Severe storm forecast guidance based on explicit identification of convective phenomena in WRF-model forecasts

The Storm Prediction Center and National Severe Storms Laboratory have collaborated on the development and implementation of high-resolution, deep convection allowing WRF model guidance since 2003. During this time the utility of these high-resolution models as guidance in severe weather forecasting has been explored. This exploration has consisted of formal evaluation efforts such as those associated with the Hazardous Weather Testbed (HWT) annual Spring Experiments, and almost daily informal exchanges based on quasi-operational real-time guidance.

Following several years of real-time, convection allowing deterministic guidance, the natural evolution of high-resolution modeling was toward a storm-scale ensemble forecast (SSEF) system. To this end, a multi-year HWT Spring Experiment was designed (in collaboration with the University of Oklahoma's Center for Analysis and Prediction of Storms, the NCEP Environmental Modeling Center, and the National Center for Atmospheric Research) to build and evaluate a WRF-based, ten member SSEF system employing a grid length of 4 km and covering a domain over the eastern three-fourths of the CONUS. Because these numerical forecasts attempt to predict convection explicitly, attributes of convective structure and mode can be gleaned directly from the output. During the 2008 Spring Experiment, the “severe” convective phenomena were identified in explicit SSEF output. The identification was based on updraft helicity (i.e., model generated supercells), lowest model level wind speeds associated with convection, and the automated detection of linear or bowing line segments. These severe storm proxies were then combined statistically using a resampling approach to produce coverage forecasts of severe weather occurrence directly from the SSEF. Although still very early and rudimentary in its development, the direct extraction of severe weather proxies from the SSEF proved surprisingly useful on a number of days in delineating the area and magnitude of the severe weather hazard. Of particular significance is the concept of identifying and using simulated convective phenomena to create a severe weather forecast, compared to the traditional forecasting approach of assessing characteristics of the mesoscale environment to determine the spectrum of convective storms types that are possible.

The concept was applied initially to the SSEF, but it can be potentially applied to deterministic convection-allowing WRF-model output as well. Details on the identification of convective phenomena and their subsequent statistical treatment will be provided. Examples from the 2008 Spring Experiment will be shown, along with verification of the probabilistic guidance over the entire six week experiment. Ideas for ongoing and future work will also be offered.