Abstract
This paper presents the mesoscale part of a two-part evaluation of thirty forecasts produced by a mesoscalenumerical weather prediction model (MASS 2.0). The general approach taken to evaluate the mesoscale predictivecapabilities of the model is to utilize observed patterns of convection as “verification” of the unfiltered forecastfields. More specifically, these fields are combined into convective predictor fields, the loci of which are thenrelated at two hourly intervals to the loci of strong mesoscale convective systems (MCSs) identifiable in nationalradar summary plots and GOES satellite imagery. Results show that the genesis of 48% of the 149 observed MCSs could be accurately (−+3 h/250 kin) relatedto coherent predictor fields with a very low false alarm rate of 13%. Convection “underforecasts” (or “misses”)were related in 67% of the instances to systematic forecast errors at the synoptic scale, many of which arediscussed in detail in Part I. This suggests that a necessary, but insufficient, condition for accurate forecasts ofmesoscale phenomena is accurate initialization and temporal integration of the larger-scale circulation patterns. Four cases are selected from the sample as demonstrations of the degree of coherent, detailed informationprevalent in the model forecasts of vertical motion and convective instability fields in a variety of convectivesituations. Examples of model “forecasts” of intense convective storm clusters, a severe squall line triggeredalong a dryline, orographically induced hailstorms, and sea breeze thunderstorms are provided. It is concludedthat the model can be used to gain insight into mesoscale convective processes in situations where synopticscale forecast errors have minimal impact.
