ABSTRACT

This is work in its early stages, so there aren't many pretty pictures. Cellular automata (CA's) yield models for spatial dynamics that are realistic, flexible and easy to use, even in situations without good deterministic models. Hence they are very popular, for example in ecology, traffic studies, and urban and peri-urban planning forecasts. However, the usual indices of agreement (such as the kappa index) will distinguish between different realisations of the same (usually probabilistic) CA, while the measures of pattern (such as the fractal dimension) tend to fail to differentiate the output of models that are known to different. In this talk, I consider a class of CA models for the spatial dynamics of species richness. The idea is to compete with a spatially explicit hierarchical Bayesian model (work by the other part of our team, led by Alan Gelfand) for predicting species richness from environmental data for the Proteaceae of the Cape Floristic Region. The question is whether a CA model, neutral rules can reproduce the spatial pattern of species richness from a random start. I present preliminary results for simpler domains on a few of the possible tests.