Abstract:

Susceptibility assessment has a great significance to predict and forcast landslide hazards in the mid and long term.The aim is to analyze the landslide susceptibility mapping combining rough sets (RS) theory and support vector machine (SVM) in the Three Gorges Reservoir area of Zigui to Badong County. Rough set theory is used to reduce the redundant information of 20 initial factors in the decision table and determine the kernel included 13 representative factors. Then, the kernel factors are used to train a SVM model, and landslide susceptibility maps were produced. The higher susceptibility zones is about 8.2% of the total study area, and primarily distributed in the right bank of Tongzhuang River, along the Guizhou River, the left bank of Qinggan River, the right bank of the Yangtze River of Shuping to Fanjiaping, Niukou to Dongrangkou and near the Badong. The stability zones are accounted for about 52.7% which mainly distribute in the Dianziwan to Badong, and the areas away from Yangtze River and the areas of high surface cover degree. Through verification and analysis of the results, it shows that the predictive power of the RS-SVM model is superior to the SVM model, and the prediction accuracy of the RS-SVM model is aoubt 85.6%. The method have advantages of excellent predict performacy and efficiency. Combining with the landslide survey data, the evaluation results are basically consistent with the status of the local landslide disasters and it is proved that the proposed method is an effective tool for landslide susceptibility assessment.

Key words: landslide, susceptibility assessment, rough set, support vector machine, Three Gorges Reservoir