基于Logistic回归模型的芦山震后滑坡易发性评价

摘要

On April 20,2013,Ms 7.0 earthquake occurred in Lushan County,Sichuan,China.The quake center was located in Longmen Country (30.3°N,102.95°E) and the earthquake source at a depth of 13 km.A large amount of material (landslide debris) produced by the shock of the earthquake were moved by the heavy rainfall.Based on the remote sensing (RS) and geographic information system (GIS),this paper uses Logistic regression to make a landslide susceptibility map of the Lushan area in reference to slope,relief amplitude,land cover,distance to fault,and peak ground acceleration.The data are obtained from various sources.A landslide inventory of 421 landslides is constructed to produce the dependent variable,taking a value of 0 as the absence and 1 as the presence of slope failures.At last,the relationship between the presence or absence of landslides and a set of independent variables are showed using the coefficients.The coefficients imply that the distance to fault and relief amplitude play an important role in the occurrence of landslides.Using Jenks Natural Breaks,the landslide susceptibility index falls into four categories:low,medium,high and very high.The high and very high susceptibility covers an area of 31.23％ and 15.40％,respectively.The performance of the logistic regression model is validated by the receiver operating characteristics (ROC).The area under the ROC curve (AUC) is 0.772,showing that the logistic regression method deployed in the assessment of the landslide susceptibility is reliable and can be used in the reconstruction process.%地震导致山体结构失衡,物质松动,在降雨条件下,滑坡等次生地质灾害极易发生.以“4·20”芦山地震区为研究对象,基于遥感(RS)和地理信息系统技术(GIS),以坡度、起伏度、土地类型、断层的距离、地震动的峰值加速度为评价因子,采用Logistic回归方法构建评价模型评估了研究区滑坡易发性,并通过受试者工作特征曲线(ROC)检验模型的效果.通过对421个滑坡灾害点的回归分析得出断层的距离、地震动的峰值加速度对滑坡的发生贡献最大,研究区域46.63％的地区滑坡极易发生.ROC曲线的线下面积(AUC)为0.772,验证结果显示评价结果与实际情况吻合.