The use of predictive modeling techniques for optimal exploitation of spatial databases: a case study in landslide hazard mapping with expert system-like methods

Abstract.

A case study is presented in which different probabilistic prediction models (Bayesian probability, fuzzy logic "and", "or", "sum", "product", "gamma" operations, and certainty factors) are used to produce landslide hazard maps for a hilly and mountainous region in the northern Apennines, Italy. Seven data layers are exploited to detect most vulnerable areas: lithology, distance from the geological lineaments, annual rainfall amount, land cover type, topographic slope and aspect, and the distance from hydrographic network segments. The results of the different predictions are compared using the prediction rate index and critically discussed, to evaluate the possibility of using readily available databases for land planning.