Performance of the RegCM5-CP in simulating a rainfall extreme event on the northern coast of São Paulo state, Brazil, in 2023

In February 2023, the northern coast of São Paulo state (Brazil) experienced the highest 24-hour rainfall ever recorded at meteorological stations operated by the National Institute of Meteorology and the National Center for Monitoring and Early Warning of Natural Disasters (CEMADEN). The Barra da Una station (23.758°S, 45.764°W), for instance, recorded 633 mm of rainfall in 24 hours between 18 and 19 February 2023 (from 1200 to 1200 UTC). Rainfall estimates from the Salesópolis radar also indicated precipitation exceeding 450 mm/day along the coast between the municipalities of Bertioga and São Sebastião. On the other hand, satellite estimates, such as CHIRPS, were unable to capture this precipitation event. In this context, the objective of the study is to analyze the performance of the Regional Climate Model version 5 (RegCM5) in convection-permitting (CP) mode in simulating this extreme daily precipitation event on the northern coast of São Paulo state. A simulation with a horizontal resolution of 4 km, nested in the ERA5 reanalysis, was conducted with a domain over southeastern Brazil. This simulation started  on February 17 at 00 UTC. The simulation results were compared with rainfall data measured at CEMADEN stations and precipitation estimates obtained from the Salesópolis radar. The comparisons indicated that RegCM5-CP underestimates precipitation along the coast and misplaces the maximum precipitation. For instance, radar data of accumulated precipitation from 0000 UTC on February 18 to 0000 UTC on February 20, 2023, show the maximum precipitation occurring at the sea-land interface, whereas RegCM5-CP shifts it to the mountainous region inland. Additionally, the model displays other cores with maximum precipitation, such as along the border with Minas Gerais state, which are not observed in the radar data. New experiments are being conducted to test different physical parameterizations to improve the representation of the studied event. The authors thank CNPq, FAPESP and FAPEMIG for the financial support.