Order this book
CRESS, University of Surrey
The findings for the five models are reflexive and interesting taken as a whole. However, there is a lack of transparency, caused by the brevity in which each model is discussed, meaning the findings relevant to JASSS readers are less in depth than one might hope. As a doctoral thesis, this book does not make for an entertaining read, and of course it should not. It is also not particularly compelling; the case for the significance of the work is not persuasive. The value in reading this book (for which you or your organization will have to pay an eye watering US$106) is in considering several traditional methods applied to the problem in a strongly reflexive manner alongside the ABM. No real interest is shown in policy or real world repercussions, rather the thesis aims to answer strictly methodological questions; this applied look at the various methods is the real strength of the work.
Arsanjani implements four traditional models (a cellular automata model, a Markov chain model, a cellular automata-Markov model and a logistic regression model) and an ABM; they are applied to the LUCC issue of rapid urban expansion in Tehran. The four traditional models are all used to help city planners with their forecasts and understanding of land use patterns; here, the aim is to reflect on their suitability and make comparisons with the ABM. The policy problem itself (urban expansion) is discussed in a somewhat normative and emotive way, which feels strange in a thesis; we are told that rapid expansion is becoming a "national disaster" (pg., 3) and "massive immigration towards the city must be stopped" (pg., 3).
Perhaps we can forgive this lack of detail on the real world problem as the focus is on the methods; however the background rationale for any type of modelling here is not explored in enough detail. How the proposed benefits of using an ABM would translate to policy is not dealt with sufficiently. The literature review itself gives a very useful description of many concepts and issues in the relevant areas of LUCC and simulation; however it feels like something you might expect from a textbook, rather than a thesis. There is not enough space given to the many studies/papers around these kinds of problems using ABM approaches. Furthermore there is no critical engagement with the literature. The reader is left with a solid basic understanding of the topic, but with less sense of the position of this work within that field.
All five models are given fairly equal amounts of methodological discussion. For the ABM, the methods sections give a good overview of the model, but lack real detail. This means that some key questions for understanding the model are not touched upon, such as, how data is used, or how agents make decisions. This leaves the feeling of a lack of transparency, a common problem for many methodologies, but a potentially fatal flaw for a method still not fully mature and understood in mainstream social science. The lack of a formal description of the model leaves its presentation ill-defined; perhaps the use of some pseudo-code, or an ODD protocol may have helped here. These formal descriptions can be clunky but ensure that the model is comprehensively presented. As always, there are arguments to be had about the assumptions underlying the model, the behaviour rules and interactions are all setup and justified quite intuitively. Though they feel reasonable enough, I am sure they could be challenged; but this is not the point of the work, so it is not discussed by the author.
The headline conclusion that the ABM is superior to the traditional methods for these kinds of problems is made unassumingly, hidden in a long general discussion and the listing of strengths and weaknesses of each model. Arsanjani suggests that the ABM benefits from taking the best bits of all the other models, but without the drawbacks. This makes sense on first thought, but in fact, the suggestion is made throughout the thesis that the results of the traditional models are used as inputs for the ABM. The author states "the strengths of the traditional methods will be imported into the [...] ABM" (pg., 95), it is not made explicitly clear how this is done. Furthermore, in this sense the ABM is not superior, but just a further step in the modelling cycle, reliant on the modelling done prior to it; this contradiction is not really dealt with.
The other key objective, the task of promoting building an ABM inside a GIS environment as a method is technically difficult, and achieved by Arsanjani (who is clearly a skilled modeller). However, this is an increasingly common approach, with various options open on how to implement such a model. This means the work loses some novelty, furthermore the case for the value added by this approach is never made in enough depth.
Overall, this book is an excellent source for those working in similar areas, familiar with ABM, and concerned with methodological reflection and innovation; however it is less rewarding for those looking for a wider discussion and/or an interest in the real world problem of urban expansion.
Return to Contents of this issue
© Copyright JASSS, 2012