GIS-Based Landslide Susceptibility Evaluation Using Certainty Factor and Index of Entropy Ensembled with Alternating Decision Tree Models

Up to now, numerous models have been developed and put to use by modelers to portray susceptibility of an area to landsliding. What keep them going might be the slightest differences in performance. These differences, however small, still would surprisingly make huge progress in identifying well suited areas for strategic planning. This kept in mind, we aimed to map landslide susceptibility over a critical landslide prone area, the Longhai Region, Baoji City, in China, using two models namely certainty factor (CF) and index of entropy (IOE) ensemble with alternating decision tree (ADTree). As inputs, 93 landslides together with 14 predisposing factors were mapped. Both CF and IOE models pointed at three main factors as the most important ones including residential land use, areas nearby roads, and normalized difference vegetation index (NDVI). Although obtained ADTrees for both models were similar, slightly different results were obtained. IOE-ADTree was more practical, since it better predicts highly susceptible areas. The receiver operating characteristic (ROC) curve cleared further the differences so that IOE-ADTree with 84% fitting ability and 85.3% generalization capacity outperformed CF-ADTree with the respective values of 83.9 and 83.8%. Therefore, the IOE-ADTree exhibits as a promising ensemble model for the study area.