Reviewed by
Mike Bithell
Department of Geography, University of Cambridge
My first thought about this book was that its heart lay in the right place: cities are positioned as far-from-equilibrium complex systems where flows not just of material goods, but of information and ideas, on multiple interacting networks, play a key role in structuring and determining city evolution. Developing tools for understanding processes using simple models is emphasised, while observing that the problem of equi-finality (where many different model formulations may provide equally good fits to data) is one that remains to be resolved. A good feeling for the extensive history of the subject is provided throughout, while giving an overview of various different aspects, but comprehensive description of specific city models is mostly left to the literature. It is stressed the book cannot as yet provide a completed prescription for building city models: While multiple different approaches and views may need to be taken depending on the problem at hand, there yet remains much to be understood. However, my feeling by the end of the text was that the good intentions expressed in many places were not really borne out by the bulk of the material; but more on this below. The book is divided into three sections: Foundations and Prerequisites, The Science of Cities, and The Science of Design.
Section I begins with an introductory chapter that touches on some basic aspects of complexity, dynamical systems and hierarchies, and then scaling relationships, introducing “seven laws of scaling” for cities, before briefly introducing gravity models. Chapter two then expands on the idea of flows, starting with a historical overview before elaborating more on gravity models viewed as flows between origins and destinations. The third chapter introduces some standard network ideas including threshold behaviour, small worlds, and measures of centrality. The idea of regarding flows as processes on networks that can be represented as Markov chains is briefly expounded. Some of the network measures are applied to the City of London, regarded as a planar graph. One departure from the standard treatment is to introduce bi-partite graphs, in this case looking at the coupling between origins and destinations for residence and employment zones.
Section II starts with two chapters on rank-size relationships. The reader is advised to take a look at the colour figures on www.complexcity.info/newscience here, as they make the black-and-white figures in the printed version somewhat easier to interpret. An initial discussion of simple one-dimensional random processes that can generate the observed rank-size power law is followed by a nice exposition of a number of different time series datasets for city size, but also for business sizes and building heights, using the idea of the “rank clock”. This gives a good feel for the way in which cities grow and fade in size relative to others globally, but also separately in diverse locations such as the US, UK and Italy. Subsequently some more elaborate 2-D simulation treatments for trying to model the form of the rank -size relationship are given, including a network-based approach, the point being that the simpler models do not seem to capture the volatility of the time series very well. Here change over time begins to make an appearance, but chapters 6 and 7 switch to a discussion of space syntax viewed through the medium of bipartite graphs on static street networks and their junctions. A broad range of of distance and accessibility measures are looked at, emphasising the difficulty of interpretation that can arise when trying to summarize interacting but static network patterns. Chapter 8 gives a brief overview of fractal growth models, including diffusion-limited aggregation and the use of cellular automata, and the section concludes with an overview of a specific entropy maximizing model of London. This chapter comes closest to displaying a comprehensive city model, although the main focus is on land-use transport interaction models in equilibrium, and comparative statics for looking at policy “what if” analyses. A case is made for models that operate rapidly in an interactive dialogue with stakeholders, as a means for exploration and learning, rather than emphasizing strict accuracy.
Section III is devoted almost entirely to linear algebra and Markov-chain models on rather abstract networks, concentrating on the resulting (normative) equilibria as guides to planning. A good deal of stress is laid on the idea that models here are not looking for right answers or correct decisions, but can be used as exploratory tools. The way in which networks can be used to integrate the effects of multiple actors, factors for change, policies and degrees of power and interest is intriguing although based heavily on Coleman's theory for collective action: the focus remains on an approach that views consensus building as linear sequential averaging, with the result that (at least in strongly connected networks without commensurate cycles) Markov processes lead to an equilibrium that depends only on weighted network structures, and not on initial states.
Overall, for a book that at the start began to talk about complex systems, the whole text has a decidedly linear equilibrium feel. There is a great deal of linear algebra, especially in the final section, where neo-classical ideas about optimality become steadily more prominent. In the discussion of rank-size relationships there is no real engagement with the social and economic reasons why people might move between settlements, and it would seem to me that here is the place to look for some of the answers to the equi-finality problem: matching multiple different patterns based on the reasons why people act, rather than trying to guess aggregate population level rules I would have thought more likely to be productive. The chapters on space syntax seemed not to include any feel for how cities are actually experienced by their inhabitants, a view that would have been helped by a more individual-based approach. By chapter 9 I was hoping to see some real evidence of time-dependent complex system analysis, in which the flows of water, goods, materials, energy, land, people and information began to be coupled together to give dynamical systems that could change over time, but such a synthesis did not emerge. In the last section chapter 12 begins to hint that there might be more to decision making than just sequential averaging, but the exposition of simulation as a way forward falls short of any serious development: that there might be cases where conflict depends upon history and leads to irreconcilable differences, for example, is not considered. Ultimately a rather eclectic treatment that left me thinking that the new science of cities has a long way to travel, and that there is a great deal of space remaining in which dynamical agent-based models should be able to play a much larger and more-exciting role.
Return to Contents
of this issue