Experiences with 0-36 hour Explicit Convective Forecasting with the WRF-ARW Model: A Vision of the Future?

During the past three springs and summers, 36 hour realtime forecasts were conducted daily with WRF-ARW using a 4 km horizontal grid resolution and explicit convection over the central and eastern US, using the 00 UTC ETA model for the initial state and boundary conditions. These forecasts were evaluated on a day to day basis by NWS forecasters and researchers across the country, and were also highlighted in the SPC/NSSL 2004, 2005 Spring Program. These simulations showed an uncanny ability at times to forecast the timing and location of significant convective outbreaks, and often accurately distinguished the basic mode of convective system organization (e.g., squall lines, bow echoes, MCVs, supercell lines, ordinary scattered thunderstorms, etc.). Such results offer the community much hope that multi-day explicit convective forecasting could soon be valuable addition to existing forecast guidance. The unexpected success at forecasting a seemingly highly unpredictable phenomena out to 36 h is most attributable to the strong association between significant convective outbreaks and the more predictable meso- and synoptic-scale features. Still, a significant false alarm rate emphasizes the need for further research and model improvements.

In an effort to measure our current progress, these forecasts were compared to guidance offered by the operational 12 km ETA, which employed convective parameterization. Using various subjective measures of forecast quality, evidence exists that there is significant value-added from the high-resolution forecasts as regards forecasts of convective mode and propagation. Significant benefits may also be evident when considering the climatological aspects of convection, where representing convective processes ``explicitly'' seems to be critical to properly representing the diurnal cycle and convective episodes. However, value-added is less obvious when considering more general guidance as to timing, location and overall intensity of convective outbreaks.

This presentation will go on to highlight current known limitations as regards existing physics packages (i.e., pbl, microphysics, etc.) when applied at such high resolutions, as well as the need for a better representation of the initial state at the mesoscale, including, among other factors, a better representation of atmospheric moisture and the inclusion of ongoing convection.