2016年9月15日发射的TG-2空间实验室上搭载的MAI是我国首个在轨运行的多角度偏振成像仪, 主要用于获取云和气溶胶等大气环境信息。 星载遥感仪器的定标是观测资料定量应用的关键前提且贯穿仪器的整个寿命期。 MAI发射前已经进行了实验室定标, 且精度较高。 为了监测MAI发射后的在轨运行情况, 针对其未配置在轨定标装置的问题, 利用Metop-B/GOME-2可见光波段的高光谱分辨率和较高探测精度的优势, 提出了基于GOME-2对MAI 565, 670以及763 nm通道进行在轨监测及交叉定标的方法。 该方法首先通过时空匹配、 视线几何匹配等获取MAI与GOME-2相近时刻、 相近视线几何条件下的同目标观测数据, 再将GOME-2反射率按照MAI可见光通道光谱响应函数进行卷积, 得到各通道的参考反射率, 与MAI反射率进行对比分析, 从而实现对MAI的定标。 利用不同反照率特性目标的匹配观测数据, 该方法能够实现仪器的高、 中、 低端观测的全覆盖定标。 定标过程主要包括: （1）对2016年12月到2017年2月期间TG-2和Metop-B的运行轨道进行预报, 获取二者交叉观测的整轨数据; 设置观测时间差为900 s, 初步匹配得到8组MAI与GOME-2交叉观测样例, 包含2 455组匹配像元; （2）对匹配像元空间位置进行检验, 保留单个GOME-2像元覆盖的MAI像元数超过338的交叉样本, 以确保单个GOME-2像元尽可能被MAI观测充满; （3）给定GOME-2观测天顶角小于30°的限制条件, 同时设置视线几何检验条件为两仪器观测天顶角余弦的比值接近于1, 且相差不超过0.05, 并充分利用MAI的多角度观测优势, 对每一个MAI像元采用最多14个方向的视线几何进行匹配, 从而选择最优的视线匹配方向; （4）设置观测目标均匀性检验条件为一个GOME-2像元覆盖的全部MAI像元反射率的标准差和均值之比小于0.5, 对匹配像元进行检验, 得到469个匹配的GOME-2像元。 （5）将以上GOME-2像元对应的各个波长的反射率按照MAI可见光通道的光谱响应函数进行积分, 即可得到MAI各通道对应的GOME-2参考反射率。 （6）利用GOME-2像元空间分辨率显著大于MAI分辨率的特征, 对每个GOME-2像元覆盖的全部MAI像元反射率进行平均作为MAI反射率, 显著降低了定标结果对观测目标均匀性的依赖程度。 （7）将GOME-2参考反射率与MAI反射率进行回归分析, 得到定标系数, 实现对MAI的在轨交叉定标。 为了分析各匹配条件对定标结果的影响, 利用单一变量法对像元匹配过程中各检验条件阈值进行调整并开展了分析试验。 结果表明, 当进一步严格匹配筛选条件时, 定标结果不会产生显著变化。 基于该方法对MAI三个通道反射率和GOME-2参考反射率进行对比分析, 结果表明二者之间存在显著地线性关系, 且相关系数均优于0.97, 对比差异的均值分别为1.6%, 4.2%和2.3%, 标准差分别为3.1%, 4.1%和2.4%。 总体来看, 利用在轨交叉定标方法能够实现MAI可见光波段的在轨监测及定标, 为MAI数据的定量应用奠定了基础。

The MAI on TG-2 space laboratory, which was launched on 15 September 2016, is the first on-orbit Multi-angle Polarization Imager in China. The capability of MAI is mainly used to obtain macroscopic and microphysical features of clouds. On-orbit calibration of spaceborne remote sensing instruments is a key prerequisite for the quantitative application of observational data and extends throughout the life of the instrument. Laboratory calibration has been performed prior to MAI launch with high accuracy. In order to monitor the status of MAI after launched, aiming at the problem that MAI has no onboard calibration system, a method of on-orbit monitoring and inter-calibration of TG-2/MAI 565, 670 and 763 nm channels based on Metop-B/GOME-2 hyperspectral data has been presented. The method first obtains the data of same observation target at the similar time and near geometric condition of MAI and GOME-2 based on spatial, temporal and geometric collocation criterion. Then, the GOME-2 reflectance is convoluted with the spectral response function of the MAI visible channels to obtain the reference reflectivity of visible channels. Finally, compare the reference reflectivity with the MAI reflectivity to achieve the onboard calibration of MAI. The process of calibration mainly includes: (1) Forecasting the orbit of TG-2 and Metop-B from December 2016 to February 2017 to obtain the collocated observations between MAI and GOME-2. The temporal matching interval is set to 900 s, and 8 collocated samples of MAI and GOME-2 are obtained, including 2 455 matched pixels. (2) The spatial location of matched pixels is checked, and the cross samples with MAI pixels over 338 covered by a single GOME-2 pixel is reserved to ensure that a single GOME-2 pixel is filled as completly as possible by the MAI pixels. (3) The limit of GOME-2 observation zenith angle is set to 30°, and the geometry of the observation sight detection condition matching pixels is set to the ratio of cosine of the two instruments observed zenith angle is close to 1, and the difference is not more than 0.05, and takes full advantage of MAI multi-angle observation, which allows each MAI pixel with up to 14 viewing angles. Therefore, the optimal matching viewing angle could be chosen; (4) In the target uniformity checking, the condition of uniformity detection for matched pixels is set to the ratio of the reflectance of all MAI piexls coveraged by a GOME-2 piexl standard deviation and the average is less than 0.5. And 469 GOME-2 pixels are reserved. (5) The reflectance of each wavelength corresponding to the above GOME-2 pixels is convoluted with the spectral response function of the MAI visible channel to obtain the corresponding GOME-2 reference reflectance of each MAI channel. (6) Based on the large difference of the spatial resolution of GOME-2 and MAI pixel, the reflectance of all MAI pixels covered by each GOME-2 pixel is averaged and taken as MAI reflectivity, which significantly reduces the dependence of calibration results on target uniformity. (7) And the inter-calibration coefficients are derived by regression analysis of the GOME-2 reference reflectivity and the MAI reflectivity. Onboard inter-calibration of the MAI is achieved. To analyze the influence of matching and screening conditions on the calibration results, the simple variable method is used to adjust the threshold of each test condition in pixel matching and screening process. The results show that the calibration results do not change significantly when the matching and screening conditions are more stringent. The MAI reflectance and the GOME-2 reference reflectance are compared, and the results indicate that both reflectivities have a significant linear relationship with the correlation coefficients all better than 0.97. The mean values of their differences are 1.6%, 4.2% and 2.3%, and the standard deviations are 3.1%, 4.1% and 2.4% for the three channels, respectively. Therefore, on-orbit monitoring and vicarious calibration of the MAI visible bands can be achieved by the inter-calibration method, which lays the foundation for the quantitative application of MAI data.