基于信息量-逻辑回归模型的泥石流易发性评价——以四川省石棉县为例  

Debris flow susceptibility assessment based on information value and logistic regression coupled model:case of Shimian County,Sichuan Province

作　　者：于淼 邢会歌[1] 胡士瑜 YU Miao;XING Huige;HU Shiyu(College of Architecture and Environment,Sichuan University,Chengdu 610065,China)

机构地区：[1]四川大学建筑与环境学院,四川成都610065

出　　处：《人民长江》2021年第12期107-114,共8页Yangtze River

基　　金：四川省青年科技创新研究团队项目(2020JDTD0006)。

摘　　要：泥石流是中国西南地区常见的地质灾害,通常会造成巨大破坏和人员伤亡,泥石流易发性评价可以为地区防灾减灾及工程选址规划提供理论依据。以四川省石棉县为研究区,获取该县288个历史泥石流作为泥石流数据集,选取高程、坡度、坡向、水系密度、地层岩性等13个评价因子,使用信息量模型、逻辑回归模型和信息量-逻辑回归耦合模型对石棉县进行泥石流易发性评价,并利用受试者工作特征曲线(ROC)对模型进行对比和评估。结果表明:四川省石棉县泥石流高易发区集中分布在该县纵向中部地区,多处于河流、道路及断层附近;信息量模型、逻辑回归模型和信息量-逻辑回归耦合模型的ROC曲线下面积(AUC)分别为0.917,0.928和0.931,可见耦合模型的精度最高,分区结果更加合理可靠。因此,信息量-逻辑回归耦合模型在泥石流易发性评价上具有优越性,其分区结果对泥石流防治具有实践意义。Debris flow is a common geological disaster in southwest of China,usually causing huge damage and casualties.The assessment of debris flow susceptibility can provide a theoretical basis for regional disaster prevention and mitigation and engineering site selection planning.Shimian County,Sichuan Province was used as the study area,and 288 historical debris flows in the county were obtained as the debris flow data set.Thirteen assessment factors including elevation,slope,aspect,topographic relief,water system density,lithology,NDVI,land use,fault distance,residential density,population density,road density and rainfall were selected.Debris flow susceptibility was assessed by adopting information value model,logistic regression model and information value-logistic regression coupled model.And the models were compared and evaluated by using receiver operating characteristic(ROC)curves.The results showed that the debris flow-prone areas in Shimian County were concentrated in the central and central-eastern parts of the county,mostly near rivers,roads and faults.The area of ROC curve(AUC)obtained by the information value model,logistic regression model,and information value-logistic regression coupled model were 0.917,0.928 and 0.931,respectively.It can be seen that the coupled model has the highest accuracy and the results is more reasonable and reliable.These results suggest that the information value-logistic regression coupled model is superior in the assessment of debris flow susceptibility,and the zoning results have practical implications for debris flow prevention in Shimian County.

关 键 词：泥石流 易发性评价 信息量模型 逻辑回归模型 信息量-逻辑回归耦合模型 ROC曲线 

分 类 号：P642.2[天文地球—工程地质学]