关键词: 泥石流;易发性评价;信息量模型;逻辑回归模型;信息量-逻辑回归耦合模型;ROC曲线;

Abstract: Debris flow is a common geological disaster in southwest of China,usually causing huge damage and casualties. The assessment of debris flow susceptibility can provide a theoretical basis for regional disaster prevention and mitigation and engineering site selection planning. Shimian County,Sichuan Province was used as the study area,and 288 historical debris flows in the county were obtained as the debris flow data set. Thirteen assessment factors including elevation,slope,aspect,topographic relief,water system density,lithology,NDVI,land use,fault distance,residential density,population density,road density and rainfall were selected. Debris flow susceptibility was assessed by adopting information value model,logistic regression model and information value-logistic regression coupled model. And the models were compared and evaluated by using receiver operating characteristic(ROC) curves. The results showed that the debris flow-prone areas in Shimian County were concentrated in the central and central-eastern parts of the county,mostly near rivers,roads and faults. The area of ROC curve(AUC) obtained by the information value model,logistic regression model,and information value-logistic regression coupled model were 0. 917,0. 928 and0. 931,respectively. It can be seen that the coupled model has the highest accuracy and the results is more reasonable and reliable. These results suggest that the information value-logistic regression coupled model is superior in the assessment of debris flow susceptibility,and the zoning results have practical implications for debris flow prevention in Shimian County.

Key words: debris flow; susceptibility assessment; information value model; logistic regression model; information value-logistic regression coupled model; ROC curve;