Integrated Urban Models Volume 2: New Research and Applications of Optimization and Dynamics (Routledge Revivals)

Stephen H. Putman

• Publisher: Routledge
• Year: 2014
• ISBN:
  9781317748182
• Edition: 1
• Topic: Architecture
• Subtopic: Urban Planning & Landscaping

Following on from Integrated Models Volume 1: Policy Analysis of Transportation and Lane Use ( Routledge Library Editions, 2006), this book bridges the gap between the scholars and the practitioners of transportation and land-use modelling. First published in 1991, chapters discuss model-calibration and model-solution problems, describe a series of numerical and policy analyses, and propose potential directions for location and land-use research. This reissue will be of particular value to undergraduate and postgraduate geography students with an interest in integrated urban modelling; in particular, the research conducted in the field over the past two decades.

1

Introduction and overview

1.1 Introduction

This book is as much a complement as it is a successor to Integrated Urban Models (Putman, 1983a). That book was an account of the development and testing of a set of mathematical models of the interrelationships between transportation, location, and land use. The research described there began in 1971. Prior to that time there had been virtually no work done towards quantifying the forwards and backwards linkages between the metropolitan patterns of employment and population location and the networks of transportation facilities which connected them.

There had, of course, been considerable development of mathematical models of location (land use) and of mathematical models of transportation. Even so, the models were often improperly specified and their workings only poorly understood. The interactions between such models were largely unexplored. In the twelve years’ research described in Integrated Urban Models considerable work was done to develop and test improved, practical, versions of the models and to examine and understand their interactions.

One of the results of that research was the development of a set of computer versions of the mathematical models. Taken together these are known as the Integrated Transportation and Land Use Package (ITLUP). The employment location model (EMPAL) and the residential location model (DRAM) are two of the major components of ITLUP. The overall model system was the first operational integrated transportation and land-use model package. The ITLUP package and its component models have been tested with data for nearly four dozen different urban/metropolitan regions worldwide. EMPAL and DRAM, in particular, have seen more actual agency policy forecasting applications than any other spatial models. They have been used in regions as different as Seattle, Houston, Washington (DC), Los Angeles, Taipei, and Sarajevo.

Another of the results of that same research was the evolution of the various components of ITLUP into a computerized ‘test bed’, as well as the development of a substantial collection of metropolitan area data sets. It was thus possible to engage in a continuing series of experiments. New theoretical questions could, relatively quickly, be tested against actual data sets. Results, successful or unsuccessful, could be confirmed by repeating tests on numerous data sets.

Not long after the ITLUP research began there started an extensive and impressive development of new theory regarding individual model structures as well as linked model structures. One major theme of these developments was that of discrete choice models, often with logit formulations—though these models were first used with reference to mode choice, they were later extended to location choice as well. Another theme has been that of equilibrium analysis, and along with it, some considerable speculation on dynamics. A third theme has been the notion of recasting spatial interaction models, derived from entropy maximization approaches, as mathematical programming models of, say, consumer surplus maximization. Simultaneously, and in a sense cutting across the other lines of endeavor, there has also been some work in other disciplines on model systems and on more general issues of system complexity.

All this, the development of the modelling ‘test bed’, the acquisition of data archives, the lessons learned and questions raised by repeated model applications, and the new progress in theory development, set the stage for the work described here.

1.2 Purpose

The purpose of the work presented in this book is to describe some of the more important theoretical developments, as well as the numerical experiments and empirical comparisons which were made to evaluate them. This has led to the resolution of some of the outstanding issues from previous integrated model work, as well as pointing the way for future research.

The problem of assigning trips to networks has received considerable research attention in the past decade. In particular, the theoretical development of the user-equilibrium assignment model along with algorithms for its solution, has attracted a good deal of attention. The stochastic multipath assignment model, developed a few years earlier was still considered by some to be a valid approach as well. There was work towards integrating the two concepts into the stochastic user-equilibrium method. An important goal in this book is to compare these methods, as well as the more traditional all-or-nothing, incremental assignment, and quantal assignment methods. As well as dealing with new developments in the theory, these tests help to resolve some questions remaining from the Integrated Urban Models work, where only very limited tests were possible, and none of those were made with the newer model constructs.

The problem of ‘simultaneous’ trip distribution and trip assignment has also received considerable attention in the past decade. The problem of achieving an equilibrium solution to ITLUP was left unresolved in Integrated Urban Models. These are two closely related problems. A major goal in this book was to resolve the ITLUP equilibrium problem. A theoretically sound and eminently practical solution to the ITLUP equilibrium problem is given in chapter 8. What is more, its development illuminates several important model system issues as well as setting the stage for additional work. In particular, many questions regarding the meanings of, and differences between, intermodel and intramodel equilibrium solutions become more clearly defined.

