An activity-based CA model for Flanders, Belgium. Robustness analysis and improved distance computation in a variable grid representation

Cellular automata (CA) models are increasingly applied for simulating land-use change in urban areas. However, in areas with strongly mixed land uses, like Flanders, Belgium, different types and intensities of human activity occur within a single dominant land use. This is in conflict with the discrete and dominant land-use states applied in CA. The direct modelling of the intensity of activities (population density and employment in different sectors) within a CA grid environment is an interesting alternative to model mixed and multifunctional land use.
In this research, an activity-based cellular automata (ACA) model, developed by White et al., will be further enhanced, applied and calibrated for Flanders. It also uses a variable grid approach: linking with a regional model is not necessary because the neighbourhood is expanded to the entire modelling area. The model should be able to cope with the complex multi-nodal structure and messy morphology of Flanders, typified as it is by multifunctional land use and diffuse, fragmented urban development strung out along roads.
In this paper we firstly show the results of a robustness analysis carried out to investigate whether the model behaves as expected under extreme circumstances. Secondly we propose a method to compute and store distances between cells in the variable grid ACA. This calculation should be based on the existing transportation network rather than on simple Euclidian distances applied in classical local CA neighbourhoods.