GIS-Based Logistic Regression Application for Landslide Susceptibility Mapping in Son La Hydropower Reservoir Basin

Landslide susceptibility map is an important tool for planning and management of landslide prone areas in better way. Logistic Regression (LR) based machine learning model has been successfully applied in many parts of the world for landslide susceptibility mapping. In this study, we have applied the LR model combined with GIS to create landslide susceptibility map of the basin area of Son La hydropower plant catchement, Vietnam. For this, a total of 186 landslide locations identified in the basin area were used to construct landslide inventory. In total, 12 landslide conditioning factors (elevation, aspect, slope, curvature, deep division, fault density, river density, road density, weathering crust, rainfall, aquifer and lithology) were used for training and validating the model. Various standard statistical indices including theROCcurve were used to evaluate performance of the LRmodel. Results show that predictive capability of model performance is very good (AUC = 0.832) in accurately mapping landslide susceptibility of the study area. Thus, it can be concluded that the LR model is a great tool in constructing a reliable landslide susceptibility map of the study area, which can be used in better landslide hazard management.