作者：

孙才 ;铁永波 ;宁志杰 ;徐伟 ;熊小辉

单位：

重庆交通大学河海学院;中国地质调查局成都地质调查中心(西南地质科技创新中心);自然资源部地质灾害风险防控工程技术创新中心;自然资源部成都地质灾害野外科学观测研究站;OB欧宝娱乐官方平台 (北京)中国地质科学院

关键词：

滑坡易发性;因子分级;频率比模型;支持向量机模型;耦合模型

英文关键词：

landslide susceptibility;factor grading;frequency ratio model;support vector machine model;coupling model

摘要：

针对滑坡易发性评价中因子分级基础数据与评价模型的选取问题，本文以滑坡灾害频发的四川省喜德县为研究区，采用斜坡单元为评价单元，通过对评价因子进行相关性分析，选取高程、坡度、曲率、NDVI、SPI、距水系距离、距道路距离、距断层距离、斜坡结构、工程地质岩组、土地利用类型11个评价因子，分别对区域点属性和滑坡点属性两类基础数据采用自然断点法进行因子分级，代入频率比模型和频率比–支持向量机耦合模型来评价滑坡易发性，并使用受试者工作特征（ROC）曲线与典型斜坡来验证模型精度。结果显示：以滑坡点属性作为分类基础数据并运用耦合模型得到的评价精度最高，对应的曲线下面积（S_AUC）值为0.752，能更好地预测滑坡易发性；模拟结果显示，研究区极高、高易发区面积占比分别为4.65%和23.73%，主要分布在地形起伏较大、断层发育、人类工程活动强烈的区域。相反，断层稀疏、人口分散的地区属于中、低易发区，其面积占比分别为44.20%和27.42%。结果将为喜德县及其类似地区滑坡易发性评价工作提供科学参考。

英文摘要：

This study addresses the critical issue of selecting factor classification base data and evaluation models for landslide susceptibility mapping,focusing on Xide County in Sichuan Province,a region frequently affected by landslide hazards.Utilizing slope units as evaluation units,a correlation analysis of the evaluation factors was conducted,ultimately selecting eleven key factors:elevation,slope angle,curvature,normalized difference vegetation index (NDVI),stream power index (SPI),distance to watercourses,distance to roads,distance to faults,slope structure,engineering geological rock groups,and land use types.Factors were classified using the natural breaks method for both regional point attributes and landslide point attributes.These classified factors were then incorporated into the frequency ratio model and the frequency ratio-support vector machine coupled model to evaluate landslide susceptibility.The precision of these models was validated using receiver operating characteristic (ROC) curves and typical slope analysis.The findings revealed that using landslide-specific attributes as the classification base data within the coupled model framework yielded the highest evaluation accuracy,with an area under the ROC curve (S_AUC) value of 0.752,indicating a superior predictive capability for landslide susceptibility.The simulation results indicated that areas of extremely high and high susceptibility constitute 4.65%and 23.73%of the study area,respectively,predominantly located in regions characterized by significant topographic relief,well-developed faults,and intense human engineering activities.Conversely,regions with sparse faults and low population density were categorized as medium and low susceptibility zones,accounting for 44.20%and 27.42%of the study area,respectively.These findings provide essential scientific insights and references for the effective assessment and management of landslide susceptibility in Xide County and other similar regions.

基金：

乌蒙山区重大地质灾害易灾机理和风险防控研究（202403AA080001）;中国地质调查局地质调查项目（DD20221746,DD20230086）

语种：

中文

DOI：

10.19826/j.cnki.1009-3850.2024.08002

来源：

沉积与特提斯地质,2024(03):547-559.

中图分类号：

P642.22

出版日期：

2024-09-15

提交日期：

2024-10-12

引用参考：

孙才;铁永波;宁志杰;徐伟;熊小辉. 基于频率比-支持向量机耦合模型的四川省喜德县滑坡易发性评价[J]. 沉积与特提斯地质,2024(03):547-559.