John E Abraham

Microsimulating urban systems

Abstract This paper presents a status report concerning on-going research and development work by a team of Canadian researchers to develop a microsimulation, agent-based, integrated model of urban land use and transportation. It describes in some detail the overall design and current status of the ILUTE (Integrated Land Use, Transportation, Environment) modelling system under development. The overall purpose of ILUTE is to simulate the evolution of an entire urban region over an extended period of time. Such a model is intended to replace conventional, aggregate, static models for the analysis of a broad range of transportation, housing and other urban policies. Agents being simulated in the model include individuals, households and establishments. The model operates on a “100% sample” (i.e., the entire population) of agents which, in the base case, are synthesized from more aggregate data such as census tables and which are then evolved over time by the model. A range of modelling methods are employed within the modelling system to represent individual agents’ behaviours, including simple state transition models, random utility choice models, rule-based “computational process” models, and hybrids of these approaches. A major emphasis within ILUTE is the development of microsimulation models of market demand-supply interactions, particularly within the residential and commercial real estate markets. In addition, travel demand is modelled explicitly as the outcome of a combination of household and individual decisions concerning the participation in out-of-home activities over the course of a day. Spatial entities in the model include buildings, residential dwelling units and commercial floorspace, as well as aggregate “spatial containers” such as traffic zones, census tracts or grid cells.

Specification and Estimation of Nested Logit Model of Home, Workplaces, and Commuter Mode Choices by Multiple-Worker Households

Household behavior in the selection of home location and the selection of workplace locations and commuting modes for employed members involves trade-offs among the attributes of the available alternatives for the different household members. A modified form of nested logit model representing this behavior has been developed and estimated using disaggregate revealed preference observations collected in Calgary, Alberta, Canada. Three categories of choice—choice of home location for the household, choice of workplace location for each worker in the household, and choice of mode for the trip to work for each worker in the household—are treated as a joint choice made by the household, allowing for differing numbers of workers in different households. A nesting structure that takes into account the greater similarity among mode alternatives is combined with a system for weighting, by age and gender, the contributions of different individual workers’ utilities to the total household utility. This leads to a nested logit model in which each household has its own nesting structure that is based on the age and gender of the household members. The utility function coefficients and weighting function parameters were estimated with full-information maximum likelihood by using purpose-built software. The resulting model extends consideration of household spatial behavior at the disaggregate level beyond the one-worker and sequential-conditional choice paradigms and provides various insights into the nature of this behavior.

FIRM LOCATION IN THE MEPLAN MODEL OF SACRAMENTO

MEPLAN land use and transportation-interaction models traditionally include models of business-location choice. The mechanisms in the model that allow for realistic aggregate assignment of firms to zones, and the resulting impact that firm location has on the entire modeling system, are described. The models use a form of logit function to allocate production activities to zones. Interdependencies in the Social Accounting Matrix, together with trip- and land-cost data and usage rates, generate a utility function for the attractiveness of purchasing a given sector’s output from a given zone. The alternative specific constants and the dispersion parameter are estimated in calibration, based primarily on cost data, trip-length distributions, and arrangement of activity in a calibration year. Results from a model of Sacramento show the various features and strengths of the model, while pointing out some weaknesses and potential pitfalls. The strengths include the realistic representation of current and continuing patterns, the close linkages between different industry and household types, and the market-based nature of the model. The weaknesses include the aggregate nature of the model and the possibility of difficulties in establishing realistic, alternative, specific constants.

Driver Behavior at Rail-Highway Crossings

A study of driver behavior at 37 rail-highway crossings in Michigan revealed the possible association between past crash histories and violations. Data collection included recording license plate numbers for violating vehicles, driver gender, approximate age of the driver, and the vehicle make and model. Driver violations were categorized into five different levels of severity ranging from routine to critical. The 37 study sites were subdivided into four groups based on crossing geometry and traffic control. The number of sites in the groups ranged from 5 to 18. Seven years of crash data on the study sites were considered for significance testing. Observed violation data for the same groups were calculated, and tests for statistical significance were performed on them. The results of this study indicated promise for the use of the violation data in determining the relative hazardousness of rail-highway crossings in combination with crash histories. The violation data may also be used to develop countermeasures that would help alleviate violations and eventually traffic crash problems at rail-highway crossing sites. Targeted enforcement as well should assist in driver behavioral modifications. Additionally, the timely arrival of trains after the warning devices are triggered is an essential element that motorists assess when considering taking risks.

Comparisons from Sacramento Model Test Bed

Three land use and transport interaction models were applied to the Sacramento, California, region by various teams of researchers. The results of these efforts were compared with each other and with the traditional transport demand model used by the regional government. The results of the modeling efforts are compared, with the focus being on how the design of the modeling frameworks and their application influenced the modeling results. A trend scenario was compared with three different policy scenarios: one that involved high-occupancy vehicle (HOV) lane construction, one that added beltway construction as well as HOV construction, and a third that involved light rail construction and limited pricing of automobile use. The results differ among the different models for the trend scenario, as well as for each model with respect to scenario-to-trend comparisons. The results show some of the limitations of aggregate models calibrated to cross-sectional data. The differences between the models provide important insight into how models should be calibrated and how their results should be used. Uncertainty in land use transport interaction models seems inevitable, and further research should investigate how such modeling frameworks should best be used to understand the influence of policy in the face of uncertain futures.

