Exploration of the use of short‐wave infrared radiances in weather forecasts: Part I. Methodologies for bias correction and quality control

Abstract Short‐wave infrared (SWIR) radiances around the 4.3 μm CO 2 absorption band from polar‐orbiting hyperspectral sounders provide useful thermodynamic information for numerical weather prediction (NWP) models. They are not assimilated in any of the operational NWP models because the Non‐Local Thermodynamic Equilibrium (NLTE) effects can increase the brightness temperature by more than 10 K. Directly assimilating NLTE‐affected SWIR radiances is challenging because of two reasons: (1) the radiative transfer model like the Community Radiative Transfer Model (CRTM) underestimates NLTE effects by 0.76 K from the old CRTM NLTE coefficients and by 0.46 K from the new CRTM NLTE coefficients, leading to day/night discrepancies in observation minus background (OMB) bias; and (2) CRTM does not simulate aurora‐related NLTE effects, which can happen during day and night. In this study, methodologies are developed to bias correct CRTM NLTE simulations to minimize the day/night discrepancies in OMB biases and to quality control SWIR radiances that cannot be well simulated by CRTM. The NLTE estimates from the Spectral Correlations to Estimate Non‐local Thermal Equilibrium (SCENTE) method — which exhibit better agreement between observations and background than those from CRTM simulation—were used as a reference to develop the linear regression‐based bias correction scheme. Extensive evaluations were carried out to understand the performance of the bias correction and the quality control schemes. Results showed that the schemes reduced day/night discrepancies in OMB bias to less than 0.1 K for different seasons and 0.14 K for different satellite zenith angles. These small discrepancies open the potential to assimilate daytime and nighttime SWIR radiances simultaneously. In addition, the quality control procedure is effective in screening out SWIR radiances affected by aurora‐related NLTE effects. A larger percentage of nighttime data were filtered out compared to daytime, underscoring the importance of addressing nighttime SWIR radiance assimilation. Lastly, the large OMB biases in high latitudes reported in previous studies are eliminated after the bias correction and quality control.