不确定性层次聚类滑坡灾害危险性评价方法及应用研究

发布时间：2018-02-05 00:10

本文关键词： 不确定数据层次聚类滑坡危险性评价 GIS　出处：《江西理工大学》2016年硕士论文　论文类型：学位论文

【摘要】：中国地貌结构丰富繁杂是全球滑坡灾害非常严重的国家之一。滑坡作为地质灾害的主要类型,危害着乡村、工业矿区、城镇。滑坡灾害拥有区域性、群发性和多发性等特性以及国民经济总体水平不高的情况下,决定了我国不可能有富足的经济和技术实力在短期内对全部潜在危害的滑坡灾害点进行全部普及的工程治理,因此滑坡危险性评价模型的准确性、可靠性和有效性已经成为目前地质灾害预测的难点和迫切需要解决的问题之一。区域滑坡危险性评价系统是一个开放的非线性耗散结构,因此需要借助非线性科学等建立评价数学模型。数据挖掘聚类分析是多元统计分析中的一种常用的非监督分类法,它以统计学形式将具有相似特征的数据进行归类,预测与分析算法含义清晰。但由于滑坡灾害相比与其他地质灾害的独特性质,往往存在很多隐性的和不确定的因素,如降雨、人工破坏等,传统的聚类分析方法无法对其进行有效刻画,从而导致滑坡危险性评价精度的下降,这也使滑坡灾害的预测和分析困难重重。因此,新的滑坡灾害的评价方法的研究刻不容缓。本文以陕北黄土高原北部的延安市宝塔区地质灾害详细调查资料为数据依据,以自然作用、人类活动以及自然作用于人类活动双重作用下的黄土地质灾害形成机理的理论依据为支撑,通过对研究区地质环境背景及滑坡概况的深度研究,创建评价因子与决策因子相互作用的区域滑坡危险性评价指标体系,引入数据挖掘聚类分析高效算法,总结前人在滑坡等地质灾害研究中所存在的不足,进行改进,提出一种基于不确定性层次聚类分裂算法(U-DIANA)的区域滑坡危险性评价方法。该算法在针对滑坡等地质灾害中降雨等评价因子具有的不确定性和极强参与性,通过在数据挖掘层次聚类的理论上构建了不确定性层次聚类分裂算法,解决了对降雨等不确定性评价因子刻画不足的问题;运用ArcGIS对该研究区滑坡危险性评价因子与决策因子进行叠加分析、空间统计分析,可视化分析等系列功能相互协作,重现地理实体;设计"搜索法"与"专家评价法"相结合的滑坡危险性等级确定准则,建立滑坡危险性等级评价模型,并将模型应用于延安市宝塔区进行验证。实验结果表明,该评价模型在聚类有效性分析中取得了较高的精度,且精度高于传统层次聚类评价方法,达到区域滑坡评价精度标准。为防灾减灾提供相关技术保障。
[Abstract]:The landform structure of China is rich and complicated, which is one of the most serious landslide disasters in the world. As the main type of geological disasters, landslide endangers villages, industrial mining areas, towns and towns. Landslide disasters have a regional nature. The characteristics of mass occurrence and multiple, as well as the overall level of the national economy are not high. It is decided that it is impossible for our country to have abundant economic and technical strength in a short period of time to control all the potential hazards of landslide, so the model of landslide risk assessment is accurate. Reliability and effectiveness have become one of the most difficult and urgent problems in geological hazard prediction. The regional landslide risk assessment system is an open nonlinear dissipative structure. Therefore, it is necessary to establish evaluation mathematical model with the help of nonlinear science. Data mining clustering analysis is a common unsupervised classification method in multivariate statistical analysis. It classifies the data with similar characteristics in the form of statistics, and the prediction and analysis algorithm has a clear meaning, but because of landslide disasters compared with other geological hazards of unique properties. There are many hidden and uncertain factors, such as rainfall, artificial destruction, etc., which can not be effectively described by the traditional clustering analysis method, resulting in the decline of landslide risk assessment accuracy. This also makes it difficult to predict and analyze landslide disasters. It is urgent to study the new evaluation method of landslide disaster. Based on the detailed investigation data of geological hazards in Baota District of Yanan City in the northern part of the Loess Plateau in northern Shaanxi Province, this paper takes natural action as the basis. Based on the theoretical basis of the formation mechanism of loess geological hazards under the dual action of human activity and natural action on human activities, the geological environment background and landslide profile of the study area are studied in depth. The evaluation index system of regional landslide risk is established, and the efficient algorithm of data mining clustering analysis is introduced to sum up the shortcomings of previous researches on landslide and other geological hazards. Make improvements. A hierarchical clustering splitting algorithm based on uncertainty (U-DIANAA) is proposed. This algorithm has uncertainty and strong participation in evaluating factors such as rainfall in geological hazards such as landslides. Based on the theory of hierarchical clustering of data mining, an uncertain hierarchical clustering splitting algorithm is constructed, which solves the problem of inadequate characterization of uncertain evaluation factors such as rainfall. The landslide risk assessment factors and decision factors in the study area were analyzed by ArcGIS. The spatial statistical analysis and visual analysis were used to reproduce the geographical entity. The criteria of landslide hazard grade determination combined with "search method" and "expert evaluation method" are designed, and the evaluation model of landslide risk grade is established. The model is applied to validate the landslide hazard grade in Baota District of Yan'an City. The experimental results show that the model can be used in Baota District of Yanan City. The accuracy of the evaluation model is higher than that of the traditional hierarchical clustering method in the cluster validity analysis, which meets the precision standard of regional landslide evaluation and provides the relevant technical guarantee for disaster prevention and mitigation.
【学位授予单位】：江西理工大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：P642.22

【相似文献】