Assessment Of Co-seismic Landslide Susceptibility And Post-earthquake Debris Flow-prone Catchments Identification

A large number of landslides were triggered by the earthquake and then loose source was increasing rapidly.These loose soils may cause many debris flows in the rainfalls.The aim of this paper is to evaluate the landslides susceptibility and identify debris flow-prone catchments in the area of Wenchuan-Dujiangyan,Sichuan Province.Nine impact factors,elevation,the slope angle,slope aspect,plan curvature,profile curvature,geology,distance to faults,distance to rivers and distance to roads,were selected as the most important conditioning factors.Then we calculated the contribution degree of the nine conditioning factors for landslides by the information gain ratio model.According to the information gain ratio,plan curvature and profile curvature were deleted.The data of seven impact factors was normalized by the Min-max model and the Z-score model.The landslides probability was calculated by the logistic regression model.Five landslide susceptibility maps were developed.The performance was evaluated using the receiver operating characteristic(ROC)curve.Shape factor ratio,landslide area ratio and relief ratio were selected as the conditioning factors for identifying debris flow-prone catchments.We used 3 random 70% data of the catchments to establish the critical value and 30% data to validate the critical value based on the logistic regression model.The performance was evaluated using Accuracy,Sensitivity,Specificity,Positive predictive value and Negative predictive value.The main conclusions:(1)As the higher the altitude,the less landslides distribution and landslides may become the collapses or other geological hazard.Landslides mainly distributed in granite and diorite area,within 3km from distance to faults,600 m from distance to rivers and 5km away from distance to roads.(2)The information gain ratio of the nine impact factors from large to small is elevation,distance to rivers,the slope angle,geology,distance to faults,slope aspect,distance to roads,profile curvature and plan curvature.So we removed the profile curvature and plan curvature.(3)The results of the landslide susceptibility assessment by the Min-max and Z-score model can be regarded as the same.(4)The values of the AUC for training data of landslide susceptibility assessment by normalized model and logistic regression model are 0.729 and 0.65,respectively.And the AUC for validated data of normalized model and logistic regression model are 0.706 and 0.625,respectively.It indicates that the accuracy of the normalized model is better than the logistic regression model.(5)The accuracy of the model used dataset a is 83%,which is highest than other datasets.And the Positive predictive value is 89.7%.It indicates that the 89.7% of debris flow-prone catchments are correctly classified to the debris flow-prone catchments class.And the Negative predictive value is 50%,which indicates that 50% of non-debris flow-prone catchments are correctly classified to the non-debris flow-prone catchments class.The result from this study demonstrates the benefit of identifying debris flow-prone catchments.