Abstract: Landslide is the result of the interaction among several nonlinear factors; and a large amount of field data is needed to predict it. The Mamdani fuzzy inference system(FIS) model is applied to the Badong-Zigui area in Three Gorges Reservoir Region; and the results are analyzed. Based on the 3 kinds of factors (geomorphology, eco-environment and geology), which are collected by GIS and RS, 192 rules of inference are established. The regional landslide prone zoning is mapped and the susceptibility indices are derived using the Mamdani FIS, which is developed and employed in Matlab. The area under curve (AUC) obtained from the receiver operating characteristics (ROC) based on results is 82.8%, indicating that this landslide susceptibility map is successful. With the parameters selected by spatial information technology, the Mamdani FIS using its nonlinear analysis capabilities, based on expert opinion and specific inference rules, is free from prior knowledge, simplifying the data requirement. In addition, the model could suit to different areas with simply changes in reference rules based on expert opinion, demonstrating a strong adaptability.

Key words: landslide susceptibility, Mamdani-FIS(fuzzy inference system) model, geographic information system(GIS), remote sensing(RS), Three Gorges