| 信息量模型、确定性系数模型与逻辑回归模型组合评价地质灾害敏感性的对比研究 |
| |
| 引用本文： | 张晓东,刘湘南,赵志鹏,吴文忠,刘海燕,张勇,高宇亮.信息量模型、确定性系数模型与逻辑回归模型组合评价地质灾害敏感性的对比研究[J].现代地质,2018,32(3):602. |
| |
| 作者姓名： | 张晓东 刘湘南 赵志鹏 吴文忠 刘海燕 张勇 高宇亮 |
| |
| 作者单位： | 1中国地质大学(北京) 信息工程学院，北京100083；2宁夏回族自治区地质调查院，宁夏 银川750021；
3山东科技大学 地球科学与工程学院,山东 青岛266590 |
| |
| 基金项目： | 宁夏回族自治区国土资源厅“宁夏盐池县地质灾害详细调查项目”(XC(2012)-05)；宁夏回族自治区地质局水环创新团队后补助资金研究项目(2017-水环团队04)。 |
| |
| 摘 要： | 为探索区域地质灾害敏感性评价方法，以宁夏盐池县为研究区域，选取坡度、坡向、坡高、高程、地层、距河流距离、距道路距离、植被覆盖度等8个影响地质灾害发生的评价因子，分别采用信息量模型+逻辑回归模型(I+LR)和确定性系数模型+逻辑回归模型(CF+LR)2种组合模型对盐池县地质灾害敏感性进行评价，将该区域地质灾害划分为极低、低、中和高敏感区4类，并完成结果检验。结果表明：(1)2种组合模型得到的低、中敏感区面积基本相当，而高敏感区面积相差较大，CF+LR模型较I+LR模型高敏感区面积增加约5336 km2，而极低敏感区面积减少约6%；(2)2种组合模型的合理性均符合检验要求，且ROC精度检验AUC值分别为0868和0829，渐进Sigb均小于005，表明2种组合评价模型都能较为客观准确地评价盐池县地质灾害敏感性；(3)ROC检验精度与盐池县地质灾害发育情况均表明I+LR模型精度更高。
|
| 关 键 词： | 地质灾害 敏感性 确定性系数模型 逻辑回归模型 宁夏盐池 |
Comparative Study of Geological Hazards Susceptibility Assessment:
Constraints from the Information Value+Logistic Regression Model
and the CF+Logistic Regression Model |
| |
| ZHANG Xiaodong,LIU Xiangnan,ZHAO Zhipeng,WU Wenzhong.Comparative Study of Geological Hazards Susceptibility Assessment:
Constraints from the Information Value+Logistic Regression Model
and the CF+Logistic Regression Model[J].Geoscience——Journal of Graduate School,China University of Geosciences,2018,32(3):602. |
| |
| Authors: | ZHANG Xiaodong LIU Xiangnan ZHAO Zhipeng WU Wenzhong |
| |
| Institution: | 1School of Information Engineering, China University of Geosciences, Beijing100083, China;
2Ningxia Geological Survey Institute,Yinchuan, Ningxia750021, China;
3College of Earth Science and Engineering,Shandong University of Science and Technology,Qingdao, Shandong266590, China |
| |
| Abstract: | Susceptibility assessment constitutes an important part of geological hazard research. This paper discussed two integrated approaches, namely the Information Value+Logistic Regression model (I+LR) and CF+Logistic Regression model (CF+LR), to evaluate the geological hazard susceptibility of Yanchi County, Ningxia Hui Autonomous Region. 462 samples (231 hazardous samples and 231 non hazardous samples) were divided into two groups: 75% and 25% of the samples were used for model training and validation, respectively. Nine influencing factors, including slope dip angle and direction, slope height, elevation, strata, distance to rivers and roads, and the normalized difference vegetation index (NDVI) were considered in this evaluation. Based on such information, the information value and CF of the training samples were calculated, which were then fed into the SPSS for analysis via logistic regression. After collinearity diagnostics and correlation analysis, six factors were incorporated into the model eventually. Regression equation was determined using the obtained constants and coefficients, and the Receiver Operating Characteristic (ROC) curve was used to assess the accuracy of the two models. The results show that: (1)Susceptibility maps reveal four susceptibility classes, ie, very low, low, moderate and high. The area percentage of the (I+LR) model for the four class accounts for 6079%, 2344%, 1134% and 443%, respectively, with that of the (CF+LR) model being 5449%, 2289%, 1030% and 1232%. This indicates that the areas of the low and moderate classes are basically the same. The high class area of the (CF+LR) model has increased for 5336 km2 more than the (I+LR) model, but the very low class area has dropped by as much as 6%. (2) The area under the curve for the successful rates are 0868 and 0829, respectively, and asymptotic Sigb is lower than 005 for the two models. Both integrated approaches can produce reasonable accuracy. (3) The ROC accuracy and the geological hazard development conditions at Yanchi both indicate that the (I+LR) model has higher accuracy over the (CF+LR) model. |
| |
| Keywords: | geological hazard susceptibility CF model logistic regression model Yanchi of Ningxia |
| 本文献已被 CNKI 等数据库收录！ |
| 点击此处可从《现代地质》浏览原始摘要信息 |
| 点击此处可从《现代地质》下载免费的PDF全文 |
|
