A novel automated channel-selection method based on the gas sensitivity and weighting function characteristics has been applied on simulated ultra-spectral thermal infrared data for CO profile retrieval in our previous work. The method consists of two steps: 1) channels with abundant gas information and insensitivity to other gases are selected as the initial channel group, 2) the optimal channel group is then obtained by optimizing the distribution of the weighting function. The retrieval results show that the method can reduce the redundancy of channels and improve the retrieval accuracy and efficiency of CO profiles. In this paper, the proposed method is assessed by applying on the retrieval of O3 and CH4 profiles from the ultra-spectral data and then a set of channels are selected for each gas and atmospheric situation. By comparing to the Optimal Sensitivity Profile (OSP) method, which suggests good performance in the literature, it shows that the selected channels by the proposed method in all the sets are less correlated and some channels with special information but relatively low sensitivity are screened. The root mean square errors (RMSEs) of the most retrieved gas profiles by the novel method are smaller than these by the OSP one. The results indicate that the automated channel-selection method can facilitate the retrieval accuracy for different gas profiles from ultra-spectral data and may have application in the ultra-spectral feature selection and data compression.