VIIRS Radiance Cluster Analysis in CrIS Observations for Enhanced Data Assimilation in NWP Models

Abstract The Cross‐track Infrared Sounder (CrIS) radiance data plays a crucial role in numerical weather prediction (NWP) models by providing essential atmospheric sounding information through data assimilation. However, challenges arise in handling subpixel cloud contamination within CrIS fields of view (FOVs), which can impact the accuracy of radiance simulations. To address this, the Visible Infrared Imaging Radiometer Suite (VIIRS) Radiances Cluster analysis within the CrIS FOVs is developed to characterize subpixel scene homogeneity. This paper describes the algorithms and data processing procedures for this cluster analysis. A fast and accurate collocation method was developed to directly align VIIRS measurements within CrIS FOVs using line‐of‐sight (LOS) pointing vectors. This method supports both terrain‐corrected and non‐terrain‐corrected VIIRS geolocation data sets as inputs. The K‐means clustering method is used to group collocated VIIRS radiance within CrIS FOVs into seven (7) clusters based on their radiance values. The mean, standard deviation, and coverage of each cluster are output for each CrIS FOV. Comparisons with the Infrared Atmospheric Sounding Interferometer cluster analysis demonstrate similar performance, confirming the validity of the CrIS‐VIIRS approach. Data assimilation experiments at the European Centre for Medium‐Range Weather Forecasts indicate that the VIIRS radiance cluster data can be effectively integrated into NWP models, aiding in cloud detection and improving data quality. These findings highlight the potential of CrIS‐VIIRS clustering for enhancing data thinning, quality control, and assimilation of cloudy radiance observations in operational NWP systems.