信息量与多模型耦合的滑坡易发性评价研究  

Landslides susceptibility evaluation by using coupling models of information value with multiple machine learning models

作　　者：仪政 宋琨[1] 黄海峰[1] 李辉 YI Zheng;SONG Kun;HUANG Haifeng;LI Hui(Hubei Key Laboratory of Disaster Prevention and Mitigation,China Three Gorges University,Yichang443002,China;Hebei Geological Environment Monitoring Institute,Shijiazhuang 050011,China)

机构地区：[1]三峡大学防灾减灾湖北省重点实验室,湖北宜昌443002 [2]河北省地质环境监测院,河北石家庄050011

出　　处：《人民长江》2021年第10期146-151,共6页Yangtze River

基　　金：国家自然科学基金项目(41702378)。

摘　　要：滑坡易发性评价是识别滑坡灾害危险性"靶区"的基础和关键。三峡库区秭归县的侏罗系地层区是滑坡灾害易发区,选取坡度、地表切割深度、曲率、距水系与路网距离等8个影响因子,通过信息量模型(I)与人工神经网络(ANN)、随机森林(RF)、XGBboost、支持向量机(SVM)等4种模型耦合,进行了滑坡易发性评价。通过受试者工作特征(ROC)曲线对比分析发现:信息量和支持向量机耦合模型的ROC曲线AUC值最大(0.848),滑坡易发性的分区图显示高易发区主要集中在水系和路网两侧,其结果与实际滑坡分布基本一致。研究成果可为区域滑坡易发性评价和滑坡灾害的防治提供参考。Landslide susceptibility evaluation is the basis and key to identify landslide hazard target area.The Jurassic stratum area in Zigui County,the Three Gorges Reservoir area,is a landslide hazard prone area.Eight influencing factors including slope,surface cutting depth,curvature,and distance from water system to road network were selected.The information model(I)was coupled with four common machine learning models,namely artificial neural network(ANN),random forest(RF),XGBboostand support vector machine(SVM)to carry out the evaluation of landslide susceptibility.Through a comparative analysis on the receiver operating characteristic curve,it was found that the AUC valuein the ROC curve of the model coupled with information valuemodel and SVMmodel was the largest(0.848).The zoning map of landslide susceptibility showed that the high-prone areas were concentrated on both sides of the water system and the road network,and the results were basically consistent with the actual landslide distribution.The research results can provide reference for regional landslide susceptibility evaluation and landslide disaster prevention.

关 键 词：滑坡易发性评价 滑坡易发性分区 信息量模型 多模型耦合 侏罗系地层 三峡库区 

分 类 号：P642[天文地球—工程地质学]