Abstract
The performance of three disparate cumulus parameterization schemes (Arakawa-Schubert, Kreitzberg-Perkey, and Kuo) was tested semi-prognostically in the mid-latitude environment of severe convective storms. With the aid of the analysis package of a mesoscale model, three independent rawin data sets were analyzed on different grid sizes (120 km, 45 km) to provide the larger scale variables and the verifying heat and moisture budgets. The terminology of dynamic control, static control, and feedback was used to separate different assumptions used by various cumulus parameterization schemes. These assumptions were then tested rigorously to find systematic errors and show possible improvements. For the Arakawa-Schubert parameterization heating and drying in the lower troposphere were systematically overpredicted. This problem was fixed by adding a simple parameterization of moist convective-scale downdrafts to the dynamic control, static control and feedback of the scheme. For the Kuo-type parameterizations, assumptions used by the feedback led to serious errors in predicting the vertical heating distribution. For the Kreitzberg-Perkey scheme assumptions used by the dynamic control led to an erroneous reproduction of rainfall patterns. Furthermore, assumptions used by the feedback of this scheme led to too much moistening and cooling in the middle and upper troposphere. For all schemes, it was shown that moist convective scale downdrafts, which play a vital role in the mid-latitude environment of severe convective storms, should probably be a part of the feedback in any cumulus parameterization scheme.