Deriving Hourly Severe Weather Probabilities from Convection Allowing Ensembles

As part of the 2014 NOAA Hazardous Weather Testbed Spring Forecasting Experiment (SFE), convection-allowing ensembles were used to generate first-guess fields for hourly total severe weather (i.e., combined tornado, hail, wind) probabilities. The goal of this experiment was to begin thinking critically about providing high temporal resolution forecasts, particularly the notion of forecasting events as opposed to environments favorable for severe weather occurrence.

We will review the probabilities generated by forecasters and models during the 2014 SFE to highlight the inherent difficulties creating event-driven probabilities at various time/space scales . Given this new approach, there is a need for reliable guidance, but challenges remain. Existing neighborhood approaches account for location errors via spatial smoothing, but these smoothers have primarily been developed for 24-hour probabilities. Therefore, we examine multiple spatial smoothers and using multiple proxy-severe variables. In addition, higher temporal resolution would require a time-based neighborhood approach. In order to examine how to construct a temporal smoother, we use a long-term climatology from a convection-allowing ensemble to inform our options.