基于卫星偏振遥感观测的气溶胶和地表多参数最优化联合反演模型和算法研究

To focus on the urgent demanding of domestically-made polarimetric satellite for the aerosol remote sensing application, and in combination with the international research frontier and future development trend, the optimized inversion models and algorithms for the simultaneous retrieval of aerosol and surface multi-parameters are proposed. For the domestic polarimetric satellites adopting the single-viewing observation (such as Polarized Scanning Atmospheric Corrector on HJ-2) and multi-angle observation (such as Directional Polarimetric Camera on GF-5) respectively, almost all of the measurements of spectral, angular, scalar and polarization are comprehensively utilized for inversion modeling in this study. For the surface-atmosphere contribution decoupling, principal component analysis (PCA) is used to reconstruct the multispectral surface reflectance with normalized difference vegetation index (NDVI) self-consistent bidirectional polarized reflectance distribution function (BPDF) for single-viewing case, while improved bidirectional reflectance distribution function (BRDF) and BPDF model are employed for multi-viewing case. Besides, information content analysis is employed to evaluate the inversion capabilities and retrieved parameters. By introducing multi-source errors and a priori constraint, the non-linear optimized inversion modeling can be carried out, and the quasi-Newton method is used for solutions with optimization iteration, so as to realize the joint retrieval of aerosol optical and micro-physical parameters, as well as the corresponding surface parameters. By this means, the complete optimized inversion models and algorithms could be established for the retrieval of aerosol and surface multi-parameters, and thus the ill-posed problem could be effectively solved. Based on the inversion study, the retrieval results of multi-parameters could be further verified and analyzed by the ground measurement data, which can further provide the key support for the related applications of domestic polarimetric satellites.

面向国产偏振卫星在气溶胶遥感领域应用的紧迫需求，结合国际前沿和未来发展趋势，本项目拟开展气溶胶和地表多参数最优化反演模型和算法研究。综合利用光谱、角度、标量和偏振观测最为全面的信息，分别针对单角度(如HJ-2同步大气校正仪)和多角度(如GF-5卫星DPC)观测的典型偏振传感器，采用主成分分析(PCA)光谱重建、改进的地表二向性反射分布函数(BRDF)及归一化植被指数(NDVI)自洽约束的地表偏振模型(BPDF)进行地气解耦合，并结合信息量分析来定量评价传感器的反演能力和可反演参数。通过引入多源误差和先验约束，开展非线性最优化反演建模和利用拟牛顿法迭代求解，进而实现气溶胶光学、微物理参数和地表模型参数的联合反演，建立一套完整的气溶胶和地表多参数最优化反演模型和算法，可有效解决病态反演问题。在此基础上，利用地基观测数据对多参数反演结果进行验证分析，为国产偏振卫星的遥感应用提供关键支撑。

面向我国最新发展的偏振系列卫星在气溶胶遥感领域应用的紧迫需求，结合国际前沿和未来发展趋势，针对利用国产偏振卫星传感器进行气溶胶和地表多参数遥感反演存在的关键问题，重点研究偏振卫星遥感反演中的地气解耦合及大气气溶胶和地表多参数非线性最优化反演方案，并利用地基观测数据对气溶胶等关键参数进行验证及分析。在针对不同观测模式下气溶胶和地表关键参数的信息量分析和最优化反演测试的基础上，建立了一套较为完整的通过国产偏振卫星观测快速获得气溶胶等多参数的模型和算法，并对实际反演结果进行了系统的精度验证。本项目主要研究内容包括三部分：模型和算法研究、结果反演及验证、相关拓展应用研究。模型和算法研究主要包括偏振遥感地气解耦中的地表模型、偏振遥感观测前向模拟及敏感性分析、气溶胶和地表多参数最优化反演模型等核心内容。基于国际上PARASOL卫星和我国最新发展的偏振遥感系列卫星传感器（DPC/GF-5、SMAC/GFDM和PSAC/HJ-2等）实际观测数据，开展了空间分布的大气关键参数（气溶胶和水汽）反演研究，系统地精度验证结果显示所开发的反演算法可满足相关卫星载荷在轨测试的精度要求。在此基础上，进一步拓展应用到我国宽波段多光谱卫星（GF-1）和风云系列卫星（FY-4A）气溶胶光学厚度的反演上，研究覆盖的波段范围也从可见光-近红外-短波红外拓展到了紫外和中红外波段。本项目的研究工作有效支撑了我国高分多模卫星大气同步校正仪（SMAC/GFDM）、环境减灾二号A/B卫星大气同步校正仪（PSAC/HJ-2）和高分五号02卫星偏振交火传感器（PCF/GF-5 02）的部分在轨测试工作，为气溶胶光学厚度和水汽柱浓度等大气关键参数产品的业务化生产提供了反演算法保障。