Go to Paper
Return to GeoComputation 99 Index

Empirical CA Simulation from a High Resolution Population Surface

MARTIN, David (D.J.Martin@soton.ac.uk) and WU, Fulong, University of Southampton, Department of Geography, Southampton, SO171BJ, U.K.

Key Words: population density surface, cellular automata, SE England, urban growth, simulation

There is growing literature on the application of cellular automata (CA) to simulate the growth of urban settlement forms (e.g. Batty, 1998). CA allows researchers to view the city as a self-organizing system in which the basic land parcels are developed into various land-use types. A model of the urban system is constructed by the aggregation of uncoordinated local decision-making processes. One of the most important potential uses for such simulations is their ability to model the impact of alternative planning regimes on the development process. CA applications, based on hypothetical urban forms, can provide valuable insights, but the interpretation of such modelling is hampered by difficulties in relating the modelled form to empirical combinations of settlement and constraints. The use of CA methods to model the future development of real urban systems is made particularly complex by the tension between self-organization and the application of empirical constraints.

This paper describes the application of CA to the simulation of urban growth in the southeast region of the U.K., an area currently subject to considerable development pressure. The actual settlement pattern is initially modelled as a fine resolution grid using a population surface modelling technique originally developed for use with census area centroid data (Bracken and Martin, 1995), but here applied to unit postcode information that offers greater spatial and temporal resolution than that available from the population census or conventional land-use mapping. This application is a further development of SimLand (Wu, 1998), which makes use of Arc/Info for spatial data management, with AML programs to permit the evaluation of a range of alternative local and regional constraints on the development process. We classify the wide variety of factors affecting development into static and dynamic ones. The success or failure of a seed becoming a developed land use depends on their combined effect on the self-organised process of local growth. This is further dependent upon the threshold that allows such a process to proceed. From the observation of land use states in two time periods, the distribution of land use changes is identified. In general, the threshold and its transformation are used to reflect three types of inputs: the growth rate that is related to economic activity, regional variation, and policy control. With different thresholds applied, simulation can generate a series of scenarios of urban development. Development scenarios are treated not as place-specific predictions, but as possible realizations of the development process from which a number of structural indicators can be derived.

This research has a number of original features: it uses detailed empirical spatial data on a fine resolution grid; integrates global effects with the local self-organisation mechanism of urban growth in a more explicit and parameterised way; uses GIS functionality, thus, provides a closer integration with other decision-making tasks; searches the parameter space through a computationally intensive approach; and uses structural indicators to compare the simulation results with the reality.


Batty, M. (1998) Urban Evolution on the Desktop: Simulation with the Use of Extended Cellular Automata Environment and Planning A 30, 1943-1967.

Bracken, I. and D. Martin (1995) Linkage of the 1981 and 1991 U.K. Censuses Using Surface Modelling Concepts, Environment and Planning A 27, 379-390.

Wu, F. (1998) SimLand: A Prototype to Simulate Land Conversion through the Integrated GIS and CA with AHP-Derived Transition Rules, International Journal of Geographical Information Science, 12, 63-82.