支持向量机与Newmark模型结合的地震滑坡易发性评估研究

发布时间：2018-06-26 01:20

本文选题：滑坡易发性评估 + 地震滑坡　；参考：《地球信息科学学报》2017年12期

【摘要】：Newmark位移模型是研究地震滑坡易发性的经典模型,机器学习方法支持向量机模型也越来越多的应用到滑坡易发性评估研究。本文将Newmark位移模型与支持向量机模型相结合,建立基于物理机理的地震滑坡易发性评估模型并应用于2008年汶川地震重灾区汶川县。从震后遥感影像目视解译出汶川县1900处地震诱发滑坡,并将其随机划分为70%的训练数据集和30%的验证数据集。选择地形起伏度、坡度、地形曲率、与构造断裂带距离、与水系距离、与道路距离6个因子与Newmark位移值共同作为地震滑坡易发性影响因素。利用ROC曲线和模型不确定性等指标对模型结果进行评估,并与二元统计模型频率比和多元统计模型Logistic回归的结果进行对比。结果表明:与频率比和Logistic回归模型相比,支持向量机模型的正确率最高,训练集和验证集ROC曲线下的面积分别为0.876和0.851。将模型应用于绘制汶川县地震滑坡易发性图,结果显示滑坡易发性图与实际的滑坡点位分布一致性较高,有80.4%的滑坡位于极高和高易发区。这说明支持向量机与Newmark位移方法结合建立的地震滑坡易发性评估模型有较高的预测价值,可以为滑坡风险评估和管理提供依据。
[Abstract]:Newmark displacement model is a classical model to study the vulnerability of earthquake landslide. The machine learning method support vector machine model is applied more and more to the evaluation of landslide vulnerability. In this paper, the Newmark displacement model and support vector machine model are combined to establish the earthquake landslide vulnerability assessment model based on physical mechanism and applied to Wenchuan county in Wenchuan earthquake disaster area in 2008. 1900 earthquake induced landslides in Wenchuan County were visually interpreted from the remote sensing images after the earthquake, and were randomly divided into 70% training data set and 30% validation data set. Six factors, such as terrain fluctuation, slope, topographic curvature, distance from tectonic fault zone, distance from water system, distance from road to road, and displacement value of Newmark, are selected as influencing factors of earthquake landslide susceptibility. ROC curve and model uncertainty were used to evaluate the model results, and the results were compared with the frequency ratio of binary statistical model and logistic regression of multivariate statistical model. The results show that the accuracy of SVM model is the highest compared with frequency ratio and logistic regression model. The area under ROC curve of training set and verification set are 0.876 and 0.851 respectively. The model is applied to draw the landslide susceptibility map of Wenchuan County. The results show that the landslide susceptibility map is consistent with the actual landslide location distribution, and 80.4% of the landslides are located in extremely high and high prone areas. This shows that the model of earthquake landslide vulnerability assessment based on support vector machine and Newmark displacement method has high predictive value and can provide basis for landslide risk assessment and management.
【作者单位】：北京师范大学环境演变与自然灾害教育部重点实验室;北京师范大学减灾与应急管理研究院;
【基金】：国家自然科学基金项目(41271544) 地表过程模型与模拟创新研究群体科学基金(41621061) 国家重点研发计划专项项目(2016YFA0602403)
【分类号】：P642.22

【相似文献】