隐伏断层地震诱发滑坡易发性多尺度评价

发布时间：2018-03-04 00:30

本文选题：地震滑坡　切入点：易发性评价　出处：《浙江大学》2017年硕士论文　论文类型：学位论文

【摘要】：地震滑坡常见于地震灾害链中,不仅带来巨大的生命财产损失,还严重影响了震后救援工作。地震滑坡物理机制复杂,影响因素繁多,基于数据挖掘和GIS的地震滑坡易发性评价有效规避了基于物理模型评价存在的物理参数和监测数据获取困难的问题,成为地震滑坡易发性评价的重要手段。本文以2004年日本新o_中越地震为研究对象,本次地震主震震级为6.8级,其后陆续发生了 4次震级大于6.0级的余震,诱发了数以千计的滑坡,此次地震为隐伏断层地震。本文采用了三种数据挖掘方法,同时为了弥补隐伏断层地震缺乏地表破裂带这一重要影响因子的不足,新增了同震地表变形作为影响因子,分别针对大面积区域和震中区域两个尺度开展了基于数据挖掘和GIS的地震滑坡易发性评价工作。主要的研究工作和研究成果如下:(1)针对大面积研究区域,采用了逻辑回归、支持向量机和人工神经网络等三个数据挖掘模型开展了地震滑坡易发性评价,并对研究区域进行了合理的易发性等级分区。结果表明三种模型方法都取得了较好的结果,ROC曲线下面积都达到了 80%以上,进行易发性分区后随着易发性等级提升,区域内滑坡比例呈较为明显的梯度提升。综合比较确定人工神经网络效果最好。(2)针对震中区域,在原有影响因子的基础上新增同震地表变形作为影响因子,采用模拟效果最好的的人工神经网络进一步开展地震滑坡易发性评价,获得了震中区域的精细化地震滑坡易发性等级分区图。研究结果显示考虑同震地表变形对结果有一定提升,其贡献大于常见的坡向和距道路的距离等影响因子。因此,尤其是针对隐伏断层地震影响因子资料不足的情况下,同震地表变形具备较大的应用价值。(3)多尺度区域评价结果对比分析表明,相较于大面积研究区域地震滑坡易发性评价,震中区域研究在同一区域取得更为精细的地震滑坡易发性分区图,但模拟结果中ROC曲线下面积普遍较低。大面积区域和震中区域地震滑坡易发性评价相结合,实现了多尺度地震滑坡易发性分区,满足地震滑坡灾害的不同层面的防控治理的需要。
[Abstract]:Earthquake landslide is common in earthquake disaster chain, which not only brings huge loss of life and property, but also seriously affects post-earthquake rescue work. The evaluation of seismic landslide vulnerability based on data mining and GIS effectively avoids the problems of obtaining physical parameters and monitoring data based on physical model evaluation. It has become an important means to evaluate the vulnerability of earthquake landslides. In this paper, the earthquake of 2004 in Japan was studied. The magnitude of the earthquake was 6.8, and four aftershocks with magnitude greater than 6.0 occurred one after another. Thousands of landslides have been induced, and the earthquake is a concealed fault earthquake. In this paper, three methods of data mining are used to make up for the deficiency of the hidden fault earthquake lacking the important factor of surface rupture zone. The coearthquake surface deformation is added as the influence factor, The evaluation of seismic landslide vulnerability based on data mining and GIS is carried out on two scales of large area area and epicenter area respectively. The main research work and research results are as follows: 1) for large area study area, logical regression is adopted. Three data mining models, support vector machine (SVM) and artificial neural network (Ann), are used to evaluate the vulnerability of earthquake and landslide. The results show that the area under the ROC curve is more than 80%, and the area increases with the grade of susceptibility. Compared with each other, artificial neural network has the best effect. Aiming at the epicentral area, the coseismic surface deformation is added as the influence factor on the basis of the original influence factors. The artificial neural network, which is the best simulation method, is used to further evaluate the vulnerability of seismic landslides. The fine zoning map of landslide susceptibility in epicentral area is obtained. The results show that considering the coseismic surface deformation, its contribution is greater than that of common influencing factors, such as slope direction and distance from road, and so on. Especially in view of the lack of data on the influencing factors of concealed faulted earthquakes, the results of multi-scale regional evaluation of coearthquake ground deformation have great application value. The results show that compared with the large area study area, seismic landslide susceptibility evaluation is better than that of coearthquake ground deformation. The regional study of epicenter obtained a more precise zoning map of seismic landslide susceptibility in the same area, but the area under the ROC curve is generally lower in the simulation results, and the large area is combined with the seismic landslide susceptibility evaluation in the epicenter region. The multi-scale seismic landslide prone zone is realized, which meets the need of prevention and control of different levels of earthquake landslide disaster.
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：P642.22

【参考文献】