Classifying urban models

Classifying urban modelsFifty years ago, Ira S. Lowry (1965) published his 'Short Course in Model Design' in the Journal of the American Institute of Planners as part of the special issue edited by Britton Harris on Urban Development Models.Lowry's paper became a classic for it provided a superb overview of the domain which we now call 'urban models', abstractions of city systems that are represented in various types of mathematical and logical formalisms that are computable; that is, systems that can be implemented on digital computers and which are capable of enabling 'scientific' prediction.The term model was rapidly becoming part of our scientific vocabulary during those years (see Batty, 2007) but Lowry's particular contribution was to sketch out a domain that could be approached through many different styles and types of model, also paying tribute to different kinds of representation of cities from various perspectives that could be articulated in computation.In this sense, he bounded the field and provided us with a map of the terrain that we could then explore.Such explorations have occupied us since that time and half a century on it is worth exploring not only how far have we come but also whether or not Lowry's classification has stood the test of time.We know there are many modelling types such as agent-based, micro-simulation and so on which did not feature as such in Lowry's classification but it is also clear that many of the origins of developments since then do appear at least implicitly in his prescient thesis.If you examine the evolution of different models since their inception in the Chicago Area Transportation Study in 1955(Plummer, 2003), aggregative models based on a fusion of social physics and macro-economic structures such as input-output relations and urban economics dominated the field for the first 20 years.Lowry's (1964) own model itself became the classic exemplar and there were many ways in which this model structure was thence elaborated through disaggregating its activities and adding a more explicit temporal dimension.Although dynamics provided the momentum for different model types, the notion of disaggregation to the point where individual objects comprising the aggregates could be simulated was encapsulated in the development of agent-based models and modelling (ABM).These were considerably more generic in structure than aggregate urban models in that ABM are completely general and can be applied to any kind of system whose objects can be assumed distinct, not just activities in cities.A particular form of ABM rooted in the physical domain emerged called cellular automata (CA) models which essentially articulated the 2D space as a grid of cells, regular or otherwise, and in both ABM and CA, the focus was on how the system behaves, usually in time thus embracing movement and complexity.More recently an earlier set of models called microsimulation have begun to merge with ABM and CA although this particular usage in its strict form relates to a long-standing tradition in the economic and social sciences for simulating individual behaviour described by aggregate probability functions stemming from the work of Orcutt et al. (1961).These classes of disaggregate model are usually Environment and Planning B: