Attribution of flood event: a case study of the April/May 2024 floods in Southern Brazil

The flood event of April/May 2024 that hit the Southernmost State of Brazil, Rio Grande do Sul, broke local records, with rivers reaching their highest level in recorded history. Around 2.4 million people are estimated to have been affected by the flood, with hundreds of thousands displaced and/or without access to potable water and electricity in their homes. Extreme precipitation, linked to a negative surface pressure anomaly, as suggested in the ERA5 dataset, was the primary driver. 

While attributing extreme precipitation events is an established practice, less work has been done to directly attribute river flood events to climate change, often due to a lack of long-term data in the region of interest. This study explores the feasibility of employing the existing attribution framework for attributing extreme discharge events.

We analyzed daily precipitation and river discharge data from over 40 stations (1960–2023, <10% missing) using two approaches. First, a "factual" distribution was developed using all the data available. A “counterfactual” distribution was obtained by fitting a distribution with the global mean surface temperature as covariate and a constant dispersion parameter, and then deriving the distribution assuming a 1.2°C cooler world.  Second, the data was divided into two separate periods: 1960-1991 (“past”) and 1992-2023 (“present”).  In both cases, differences in extreme values between these distributions were statistically assessed. Additionally, the surface pressure anomaly in ERA5 was used for analog attribution study, to assess the significance of the changes in surface pressure, precipitation, temperature, and discharge fields, between the “past” and “present” time periods.

Hydrological simulations performed with the CETEMPS Hydrological Model (CHyM) coupled with the CORDEX (Coordinated Downscaling Experiment)-CORE models output, both for a historical period and under the rcp85 scenario, were also used.  The model validation done for the historical period, comparing CHyM outputs against discharge station data, showed quite good agreement between the two for several statistics. Both the hydrological simulations and the regional climate CORDEX-CORE simulations were used in the analog attribution study to confirm the attribution of the event to global warming. This analysis investigates the potential of integrating hydrological modeling and observational discharge data to advance the attribution of extreme flood events to climate change.