The Impact of Cumulus Parameterization on Regional Climate Simulations of Central American Climate

In this study we contrast and verify two coupled ocean-atmosphere regional climate simulations (RR-RAS and RR-KF) spanning 15 years, at 15 km grid spacing, and centered over Central America that differ only in their cumulus parameterization schemes. The Relaxed Arakawa Schubert (RAS) scheme is based on quasi-equilibrium theory and Kain-Fritsch (KF) scheme is based on an entraining/detraining plume model that triggers convection at a threshold CAPE. The model simulations are verified for their seasonal, sub-seasonal, and diurnal variations using reanalysis, and other independent observations. The RR-KF scheme shows several notable improvements in the simulation relative to the RR-RAS scheme. These include a reduction of the terrestrial wet bias, improved representation of the ITCZ location in the eastern Pacific, reduction of the dry bias of the atmospheric column over the oceans, reduction of the westerly bias of the low-level trade winds, and improvement of the cold bias of the coastal SST. Despite these improvements, the bias in RR-KF simulations is still large and significant. It is encouraging to note that both simulations produced diurnal variability of rainfall, with its amplitude and phase verified reasonably well with corresponding observations. The simulation of the mid-summer drought was comparable between the two simulations, with a verifiable gradient of its amplitude between the Pacific and the Caribbean coasts but with an underestimation of its frequency.