Assimilation of shortwave infrared radiances from NOAA-20 CrIS in GFS

Substantial day/night discrepancies in the observation minus background (OMB) bias are found for shortwave infrared (SWIR) radiance observations from Cross-track Infrared Sounder (CrIS). To assimilate SWIR radiances, daytime and nighttime innovations for SWIR channels need to have similar characteristics. Several factors that are difficult in radiative transfer modelling may contaminate the measurements and contribute to the discrepancies, including Non-Local Thermodynamic Equilibrium (NLTE) effects for CO2 absorption band around 4.3 𝜇𝑚 and the reflection of solar radiances for surface channels. The Community Radiative Transfer Model (CRTM) is found to underestimate the CrIS NLTE effects on average by about 0.76 K in daytime. An objective methodology has been developed to correct biases in daytime CRTM NLTE simulation, along with a quality control scheme to remove SWIR radiances with poor CRTM NLTE simulation after bias correction. When applying to CrIS measurements, it is found that bias correction greatly reduces the day/night discrepancies in the OMB bias, and the quality control is effective to remove observations with large OMB. It is also interesting to find that the quality control removes more nighttime SWIR radiances than daytime, contradicting the common belief that nighttime SWIR radiances can be directly assimilated without the worry about NLTE effects. The bias correction and the quality control schemes are implemented into the Gridpoint Statistical Interpolation (GSI) assimilation system. Preliminary results on the evaluation of the performance of the bias correction and quality control schemes in GSI will be presented.