An Operational Perspective on Convection-Allowing Models and Ensembles in Severe Weather Forecasting

Real-time convection allowing model (CAM) guidance has been utilized in operational severe weather forecasting for more than a decade.  Testing and evaluations during field programs such as BAMEX (2003) and annual Hazardous Weather Testbed (HWT) Spring Forecasting Experiments (SFE) since 2004 have played key roles in this transformative process.  CAM output complements traditional larger-scale environment-based forecast approaches by providing unique information on convective initiation, mode, intensity, and evolution that can relate more directly to severe weather potential and hazard type.  Our experience has indicated that CAM forecast solutions are sensitive to model configuration (e.g., dynamic core and physics) as well as initial/lateral boundary condition specifications.  And when multiple CAM solutions became available to forecasters, differing convective storm forecasts were often realized, especially in weakly forced situations.  In these situations many of the forecasts could appear as plausible solutions within the specific mesoscale environment.   These varying solutions reflect limits to predictability on the grid scale, which prompted efforts to develop storm-scale ensemble prediction systems to quantify uncertainty and provide a range of possible convective storm solutions, starting with the first CAM ensemble system developed by OU/CAPS for the 2007 HWT SFE.

Continued development efforts led to improvements in deterministic CAMs over a CONUS domain (e.g., NSSL-WRF, ESRL HRRR, EMC NAM Nest) including the generation of specialized output fields for explicit simulated storm attributes and hourly maximum history variables (e.g., reflectivity, updraft speed, updraft helicity, surface wind, hail size).  Focused exploration of experimental CAM ensembles also expanded across the community with CAPS, SPC, NSSL, NCAR, and the USAF, among others, developing systems with varying configurations and levels of complexity.  In early 2011 the SPC Storm Scale Ensemble of Opportunity became the first real-time CAM ensemble produced year-round, providing operational forecasters with guidance for severe storms, heavy precipitation, fire weather, and winter weather.  Innovative post-processed statistical guidance including probabilistic storm attribute fields is now available from many experimental CAM ensembles. 

Although CAM ensembles enable quantification of severe weather risks including generation of calibrated probabilistic fields, current operational practice at the SPC indicates that CAM ensembles are used to supplement (not replace) deterministic CAM output for convective storm forecasting.  This is based partially on the long-standing relationship between observed radar-based storm reflectivity/dynamic structure and real-time severe weather threat assessment that has been subsequently applied to the interpretation of CAM simulated storm predictions.  This approach is also analogous to the complementary use of deterministic regional/global models and larger scale ensemble systems, where a combination of detailed pattern/environment information from deterministic models and ensemble statistical output provide a broad range of information utilized by severe weather forecasters. 

Implications for future CAM development and post-processing strategies to better meet the needs of operational severe weather forecasters are discussed.