基于优化最大熵模型的黄土滑坡易发性评价:以陕西省吴起县为例  

Evaluation of Loess Landslide Susceptibility Based on Optimised MaxEnt Model:A Case Study of Wuqi County in Shaanxi Province

作　　者：张天宇 李林翠 刘凡 洪增林 钱法桥 胡斌 张淼 ZHANG Tianyu;LI Lincui;LIU Fan;HONG Zenglin;QIAN Faqiao;HU Bin;ZHANG Miao(Shaanxi Geological Environment Monitoring Station,Xi’an 710054,Shaanxi,China;College of Geology Engineering and Geomatics,Chang’an University,Xi’an 710054,Shaanxi,China;Key Laboratory of Mine Geological Hazards Mechanism and Control,Xi’an 710054,Shaanxi,China;61363 Troops,Xi’an 710054,Shaanxi,China)

机构地区：[1]陕西省地质环境监测总站,陕西西安710068 [2]长安大学地质工程与测绘学院,陕西西安710054 [3]自然资源部矿山地质灾害成灾机理与防控重点实验室,陕西西安710054 [4]61363部队,陕西西安710054

出　　处：《西北地质》2025年第2期172-185,共14页Northwestern Geology

基　　金：国家自然科学基金项目“基于滑带土水文动态响应的黄土滑坡地貌演化预测模型研究”(42201011);陕西省公益性地质调查项目“黄河支流洛河流域地貌演化及地质灾害隐患识别研究项目”(202101);国家重点研发计划资助“极端天气黄土体灾变风险防控技术装备研发”(2022YFC3003400)联合资助。

摘　　要：黄土高原地区滑坡灾害频发,严重危害人民生命财产安全和重大工程建设,进行精准的滑坡易发性评价,识别“什么地方易发生”,有助于高效预测滑坡灾害风险,为防灾减灾提供有效的科学依据。笔者以黄土高原腹地吴起县为例,采用优化最大熵模型(MaxEnt),利用505个滑坡点,选取高程、坡向、坡度、地形粗糙度、岩性、河流缓冲区、降雨、NDWI(地表湿度)及道路缓冲区作为评价因子,并引入InSAR地表形变数据作为动态评价因子,开展了滑坡易发性评价。基于Enmeval数据包调整优化的MaxEnt模型,分别随机选取90%和10%的滑坡点进行模型训练及验证,模型精度高(AUC值为0.855),模拟效果准确可信。引入InSAR地表形变速率作为动态评价因子,模型精度、评价结果均有所提升。评价结果显示:研究区较高易发区面积和高易发区面积分别占吴起县总面积10.27%和6.33%,高、较高易发区内的滑坡点占全部滑坡点的73.27%,滑坡易发性评价结果与滑坡点分布现状吻合,评价效果好。高程、坡度和地表粗糙度对模型模拟结果贡献较高,是研究区滑坡易发性重要评价因子。Landslide disasters which occur frequently in the Loess Plateau,seriously endanger the safety of people's lives and property,and affect the construction of major projects.Accurate landslide susceptibility assessment is useful for efficiently and quickly landslide risk prediction,and can provide scientific backing for disaster prevention and reduction by identifying"where landslides are prone".Taking Wuqi County on the Loess Plateau as an example,we use the optimized MaxEnt model and 505 landslide points to evaluate the landslide susceptibility.Elevation,aspect,slope,terrain roughness,lithology,river buffer,rainfall,NDWI(surface humidity),road buffer,and InSAR surface deformation data,which was introduced as dynamic evaluation factors,were selected as influencing factors.The results show:In the MaxEnt model based on Enmeval packet adjustment,when 90%landslide points were randomly selected as the training set and 10%landslide points as the verification set,the model accuracy was the highest(AUC value was 0.855),and the simulation effect was accurate and reliable.InSAR surface deformation rate was introduced as a dynamic evaluation factor,and the model accuracy and evaluation results were both improved.In the study area,the area of high and relatively high susceptibility areas accounted for 10.27%and 6.33%of the total area respectively,and the landslide points in the high and relatively high prone areas accounted for 73.27%of the total landslide points,of which the high prone areas accounted for 48.11%.The evaluation results of landslide susceptibility were consistent with the distribution of landslide points,which proves that the evaluation works well.Elevation,slope and surface roughness contribute significantly to the simulation results,and are important factors affecting the landslide susceptibility.

关 键 词：黄土地貌区 优化MaxEnt模型 黄土滑坡 易发性评价 吴起县 

分 类 号：P694[天文地球—地质学]