Numerical Evaluation of the Sensitivities of Three Cumulus Parameterization Schemes: A Case Study of a Severe Rainfall Event over the Central Mongolia

Extreme precipitation events impose substantial impacts on critical infrastructure and natural ecosystems, particularly in semi-arid continental regions such as Mongolia. Accurate simulation of these events in numerical weather prediction (NWP) models is crucial for natural disaster preparedness and climate adaptation. This study evaluates the performance of three cumulus parameterization (CP) schemes Kain Fritsch (KF), Betts Miller Janjic (BMJ), and Grell Freitas (Grell) in simulating a convective rainfall event over central Mongolia on 20–21 June 2020 using the Advanced Research WRF model. A four domain nested configuration (27–1 km) was employed, initialized with the fifth generation ECMWF reanalysis (ERA5) and validated against the integrated multi satellite retrievals for the global precipitation measurement mission (GPM-IMERG) precipitation estimates and rain gauge observations. Results indicate clear scheme dependent differences. The Grell scheme reproduced the intensity, spatial structure, and timing of observed rainfall most accurately, simulating persistent mesoscale convective bands. KF generated strong localized peaks but with fragmented organization, while BMJ substantially underestimated both rainfall magnitude and duration. Vertical hydrometeor profiles further indicated stronger convective development under Grell, with higher rainwater, snow, graupel, and cloud ice mixing ratios compared to KF and BMJ. Skill assessment using the Fraction Skill Score (FSS) confirmed Grell’s superior spatial consistency with observations (0.88), relative to KF (0.86) and BMJ (0.84). These results highlight the critical role of cumulus parameterization in shaping model fidelity under continental climates. Overall, the Grell–Freitas scheme exhibited the most robust performance, though minor biases in intensity and timing remain. Future research should incorporate ensemble approaches and improved physical parameterizations to improve precipitation predictability in data sparse semi-arid regions.