摘要
滑坡是常见的三大地质灾害之一,本文针对滑坡易发性进行评价,以小金县为研究区域,选取水系密度、与公路距离、与居民点距离、公路密度、居民点密度、地形粗糙度、植被指数、坡度及地质硬度等9个影响因子,基于GIS技术,采用SVM(支持向量机)、ANN(人工神经网络)及决策树3种评价模型进行滑坡数据分析。通过各模型ROC曲线结果对比,得到SVM、ANN及决策树模型曲线下AUC值分别为0.987、0.973及0.969,初步得出结论:在滑坡易发性评价方面3个模型均有较好表现,但SVM模型在样本数据分类处理和整体评价效果方面更具优势。最后通过ArcGIS平台对SVM模型的评价结果数据进行可视化,得到小金县的滑坡易发性分级图,表明研究区滑坡由中部线向左右两侧以树枝状分布。
Landslides are one of the three common geological hazards.In view of the landslide susceptibility assessment,9 factors such as water system density,distance from highway,distance from road,road density,residential point density,terrain roughness,vegetation index,slope and geological hardness are selected in the study area of Xiaojin County.So based on GIS technology,3 kinds of evaluation models,SVM(support vector machine),ANN(artificial neural network)and decision tree,are used to analyze landslide data.By comparing the results of the ROC curves of each model,the AUC values of the SVM,ANN and decision tree models are 0.987,0.973 and 0.969 respectively.It is concluded that the three models of landslide susceptibility are all good,but the SVM model has a better advantage in the classification of sample data and the overall evaluation effect.Finally,the ArcGIS platform is used to visualize the evaluation results of the SVM model,and the landslide susceptibility classification map of the Xiaojin County is obtained.It shows that the landslide is distributed from the central line to the left and right sides of the study area.
作者
谢明娟
蒲瑞
韩信
林莉
周春香
唐章英
XIE Mingjuan;PU Rui;HAN Xin;LIN Li;ZHOU Chunxiang;TANG Zhangying(School of Geoscience and Technology,Southwest Petroleum University,Chengdu 610500,China)
出处
《测绘与空间地理信息》
2019年第10期83-85,89,共4页
Geomatics & Spatial Information Technology
基金
西南石油大学开放实验基金项目(KSZ17034)资助