Abstract

Cities are complex systems, with many dynamically changing parameters and large numbers of discrete actors. Computer modelling is becoming an increasingly important tool for examining how such complex systems operate. This thesis explores the role of agent-based models and specifically how geospatial agent-based models can aid our understanding of urban dynamics.

The thesis begins with an introduction to urban modelling; classical theory and techniques for modelling urban systems are first briefly reviewed. This is followed by recent advances in urban modelling. In particular, the shift from studying the aggregate to the individual utilizing ideas from complexity theory and automata approaches (e.g. cellular automata and agent-based models) is identified and discussed.

A detailed discussion of agent-based modelling is then presented and how such models can be used to study a wide range of urban phenomena. It is noted that while many applications have been developed to study urban phenomena, they have not been overtly spatial. Key challenges to developing geospatial agent-based models are examined and how the use of simulation/modelling systems are allowing researchers to more easily explore a diverse range of urban problems that cities face, is described. A detailed discussion of the Repast simulation/modelling software system is then presented, specifically highlighting its core functionality and its ability to use geospatial data to build artificial worlds.

A basic model framework is then presented focusing on the integration of vector-based geospatial data and agent-based modelling for the exploration of specific urban applications. The integration of geographical information systems (GIS) and agent-based modelling provides the ability to link agents to actual ‘real’ world places. This framework allows for the testing of different hypotheses and theories of urban change, thus leading to a greater understanding of how cities work and function. The emphasis of this model framework is to provide a set of tools which allow for the rapid development of agent-based models incorporating data, space and geometry directly into the modelling process. This framework is then applied to two model applications, a residential segregation model and a location model of residents and businesses within an urban system. The models highlight how traditional urban theories can be fused with more recent approaches to modelling and demonstrate the importance of space in the modelling process. Specifically we discuss how the individual actions of a range of agents, both in space and time at the micro-scale, result in aggregate patterns emerging at the macro-scale. These models are explored under a variety of experimental conditions and we conclude with proposals for further research and applications.