Landslides Susceptibility Evaluation In Loess Mountain

Landslide susceptibility evaluation is to predict the spatial distribution and occurrence probability of landslides,which can provide important scientific basis for landslide risk management and urban planning,and has certain guiding significance for regional landslide disaster prevention and mitigation.Based on the research of the Geological Hazards Survey of the Hequ-Hancheng Section of the Loess Plateau in Shanxi,Shaanxi and China by the Geological Survey of Xi'an,the Chinese Geological Survey Bureau participated in Linxian County-Qingjian,Shaanxi Province(Wubao map,Chengjiazhuang map,Xin Field map of Jiagou and Xiasanjiao Town)field geological disaster survey,combined with the basic geological profile of the area,analyze the development and distribution of landslide hazards,using weighted information model,support vector machine model(SVM),random forest model(RF)Analyze and evaluate the landslide sensitivity in the study area and obtain the following main results:(1)The study area is located in the middle of the meandering of the Yellow River in the Loess Plateau.The loess beams are well developed,the gullies are vertical and horizontal,the soil erosion is serious,and landslide hazards occur frequently.Geomorphic factors(elevation,slope,aspect,roughness,plane curvature,profile curvature),geological environmental factors(rainfall,stratigraphic lithology,tectonic distance,normalized vegetation index,normalized water body index,river distance)and Human engineering activity factors(road distance,land use type)are used as the main factors to evaluate the sensitivity of landslide disasters.(2)Using Arc GIS as a platform,establish a landslide disaster sensitivity evaluation database,combined with MATLAB data processing tools,respectively apply weighted information quantity model,support vector machine model,random forest model to analyze the landslide disaster sensitivity in the study area,the study believes that the random forest model The evaluation accuracy is higher than the weighted information quantity model and the support vector machine model.(3)The natural breakpoint method is used to divide the study area into four types: high-sensitivity area,medium-sensitivity area,low-sensitivity area,and insensitivity area.Based on the ROC curve and the actual analysis of landslide disaster development,it is considered that the landslides divided by the random forest model are sensitive The results of zoning of zoning are more scientific than those of landslide sensitivity divided by support vector machine model and weighted information model.