摘要
本文以三峡库区巫山县为研究区,利用收集的资料,提取出9类指标因子(高程、坡度、坡向、地形湿度指数TWI、地表粗糙度指数TRI、地层岩性、水系距离、构造距离、植被覆盖指数NDVI),利用相关性分析剔除高程因子。将灾害点和指标因子数据带入支持向量机(SVM)和人工神经网络(ANN)模型,得到研究区滑坡易发性区划图。根据ROC曲线对模型的精确度进行评价,得到SVM模型的成功率和预测率曲线的AUC值分别为0.919和0.862,ANN模型分别为0.86和0.837,表明两个模型均适用于研究区滑坡易发性评价。根据以上工作,本文提出了基于Max{LSI(SVM);LSI(ANN)}函数的SVM-ANN模型,并将其应用到该区的滑坡易发性评价中。SVM、ANN和SVM-ANN模型中,历史滑坡灾害点分布在高-极高易发区的比例分别为90.06%、83.18%和94.01%,表明SVM-ANN模型更适用于滑坡灾害风险分析的实际应用。
In this paper,nine controlling factors were extracted from the landslide database in Wushan,Three Gorges reservoir area(i.e.elevation,aspect,slope,topographic wetness index,terrain ruggedness index,lithology,distance from rivers,distance from structure,NDVI).Support vector machines(SVM)and artificial neural network(ANN)models were used to obtain the landslide susceptibility map in the study area.The performance of the methods was evaluated by the receiver operating characteristic(ROC)curve.The area under success rate and prediction rate curve(AUC)values of the SVM model are 0.919 and 0.862,the ANN model are 0.86 and 0.837,respectively.It is mean that both models are suitable for the evaluation of landslide susceptibility in the study area.According to the above work,this paper proposes SVM-ANN model based on Max{LSI(SVM);LSI(ANN)}function and applies it to the landslide susceptibility evaluation.Among SVM,ANN and SVM-ANN models,the proportion of historical landslide hazard points in high to very high prone areas is 90.06%,83.18%and 94.01%respectively,which indicates that SVM-ANN model is more suitable for landslide hazard analysis and management application.
作者
夏辉
殷坤龙
梁鑫
马飞
XIA Hui;YIN Kunlong;LIANG Xin;MA Fei(China University of Geosciences,Wuhan,Hubei 430074,China;Geological Disaster Prevention Center of Chongqing,Chongqing 400015,China)
出处
《中国地质灾害与防治学报》
CSCD
2018年第5期13-19,共7页
The Chinese Journal of Geological Hazard and Control
基金
国家自然科学基金:三峡库区缓倾角地层滑坡机理与预报判据研案(41572292)
作者简介
夏辉(1992-),男,地质工程专业,硕士,主要从事地质灾害风险评价研究。E-mail:xiahui@cug.edu.cn。
