In this research, genetic algorithm is used to evolve the framework of artificial neural networks automatically. The genetic adaptive neural networks are employed in the classification of satellite images and assessment of landslide disaster due to hillside land use. Hillside land use factors causing landslide in the watershed of eastern Taiwan are quantified. Using discriminant analysis to classify the relations of developing level, rainfall and landslide and using geographic information system, a working platform to plot disaster potential map is established. The assessment procedures for rainfall-induced landslides are constructed. This research result can be a help for sustainable management of watershed.