An Examination of Convective Parameterizations Using a Cloud-Resolving Model

Abstract

This present thesis evaluates various convective parameterizations (CPs) available in the weather research and forecasting (WRF) model as single column model (SCM) version with cloud-resolving model (CRM) as a reference model. Method of evaluation follows a conventional way to compare SCM and CRM with intensive field data. The field data is the South Great Plain during 1997 (SGP97) which includes complex process of precipitation in the large-scale model and has difficulties of representing observational signals. To control cloud and radiation in convections accurately, the thesis set up the same microphysics (MP) and radiation physics (RAD) from CRM, as well as all the simulations are run by nudging the same large-scale forcing and latent/sensible fluxes into SCM and CRM. CRM simulates reasonable performances of precipitation, while all CPs shows diverse patterns of ones mainly due to deficiencies of trigger function and closure assumption in CPs. In the part of precipitation, Peirce skill score depends on not only hit rate, also false alarm rate in models. CPs are able to divide convection-type (KF, SAS-type, ZM, BMJ, Grell-type) and grid-tyep (OKF and TIED) runs for how percentages of convective rainfall of CPs simulate. Concerning of precipitation intensity, larger percentage of simulating light precipitation in CPs tends to produce smaller percentage of strong one. In the views of vertical distributions, CPs overestimate condensation and radiation rate from the apparent heat source in the convection of upper-level. This is because CPs are more humid and cloudy than CRM. Grid-typed runs overly simulate cloud ices and rain water because of larger clouds, while convection-typed runs tend to underestimate them. Graupel and cloud water are under- and overestimates, comparing with CRM, respectively. The thing is that the precipitation intensity is not determined by the amount of hydrometeor. Net cloud radiative forcing (CRF) also shows the diversity of CPs, and grid-typed runs are higher values, driven by larger amount of clouds. Longwave (LW) and shortwave (SW) CRF of CPs are normally overestimated caused by the immense clouds. Cloud-radiation processes in a given precipitation are still unbalanced in CPs because of the deficiencies from the assumptions. In the case of diurnal precipitation, peaks on nocturnal time are improperly simulated in almost CPs, whereas only SAS-typed CPs well-reproduce the peaks because of their convective inhibition from trigger function. There are still lots of false alarms of precipitation in CPs on daytime due to surface fluxes. To minimize uncertainties from the deficiencies in CPs, multi-convective ensemble (MCE) is one of promising methods.