Empirically Derived Probability Maps to Downscale Aggregated Land-Use Data

Land-use simulation results are often provided at spatial resolutions that are too coarse to establish links with local or regional studies that, for example, deal with the physical or ecological impacts of land-use change. This chapter aims to use novel spatial statistical techniques to derive representations of land-use patterns at a resolution of 250 metres based on aggregate land-use change simulations. The proposed statistical downscaling method combines multinomial autologistic regression and an iterative procedure using Bayes’ theorem. Based on these methods, a set of probability maps of land-use presence is developed at two time steps. The method’s low data requirements (only land-use datasets are used) make it easily replicable, allowing application over a wide geographic area. The potential of the method to downscale land-use change scenarios is shown for a small area in Belgium using the CORINE land-cover dataset.