InSAR支持下数据与知识驱动的区域滑坡空间预测

It's one of the most urgent key science problems to carry out the key theory of landslide susceptibility assessment to realize scientific early warning of the landslides and to prevent and reduce the loss of the disasters in reservoir areas. The existing prediction methods couldn't deal with the issues to select the predictors and determine the index weights of predictor objectively and quantitatively and to avoid with the loss of useful information during the discretization of the geological variables. The project intends to carry out the following researches in the typical reservoir through detailed field survey, remote sensing interpretation and GIS spatial analysis and modeling:(1) obtain the regional deformation field with InSAR technology; (2) reveal the relationship between the different types of landslides and the related factors based on the research of the typical landslide deformation mechanisms and the regional spatial distribution of landslide; (3) select the predictors including the regional deformation field with InSAR technology objectively and quantitatively besed on weights of evidence method and expert knowledge; (4) establish the fuzzy weights of evidence model using integrating data-driven and knowledge-driven modeling and the data-driven model to determine the index weights of predictor objectively and quantitatively; (5) establish the weighted weights of evidence model to overcome the layer-related conditions to solve the problem of the conditional independence assumptions restrict of the weight of evidence model; (6) establish the test standards of the landslide susceptibility mapping and achieve landslide susceptibility assessment in reservoir areas using integrating data-driven and knowledge-driven modeling. The research results will provide an effective improvement of the methods of landslide susceptibility assessment in the reservoir area, the source of scientific and technical support for the practice of the regional large-scale landslide disaster prevention and mitigation.

系统开展区域滑坡空间预测的关键理论方法研究，是实现区域滑坡科学预警亟待解决的关键科学问题。现有的预测方法没有很好地解决预测指标选取、指标权重的定量取值及地质变量离散化导致的有用信息丢失等问题，影响了预测精度。本项目拟以区域滑坡为研究对象，研究利用InSAR技术获取区域变形场的方法；通过工程地质调查与遥感解译，剖析典型滑坡变形机制与成因，结合区域滑坡特征、空间分布规律及InSAR技术，揭示区域滑坡成因及变形演化规律；将基于InSAR技术的滑坡区域变形监测信息作为区域滑坡空间预测指标，补充现有预测指标体系，建立区域滑坡空间预测指标体系；建立数据与知识驱动的模糊证据权模型，确定预测指标的权重取值；建立加权证据权模型，解决证据权模型条件独立性假设制约难题；建立预测结果的检验标准，实现基于数据与知识驱动的的区域滑坡空间预测。研究成果对于区域性大范围滑坡群体的防灾减灾实践具有重要理论意义和应用价值。

本项目以典型区域（贵州省望谟县滑坡易发区）区域滑坡为研究对象，研究了利用InSAR技术获取区域变形场及变形演化规律的方法。通过遥感影像资料解译及详细的野外工程地质调查与分析，剖析了典型滑坡变形机制与成因，结合区域滑坡特征、空间分布规律及InSAR技术，揭示了区域滑坡分布规律及影响因素，建立了区域滑坡及其影响因素信息图谱。建立了区域滑坡空间预测指标体系，建立了基于数据与知识驱动的模糊证据权模型，实现了基于数据与知识驱动模型的的区域滑坡空间预测，研究成果在贵州省望谟县望谟河小流域及长江三峡库区新滩-黄蜡石区域滑坡空间预测研究得到了应用。研究成果对于区域性大范围滑坡群体的防灾减灾实践具有一定的理论和应用价值。