Record-breaking rainfall and the upcoming El Niño: Research update
A UCLA study shows that the record-breaking rainfall of 2024 was partly due to climate warming, with 40% of the increase attributed to human activity. Researchers warn of conditions favorable for a powerful new El Niño, with significant implications for global weather forecasting.
In 2024, extreme rainfall reached record levels, triggering major flooding across several continents. According to researchers at the University of California, Los Angeles, 40% of this increase is directly linked to climate warming, and current atmospheric conditions suggest the reappearance of an El Niño of comparable intensity.
What the study reveals: Precise results and key figures
The analysis by UCLA's team, recently published, shows that the 2024 rainfall event surpassed all previous records in terms of water volume measured at Copernicus network tracking stations. Although the exact figure is not communicated, the authors emphasize that the anomaly exceeds historical thresholds by 20% in tropical and subtropical zones. The report also indicates that climate models, including the ECMWF system, attribute 40% of this increase to the rise in global water vapor linked to warming, while the remaining 60% comes from natural climate variability.
Another crucial point: strong El Niño events – those exceeding the ONI index of +2.5°C – have been observed only four times in the past six decades. The last one, in 2023-2024, therefore set a worrying precedent because oceanic conditions (surface temperatures exceeding the average by 0.8°C) are already present in the central Pacific.
How: Research method and data used
To quantify the impact of warming, scientists combined several data sources: satellite observations from the Sentinel-5P mission, atmospheric analysis products from Copernicus, and outputs from the predictive model GraphCast, a neural network trained on over 40 years of atmospheric data. In parallel, they used the Pangu-Weather model, which integrates machine learning to refine short-term precipitation projections.
Researchers first calibrated precipitation observations with ECMWF model simulations to reduce prediction uncertainty due to measurement biases. They then applied a causal attribution technique based on counterfactual experiments, where the same climate scenario is simulated without increased greenhouse gases. The difference between the two simulations allowed isolating the anthropogenic warming share – hence the 40% figure.
Finally, the team used convolutional neural networks to identify spatio-temporal patterns associated with strong convection episodes, reinforcing the robustness of their conclusions. All these approaches were validated through comparisons with historical data archives, ensuring a reduced margin of error despite the phenomenon's complexity.
