Sensitivity of the Regional Climate Simulation of the Central American Climate to Cumulus Parameterization

Central America is a region whose climate is influenced by many oceanic, geographic, and atmospheric features within various temporal and spatial scales. The mountain ranges of Central America and the gaps within them allow for the creation or amplification of low-level jets (LLJ) and the modulation of precipitation regimes through orographic lift. In addition, surrounding bodies of water and local evapotranspiration act as ample sources of moisture. Beyond this region, climate modes such as the El Niño Southern Oscillation (ENSO) coexist with these previously mentioned influences. The simultaneous presence of these factors makes the regional climate of Central America particularly challenging for models to resolve. Therefore, it is worth analyzing the differences between simulations of Central American climate between two 16-year (1986-2001) regional coupled ocean-atmosphere climate models, RSM-ROMS-RAS and RSM-ROMS-KF2, which differ in terms of their cumulus parameterization schemes. Among these models, RSM (Regional Spectrum Model) denotes the atmospheric component, ROMS (Regional Ocean Modeling System) denotes the oceanic component, and Relaxed Arakawa (RAS) and Kain-Fritsch version 2 (KF2) are the two cumulus parameterization schemes. Seasonal averages of oceanic and atmospheric variables are compared, highlighting the coupled nature of these models, along with validation using corresponding observations. These variables include mean surface rainfall rate (mm day^-1), precipitable water (kg m^-2), 850 hPa winds (m s^-1), sea-surface temperatures (℃), thermocline depth (m), surface ocean currents, surface latent and sensible heat flux (W m^-2), and the diurnal amplitude of rainfall (mm day^-1) with the time of peak diurnal rainfall (LST). Data from Integrate Multisatellite Retrievals version 7 (IMERGv7; surface rainfall rate and diurnal rainfall variables), the NASA Water Vapor Project (NVAP), the fifth generation European Centre for Medium-Range Weather Forecasts reanalysis (ERA5; 850 hPa winds and surface radiation fluxes), NOAA Optimum Interpolation Sea-Surface Temperatures version 2 (OISSTv2), and Simple Ocean Data Assimilation (SODA) are used for validation of these respective variables. Results reveal significant biases within RAS and KF2 regarding the simulation of seasonal climatological variables over Central America. Comparisons of the mean surface rainfall rate and precipitable water with observations show that both models display the wet season (during boreal summer and fall) and dry season (during boreal winter and spring) characteristic of the region. However, RAS and KF2 contain a wet bias over land across all seasons, with slight improvements in the KF2 scheme. Additionally, KF2 and RAS simulate the horizontal structure of the Caribbean LLJ accurately but underestimate its strength, with similar inaccuracies appearing in other LLJs of the region. KF2 improves upon the cold SST biases of RAS, along with accurately conveying the seasonal growth of the equatorial Pacific cold tongue. However, unlike RAS, KF2 shows warm SST biases within the approximate location of the Inter-Tropical Convergence zone during the boreal summer and autumn. Slight improvements appear within KF2 for simulating thermocline depth in the Caribbean Sea, but the model estimates shallower depths than observations. In terms of surface ocean currents, both models inaccurately simulate the movement and growth of the Costa Rican Dome, keeping it along the coast throughout the year. RAS generally underestimates surface latent heat flux around the locations of the Panama Jet, Caribbean LLJ, and Choco Jet throughout all seasons, which KF2 improves on. However, KF2 overestimates surface latent heat flux within the ITCZ location during the boreal summer and autumn. KF2 also displays improvements regarding the simulation of the seasonal climatology of surface sensible heat flux, with higher values compared to RAS. Lastly, both models struggle to simulate the amplitude and phase of the diurnal cycle of rainfall throughout Central America.