Cloud Adaptive Additive Inflation for Improved Assimilation of All-Sky ABI Radiances in JEDI-MPAS

Several studies have shown that the ensemble-based data assimilation of all-sky ABI radiances in general can improve convective-scale analyses and forecasts. However, in certain localized situations the ensemble based background error covariance can be grossly deficient, limiting the benefit of the assimilated ABI radiances. For example, all ensemble members may fail to initiate convection in the background forecast where a cloudy radiance is observed, or all ensemble members may have convectively-generated cloud cover where a clear air radiance is observed. In such situations the analysis increment can be very small and physically inconsistent with the observed features. Previous studies have somewhat mitigated this issue in the case of convection initiation through the application of randomly generated posterior additive inflation. The present study aims to better address this issue of localized ensemble covariance deficiency through the implementation of a cloud adaptive approach to prior additive inflation derived specifically for the assimilation of all-sky ABI brightness temperature (BT) observations. A series of cold-start ensemble forecasts on diverse cases were used to characterize the typical covariance structures between ABI BT and model state variables in clear air and in both precipitating and non-precipitating clouds at different altitudes. These cloud-adaptive covariance structures are used in the present case study to optimize the implementation and demonstrate the impacts of prior additive inflation using perturbations with realistic vertical and cross-variable structures for different cloud categories.