The Spring 2018 data assimilation for the HRRR Ensemble (HRRRE), which is an experimental convective-scale ensemble data-assimilation and forecasting system, did not utilize cloud observations. Preliminary Spring 2018 HRRRE verification statistics indicated that clouds were over forecasted, particularly at mid- and upper-levels, and that the pre- and near thunderstorm environments were too cool at low-levels. Further, a cooler, cloudier environment impacts convective initiation, and potentially explains a lower reflectivity bias for 6-18 hour HRRRE forecasts than for HRRR forecasts. We hypothesize that using the HM-analysis for cloud clearing will improve HRRRE forecasts of the pre- and near storm environments. This presentation will show results from retrospective HRRRE experiments that take advantage of cloud clearing via the HM-Analysis.
Additional cloud observation assimilation development for the HRRR has been underway to further improve low-level cloud retention and cloud prediction overall. To facilitate these needed improvements, a transition of the HM-analysis to a hybrid ensemble-variational framework within GSI is ongoing. The HRRRE can also benefit from this development via the direct EnKF assimilation of cloud observations. Single analysis and retrospective case study experiments examining the impact of direct cloud assimilation in the HRRRE will be presented. The sensitivity to observation errors and the balance between cloud and moisture observation types will be discussed. Qualitative comparisons with satellite observations and pre-storm surface environments will be assessed, as well as storm initiation location and timing.