Urban Systems (Routledge Revivals)

Contemporary Approaches to Modelling

C S Bertuglia, G. Leonardi, S. Occelli, G. A. Rabino, R. Tadei, Alan Wilson, C S Bertuglia, G. Leonardi, S. Occelli, G. A. Rabino, R. Tadei, Alan Wilson

• Publisher: Routledge
• Year: 2013
• ISBN:
  9781134695263
• Edition: 1
• Topic: Architecture
• Subtopic: Urban Planning & Landscaping

This edited collection, first published in 1987, provides a comparative analysis of different approaches to urban modelling, and lays the foundations for the possibility of integration and a more unified field. The first part contextualises the development of the field of urban systems modelling, focusing on the variety of approaches and possible implications of this on the future of research and methodology. Next, the editors consider economic and 'non-economic' approaches, followed by an analysis of spatial-interaction-based approaches. Providing an overview to the field and research literature, the overarching argument is that there should be an integrated methodological approach to urban system modelling.

Part One

URBAN SYSTEMS MODELLING: THE CONTEXT

Chapter 1

INTRODUCTION

C.S. Bertuglia and A.G. Wilson

If Lowry's (1964) Model of metropolis is taken as an established ‘beginning’ of urban modelling, then the field is now over 21 years old. It can perhaps be said to have matured, to have come of age in a more than arithmetical sense, in that very high levels of skill have been developed, but also a range of quite different styles of approach. This has drawbacks as well as advantages. Many students of urban modelling, for example, are taught within the confines of one ‘school’. A similar point can be made about approaches to research. This makes it difficult to evaluate effectively the strengths and weaknesses of the different styles of approach. There is very little ‘comparative’ literature. The aim of this book is to fill that gap. This both helps to provide a basis for the evaluation of different approaches but also most importantly, lays the foundations for the possibility of integration in a number of respects and the development of a more unified field.

The strategy we have used to achieve this objective can be described as follows. First, in the remainder of Part One, we present a substantial historical review of the development of urban modelling, focusing in turn on the variety of approaches on the one hand and the implications of this history for future research policy on the other. Secondly, we subdivide the approaches into three main categories, two of which are then further subdivided. The first category is ‘economic’, and the next two could be labelled ‘non-economic’; the first of these labelled ‘operational research’, the next a variety of approaches to ‘spatial interaction’. The labels will ultimately be seen to be too restrictive. For example, there are economic approaches to spatial interaction, and many operational research and spatial interaction models can be given an economic interpretation. Indeed, in Chapter 10 of Part Five, arguments and theorems are presented which demonstrate formally that there is a considerable degree of equivalence at a basic level between the various approaches. In the remainder of this chapter, however, we discuss briefly the different approaches considered within each of the main Parts.

Economic contributions to urban modelling can be considered to have a long history through the ‘classical’ contributions to location theory of such authors as von Thünen and Weber. This tradition has proved rich in insight in contemporary research and Martin Beckmann (Chapter 3) is a distinguished exponent. The contribution of Beckmann is, of course, developed strictly according to the principles of the New Urban Economics, and therefore it trades off power of analysis with realism, by introducing the standard simplifying assumptions which are typical of that school.

In order to remove some of these assumptions and build operational models which remain consistent with economic theory, several approaches have been developed, such as disaggregate choice theory and random utility theory. Such a contribution is made by Leonardi (Chapter 4). Much effort in these approaches has been devoted to the behavioural side of the models, by introducing heterogeneity of preferences and imperfect information in the choices of urban actors. Interestingly enough, many of the models obtained in this way bear strong similarities with those derived by entropy maximising, a non-economic approach to be discussed in Part Four (Chapter 7). A different kind of difficulty is exposed in at least some views by another critique: that the underlying assumptions of neo-classical economics are unsound and an analysis which relates back to Marxist economic theory is to be preferred. These alternative assumptions generate new kinds of urban models which are represented here in the chapter by Sheppard (Chapter 5).

Since the second world war, there has been a largely separate set of developments under the heading of operational research with a number of techniques being obviously relevant to locational analysis. Most applications have been to problems which are smaller in scale – such as those of a single firm – than those of urban modelling, but nonetheless there are lessons to be learned. This is particularly important now that there is beginning to be some conceptual integration between some of the models of operational research and the rest of urban modelling (cf Wilson, Coelho, Macgill and Williams, 1981). The history and these more recent developments are reviewed by Colorni in Chapter 6, which constitutes the whole of Part Three since it is distinctive relative to the other approaches.

