New channel selection method of hyperspectral infrared sounders for use in numerical weather prediction

Abstract This study proposed a new channel selection method by examining explicit analysis results obtained from the assimilation of hyperspectral infrared radiances in a numerical weather prediction model. Initially, this new method was evaluated to expand the limited use of infrared atmospheric sounding interferometer (IASI) channels in the current operational Korean Integrated Model (KIM) data assimilation system. To reduce the computational cost of getting the analysis results, we divided all IASI channels into 20 channel groups using the K ‐means clustering algorithm based on the three‐dimensional Jacobian probability density function calculated for each channel. We selected the first channel group with the smallest temperature and humidity analysis errors compared to the ERA5 reanalysis. The channel group that made the smallest errors along with previously selected groups was added sequentially until the additional channel group increased the error. The IASI channel list was expanded to 138 by adding channels sensitive to upper‐tropospheric temperature and mid‐ and upper‐tropospheric water vapor. The impact of the new IASI channel set on the background, analysis, and forecast fields in the KIM was assessed by comparing the results to those obtained using the current operational IASI channels. It was confirmed that the new channels have an overall positive impact on the background and analysis and have improved the KIM forecast accuracy, particularly demonstrating significant improvements in the Northern Hemisphere. Hence, this study potentially provides a new way of channel selection for the assimilation of any future hyperspectral radiances.