摘要
滑坡易发性评价是提高区域滑坡灾害事前防治效率的重要手段。为了提高滑坡易发性评价精度,以江西省寻乌县为研究对象,以县域内716处历史滑坡点为研究基础,选取海拔高度、地形坡度、坡向、地形起伏度、工程地质岩组等10个评价因子构建滑坡易发性评价指标体系。从滑坡灾害点总样本中随机选择80%作为训练样本进行建模分析,并随机选取等量未发生滑坡的样本,共同组成因变量样本集,分别采用信息量、确定性系数、逻辑回归信息量、逻辑回归确定性系数模型对滑坡易发性进行分区。利用剩余20%滑坡样本,分别采用频率比法、Sridevi Jadi经验概率法和ROC曲线法检验评价结果的合理性,并讨论不同模型的适用性和准确性。结果表明,四种模型中,随着滑坡易发性等级的增高,其频率比值显著增大,其中低、中易发区频率比值均<1,极高易发区频率比值均>3,均可以有效评价寻乌县滑坡易发性;精度检验分析结果中逻辑回归信息量模型的AUC值最大,为0.854,其适用性和精度要明显优于其他三种模型;滑坡灾害极高易发区主要分布在各乡镇河流的沿岸、靠近主干道相对集中的区域。其评价结果可以为县域地质灾害防治工作提供数据支撑。
Landslide susceptibility evaluation is an important means to improve the efficiency of regional landslide disaster prevention and control in advance.In order to improve the evaluation accuracy of landslide susceptibility,this paper takes Xunwu County of Jiangxi Province as the research object and selects 10 factors such as elevation,terrain slope,aspect,relief intersity and engineering geology rock group to construct the evaluation index system based on 716 historical landslide points in the county.80%of the total samples of landslide disaster points were randomly selected as training samples for modeling analysis,and the same numbers of sample points without landslide were randomly selected to form the dependent variable sample set.Information volume model,deterministic coefficient model,logistic regression information volume model and logistic regression deterministic coefficient model were used to classify the landslide susceptibility.Frequency ratio method,Sridevi Jadi empirical probability method and ROC curve method were used to test the rationality of the evaluation results of the remaining 20%landslide sample points,and the applicability and accuracy of different models were discussed.The results show that the frequency ratio increases significantly with the increase of landslide susceptibility grade in four models,the frequency ratio of low and medium prone zone is less than,and the frequency ratio of extremely high prone area is greater than 3.All the models can effectively evaluate the landslide susceptibility in Xunwu County.Among the accuracy test analysis results,logistic regression information volume model(LR-I)has the largest AUC value of 0.854,and its applicability and accuracy are significantly better than the other three models.The highly prone areas of landslide disasters are mainly distributed along the banks of the main rivers and near the main roads.The evaluation results can provide data support for the prevention and control of geological disasters in counties.
作者
王成楠
吴琳伟
宋勇
贺婷
王彪
赵影
胡旭东
Wang Chengnan;Wu Linwei;Song Yong;He Ting;Wang Biao;Zhao Ying;Hu Xudong(Geological Non-ferrous Geological Brigade of Jiangxi Geological Bureau,Ganzhou,Jiangxi 341000;Ganzhou Rare Earth Mine Ecological Restoration Technology Innovation Center,Ganzhou,Jiangxi 341000;Nuclear Geological Brigade of Jiangxi Bureau of Geology,Yintang,Jiangxi 335000;Jiangxi Jiujiang Yangtze Block East Geodynamic Field Scientific Observation and Research,Jiujiang,Jiangxi 332006;Jiangxi Earthquake Agency,Nanchang,Jiangxi 330026;East China University of Technology,Nanchang,Jiangxi 330013)
出处
《资源环境与工程》
2023年第2期171-182,共12页
Resources Environment & Engineering
基金
2021中央自然灾害防治体系建设补助资金(第一批)项目(202102043)
江西省地质局公益项目2023年度地质灾害防治技术支撑项目(赣州市)。
关键词
滑坡易发性
信息量模型
确定性系数模型
逻辑回归模型
精度检验
寻乌县
landslide susceptibility
information model
deterministic coefficient model
logistic regression model
accuracy test
Xunwu County
