Examination of several different versions of the WRF model for the prediction of severe convective weather: The SPC/NSSL Spring Program 2004

In recent years, it has become especially evident that the type of severe convective weather that occurs (tornadoes, hail, or damaging winds) is often closely related to the convective mode (or morphology) that storms exhibit, such as forming in discrete cells, squall lines (or quasi-linear convective systems (QLCS)), and multicellular convective systems. In addition, some severe storms develop as dynamically unique classes of thunderstorms such as supercells and bow echoes, which are believed to produce a disproportionate number of tornado and widespread straight-line wind damage events, respectively. Thus, accurate severe weather forecasts are dependent on forecasters being able to properly predict not only where and when severe thunderstorms will develop and how they will evolve over the next 4 - 7 hours, but also the convective mode(s) that are most likely to occur.

The 2004 SPC/NSSL Spring Program is a multi-agency collaborative exercise conducted from April 19 through June 4 that explored the utility of the Weather Research and Forecasting (WRF) model for severe weather forecasting, with a primary goal of comparing three versions of very high-resolution WRF forecasts (4 km grid spacing and explicit precipitation physics) with mesoscale versions of the WRF using ~10 km grid spacing and parameterized convection. The high resolution WRF model versions consisted of: 1) WRF Mass Core run at NCAR initialized with 40 km Eta model data, 2) the WRF Mass Core run through University of Oklahoma/CAPS with a state-of-the art data assimilation system (ADAS) including Level II radar data, and 3) WRF NMM core run at NCEP/EMC initialized with 40 km Eta model data. These three versions allow a meaningful comparison of the two basic physics cores plus an examination of the impact of a dynamic data assimilation system on the model predictions.

The examination of model utility is made from the perspective of SPC severe weather forecasters, who are interested in the ability of these models to provide useful guidance about convection initiation, intensity, mode, and evolution. An experimental forecasting component included the formulation of two short-range forecasts of severe convection valid during the afternoon and evening. The preliminary forecast utilized guidance from current operational models (Eta, RUC, and NCEP SREF) to provide a control benchmark of the current state of convective forecasting; a final forecast, incorporating additional guidance from the 0000 UTC WRF model runs, followed shortly thereafter. The primary motivation for this forecasting exercise was to determine: 1) if there is new and useful information in the very high resolution WRF runs from a forecaster perspective, and 2) whether severe weather forecasts can be improved when forecasters have access to new near-stormscale models using explicit precipitation physics, compared to mesoscale models with parameterized convection. In other words, is the value-to-cost ratio high enough to justify the enormous computer and communications resources required to produce model guidance at 4 km grid spacing in an operational setting?

A next-day subjective evaluation of all high-resolution WRF runs and current operational models is also conducted. This process allowed us to document thoroughly 1) the specific elements of these forecasts that provide additional and valuable information for forecasters, 2) the sensitivity of these forecasts to a) grid resolution and b) initialization data and procedures. We anticipate that these results will provide valuable information for NWS decision makers and numerical modelers as they assess how to best utilize growing computer resources in coming years.

Results from the Spring Program 2004 will be presented at the conference.