The initial development of the solution to the ITLUP equilibrium problem, as well as the solutions to the user-equilibrium and stochastic user-equilibrium assignment problems, was through a mathematical programming approach. Another goal in this book was to discuss and explore aspects of the mathematical programming approach for both activity location and trip assignment. The trip assignment work has already been mentioned. In addition, many numerical experiments were done with mathematical programming models of location. Although some points stand out rather clearly, for example, for ‘realistic’ results to be obtained a dispersion term must be included in mathematical programming models of activity location, more questions were raised than were resolved. Overall, however, mathematical programming, or an optimizing framework, was shown to be a potentially useful approach to model development, notwithstanding a number of difficulties which are discussed in the text.

Time after time during the experimental work and the writing of this book the issue of complexity arose. In some of the more recent theoretical works in this field the reader is confronted with dizzying cascades of ever more complex equations. Even in this book where the emphasis is deliberately placed on the practical consequences of the developments in the theory some fairly complex model structures are presented. There are two worrisome points in this regard. First, there is the issue of obtaining the data necessary to test and evaluate such complex structures. The field must beware of becoming too well described by a remark attributed to Mark Twain: “There is something fascinating about science—one gets such wholesome returns of conjecture out of such a trifling investment of fact.” Second, there is the possibility that even as data become available, the systems and their requisite solution procedures may become so complex as to make it quite difficult to understand what is happening in any case. These issues will have to be faced time and time again as work in this field continues.

Much of the theory development in location and transportation modelling is done for greatly simplified examples and test cases. Frequently one theory is proposed and is subsequently deposed by another, with neither of them ever having been subjected to any empirical testing. It has been said that, “in the social sciences there may not be any simple problems” (Broad, 1983, page C1). This assertion is certainly true for the types of problem addressed by use of the models discussed here. It is not, however, a justification for ignoring the empirical testing of theory. Theory development in the complete absence of empirical testing is speculation, not science. The purpose and goals in this book have been stated. The method used here to accomplish them is a constant interplay of theory development and description, with numerical experiments and empirical testing. It seems likely that future progress in this field will become increasingly dependent upon this approach. The systems which we wish to understand are too complex for simple approaches. Even if the ‘laws’ which govern our systems are found to be relatively simple (which most researchers doubt will be the case), the hundreds, or thousands, of variables necessary to describe the systems cannot be managed without the help of some form of computer model. It may even be the case that after learning to master the complexity of these systems, simpler underlying laws will be discovered. For the present, the combination of empirical analysis with theory development which is made possible by computer modelling seems to be the most productive approach.

1.3 Setting this book in context

Throughout the 1970s when this author was doing the research described in Integrated Urban Models, other researchers were also working on the general problem of integrating transportation and location or land-use models. It is unfortunate that rather little of that work is available in book form, though the reader who cares to take the effort can find some of it in journal articles and various published working papers. The purpose in the next few paragraphs is to give some guidance as to where this other work may be found, while at the same time giving a very general notion of how it relates to the material in this book.

Several researchers have worked on integrated transportation and land-use models which are, in very general terms, rather like the author’s ITLUP package. Echenique and his associates have developed a model package called MEPLAN which has seen several actual agency applications (Echenique, 1985). Mackett spent quite some time on the development of the LILT package, which has been extensively tested on data for the city of Leeds in England (Mackett, 1980). A third major effort of this sort was conducted by Wegener, who used data for the city of Dortmund, West Germany (Wegener, 1986).

The recently completed report on the work of the International Study Group on Land-Use/Transport Interaction (ISGLUTI) describes these efforts as well as several others (Webster et al, 1988). Additional efforts described there include those by Floor and de Jong (1981), Lundquist (1973), Nakamura et al (1983), and Sharpe and Karlquist (1980). In addition, there is a useful recent review by Berechman and Small (1988).

Two books have recently been published (after most of this one had been written), both of which are about integrated transportation and land-use modelling. One, by de la Barra (1989), unfortunately gives very little information on the author’s actual models, with the bulk of the book being a review of theories which form the basis for integrated modelling. The other book, by Kim (1989), is focused solely on nonlinear mathematical programming formulations of the combined location-distribution-assignment problem as discussed here in chapter 8. Kim also presents considerable discussion of the mathematical programming approach and descriptions of several solution procedures for these models. These two books have in common a total lack of reference to each other and a virtually complete lack of reference to any of the published work on integrated models such as, for example, that reviewed in the ISGLUTI report. This is regrettable, as it is by now quite clear that no individual researcher’s efforts are going to solve the integrated transportation and location modelling problem.