In Part Four, a number of spatial-interaction-based approaches are presented. We have already mentioned entropy-maximising. When such ideas are combined with the more recent developments of dynamical systems theory, an approach is available which can be used to develop dynamic versions of most of the models needed in the urban field. These are presented by Wilson (Chapter 7). Random utility models, as an alternative to entropy-maximising, have already been discussed in relation to Chapter 4. A further alternative is the cost-efficiency theory of Smith (Chapter 8). His contribution is a good example of the restrictiveness of the classification into ‘economic’ or ‘non-economic’ approaches. Although it is true that it is closely related to entropy maximising, it is also true that it starts from economic justifications which relate it to approaches like random utility theory. The main difference is that while in the latter a disaggregate approach is used, in cost-efficiency theory aggregate assumptions about economic regularities are used. It is also possible to take a more explicitly dynamic view of flows (as well as locational behaviour) and a broad foundation for this approach is provided through compartment models and master equations. These techniques are described by de Palma and Lefèvre (Chapter 9).

As noted earlier, some formal integration is achieved in Chapter 10. However, we hope that the book as a whole will provide foundations for readers to carry out their own assessment of different approaches. Moreover, it is worth emphasising that integration is not just a formal task, nor does it leave complete freedom to assemble different elements from different approaches eclectically. It is something in-between which requires a balance between formal rigour and skilful use of the best contributions from various disciplines. This argument suggests that readers should use the book as a tool to achieve this balance in a way which is consistent with their own modelling styles. This is the way the book was conceived and the main goal underlying the research project which generated it.

Once again, it should be stressed that not all approaches are good for everything, not everything is best treated by all approaches. It is an ambition of this book to assess the achievements and limitations of such diverse approaches as classical space economics, recent behavioural economics, operational research, spatial interaction, and dynamical systems theory. It is an even more ambitious goal for the book, however, to outline paths towards consistent (that is, theoretically sound and operationally feasible) integration of the different contributions. The task of integration is of course a difficult one, and more needs to be done. In Chapter 2, this theme will be brought into a sharper focus in terms of emerging strands for future development. Chapter 10, as mentioned already, will provide some examples of what can be achieved at a purely formal level. Further steps towards building a truly integrated dynamic urban modelling framework will be reported as further research progresses.

Chapter 2

AN HISTORICAL REVIEW OF APPROACHES TO URBAN MODELLING

C.S. Bertuglia, G. Leonardi, S. Occelli, G.A. Rabino and R. Tadei

2.1 Introduction

The aim of this chapter is to offer an historical review of urban modelling involving location, transport and the relationships between them and to assess the most important directions for future research in this field. The emphasis on location-transport relationships provides a useful basis for structuring an analysis. Our intention is to: (a) describe the state-of-the-art (beginning in Section 2.2); (b) identify the main directions in which this work is advancing; and (c) look at the most promising avenues of research, with the aim of establishing how research efforts can be most effectively channelled in the future (in Section 2.3).

The interrelationships of location and transport are highly complex and consequently any description of the phenomena involved necessarily takes various forms and cannot claim to be exhaustive. Our classification of studies is clearly not the only one possible, but appears to correspond by and large to the main themes emerging from research in this field (and from the authors' experience seems the most useful). We maintain, in any case, that other classifications are likely to differ only marginally. The categories we have adopted for the purpose of this study are therefore: (a) interrelations between location of economic activities and commodity flows (Section 2.4.); (b) interrelations between services and the journeys generated by their use (Section 2.5); (c) interrelations between residential location and journeys to work (Section 2.6); (d) interrelations between location and transport in the urban system (Section 2.7); (e) interrelations between urban form and transport (Section 2.8).

There are clearly a large number of theoretical, methodological and practical problems involved in the exploration of these interrelationships. Later in the review we describe the new trends emerging in this field. It is possible, however, at the outset, to identify the three main streams which appear to be the most promising: (a) models of spatial choice behaviour; (b) mechanisms of the dynamics and evolution of location-transport systems; and (c) the application of various economic paradigms to the analysis of location-transport interrelationships.

The conclusions about research priorities are sketched in Section 2.9.

2.2 The Most Important Existing Approaches

The obligatory starting point for a study aiming to describe the most important contributions to the understanding of the interrelationships between location and transport, is that of classical urban economics of the 19th century and in particular that emanating from the German school. More precisely, we must go back to von Thünen (1826) and later Weber (1909), Hoover (1948) and Isard (1956, pages 77–142 and 221–287), who wrote specifically about the interrelationships of location and transport; and to Christaller (1933), Palander (1935), Lösch (1940) and more recently Isard (1956) – again, Lefeber (1958) and Greenhut (1963) who have contributed to the question of a general location theory which implicitly takes into account transport costs as well. The most natural ‘descendant’ of this line of research is Beckmann (summarised in his 1968 book – and see also the bibliography) who managed to combine systematically geographical theory and urban economics with the methods and techniques which operational research had made available from the late 1940s on (cf Chapter 6 below).

