, Portland Metropolitan Service District (Metro), February 1996.

---

Endnotes

1. A Manual of Regional Transportation Modeling Practice for Air Quality Analysis (Version 1.0); prepared by Greig Harvey and Elizabeth Deakin for the National Association of Regional Councils; July 1993.

2. Available Methods for Land Use/Transport Interaction Modelling; David Simmons, David Simmons Consultancy (10 Jesus Lane, Cambridge CB5 8BA, England); March 1995.

3. A Technical Review of Urban Land Use Transportation Models as Tools for Evaluating Vehicle Travel Reduction Strategies; Frank Southworth, Center for Transportation Analysis, Energy Division, Oak Ridge National Laboratory; July 1995.

4. This section is drawn from Comprehensive Modeling for Long-Range Planning: Integrated Urban Models and GIS; Robert A. Johnston and Tomas de la Barra; presented at the Transportation Research Board Annual Meeting, January, 1997.

5. Simmons, Available Methods for Land Use/Transport Interaction Modelling; p. 1.

6. Making the Land Use Transportation Air Quality Connection, Volume 1: Modeling Practices; prepared for 1000 Friends of Oregon (Portland Oregon) by Cambridge Systematics, Inc. and Hague Consulting Group: 1991.

7. Southworth, A Technical Review of Urban Land Use Transportation Models.

8. This information is derived from Travel Model Improvement Program Land Use Modeling Conference Proceedings, February 19-21, 1995; prepared for the U.S. Department of Transportation (document #: DOT-T-96 09) by Gordon A. Shunk. Patricia L. Bass, Cynthia A. Weatherby, and Lynette J. Engelke, Texas Transportation Institute: February 1995.

9. Ibid.

10. Land Use Models in Transportation Planning: A Review of Past Developments and Current Best Practice; Britton Harris; January 1996 in Review of Land Use Models and Recommended Model for Delaware Valley Regional Planning Commission (DVRPC); prepared for DVRPC by URS Consultants. Inc., September 1996.

11. Porter, Melendy, and Deakin, Land Use and Travel Survey Data, p. 2.

12. Ibid., p. 3.

13. Provided to me a copy of Simmons, Available Methods for Land Use/Transport Interaction Modelling.

14. The description presented here is taken from GIS Analysis and Data Enhancement Study, Barton-Aschman and Associates, Inc.; Prepared for the Maricopa Association of Governments (MAG); January 1996.

15. The 2015 Regional Forecast and Urban Development Patterns: Population, Households, Employment and Income, Portland Metropolitan Service District (Metro), February 1996.

16. Barton-Aschman and Associates, Inc., GIS Analysis and Data Enhancement Study; p. 1-10.

17. The description of Projections 96 presented here is taken from the Association of Bay Area Governments, Projections 96, released December 1995.

18. Porter, Melendy, and Deakin, Land Use and Travel Survey Data, p. 10.

---

DOT Home | TMIP Feedback

United States Department of Transportation
Privacy statement and legal notices
This site is hosted and maintained by the Texas Transportation Institute.