Development of operational high resolution ensemble prediction systems has been indentified in NOAA and NWS strategic plans as a key contributor to future improvements in high impact weather forecasting. During the 2012 SFE, the severe storm forecasting component examined the ability of multiple storm scale ensemble systems to provide useful guidance for the issuance of experimental, high temporal resolution probabilistic severe storm forecasts. The ensemble systems were provided by: 1) OU/CAPS which ran a 26 member multi-model, multi-physics, multi-initial condition 4 km storm scale ensemble forecast (SSEF) that included advanced physics schemes and 3DVAR cloud analysis/radar data assimilation, 2) the SPC, which processed a 7 member multi-model storm scale ensemble of opportunity (SSEO) consisting of available operational and experimental cold start 4-5 km deterministic convection-allowing models, and 3) the Air Force Weather Agency (AFWA), which provided a 10 member multi-physics, multi-initial condition 4 km WRF-ARW ensemble system that was initialized from downscaled forecasts of different global models without explicit data assimilation. Each ensemble was initialized daily at 00z and produced forecasts through 36 hrs over a CONUS domain. Specialized convective storm hourly maximum fields (HMFs), including simulated reflectivity, updraft speed, updraft helicity, and 10-m surface winds, were used to characterize model-generated storm intensity. The predicted convective storm properties were processed and viewed using a variety of ensemble display techniques, including spaghetti charts, exceedance probabilities, and maximum HMF from any member that were time matched to the experimental forecast periods.
Although all were run at similar horizontal resolution, the ensembles were configured quite differently in terms of data assimilation, source of initial conditions and perturbations, physics diversity, and the number of members. This provided an opportunity to examine comparative performance given the varying degrees of complexity in each system. Evaluation results using subjective and objective measures will be presented. The findings will offer insights on strengths and weaknesses related to ensemble design and configuration, and include an assessment of the minimum number of high resolution ensemble members needed to provide operationally useful guidance to forecasters. The latter is particularly important given current limits to high performance computing resources used to support forecast operations.