Synoptic analysis and WRF sensitivity experiments of SACZ-driven extreme rainfall over Petropolis, Brazil

Abstract Extreme rainfall in the mountainous region of Petrópolis (Brazil) is frequently associated with South Atlantic Convergence Zone (SACZ) episodes, low-level moisture, and complex terrain. Despite its operational relevance, the sensitivity of numerical forecasts to physical parameterizations under these high-impact environments remains poorly understood. This study evaluates the performance of the Weather Research and Forecasting (WRF) model for five SACZ-driven extreme rainfall events using a 27–9–3 km nested configuration and an ensemble of 64 combinations of microphysics and cumulus schemes. A two-stage verification approach was applied, combining spatial qualitative assessment and quantitative metrics based on mean error, hourly root mean squared error (RMSE), temporal correlation, covariance, and POD–FAR diagnostics. The configurations highlighted in the spatial and temporal rainfall comparisons combined Ferrier microphysics (M5) with Tiedtke (C6) for 01–02 January 2012 and with New Simplified Arakawa–Schubert (C14) for 17–18 March 2013; WRF single-moment 5-class microphysics (M4) with New Tiedtke (C16) was highlighted for 08 March 2014. The 3-km grid generally preserves the shape of cumulative rainfall, but the mean daily bias changes from − 18 mm at 27 km to + 8 mm at 9 km and + 13 mm at 3 km. The present comparisons do not isolate the physical source of the 3-km bias. The findings demonstrate that the evaluated configurations and grid spacing both affect precipitation estimates for these five SACZ-related events. This study highlights key sources of uncertainty in operational forecasting over southeastern Brazil and provides guidance for optimizing physics configurations in similar mountainous, moisture-rich environments.