Evaluation of the High-Resolution Rapid Refresh Ensemble (HRRRE) Skill and Variance for Events with Strong Convective Forcing versus Weak Convective Forcing

The High Resolution Rapid Refresh (HRRR) is an hourly-updated 3-km WRF-ARW Convective Allowing Model (CAM) developed at the Earth Systems Research Laboratory (ESRL) Global Systems Division (GSD) and run operationally at the National Center for Environmental Prediction (NCEP).  The current experimental HRRR provides deterministic guidance for 0 to 36 hours, and these forecasts can aid in the production of severe weather outlooks and watches.  The future of CAM systems like the HRRR will be to produce real-time ensemble forecasts, in order to provide improved skill and uncertainty information for forecast applications.  A prototype HRRR Ensemble (HRRRE) was run experimentally in real-time during the Spring of 2016. 

The success of ensemble data assimilation at CAM-scales is limited by a number of challenges, including the difficulties associated with creating realistic ensemble perturbations and maintaining sufficient ensemble variance.  The real-time experimental HRRRE was underdispersive, which is a common problem with CAM ensembles.  To further our understanding of the relationship between ensemble skill and variance, the HRRRE performance will be compared for cases with strong convective forcing and weak convective forcing.  Ensemble and deterministic verification of reflectivity and surface observations will be presented for several strongly and weakly forced events.  Results will aid the improvement of the future HRRRE ensemble design.