【摘要】 滑坡敏感性评价是滑坡防治中的重要一环.针对信息量模型(Ⅳ)中各影响因子等权以及逻辑回归模型(LR)中因子难以量化的问题,将逻辑回归模型(LR)与信息量模型(Ⅳ)进行耦合,构建LR-Ⅳ模型,并将逻辑回归模型及信息量模型作为对照,以湖北省为例,进行了滑坡敏感性评价.选取工程岩组、地震烈度、水系距离、构造线距离、地貌、降雨面、坡度等七个因子,结合GIS的栅格分析方法,计算各区域滑坡敏感性指数,得到湖北省滑坡敏感性区划图.将历史滑坡数据随机分为70%的训练数据与30%的验证数据,通过受试者工作特征曲线来比较各个模型之间的性能.在成功率上,LR-Ⅳ模型(91.3%)与逻辑回归模型(91.5%)、信息量模型(90.3%)相差不大,都具有极高的成功率,但在预测率上,LR-Ⅳ模型(75.7%)远胜于逻辑回归模型(65.8%),略强于信息量模型(74.9%).研究结果表明,LR-Ⅳ模型表现优异,可用于滑坡敏感性评价.更多还原

【Abstract】 Landslide susceptibility assessment is an important part of landslide prevention and control. In order to solve the problem that the factors in the Information Value Model(Ⅳ)have equal weights,and the factors in the logistic regression model(LR)are difficult to quantify,this study constructs LR-Ⅳ model by coupling logistic regression model with information value model. The logistic regression model and information value model were taken as a control. With Hubei Province as an example,landslide susceptibility assessment was conducted. In this study,we selected seven factors of engineering rock group,seismic intensity,distance of water system,distance of structural line,topography,rainfall surface,slope and so on. Combined with the grid analysis method of GIS,the landslide susceptibility index of each area is calculated,and the zoning map of landslide susceptibility in Hubei Province was finally obtained. During the model comparison,historical landslide data were randomly divided into 70% of training data and 30% of validation data,and the performance of each model was compared by the receiver operating characteristic curve. The LR-Ⅳ model(91.3%)has a very high success rate compared with the logistic regression model(91.5%)and the information value model(90.3%),but the LR-Ⅳ model(75.7%)is far superior to the logistic regression model(65.8%)and slightly stronger than the information value model(74.9%). Overall,the LR-Ⅳ model performswell and can be used for the landslide susceptibility assessment.更多还原