高分五号卫星搭载的大气环境红外甚高光谱分辨率探测仪AIUS（Atmospheric Infrared Ultraspectral Sounder）是中国研发的第一个红外波段具有甚高光谱分辨率的掩星探测仪，为大气分布状态的研究提供了强有力的数据支持。气温是表征大气热力状态的重要参数，其分布状态直接影响地—气系统长波辐射和太阳短波辐射的相互作用，进而对全球辐射能量的收支平衡产生影响。采用高光谱数据直接进行气温反演的数据量较大、存储不便，并且不同波谱信息之间存在相关性，因此需要进行通道选择。考虑到实际气温反演精度受干扰成分的影响较大，致使反演精度较低，本文基于信息熵的通道选择算法与目标成分及干扰成分的敏感性分析相结合开展试验完成AIUS掩星观测数据的通道选择，为后期AIUS的温度反演奠定基础。首先，基于RFM（Reference Forward Model）正向辐射传输模型进行目标成分及干扰成分的敏感性分析，探究通道选择的可行性及进行初步的通道选择；然后，基于信息熵的理论进行通道选取，并对结果进行分析讨论；最后，依据通道选取结果，结合最优化算法进行温度反演效果验证。研究表明，掩星观测对气温变化具有较高的敏感性，该通道选择方法在掩星观测模式下是可行的；随通道数目的增加，信息量大体呈对数型增加，在通道个数为1000时接近饱和；在保证运算效率条件下，选用AIUS的100个通道进行温度反演，可基本满足精度需求。

Atmospheric ultraspectral sounder (AIUS) of GF-5 satellite was successfully launched in May 2018. AIUS is the first very high-resolution infrared-band occultation observation sensor developed in China that provides strong data support for investigating atmospheric distribution state. Air temperature is an important parameter that characterizes the thermodynamic state of the atmosphere. Its distribution state directly affects the interaction between long- and short-wave radiation of the earth-atmosphere system and subsequently influences global balance of radiation energy. Direct inversion of temperature using hyperspectral data is limited by its large size, inconvenient storage, and requirement of channel selection due to the correlation between different spectral information. The strong influence of interfering components on the accuracy of actual air temperature inversion results in low inversion accuracy. Channel selection algorithm based on information entropy is combined with sensitivity analysis of target and interfering components in this study to carry out experiments, complete the channel selection of AIUS occultation observation data, and provide a theoretical foundation for future AIUS temperature inversion investigations. First, the sensitivity analysis of target and interference components based on the RFM model is conducted to explore the feasibility of channel selection and perform preliminary channel selection. Second, channel selection is carried out on the basis of information entropy theory, and the results are analyzed and discussed. Finally, the temperature inversion effect is verified using optimization algorithm on the basis of the channel selection results. Occultation observation is highly sensitive to temperature changes, and channel selection is feasible in the occultation observation mode. The increase in number of channels logarithmically increases the amount of information and is close to saturation when the number of channels is equal to 1000. One hundred AIUS channels are selected for temperature inversion to improve operational efficiency and meet precision requirements.