极端降雨背景下滑坡易发性评估模型集成及关键因素分析  

Integration of Landslide Susceptibility Assessment Models and Analysis of Key Factors Under Extreme Rainfall Conditions

作　　者：孔雅茜 肖正辉[2] 黄炜敏 罗伟奇 肖巍峰 曹运江[2] 罗路广 唐豪 Kong Yaxi;Xiao Zhenghui;Huang Weimin;Luo Weiqi;Xiao Weifeng;Cao Yunjiang;Luo Luguang;Tang Hao(Hunan Center of Natural Resources Affairs,Changsha Hunan 410004;Hunan University of Science and Technology,Xiangtan Hunan 411201)

机构地区：[1]湖南省自然资源事务中心,湖南长沙410004 [2]湖南科技大学,湖南湘潭411201

出　　处：《国土资源导刊》2025年第1期89-97,共9页Land & Resources Herald

基　　金：湖南省自然科学基金项目(编号:2023JJ30238);湖南省地质灾害监测预警与应急救援工程技术研究中心开放课题(编号:hndzgczx202409);湖南省教育厅研究生科研创新项目(编号:CX20240904)。

摘　　要：极端降雨事件的频发显著增加了滑坡灾害的风险,准确评估滑坡易发性对于有效管理和减轻这一风险至关重要。以湖南省平江县为例,构建了基于多模型集成的滑坡易发性评估方法,采用逻辑回归、随机森林和极端梯度提升三种机器学习算法,通过加权投票分类器进行模型集成。结果表明:集成模型在滑坡易发性评估中表现最佳,准确率达0.91,AUC值为0.86;地形湿度指数、坡向和最大一小时降雨量被识别为影响滑坡发生的三个最重要因素;滑坡易发性空间分布呈现明显差异,北部和东北部地区易发性较高,92.2%的已知滑坡点位于高、中高易发区,验证了模型的可靠性。揭示了平江县滑坡的成灾机理与关键诱发因素,为该地区滑坡风险预警体系的构建及差异化防灾策略的制定提供了地质科学依据。The frequent occurrence of extreme rainfall events has significantly increased the risk of landslide disasters.Accurate assessment of landslide susceptibility is crucial for effective management and mitigation of this risk.Taking Pingjiang County,Hunan Province as an example,this article constructs a landslide susceptibility assessment method based on multi-model integration.The study adopts three machine learning algorithms:logistic regression,random forest,and extreme gradient boosting,and integrates the models through a weighted voting classifier.The results show that the integrated model performs best in landslide susceptibility assessment,with an accuracy rate of 0.91 and an AUC value of 0.86.Topographic wetness index,slope orientation,and maximum one-hour rainfall are identified as the three most important factors affecting landslide occurrence.The spatial distribution of landslide susceptibility exhibits significant differences,with higher susceptibility in the northern and northeastern regions.92.2%of known landslide points are located in high and medium-high susceptibility areas,verifying the reliability of the model.The study reveals the disaster mechanism and key inducing factors of landslides in Pingjiang County,providing geological scientific basis for the construction of a landslide risk early warning system and the formulation of differentiated disaster prevention strategies in the region.

关 键 词：滑坡易发性 极端降雨 多模型集成 特征重要性 平江县 

分 类 号：P642.22[天文地球—工程地质学]