Next steps in automated thunderstorm nowcasting: improving performance and forecasting storm severity

The Automated Thunderstorm Nowcasting System (Autonowcaster) developed at NCAR has been running operationally at the NWS Washington D.C./Baltimore, Maryland Weather Forecast Office the past four summers. The Autonowcaster is an expert system that incorporates radar, satellite, METAR, sounding and numerical cloud model data using fuzzy logic to produce time and place specific forecasts. Selected cases from the summers of 1999 and 2000 are presented to illustrate the performance of this system for forecasting the initiation, growth and decay of thunderstorms, with special emphasis on the contribution of satellite information toward improving the forecasts and lead time for thunderstorm initiation.

As a result of running the Autonowcaster system at a WFO office, it was clear that forecasters need to know not only where new storms will form, but which forecast storms will become severe within the next 15-60 min; i.e., storms that produce hail, tornadoes, destructive winds and flooding. The second focus of this study is to explore methodologies for producing nowcasts of severe thunderstorms. Some of the interest fields currently produced by the Autonowcaster have information that directly relates to the strength or intensities of existing storms and to the environments in which storms will form or propagate in to at some later time. For example, high resolution horizontal and vertical velocity output from the numerical cloud model provide important information on boundary layer convergence fields and low-level shear relative to convergence lines. These fields can be used to address storm longevity and severity. This study will also explore using output from the NSSL Warning Decision Support System (WDSS) as 1) interest fields within the Autonowcaster to address storm longevity and severity and 2) to use as a verification source for storm severity forecasts.