Modelling spatial and temporal patterns of tropical land use change

We developed two spatially explicit models to simulate rates and patterns of tropical land use change. These models also calculate total amounts and spatial distributions of the carbon content and carbon dioxide exchange resulting from deforestation and other land use changes. We use two basic approaches: hypothesis deduction (GEOMOD1) and statistical deduction (GEOMOD2). The hypothesis deduction approach for selecting pattern drivers is based on user-supplied assumptions about how people actually use land. The statistical deduction approach analyses historical patterns of land use change and compares them to user-supplied map layers of physical and cultural attributes. The model then chooses drivers based on the best fit of the patterns. We used digitized and remotely sensed data for Southeast Asia and Africa to test these models. We found that: (1) the variation in accuracy of the model predictions (74-96%) depends on the time scale used, the number of land classes modelled and the accuracy of initializing data; (2) the drivers of land use change are scale- and terrain-dependent: specifically topographic features are more important than climate variables for large scale simulations where topography is rugged; (3) the amount of land use change is influenced greatly by population growth and land use policy in different countries, but physically features fundamentally determine the pattern of land use change; and (4) our spatial modelling approach does improve the spatial and temporal resolution of carbon release estimates from tropical land use change, although it does not yet fundamentally change the magnitude of carbon exchanged relative to previous national level studies. Application of these models over longer periods of time and larger land areas, however, will eventually refine our estimates of global carbon exchange due to land use change throughout the tropics.