ASSESSMENT OF REGIONAL LANDSLIDE SUSCEPTIBILITY BASED ON COMBINED MODEL OF CERTAINTY FACTOR METHOD

The research area of this paper is located from Diexi Township to Shidaguan Township, Maoxian County, Sichuan Province. Ten impact factors are selected according to the field data and the basic conditions of the study area. They are the slope, slope direction, section curvature, elevation, undulation, stratum, distance from the river, distance from the fault, land type and vegetation coverage. Using GIS technology as the operating platform, three methods are used to evaluate the sensitivity of landslide hazard in the study area. They are the deterministic coefficient+analytic hierarchy process(CF-AHP), the deterministic coefficient+logistic regression method(CF-LR) and the deterministic coefficient+multi-layer perceptron method of neural network(CF-MLP), The landslide hazard in this area is divided into four categories:extremely low, low, medium and high sensitive areas. The effect of the model is tested by receiver operating characteristic curve(ROC). The under-line area(AUC) of the ROC curves of the CF-AHP, CF-LR and CF-MLP combined models are 0.850, 0.884 and 0.867, respectively. The CF-LR combination model works best. In the CF-LR combination model, the areas of high, medium, low and extremely low sensitive area account for 11.3%, 25.1%, 22.5% and 41.1% of the total area of the study area, respectively. The results show that the high sensitive area is mainly located around the main water system and fault-developed area. The calculated sensitivity zoning results are close to the actual situation in field. Thus, the results are valuable for risk assessment of geological disasters.