InSAR技术支持下的三峡库区库岸滑坡早期识别与面状变形预测研究

Effective landslide early identification and deformation prediction are of great significance for improving the active prevention ability of landslide disasters. As the early identification and the point-based deformation prediction model of landslides are not fine enough, in this project, the Wanzhou District of the Three Gorges Reservoir Area is selected as the case to carry out the early identification and area deformation prediction of landslide based on InSAR. Firstly, the characteristics of the displacement-time curves of the reservoir landslides will be statistically summarized and generalized, and the identification index database under InSAR technology will be established to quickly identify landslides in the study area. For the major reservoir landslide, fine identification of the spatial distribution and time characteristics of the landslide surface deformation is going to carry out using the time series 3D surface deformation information extracted from the multi-track high-resolution SAR images. The data of rainfall and reservoir water level will be used to mine the response relationship between landslide deformation (area deformation) and the influencing factors in different positions. On this basis, the landslide area deformation will be predicted based on response relationship of induced factors, using the GIS technology and the kernel extreme learning machine algorithm, and the prediction performance will be verified. The research results are expected to improve the fine identification and deformation prediction ability of the reservoir landslide, which can provide support for landslide early warning and risk control.

有效的滑坡早期识别与变形预测对提升滑坡灾害的主动防范能力具有重要意义。针对滑坡早期识别和以往基于特征点的变形预测模型存在精细化程度不高的问题，本项目以三峡库区万州区为研究对象，开展合成孔径雷达干涉测量技术（InSAR）支持下的库岸滑坡早期识别与面状变形预测研究；首先，统计并概化库岸滑坡的位移时间曲线特征，建立InSAR技术下的识别指标库，对研究区的库岸滑坡进行快速的早期识别；选取典型重大滑坡，应用多轨道高分辨率雷达影像提取滑坡地表的时序三维形变信息，精细化识别滑坡变形的空间分布与时间变化特征；结合降雨和库水位数据，挖掘滑坡变形（面状）与诱发因素在不同空间位置的响应关系，在此基础上，应用GIS技术和核极限学习机算法，开展基于诱发因素响应关系的滑坡面状变形预测研究。研究成果有望提升库岸滑坡的精细化早期识别与变形预测能力，可为库岸滑坡的预警预报和风险管控提供支持。

针对库岸堆积层滑坡早期识别和以往基于特征点的变形预测模型存在精细化程度不高等问题，以“区域滑坡早期识别-重大滑坡精细化识别-面状变形预测”为框架，从三峡库区库岸滑坡的成灾机理出发，基于时间序列InSAR位移、勘查资料和地面专业监测等数据的统计分析特征，开展 InSAR 技术支持下的三峡库区库岸滑坡早期识别与变形预测研究，主要研究进展包括：1）揭示了库水位波动和降雨条件下的三峡库区动水压力型和浮托减重型滑坡的变形特征和成灾机理，统计了累计位移、几何形态和滑体渗透性等特征，提出了动水压力型和浮托减重型滑坡的类型划分指标方法；2）应用时间序列InSAR技术开展了万州和云阳的地表变形速率反演与投影，结合InSAR点目标变形强度，空间分布密度和滑坡类型划分指标，提出了三峡库区库岸滑坡早期识别的技术方法；3）分析了万州区滑坡影响因素与空间发育的量化关系，耦合集成学习与机器学习算法构建了易发性建模方法，同时采用时序InSAR反演的地表形变速率，构建判定矩阵，提出了耦合集成机器学习和时序InSAR技术的区域滑坡易发性评价方法；4）基于多时相InSAR方法提取的高精度形变信息，精确表征了典型动水压力型和浮托减重型滑坡的地表变形的空间差异和时间变化特性，揭示了诱发因素与库岸滑坡变形的动态响应机制，应用机器学习算法构建了不同类型的大型库岸滑坡变形预测模型。