Susceptibility Appraisal Of Landslide Adjacent To The Dam Site Of Yamangshu Hydro-power Station Based On3S Technology

With the rapid economic growth and the increase of human engineering,geological disasters have caused more and more damages to people’s lives andproperties, which restrict the human’s sustainable development. Landslide riskevaluation, which is difficult but important for landslide research, is the most effectivemethod to mitigate the loss led by landslide disasters. The susceptibility analysis,which is essential for prevention and early warning, is the foundation of the riskevaluation and management for landslides.In spite of achieving fruitful results in thefield of geological disasters for our country during the past several decades, on thewhole, in comparison to the international advanced level, there are large gapsespecially in the selection of the influencing factors and the theory application of thelandslide disasters.The Southwest China is located in the transition zone between the first step andthe second step of our country’s landform. In this region, the active neotectonics,water syster were developed, the geological conditions are complicated, the slopes arelarge, and the rainfall is abundant. Therefore, large amounts of landslides weredeveloped in this area, which offer unexceptionable examples for feature analysis andsusceptibility assessment. Based on the abovementioned details and the3Stechnologies, the landslides in Yansangshu reserior region in Nujiang basin wereinterpreted and investigated; then, the characteristic parameters were obtained.Ultimately, the data-base of the landslide disasters can be generated. Based on methodologies of deterministic coefficient, variable dimension fractal, analytichierarchy process, entropy method, and logistic regression, the landslides in theresearch area were qualitatively and quantitatively researched and the susceptibilityresults were regionalized. A method of work was formed as “data collection-remotesensing image processing-landslide information extraction-landslide susceptibilityanalysis”. The main research contents and the productions were listed as follows：1. Based on remote sensing images in high resolution and konkret field survey,the spatial data-base of the concentrated landslide disasters can be generated by usingARCGIS and the geological metainformation data standard. In addition, thedevelopment types, scales, grain composition, and genetic mechanism were analyzedbased on the data-base information, regional geology, hydrogeololgy, geomorphology,and landform of the landslides.2. The selection of mapping unit was discussed and the mapping unit comesdown to five kinds: the grid unit, regional unit, uniform conditions unit, slope unit,and terrain unit. According to engineering needs, the evaluation of the geologicalhazard in this paper chose slope unit and grid unit, and the features of extractedfactors for the mapping units were compared.3. Through summing up the former researcher’s experiences, the influencingfactor system of landslide susceptibility analysis was generated, and the size effect ofthe topographic relief was researched. Based on the aforementioned researches, theinfluencing factor extraction was investigated based on the density of landslides anddeterministic coefficient, and the most applicable condition for the landslide developwas summed up. The theory of the variable dimension fractal was used for evaluatingthe relationship between the inducing factors and the occurrence of the landslide toprovide the fractal basis to establish the index system of landslide susceptibilityassessment.4. According to the qualitative and quantitative evaluation actuality, this paperapplied the combination weighting and logistic regression method combining theregular mapping uint and natural slope uint to assess the susceptibility of the landslidedisasters. The significance and methods of comparison for landslide susceptibility were discussed. Accuracy statistics of model were divided into cutoff-dependentperformance criteria and cutoff-independent performance criteria according towhether the statistics are constrained by the critical value. By using the empiricalprobability and ROC curve, the generated susceptibility models were analyzed. Theresults indicated that the calculation by using the logistic regression method based onregular grid cell is the most applicable for the research area.