High-resolution modeling and forecasting of sealed surface cover distribution in flanders

This study explores the potential of modeling current and forecasting future sealed surface fraction within residential/commercial areas, using neighborhood descriptors as predicting variables. Composition of the surrounding land use and spatial positioning with respect to the urban centers and transport networks are assumed to function as attraction/repulsion factors for soil sealing. Global estimation models are compared to regression tree models, which partition space based on explanatory variables. The latter models yield significant improvements in estimation accuracy compared to global models. Linking the best performing model to a cellular automata land-use change model predicting the future distribution of residential/commercial areas at the level of the Flanders region resulted in an increase of 21.1% and 39.5% of sealed surface cover by 2030 and 2050 respectively.