基于确定性系数、逻辑回归及其耦合模型的滑坡危险性评价——以云南溪洛渡水库为例  

Landslide Risk Assessment Based on Certainty Factor,Logistic Regression and Coupled Model:A Case Study for Xiluodu Reservoir

作　　者：凡净 钱洪伟[2] 邓昌宏 严佩升 马殿妍 李飞 FAN Jing;QIAN Hongwei;DENG Changhong;YAN Peisheng;MA Dianyan;LI Fei(College of Geography and Tourism,Zhaotong University,Zhaotong Yunnan 657000,China;School of Emergency Management,Henan Polytechnic University,Jiaozuo Henan 454150,China;Zhaotong Fire and Rescue Detachment,Zhaotong Yunnan 657000,China)

机构地区：[1]昭通学院地理科学与旅游学院,云南昭通657000 [2]河南理工大学应急管理学院,河南焦作454150 [3]昭通市消防救援支队,云南昭通657000

出　　处：《安全》2025年第2期23-29,共7页Safety & Security

基　　金：云南省教育厅科学研究基金项目一般项目(2024J1071);2024年度云南省本科高校基础研究联合专项—青年项目(202401BA070001-008)。

摘　　要：为精确评价滑坡危险性,本文采用地理信息系统(GIS)、确定性系数(CF)、逻辑回归(LR)及其耦合模型,结合坡度、坡向、高程等10个关键指标,对云南溪洛渡水库及库周进行滑坡危险性分区,并将其划分为极高、高、中、低和极低危险5个等级。研究发现:土壤岩性、坡度及与道路和河流的距离是滑坡风险的主要影响因素。水库库周特别是碎屑岩、页岩软质岩区、坡度在30~40°、距离断层0~800 m、河流601~900 m范围内,滑坡风险尤为显著;极高和高危险区主要分布在金沙江两岸大坝核心区、右岸25~35 km范围以及库区末端左岸20 km的人口密集区;而低和极低危险区主要位于水库中心远离主河道的区域。CF-LR耦合模型在极高和高危险区识别及分区合理性和精度上显著优于单一的CF和LR模型。To accurately assess landslide hazard,this study employs Geographic Information Systems(GIS),the Certainty Factor(CF),Logistic Regression(LR),and its coupling model,combined with 10 key indicators such as slope,aspect,elevation,etc.,to zonate the landslide hazard of Xiluodu Reservoir in Yunnan and its surrounding areas,and classify them into five levels:extremely high,high,medium,low and extremely low.The findings reveal that soil lithology,slope,and distance from roads to rivers are the main influencing factors of landslide risk.The reservoir periphery,particularly areas with clastic rocks and shale soft rock formations,slopes of 30-40°,within 0-800 m of faults,and 601-900 m from rivers,has a particularly high landslide susceptibility.Extremely high and high risk areas are mainly distributed in the core areas of the dams on both sides of the Jinsha River,the 25-35 km range on the right bank,and the densely populated area 20 km on the left bank at the end of the reservoir.Conversely,low and extremely low susceptibility areas are mainly located in the central reservoir area,far from the main river channel.The CF-LR coupling model is significantly superior to single CF and LR models in the identification and zoning rationality and accuracy of extremely high and high risk areas.

关 键 词：滑坡 危险性 确定性系数模型 逻辑回归模型 CF-LR耦合模型 

分 类 号：X913[环境科学与工程—安全科学]