利用SAR极化相位差反演北疆典型区积雪累积期的雪深

Synthetic Aperture Radar (SAR) technology has great potential in detecting snow signal in high resolution. However, the main factor of determining back scattering is surface scattering underlying snow. Hence, how to acquire signals of snow itself is a key problem. According to the observation of snow microstructure, the shapes of ice grains change and the depth of snow pack decreases during the metamorphic process. Fortunately, the polarimetric status of SAR is sensitive to the changes of shapes and orientations of scattering object. Thus, the proposal focuses on the theory and method of retrieving snow depth using polarimetric information based on high spatial-temporal resolution X-band full-polarization SAR satellite and field survey data during accumulation period. The study area is located in the flat area of middle-reach in Kelan River Basin, Altay Region, North Xinjiang Province, China. First, the physical relation between the change of co-polarized phase difference (CPD), snow depth and snow microstructure during the metamorphic process will be analyzed. Second, a simulated model of CPD will be established and verified to conduct sensitivity analysis. Third, the relationship between CPD and snow depth will be used to establish a semi-empirical inversion model of snow depth. Finally, the retrieval results will be verified and the adaptable conditions of the model will be analyzed based on the field synchronous observation data. The project is expected to clarify the mechanism of snow metamorphic process to CPD and snow depth, and to build the inversion model of snow depth based on CPD, and to continuously monitor depth of seasonal snow during accumulation period. The research findings will have scientific values for regional climate change researches and water resource managements of river basin.

合成孔径雷达（SAR）对探测高分辨率积雪信号具有较大潜力，但影响干雪后向散射的主要是雪下地表的散射，如何获取积雪本身的信号成为关键问题。对积雪微结构的观测表明，积雪变质过程中冰粒形状发生变化，雪深随之降低，而SAR的极化状态对散射体的形状、取向敏感，因此，本项目拟研究利用SAR极化信息反演雪深的理论与方法。在北疆阿勒泰地区克兰河流域中游平地选择研究区，利用高时空分辨率X波段全极化SAR卫星和地面观测数据，首先分析积雪变质过程中积雪微结构变化与同极化相位差（CPD）、雪深变化的物理联系，然后建立CPD模拟模型并进行验证和敏感性分析，最后根据CPD与雪深的关系建立半经验的雪深反演模型，开展真实性检验与适用条件分析。该项目有望阐明积雪变质过程对CPD和雪深的作用机制，建立基于CPD的雪深反演模型，实现季节性积雪累积期的雪深持续监测，研究成果对区域气候变化研究和流域水资源管理具有科学价值。

积雪深度是积雪的重要结构参数，获取高精度雪深空间分布信息对水资源管理、气候变化研究和灾害预报等具有重要意义。合成孔径雷达（Synthetic Aperture Radar, SAR）因其有较高的空间分辨率且对积雪特性有着较高的敏感度等优势，在流域尺度的雪深反演中具有较强适用性。本项目以北疆克兰河流域为研究区，利用2018年1月7日与18日两景X波段双极化TerraSAR-X数据以及地面同步观测数据，根据VV与HH极化信号在积雪中折射率不同导致相位差异的原理，使用Maxwell-Garnett方程构建同极化相位差（Co-polar Phase Difference, CPD）的正演模型，并对模型进行敏感性分析；研究CPD与含水量、地形的关系及其作用机制，明确模型的积雪和地形适用条件；根据CPD与雪深的关系建立半经验的雪深反演模型并验证模型的有效性；分析下垫面和地形对反演模型的影响并基于下垫面、地形改进反演模型。研究结果表明：（1）CPD模型对各向异性强、密度大、深度大的雪层更加敏感，对入射角较大和频率较高的SAR电磁波更加敏感；（2）忽略局部入射角差异，当积雪介电常数恒定时，CPD仅是雪深的函数，可利用实测点雪深与从SAR数据获取的CPD来拟合线性模型反演雪深，雪深反演精度与计算CPD时的滤波器窗口大小有关。利用2018年1月7日和18日的TerraSAR-X影像反演雪深的最佳滤波窗口分别是57×57像元（R=0.392，RMSE=8.01cm）和69×69像元（R=0.688，RMSE=3.40cm）;（3）根据下垫面类型划分研究区，分别使用基于地表类型的自适应滤波计算CPD进行雪深反演，可以有效提高雪深反演精度；（4）考虑地形对局部入射角的影响，计算冰粒轴间比并将其作为雪深反演算法的输入来改进雪深反演模型，可有效提高反演精度。利用2018年1月7日和18日的TerraSAR-X影像反演的雪深精度分别是R=0.557，RMSE=2.51cm和R=0.651，RMSE=2.57cm。本项目研究表明CPD模型反演山区雪深空间分布是有效和可行的，研究成果可为实现山区季节性积雪累积期的雪深持续监测提供科学支撑。