中高分辨率陆地上空大气气溶胶光学厚度与地表二向性反射率卫星遥感反演研究

It is difficult to separate the contributions of atmosphere and land surface from the satellite optical remote sensing signal, which is the key to improve the level of remote sensing information quantification. The early aerosol retrieval algorithms over land are based on the dark surface covered by the dense vegetation. The surface bidirectional reflectance properties, which are usually described as the bidirectional reflectance distribution functions (BRDF), are ignored. The disadvantages of the algorithms include that the spatial resolution of AOD are very low, the feasibility of the algorithm is so weak over the bright surface such as desert and urban area that the AOD retrieval is much uncertain. In this study, a new synchronous retrieval algorithm will be proposed of the higher spatial resolution (about 30 m) aerosol optical depth over land and the surface bidirectional reflectance distribution function parameters from the multi angular dataset sourced from the China HJ-1A/1B of the Environment and Disasters Monitoring Microsatellite. The basic assumptions are the different spatio-temporal change features of the atmospheric aerosol and the land surface BRDF parameters. Normally, atmospheric aerosols have great changes in the time dimension but little changes in the spatial dimension, which is opposite to the land surface BRDF. According to the assumptions, the time series satellite observations of 30m HJ-1A/1B CCD data over the same area are applied to reduce the number of unknowns and retrieve the AOD and BRDF synchronously by solving the radiative transfer equations with the look-up table. To improve the AOD retrieval accuracy, a prior knowledge database of aerosol model will be established and some post processing will be applied. The AOD with higher spatial resolution and better quality control can be used for the regional environment monitoring and the BRDF can be used for other parameters remote sensing modeling and inversion, in order to improve the level of the land quantitative remote sensing applications.

地气解耦问题一直是影响陆地光学卫星遥感信息定量化的重点和难点。早期的陆地气溶胶遥感反演算法大都针对浓密植被等暗地表，且忽略了地表的二向性反射特性（BRDF），其结果是反演得到的气溶胶产品的空间分辨率较低，覆盖范围有限，并且存在较大误差。本研究将以30 m分辨率的国产环境卫星CCD数据为主要数据源，充分考虑大气气溶胶和地表二向性反射参数的时空变化特征，通过合理假设来减少未知数的个数，进而利用多天多角度的卫星遥感数据构建大气AOD和地表BRDF联合反演算法。同时，为了提高反演精度，将建立气溶胶模型先验知识库，并对反演结果进行后期处理，以获得具有较高空间分辨率和质量控制的AOD和地表BRDF产品。其中，AOD可用于区域空气质量监测研究，而BRDF则可以精确刻画地物的方向性反射特征，进而可应用于其他地表参数的建模与反演，提高陆地定量遥感应用水平。

大气气溶胶对人体健康、空气质量、辐射平衡、云降水、全球变化等方面都起着重要作用。卫星遥感仍然是目前唯一能够获取大气气溶胶大尺度时空分布的有效手段。然而，传统的气溶胶卫星遥感算法大多只适用于浓密暗植被（Dense Dark Vegetation，DDV），无法应用到高亮反射率的地表。同时，传统的气溶胶卫星遥感产品的空间分辨率也较低，难以满足更多应用的需求。本项目提出了基于多角度卫星遥感数据集的大气气溶胶光学厚度（Aerosol Optical Depth，AOD）和地表二向性反射率分布函数（Bi-directional Reflectance Distribution Function，BRDF）参数联合反演算法。该算法基于地表反射特性和气溶胶浓度时空分布的合理假设，通过构建地气耦合辐射传输方程和积累必要的气溶胶模型先验知识，借助查找表（Look-Up Table，LUT）来实现大气AOD和地表BRDF参数联合反演。应用该算法，能够从连续20天4次成像的欧洲哨兵二号（Sentinel-2）卫星多光谱成像仪（Multi-Spectral Instrument，MSI）数据中获取高分辨率的大气AOD和地表BRDF参数。真实性检验结果表明，新算法反演得到的AOD与气溶胶自动观测网（Aerosol Robotic Network，AERONET）的资料的相关性达到0.8以上，不确定性为Δτ = ±0.05 ± 0.25τ；而反演得到的BRDF参数计算的反照率，与中等分辨率成像光谱仪（Moderate Resolution Imaging Spectroradiometer，MODIS）同类产品的差值大部分在±0.05以内。这些都表明新算法具有较高的精度。研究还利用得到的AOD和BRDF数据集进行了相关研究，包括分析AOD与对流层臭氧和二氧化氮的相关性和时空变化趋势、利用地表反射率数据估算植被结构参数和叶绿素含量等。项目已发表论文9篇，其中SCI收录3篇（1篇已接收），EI收录3篇，ISTP收录3篇）。更多的后续研究工作还在进行中。