Abstract: Landslide susceptibility mapping is an effective means of landslide prediction. We built a hybrid model integrating RBFNN (Radial Basis Function Neural Network) with Certainty Factor (CF) for the mapping of landslide susceptibility in Chenggu County, Hanzhong City of Shaanxi Province. First of all, we selected slope, aspect, plane curvature, profile curvature, elevation, mean annual precipitation, distance to road, distance to river, distance to fault, NDVI and lithology as landslide's triggering factors and then quantified such factors by calculating the corresponding CF. Secondly, we divided the field survey data of 184 landslides into training data and test data with a ratio of 7∶3, and generate the landslide susceptibility maps using RBFNN-CF and RBFNN models, respectively. Finally, we evaluated and compared the mapping results and the classification ability of the models according to the area under the ROC curves. The results suggest that the classification and generalization ability of RBFNN-CF model are both superior to those of RBFNN model. The hybrid model is worth popularizing in the study area, and the landslide susceptibility maps obtained in this study could also provide references for local landslide prevention and control.

Key words: landslide susceptibility, RBFNN, certainty factor, hybrid model, GIS, ROC curve