Attributing the 2023 São Sebastião Record Rainfall Using a Forecast-Based Framework

Extreme weather events are becoming more frequent and intense worldwide because of anthropogenic climate change, primarily driven by the increase in greenhouse gas concentrations.  However, these changes are not spatially uniform: regional atmospheric dynamics, physiographic characteristics, and climatological regimes strongly influence the rates and patterns of change in extreme precipitation. In this context, the present study investigates the anthropogenic influence on the extreme rainfall event in São Sebastião, São Paulo (Brazil) on 18–19 February 2023. It was the largest rainfall event recorded in 24 hours in the modern history of Brazil, with more than 600 mm of rain in less than a day, according to rain gauges nearby. The event triggered widespread flooding, landslides, and debris flows, which led to severe socio-economic losses, including 65 fatalities and damages exceeding USD 120 million. Therefore, to better understand how anthropogenic influence has changed the magnitude and likelihood of this event, Extreme Event Attribution (EEA) provides a scientific framework for quantifying these changes.  We use an innovative forecast-based attribution methodology using ensemble forecasts from the Integrated Forecasting System (IFS) of the ECMWF. Unlike traditional probabilistic attribution approaches—which evaluate changes in probability across broad classes of events—this method allows the analysis of the event itself, isolating the contribution of anthropogenic forcing while preserving the atmospheric dynamics that shaped the storm. This approach also incorporates a more traditional risk-based analysis using the HadGEM3-A climate model. The results show that the rainfall responsible for triggering the landslides was accurately predicted three days in advance by the ECMWF forecast system, demonstrating strong skill in anticipating the event’s magnitude and associated impacts. Considering the debris-flow and mudflow hazard threshold (>220 mm), both the Historical (40%) and Increase CO₂ (42%) scenarios show substantially higher exceedance probabilities compared to the Pre-Industrial (22%). When comparing the forecast-based and probabilistic approaches the largest absolute increase in precipitation occurs between Increase CO₂ (600 ppm) and Pre-Industrial (285 ppm), with approximately 57 mm (95% CI: 16.84, 97.09) of additional rainfall. The Historical – Pre-Industrial scenario results in an increase of about 46 mm (95% CI: 12.47, 79.10), while in the probabilistic approach, the anthropogenic forcing signal is estimated at approximately 50 mm (95% CI: 34.11, 65.87) and a narrower distribution of the bootstrap ensembles. Overall, the results demonstrate that human-induced climate change intensified the São Sebastião extreme rainfall event, increasing its magnitude and amplifying the associated impacts. The study also shows that results across different methods are similar, which suggests that the changes are consistent.