Stochastic Simulation of Land-Cover Change Using Geostatistics and Generalized Additive Models

change is important because it is the only way to evaluate the Abstract consequences of current and recent land-cover trends for the An approach to simulating land-cover change based on pairs of future fragmentation of the landscape. classified images is presented. The method conditions the Transition probability models have been used extensively simulations on three sources of information: an initial land-cover for analysis and stochastic modeling of land-use and landmap, maps of the probabilities of each possible class transition, cover change (Bell, 1974; Turner, 1987; Muller and Middleton, and a description of the spatial patterns of changes (e.g., 1994). Increasingly, these models use spatially variable transisemivariograms). The method can produce multiple simulated tion probabilities to account for the effects of exogenous variland-cover maps that honor each of these sources of information. ables on the transition process (Baker, 1989; Brown et al., The approach is demonstrated for data on forest-cover change 2000b). To estimate probabilities of land-use transition, landnear Traverse City, Michigan. The discussion describes use change is typically modeled as a function of variables extensions to the method and an approach to generating future describing (1) biophysical land quality (e.g., soils and terrain) land-cover scenarios based on socioeconomic information. and (2) location relative to jobs, markets, and amenities (Landis