Simulation of an anomalously high wind gust event in São Paulo: A comparison of the MPAS-A and WRF models

Abstract Local atmospheric processes are significantly influenced during the passage of tropical and extratropical cyclones, which rapidly modify local weather conditions. Consequently, there is a growing need to better understand the physical behavior of mesoscale weather models, especially considering the evolution of forecasting systems from WRF (quadrilateral grid) to MPAS-A (hexagonal mesh). In this context, this study evaluates the performance of the WRF and MPAS-A atmospheric models to simulate a storm event that occurred in October 2024 in the state of São Paulo, Brazil, associated with the influence of an extratropical cyclone. To achieve this, high-resolution simulations (3 km) were conducted and compared with observational data from INMET weather stations and ERA5 dataset. The results show that both models reasonably reproduce the temporal evolution of wind speed. MPAS-A showed better spatial agreement with ERA5 in some parts of the domain; however, the present work does not allow this behavior to be attributed to a single factor, such as mesh structure alone. Both models captured wind speed peaks exceeding 25 m/s at 850 hPa, but showed low temporal correlation with the observed data, indicating difficulties in representing the hourly variability of the event. In 2-m air temperature simulations, WRF tended to overestimate and MPAS-A to underestimate, especially during daytime periods; however, both models successfully reproduce the diurnal temperature cycle. The results highlight the importance of advances in physical parameterizations, expansion of the observational network in the Southern Hemisphere, and the incorporation of complementary techniques, such as data assimilation and post-processing, to enhance the representation of severe weather events in mesoscale models.