A comparison of cumulus parameterization schemes in the WRF model

Simulations using the WRF model provide insight into the effectiveness of cumulus parameterization schemes in representing isolated convection at various grid spacings. Idealized modeling demonstrates that, at a fine resolutions such as 2 or 4 km, classic supercellular convective features including a reflectivity hook, mid-level rotation and storm splitting can be resolved explicitly and a CPS is not needed. The Kain-Fritsch and Betts-Miller-Janjic schemes also were able to trigger convection at 4 km, however the resulting vertical velocities and rain mixing ratios were smaller in magnitude than the no-CPS simulation. The BMJ simulation especially had difficulty representing vertical velocities accurately. This could be due to the early release of latent heat and the resulting struggle to regain values similar to the other simulations. Another possibility is the lack of low-level convergence to support the continued growth of the initial and split updraft. The Grell-Devenyi ensemble scheme is currently unable to handle idealized convection at small grid spacings. Possible modification to the code could alleviate this problem but this was not undertaken for this study. These simulations demonstrated that caution must be used when employing a CPS. There are two main conclusions resulting from the simulations for the 28-29 July 2005 case study. First, it is not always appropriate to assume that model simulations using small grid spacings (< 5 km) will not need a CPS. Depending on the strength of the synoptic-scale forcing and time of year, certain atmospheric settings warrant the use of a cumulus parameterization scheme to represent the effects of sub-grid scale convective processes. The convective system on 28-29 July 2005 is an example of an environment that needs a CPS to represent accurately the small sub-grid scale processes that occur with isolated, discrete convective cells. The atmospheric setting of this case made representation of precipitation challenging and it was shown that the model could not represent, even with a complicated microphysics package, the convection explicitly. Secondly, a strong limitation of the Betts-Miller-Janjic scheme is its strong dependence on available moisture. While Janjic (1994) made improvements to the original Betts-Miller scheme to make it more applicable to non-tropical convection, its simple representation of cloud processes still apparently limit in its effectiveness in certain environments.