基于FY-4A卫星探测区域模式背景误差和观测误差估计的长波红外通道选择研究

中国新一代静止气象卫星风云四号A星（FY-4A）上搭载的干涉式大气垂直探测仪GIIRS（Geostationary Interferometric Infrared Sounder）是世界上首台在静止轨道运行的高光谱红外探测仪器，拥有1650个通道，其中长波通道689个，中波通道961个。由于计算机存储能力、数据传输和变分同化时效性等限制，目前很难在业务环境中同化全部通道，并且多通道卫星信息往往存在一定的空间相关和光谱相似，故在实际应用中必须降低高光谱资料的通道维数，去除数据冗余和观测相关性，挑选出对特定目标起主要作用的通道子集，进而利用有限的通道来最大限度地提供观测信息。考虑GRAPES（Global/Regional Assimilation and Prediction Enhanced System）数值预报系统估计得到的FY-4A观测范围内的背景误差协方差矩阵，并且在GIIRS观测误差后验重估计的基础上，结合通道权重函数峰值和信息熵分步迭代法对其长波通道进行了通道选择研究和试验。结果表明，在给定选择通道个数时，权重函数法和信息熵方法得到的通道组合进行温度和湿度误差分析时误差整体比单一方法小，在给定通道个数（40个）少于单一方案（50个）时，误差分析结果相当。GIIRS最优通道选择的研究为该资料的同化应用建立了基础。 

A study on longwave infrared channel selection based on estimates of background errors and observation errors in the detection area of FY-4A

Geostationary Interferometric Infrared Sounder (GIIRS) is on board FY-4A, the new generation geostationary weather satellite of China. It is the first high-spectral-resolution advanced infrared sounder on board a geostationary weather satellite in the world. It has 1650 spectral channels to cover long-wave infrared (689 channels) and middle-wave infrared (961 channels) bands. Due to the limitations of computer storage capacity and data transmission and variational assimilation timeliness, it is difficult to assimilate all the channels in the operational assimilation system. On the other hand, the multi-channel satellite information also exhibits certain degrees of spatial correlation and spectral similarity. Therefore, it is necessary to reduce the channel dimension of hyper-spectral data and remove data redundancy in practical application. The subset of channels that play a major role in a particular target is used to provide maximum observation information. For the long-wave channels of GIIRS, an experiment on its channel selection is carried out by combining the weighting function peaks and information entropy. Re-estimates of observation errors and background error covariance in the detection area of FY-4A based on operational Global/Regional Assimilation and Prediction Enhanced System(GRAPES) are also considered. Research results show that it is very important to reasonably choose the channels when retrieving atmospheric temperature from the GIIRS data. On the other hand, the method that combines the weighting function peaks and information entropy is better than the method that only considers a single factor when the number of channels is the same. And the decreases in atmospheric temperature and humidity errors are almost the same when the number of channels obtained from the combination method (40 channels) are smaller than that in the other method (50 channels). This study lays a foundation for the application of GIIRS in the variational assimilation system.