This alliance of geographic theory and urban economics forged ahead thanks to the work of, among others, Wingo (1961), Alonso (1964-A), Richardson (1969-A, 1969-B, 1973-A, 1973-C, 1977-B, 1978), Papageorgiou (ed) (1976-A), Fujita (1978), Puu (1978, 1979-B, 1981-A, 1981-B, 1981-C, 1982-C) and Kanemoto (1980-B). In all these cases, additional references are cited in the bibliography.

From the 1960s onwards there was a parallel development of powerful mathematical models. Unlike the types of study referred to above, each based on specific interpretative paradigms, usually from economics, this branch had its roots in quantitative formulations of empirical regularities which (at least initially) did not have economic underpinnings. These developments were built on two objectives: the building and testing of tools for direct application in town planning; and the construction of an alternative to strictly neo-classical approaches. The models constructed for planning purposes were inspired initially by the pioneering work of Hansen (1959) and found their mainstay in the Lowry model (1964). In fact Lowry's model can be taken as the foundation stone and point of departure for this whole style of model development. The models developed by the Perm Jersey Transportation Study (Seidmann, 1964, 1969) can be considered as having originated from Lowry's model, together with sectoral studies such as Herbert and Stevens' model of the housing market (1960), the industrial and service locations model and those related to transport infrastructure (cf Merlin, 1968, pages 37–39). Other models developed at the same time as Lowry's, though less global in their range, were those of Huff (1963, 1964), Harris (1964-A) and Lakshmanan and Hansen (1965). None of these models, however, provided anything like the impulse generated by Lowry's model, the influence of which has been described fully by Goldner (1971). Garin (1966) gave a matrix version of Lowry's model which is frequently referred to. Although Garin's version facilitated computation and hence accelerated its diffusion, it overshadowed for a long time certain potential general developments of the model such as the non-linearity deriving from the presence of spatial constraints (which cannot be easily dealt with in a linear algebra version). Lowry's model itself was in fact immediately applied and extended. Among the early modifications were those of Crecine (1964), Brotchie (1965), Goldner and Graybeal (1965), the Bay Area Simulation Study (1968), Crecine (1968), Goldner (1968), Echenique, Crowther and Lindsay (1969), Wilson (1971-B), Batty (1971-A), Echenique, Crowther and Lindsay (1971) and Goldner, Rosenthal and Meredith (1971). For a review of the first modifications and developments of Lowry's model, see Goldner (1971) and with reference to Great Britain, Batty (1972-C). Many other models, including certain of Wilson's (cf 1969-A), can be seen as having originated from that of Lowry. In listing them in this way we do not wish, however, to over-emphasise the aspect of ‘continuity’ or to lose sight of important innovations which certain models introduced. The applications of Lowry's model (and its variants) were numerous and of different levels of complexity. Among the first, we have those of Batty (1969-A, 1969-B), Echenique, Crowther and Lindsay (1969), Echenique et al (1969), Batty (1970-A), Cripps and Foot (1970), Echenique and Domeyko (1970), Masser (1970), Stubbs and Barber (1970), Barras et al (1971), Echenique et al (1973), Batty et al (1974), Bertuglia and Rabino (1975), Christiansen (1975), Ayeni (1976-A), IRES (1976), Piasentin, Costa and Foot (1978). Certain of the modifications and applications cited tend towards a dynamic version and are therefore dealt with later on. Alongside the models which were directly or indirectly inspired by Lowry there are others which are often forgotten. We wish here to mention at least one of these, the statistical/econometric approach exemplified by the POLIMETRIC model (Traffic Research Corporation, 1964) and EMPIRIC (Hill, 1965, Hill, Brand and Hansen, 1966). Both of these models are described more fully in Merlin (1968, page 39). Criticisms have inevitably been levelled at Lowry's model and with the experience of those who have applied it certain limitations and oversimplifications have been brought to light and suggestions for overcoming them proposed. The earliest modifications came from Cripps and Foot (1969), Batty (1970-B) and Broadbent (1970).

The second aspect of model development referred to (that of the formulation of alternatives to the traditional neo-classic approaches), was stimulated by the work of Wilson. The impact of his work can be attributed to his...