基于信息量和逻辑回归耦合模型的黄土地震滑坡危险性分析

Risk analysis of loess seismic landslides based on an information value-logistic regression coupling model

黄土地震滑坡危险性分析对黄土地区城镇化、工程建设的规划和地震灾害预防具有重要意义。以甘肃省定西市岷县—漳县交界处为研究区域,通过统计分析该区历史地震滑坡灾害数据,归纳并建立包含地震、坡度、坡高、坡向、地层岩性、年平均降雨量、河流流域和地貌类型等8个影响因子的评价指标体系,采用信息量模型、逻辑回归模型和信息量-逻辑回归耦合模型分别分析该区域黄土地震滑坡危险性。结果表明:(1)地震、河流和降雨是诱发黄土滑坡灾害发生的主要因素,其中地震因子贡献率最大;(2)研究区可划分为高、较高、中、低和极低危险区五个等级,其中高危险区主要集中于岷县、漳县与陇西县等地;(3)根据受试者工作特性(ROC)曲线精度检验结果,三种模型的AUC值分别为0.889、0.617和0.898,信息量-逻辑回归耦合模型结果的精确性相比其他两个模型更高。

Risk analysis of loess seismic landslides is critical to urbanization,engineering construction planning,and earthquake disaster prevention in loess areas.In this study,data on historical earthquake landslide hazards in the Minxian-Zhangxian junction of Dingxi City,Gansu Province,were collected and analyzed.Then,an evaluation index system containing eight impact factors,that is,earthquake,slope angle,slope height,slope direction,stratigraphic lithology,average annual rainfall,river basin,and geomorphological type,was summarized and established.Based on the proposed index system,the information value model,logistic regression model,and information value-logistic regression coupling model were used to assess the risk of loess earthquake landslides in the study region.Results show that earthquakes,rivers,and rainfall are the main impact factors of loess landslide disasters,with earthquakes having the most contribution.In addition,the study area can be divided into five risk levels:very high-,high-,medium-,low-,and very-low-risk areas;the very-high-risk areas are primarily concentrated in Minxian,Zhang-xian,and Longxi Counties.Based on the accuracy test results of the receiver operating characteristic curves,the AUC values of the three models are found to be 0.889,0.617,and 0.898,respectively,and the results of the information value-logistic regression coupling model are more accurate than those of the other two models.