面向滑坡监测的InSAR时序分析和融合方法研究

Spaceborne radar interferometry (InSAR), as a new earth observation technique, possesses many advantages and has in recent years been widely used to monitor surface deformation associated with different geological, geophysical, volcanological and hydrological phenomena. InSAR also shows great potential in landslide monitoring and has made a number of successful applications. However, due to the fact that landslides often occur in mountainous areas with heavy vegetation coverage and complicated atmospheric environment, the application of InSAR appears to be quite challenging. Currently, the existing methods cannot effectively resolve the issues raised in the InSAR processing over these areas (such as optimal selection of SAR data, effective correction of errors, reliable estimation of landslide movement et al.). To tackle these inherent issues, the project will dedicate to the development of advanced InSAR for landslide monitoring and its deformation feature analysis. It focuses on: (1)the optimal selection of SAR data (e.g. wavelength, incidence angle et al.); (2) the development of multi-temporal InSAR method optimized for landslide monitoring; (3) integration of multi- platform, multi-track InSAR measurements for resolving landslide movements along slope direction; and (4) the study of triggering factors of landslides and their spatial-temporal response from the InSARmeasurements. In this project, we will select landslides in Gansu Zhouqu proven to be vulnerable to geo-hazards and the Berkeley area, California, the United States as typical test sites. This project is expected not only to be able to overcome technical bottlenecks of InSAR based landslides monitoring but also to be helpful to build up an early warning system for landslides monitoring and therefore contribute to the prevention of landslide disaters.

雷达干涉测量（InSAR）近年来被广泛用于地表形变监测。对于滑坡监测，InSAR 同样展现出巨大潜力并取得一些成功的应用。然而，由于滑坡常发生于植被覆盖严重、地形起伏大和大气环境复杂的山区，使得 InSAR 应用极具挑战，而现有方法未能有效解决其中的问题（如 SAR 数据的优选，误差的有效改正及滑动量的可靠估计等）。本项目将着力于面向滑坡监测及其形变特征分析的 InSAR 关键技术研究，包括：（1）适用于滑坡监测的 SAR 数据优化配置；（2）面向滑坡监测的多时域 InSAR 优化模型；（3）融合多平台多轨道 InSAR 的滑坡坡向形变分解；和（4）基于InSAR形变资料的滑坡时空特征规律分析及稳定性评估，并将选择滑坡灾害频发的甘肃舟曲和美国伯克利地区作为试验区。本项目不仅可突破 InSAR 监测滑坡的技术瓶颈，还有助于建立滑坡灾害预警模型，从而提高我国滑坡灾害的防治水平。

雷达干涉测量已经发展成为一项日趋成熟的变形监测技术。 在滑坡监测中， InSAR 同样也逐 渐展现出了巨大潜力，并取得了一些成功的应用。 然而目前 InSAR 技术监测滑 坡很多时候结果并不十分理想，主要原因在于滑坡往往位于植被覆盖严重、地形起伏较大和大气复杂多变的地区， 地物地形及环境因素使得 InSAR 固有误差（如时空失相关噪声、 DEM 误差、 大气延迟和几何畸变等） 更加难以消除。为解决这些问题，在基金支持下，我们重点进行了如下研究：（1）基于DEM误差的空间特性及基线依赖关系，发展了一种DEM误差非参数改正方法，实验表明该方法需要的数据集更少对大气延迟误差的抑制能力更强；（2）顾及大气延迟误差的空间变化性，发展了一种区域分块时序InSAR模型，在该模型中大气延迟得到更为精准的描述，显著提高了大气延迟到改正效果；（3）针对山区高植被下相干性较低的现实，分别发展了基于相位残差分析的相干点选取方法及可克服相干性估计偏差的相位增强方法，提高了InSAR在高植被区的观测能力。与多个研究机构分享了算法原型。基于自研的算法，对舟曲及台湾的三个滑坡进行了数据处理及机理分析。此外，针对滑坡可监测性，大形变提取，InSAR模型误差传播，相位空间相关性约束等方面也开展了研究。项目执行期间共发表SCI论文14篇，其中半数发表于遥感类前五的期刊;参加国际会议5次，国内会议2次；开展专题讲座4次;培养博士毕业生2名，硕士3名。项目取得的一系列算法均围绕滑坡形变监测，有效提高了InSAR技术的监测能力，可为国家地质灾害隐患识别提供技术支撑。