【摘要】 以陕西省洋县为研究区,通过搜集资料、实地调查获得研究区滑坡分布状况。结合研究区地质环境特征与前人研究经验,初步选取高程、坡度、坡向、地形起伏度、曲率、距水系距离、距道路距离、降雨量及岩土体类型,共9种滑坡影响因子展开滑坡易发性研究。首先,采用皮尔森相关系数法对各因子间的相关性进行分析。其次,按照70∶30的比例将滑坡数据随机划分为模型训练集与模型验证集。然后,采用模型训练集对频率比模型(FR)、随机森林模型(RF)及两者的耦合模型(FR-RF)进行训练,利用模型验证集对模型训练结果进行检验,并绘制ROC曲线。最后,利用验证后的模型绘制研究区滑坡易发性分区图。结果表明:(1)所选取的9个滑坡影响因子是相互独立的;(2)采用的三个模型均表现良好,其中FR-RF模型预测准确度最高(0.901),其次为RF模型(0.863),最后为FR模型(0.833);(3)绘制的滑坡易发性分区图可为当地政府制定土地利用规划、预防滑坡等方案提供参考借鉴。更多还原

【Abstract】 To understand the distribution of landslides in Yangxian County, Shaanxi Province, field investigation was conducted, and concerning data were collected. According to the local geological setting and available studies, nine factors affecting landslide were selected, including elevation, slope, aspect, topographic relief, curvature, distance to rivers, distance to roads, rainfall, and geotechnical engineering, and the landslide susceptibility was studied. First, the Pearson correlation coefficient method was introduced to analyze the correlation among these factors. Next, the landslide points were randomly divided into two groups(70∶30) for training dataset and validation dataset, respectively. In addition, the frequency ratio(FR) model, random forest(RF) model, and their ensemble model(FR-RF) were trained by the training dataset, individually. The training results and the receiver operator curve were verified by the validation dataset. Finally, using the verified models, landslide susceptibility maps of study area were generated. Results show that the nine landslide factors are independent from each other. All the three models performed well, and the order of accuracy is FR-RF(0.901), RF model(0.863), and FR(0.833). The landslide susceptibility maps generated in this study can provide references for local government in land use planning and landslide prevention.更多还原