Since NCEP'?s Storm Prediction Center moved into the National Severe Storms Laboratory facility in 1996, groups of research scientists and forecasters from each institution have conducted numerous collaborative research programs. In the spring of 2000, one of the largest elements of the program was an assessment of numerical guidance for convective initiation and evolution. In this paper, we focus on a comparison of the operational Eta model with an experimental version that is run at the NSSL/SPC facility. The experimental version utilizes the Kain-Fritsch convective parameterization in place of the operational Betts-Miller-Janjic scheme, and it also uses a higher order diffusion formulation, which introduces less smoothing of mesoscale structures than the operational diffusion parameterization.

Averaged over the 6-week period of the experiment, the two versions of the Eta model produced very similar scores on traditional precipitation verification statistics, i.e., equitable-threat and bias scores. However, individual forecasts were often very different. Both versions of the Eta model had significant advantages and disadvantages compared to the other. In this paper, detailed elements of selected model forecasts are compared, contrasted, and explained as a function of the different parameterization schemes. Recommendations for combining the strengths of the two schemes and ameliorating the weaknesses are provided, with the goal of improving convective parameterization techniques in mesoscale models.