The ISGLUTI report makes an excellent starting point for a review of a number of the integrated model systems which have seen substantial empirical testing, and the list of references in the report is invaluable. Each of these modelling efforts has its own differences from and similarities to the others. All attempt, generally, to link models of activity location and models of the assignment of trips to transportation networks. In some cases an attempt is made to represent explicitly both demand and supply in activity location, whereas in other cases these phenomena are subsumed into a single model-process. To some extent, these model system configurations evolved partly in response to the theory-data interaction. The availability of a particular data source for the development of a model inevitably helps shape the model. This serves to explain some of the differences between the models. Other differences result from the perspectives of the individual developers of the models and simply underscore the fact that there remain many unsolved problems in this general area and many possible approaches to solving them. Common to all of the modelling efforts discussed in the ISGLUTI report, ITLUP included, is a constant testing of hypotheses against real data. This results in a rather persistent and productive cycling between model development and analysis of data. In general, the output of this process is an increased understanding of the phenomena being studied, and, one hopes, improved models.

The particular virtue of integrated transportation and location models is their ability to represent a much broader range of phenomena than can be done by either type of model in isolation. Most of these model systems are run in a recursive mode, with the output of one run of the model being the input to the next, and with the output of one time period being input to the next. This approach, along with the location – transportation interaction, results in integrated models being uniquely able to address indirect and induced response effects. Thus, for example, a policy which might be shown by a transportation model to reduce travel cost, could be shown by an integrated model system to have the longer-term effect of increasing travel cost. This contradictory result would be a consequence of the spatial relocations produced by the initial travel-cost reduction, and the possible tripmaking increase attendent upon those spatial relocations.

The work of Bertuglia et al (1987) comes from a somewhat different perspective. In chapter 7 of that work Wilson articulates a very sophisticated framework “which contains the main submodels of an urban and regional model system”. Although there is no attempt at empirical analysis in that discussion, the consideration of the elements of the framework is a useful goad towards further model development and empirical testing. Wilson discusses this framework again in a somewhat later paper (1989). These discussions, along with those, say, of Bertuglia et al (1987) in the last chapter of their book, hold out possibilities, or possible paths, for future work.

The work described in Integrated Urban Models is in the same class of work as that discussed in the ISGLUTI report. That class of integrated models are the only reasonable option for agency application at present. The work described in this book takes that of its predecessor as a starting point. In order to explore questions of equilibrium, research is described here which addresses the virtues and the faults of mathematical programming formulations of location models, transportation models, and integrated models. This is the point of connection to work such as that of Boyce et al (1983), Kim (1989), Prastacos (1983), Sharpe et al (1984), and Wilson et al (1981). At the same time, the research described here is firmly grounded in empirical analysis and verification, and thus retains its intimate connection to the previous development of ITLUP. A third facet of the material given here is the beginning of what will undoubtedly be a lengthy effort to incorporate the work done in economics on disequilibrium and partial adjustment models (Bowden, 1978a; 1978b; Fisher, 1983; Maddala, 1983). Many of the comments on various methods discussed and tested here, some given en passant, will be of use to agency or consultancy practitioners. The other purpose is to try to define some major paths of endeavor for the research that will take us into the twenty-first century.

1.4 The plan of this book

After this introductory chapter, there are two chapters to set the context. In chapter 2 simple numerical examples of location, distribution, mode split, and assignment models are given. These are the types of models which are discussed and extended throughout the remainder of the book. In chapter 3 a brief introduction to optimization is given and, again through the use of simple numerical examples, some of the concepts and algorithms which will be used later in the book are illustrated.

Chapters 4 and 5, and 6 and 7, are parallel pairs. In the first two, chapters 4 and 5,...

Table of contents

• Cover
• Half Title
• Title Page
• Copyright Page
• Original Copyright Page
• Preface
• Acknowledgements
• Dedication
• Table of Contents
• 1 Introduction and overview
• 2 Location and transportation model examples
• 3 Optimization and optimizing models
• 4 Location models in optimizing frameworks: 1
• 5 Location models in optimizing frameworks: 2
• 6 Transportation models in optimizing frameworks: 1
• 7 Transportation models in optimizing frameworks: 2
• 8 Simultaneous location-transportation models: 1
• 9 Simultaneous location-transportation models: 2
• 10 Equilibrium solutions to location models
• 11 Preliminary development of dynamic spatial models
• 12 Conclusion: next research steps for dynamic spatial models
• Appendix
• References
• Author Index
• Subject Index