关键词: 滑坡易发性, ArcGIS, 粗糙集, BP神经网络, 重庆市

Abstract: The susceptibility zoning assessment of landslide is studied based on comprehensive analysis of the detailed survey data of geological disasters in Longjuba area, Wanzhou, Chongqing. Firstly, we extract and analyze the influential factors of landslide in the study area, and obtain six indicators that falls into three categories by the rough set analysis. Then the indicators' attribute states are classified according to the distribution condition of landslide, and the corresponding values are calculated. Taking the BP neural network analysis model as the method, the landslide susceptibility is assessed through the sample data selection, neural network training and susceptibility calculation. The susceptibility results are divided into four levels by the classification function in ArcGIS software: low susceptibility zone (0-0.3), medium susceptibility zone (0.3-0.5), high susceptibility zone (0.5-0.7) and extremely high susceptibility zone (> 0.7). It is revealed that the very high susceptibility and high susceptibility zones are mainly located at the core of geological structure, which is consistent with the actual investigation. Therefore, the study is of practical value for the disaster prevention in the study area.

Key words: landslide susceptibility, ArcGIS, Rough set, BP neural network, Chongqing City