Policy Analysis Using the Sacramento MEPLAN Land Use-Transportation Interaction Model

A land use and transportation interaction model was developed for the Sacramento, California, region, using the MEPLAN framework. The model represents land markets directly together with location-choice processes of different industry and household categories and the interactions among them. These resulting economic interactions lead to trip matrixes that are assigned to a multimodal, capacity-restrained transportation network, generating travel times and disutilities that influence the location-choice process in the period from 1990 to 2015. A range of different policy options is simulated for that period. Five different scenarios are compared with a trend scenario, three representing different long- and short-term government transportation supply scenarios, one representing incentives and disincentives to encourage land use patterns complementary to an improved transit system, and one combining the land-incentive scenario with higher prices for private vehicle use. The scenario predictions differ in mode choice, firm location, residential location, development patterns, vehicle kilometers traveled, and land rents. Various land uses are seen to be bidding against each other, and the differences in spatial arrangements between scenarios are quite complex and demonstrate the richness of the model. Certain changes take time to develop, demonstrating the need for a long-term approach to planning.

Design and Implementation of PECAS: A Generalised System for Allocating Economic Production, Exchange and Consumption Quantities

Production, Exchange and Consumption Allocation System (PECAS) is a generalization of the spatial input-output modelling approach used in earlier land use transport modelling systems. PECAS has two component modules. One is the space development module that represents the actions of developers in the provision of space, land and floor space, where activities can locate, including the new development, demolition and re-development. The other is the activity allocation module, which represents how activities locate within the space provided by developers and how these activities interact with each other at a given point in time.

Random Utility Location, Production, and Exchange Choice: Additive Logit Model; and Spatial Choice Microsimulations

A land use modeling system has been developed and applied on the basis of random utility theory. The system abstracts the decisions of the actors located in and traveling around a city or region as a series of logit models: (a) the choice of where to locate the home; (b) the choice of technology, lifestyle, or production option, being the choice of the quantities of “commodities” (consisting of goods, services, labor, and space categories) to consume or produce and hence what interactions will occur; and (c) the location of the “exchange” (transaction or interaction) for each commodity consumed or produced (i.e., for each interaction). These logit models can be combined in a nesting structure, but an additive logit formulation is required because the exchange choices are not mutually exclusive. The additive logit model, which is developed by combining the central limit theorem with random utility theory, can be applied to any choice situation in which several independent choices are conditional on a higher level choice. The resulting choice probabilities can be applied in an aggregate allocation system (as in software for the Production Exchange Consumption Allocation System, PECAS) and result in supply-and-demand equations for each commodity in each exchange that can be solved for a short-term equilibrium. In future research, microsimulation versions of the system could allow a fully integrated and dynamic representation of land use-transport interactions.

Design and Development of a Statewide Land Use Transport Model for Alberta

Abstract The fast growing of economy in Alberta, Canada makes a well-ahead statewide planning a pressing task. A statewide land use transport model is being developed at the University of Calgary for responding such a need. The model is based on the framework of PECAS, which represents Production, Exchange and Commodity Allocation System. The virtue of such a system is that it brings the transportation planning in the context of the whole economy by considering the fundamental root of transport demand, which is the exchange of goods, services, and labor among various activities, such as industries, services, households, and government. The PECAS framework is a spatial economic simulation system with a focus on transportation planning and land development planning, which has recently been developed and applied in a number of states and metropolitan regions in the United States. It draws from previous modeling experience, especially MEPLAN and TRANUS, and provides significant improvements in theoretical basis and behavior representation. This paper describes the design and development of such a model for the Canadian province of Alberta in terms of its data sources, novelty, and strategies for further calibration.

Parameter Estimation Strategies for Large-Scale Urban Models

Large-scale urban models often are subdivided into simpler submodels. The parameters of these models can be estimated using approaches that differ in regard to whether the full modeling system is run during an estimation procedure or whether that overall estimation is performed simultaneously with the estimation of the individual submodels. There are also ways in which extra data or extra models can be used to further inform parameter values. Five different techniques are presented (“limited view,” “piecewise” “simultaneous,” “sequential,” and “Bayesian sequential”), and the statistical theory necessary to justify each technique concurrently is described. The practical advantages and disadvantages are discussed, and each technique is illustrated using a simple nested logit model example. The concepts then are further illustrated by describing the sequential parameter estimation process for a land use/transport interaction model of the Sacramento, California, region. The ideas and examples should help modelers place more of an emphasis on overall calibration, allow them to follow a more rigorous approach in establishing the parameters of large-scale urban models, and help them understand the theory and assumptions that they are implicitly adopting. Two techniques in particular are noted as worthy of future research in large-scale urban modeling: (a) establishing the likelihood function based directly on the structural equations of the model, eliminating or reducing the need to “solve” for the model outputs during parameter estimation; and (b) using Bayesian techniques to adjust parameters in an overall estimation without discarding what is already known about those parameters.