| Geological disasters occur frequently in the Three Gorges Reservoir(TGR)area,and once landslides in the slope of the fluctuating zone bank are unstable,it will cause huge loss of life and property to the people in the reservoir area.And a large number of earth or rocks will flow into the river,which may cause secondary disasters such as surge,blocking the channel even destroying the dam.The risk assessment of the bank slope is very important to the prevention and control of landslide disaster,and has a certain guiding role in the development and utilization or ecological restoration of the reservoir area.This paper takes fengjie area as the representative of the typical fluctuation belt,and makes statistics on the density of the riverbank landslide in fengjie section,then obtain the influence range of the bank slope.And cataloge historical landslides,select the elevation,slope,Profile curvature,NDVI,slope direction,slope position,micro landform,slope structure type,land use,distance from the road,distance from the river surface and Lithology,concave-convex slope type and branch trunk flow intersection as disaster causing factors to to construct the evaluation index system.Use machine learning models to quantify the probability of landslides,and assess regional risk in combination with the likely loss.The main research work and achievements are as follows:(1)The research shows that the influence range of water level fluctuation on the bank slope is within 400 m from the river surface through the statistical analysis of the geological disaster density of the reservoir bank,determining that water level 175 buffering range of 400 meters and falling zone as the study range.(2)Combined with the methods of data query,field survey and remote sensing interpretation,historical landslides in the study range are catalogued.The morphological characteristics,landslide scale,material composition and instability inducement of some landslides within the scope of the study were analyzed.The landslide in the study area can be divided into bedrock landslide and loose accumulation landslide.The main inducements are water level fluctuation,rainfall and engineering activities(3)According to the characteristics of landslide and the principles of logicality,feasibility and representativeness,14 disaster causing factors are selected.The continuous class factors are graded by natural breakpoint classification and equal interval classification and are classified by discrete class and derived discrete class factor,two evaluation systems were constructed after the reclassification.The results show that the final prediction accuracy of natural breakpoint classification system is higher than that of equal interval classification.(4)Use the random forest model and support vector machine model to map the landslide susceptibility area in the research scope.By comparing ROC,AUC,error rate,accuracy rate and recall rate and other evaluation indexes,get results are as follows: RF model is better than SVM model in the same evaluation system.(5)Based on the results of landslide susceptibility zoning with the highest accuracy,and combined with the residential area,road infrastructure and waterway in the study area,grid calculation is used for danger assessment.The high and higher dangerous areas are mainly distributed in Anping town on the south bank of the Yangtze river,from the Zhuyi estuary to the Meixi estuary on both sides of the Yangtze river,Meixi river and caotang river estuary and other areas. |
PDF Full Text Request