Assessment of the sensitivity of the RegCM5 model to physical parameterization schemes in precipitation simulation in the Bandama watershed (Côte d’Ivoire)
Abstract
This study evaluates how the choice of physical parameterization schemes affects precipitation simulation over the Bandama watershed (Côte d’Ivoire) using the fifth-generation Regional Climate Model (RegCM5). Sixteen configurations were built by combining four convection schemes (Emanuel, Kain–Fritsch, Grell, and Tiedtke), two cloud-microphysics schemes (the Subgrid Explicit moisture scheme, SUBEX, and the Nogherotto–Tompkins scheme, NoTo), and two planetary boundary layer (PBL) schemes (Holtslag and the University of Washington scheme, UW), driven by the ERA5 reanalysis over 2001–2009; thirteen simulations completed successfully and were analysed. Simulated precipitation was evaluated against two independent satellite products, CHIRPS and IMERG, using continuous and categorical metrics of bias, temporal correlation, and distribution- and event-based skill. The choice of convection scheme had the largest impact. The Tiedtke and Emanuel schemes, especially when paired with the UW boundary layer, performed best, reproducing the main phases of the seasonal rainfall cycle with moderate annual biases (generally below 30%). In contrast, the Grell scheme (in all pairings) and Kain–Fritsch without UW produced large overestimates (annual biases exceeding + 440% for Grell) and should not be used without prior calibration or validation. The boundary-layer scheme was also influential, UW outperforming Holtslag, whereas microphysics had a secondary, non-systematic effect. All configurations nonetheless shared a systematic drizzle excess, a generic limitation of parameterized convection at this resolution.