| 浙西梅雨滑坡易发性评价模型对比 |
| |
| 引用本文： | 冯杭建,周爱国,俞剑君,唐小明,郑嘉丽,陈秀秀,游省易.浙西梅雨滑坡易发性评价模型对比[J].地球科学,2016,41(3):403-415. |
| |
| 作者姓名： | 冯杭建 周爱国 俞剑君 唐小明 郑嘉丽 陈秀秀 游省易 |
| |
| 作者单位： | 1.中国地质大学地质调查研究院，湖北武汉 430074 |
| |
| 基金项目： | 国土资源部公益性行业科研专项“突发地质灾害应急响应支撑关键技术研究”201211055浙江省公益技术应用研究项目“降雨型滑坡泥石流地质灾害区域风险评估关键技术”2016C33045 |
| |
| 摘 要： | 我国目前滑坡易发性评价研究主要集中在西南地区，对东南部降雨引发特别是梅雨引发的滑坡研究较少.选取浙江省西北部梅雨控制区淳安县为研究区，通过遥感解译结合野外详细调查，共确定滑坡596处，并建立滑坡编录数据库.选取高程、坡向、坡度、曲率、工程岩组、断层、道路、建设用地、植被等9个滑坡影响因子，基于GIS栅格分析方法，采用人工神经网络(ANN)、logistic回归和信息量3种评价模型，分别对32种不同影响因子组合进行滑坡易发性对比评价，得到滑坡易发性指数图.应用评价曲线下面积AUC(area under curve)对评价结果进行检验，ANN、logistic回归和信息量3种模型的正确率分别是93.75%、89.76%和90.06%；采用淳安县2014年梅汛期发生的13处滑坡作为预测样本，3种模型预测率分别是94.75%、94.33%和77.21%.上述分析结果表明:ANN模型优于其他两者.以ANN模型评价结果指数图为基础进行易发性分区，采用滑坡强度指标进行分区结果检验，滑坡强度值由易发性低、较低、中和高依次递增，说明分区结果合理.研究成果可以为浙西降雨型滑坡特别是由梅雨引发滑坡的易发性评价提供参考.
|
| 关 键 词： | 易发性评价 滑坡 梅雨 人工神经网络 logistic回归 信息量 |
| 收稿时间： | 2015-09-10 |
A Comparative Study on Plum-Rain-Triggered Landslide Susceptibility Assessment Models in West Zhejiang Province |
| |
| Abstract: | A plenty of landslide susceptibility mapping studies in south west China have been reported in literatures. However, the assessment studies of rainfall-triggered landslides in south east China are still limited, particularly for those dominated by plum rains. Based on GIS and grid analysis, a study case in Cunan county is selected for demonstrating the comparison of applying three methods for landslide susceptibility assessment. They include artificial neural networks (ANN), logistic regression (LGR) and information model (IFM). The landslide inventory includes totally 596 landslides, which is established based on the results of remote sensing interpretation and detail survey. Totally 32 models are established by altering different combinations of controlling factors (CFs) out of totally 9 factors, including elevation, slope angle, slope aspect, slope curvature, lithology, distance from faults, distance from roads, distance from construction lands and vegetation. The indicator of Area Under Curve (AUC) is used for model evaluation. The ANN model could achieve the AUC of 93.75%, which outperforms LGR and IFM with the AUC of 89.76% and 90.06%, respectively. It also performs well in prediction to achieve the AUC of 94.75% compared to those (i.e. 94.33% and 77.21%) from LGR and IFM, where 13 landslides occurred in 2014 during plum-rain season are used for verification. The results of susceptibility zoning based on the derived susceptibility map using ANN also show reasonable, indicating the increases of landslide intensity with the increasing of susceptibility levels. Overall, this study demonstrates the best practices of applying different methods in rainfall-triggered landslide susceptibility assessment, which could be the reference for similar studies elsewhere in west Zhejiang province. |
| |
| Keywords: | |
| 本文献已被 CNKI 等数据库收录！ |
| 点击此处可从《地球科学》浏览原始摘要信息 |
| 点击此处可从《地球科学》下载免费的PDF全文 |
|
