Landslide susceptibility mapping based on the Weights of Evidence model for mountainous areas of Quang Nam province, Vietnam

Tóm tắt

This study shows the results of landslide susceptibility mapping for the southwest region of Quang Nam province using the Weights of Evidence (WoE) model. The input data consists of a landslide inventory and ten influencing factors, i.e., geology, distance to fault, elevation, relief amplitude, slope, aspect, rainfall, soil type, land use, and distance to road. The landslide inventory was constructed from three principal sources:  fieldwork survey, legacy data from previous studies, and additional analytical data from high-resolution Google Earth satellite imagery. The landslide locations were randomly categorized into two parts in the ratio 70/30: 70% (811 landslides) for modeling and 30% (348 landslides) for verification. All input data are normalized and constructed into the GIS landslide database. The results of the multicollinearity test show that no collinearity existed between ten input variables. The computation of the weights for classes of influencing factors from 70% of the landslide data using the WoE model has allowed the establishment of the landslide susceptibility map. The model performance was evaluated by using the receiver operating characteristic (ROC) analysis. The area under the curve (AUC) was computed for the success rate curve (using 70% landslide data) and the prediction rate curve (using 30% landslide data) at 0.855 and 0.844, respectively. Thus, it can be confirmed that the landslide susceptibility mapping based on the WoE model was very reliable in the study area.