Abstract In this study, we assimilated all‐sky infrared radiance (ASR) data from geostationary satellites over global regions. The impact of assimilating radiance data from GOES‐16, Meteosat Second Generation‐1, ‐4 and Himawari‐8 under all‐sky conditions was compared with the impact under clear‐sky conditions. This study builds upon a previous study by Okamoto et al. (2023), who developed cloud‐effect parameters for observation error modelling, bias correction, and quality control procedures. In this study, three key improvements were made to extend the applicability of their method to other satellites. The first improvement considers the difference in cloud effects between the model and observations, which is used to inflate observation errors and discard data with large discrepancies. The second method helps prevent inappropriate analysis increments in low‐cloud regions over oceans due to the model's inability to accurately reproduce such clouds. We discarded data in such situations with a stricter threshold of the cloud‐effect difference to preserve the analysis accuracy. The third improvement addresses the increased impact of diurnal variation in model bias due to the increasing usage of geostationary satellite data by adding the solar zenith angle as a predictor for bias correction. These enhancements enabled the assimilation of ASR data to achieve a positive impact that surpasses that of clear‐sky radiance assimilation in the tropics while mitigating regional deterioration, albeit not entirely eliminating it.
Okabe et al. (Sun,) studied this question.