人工神经网络在滑坡敏感性评价中的应用  

Application of artificial neural network in landslide susceptibility assessment

作　　者：谭龙[1] 陈冠[1] 曾润强[1] 熊木齐[1] 孟兴民[1] 

机构地区：[1]兰州大学西部环境教育部重点实验室,兰州730000

出　　处：《兰州大学学报（自然科学版）》2014年第1期15-20,共6页Journal of Lanzhou University(Natural Sciences)

基　　金：国家科技支撑计划项目(2011BAK12B06);甘肃省科技重大专项计划项目(1102FKDA007)

摘　　要：以边坡为基本研究单元,经过主成分分析和独立性检验得到白龙江流域对滑坡形成贡献最大的6个因子:人口密度、坡度、坡向、断裂距离、岩性和高程.使用人工神经网络对白龙江流域进行滑坡敏感性评价,采用ROC曲线对模型精度进行验证.研究结果表明,人工神经网络能有效地对该区域进行滑坡敏感性评价,且能将研究区划分成5个区:极低危险区、低危险区、中等危险区、高危险区、极高危险区,各区面积占研究区面积的比例分别为9.53%,41.46%,12.12%,25.33%,11.58%.Slope units were used as the basic assessment units. Six conditional independent environmental factors were selected as the explanatory variables that contribute to landslide occurrence, i.e., elevation, slope, aspect, distance from fault, lithology and settlement density. Then, the methods of artificial neural network （ANN） were conducted for landslide hazard mapping. ROC curves were plotted as a means of evaluating the quality of the susceptibility zonations for the ANN model. The results show that ANN can effectively evaluate the hazards of landslides in the region. According to the result of the model, the study area could be classified into five categories, i.e., very high dangerous zone, high dangerous zone, moderate dangerous zone, low dangerous zone and very low dangerous zone, taking an area proportion of 9.53%, 41.46%, 12.12%, 25.33%, 11.58%, respectively.

关 键 词：人工神经网络 滑坡 敏感性评价 白龙江流域 

分 类 号：P642.22[天文地球—工程地质学]