El Niño: Why climate scientists disagree on 'very powerful' or 'unprecedented'

Over half of climate scientists classify the current El Niño as 'very powerful', while a significant minority describe it as 'unprecedented'. This division is not merely semantic; it reflects deep methodological differences in interpreting oceanic and atmospheric data. Without consensus, weather forecasts, simulation models, and political decisions face increased uncertainty, making risk communication more complex for the public.

What climate scientists say: criteria for power and exception

The term 'very powerful' traditionally relies on the Niño3.4 index, measuring the surface temperature anomaly (STA) in a central equatorial Pacific zone. An excess of +2.5°C above the climatological mean typically qualifies an El Niño event as 'strong'. Some researchers, using Copernicus satellite archives and in-situ Argo fleet data, find that the current STA has consistently exceeded this threshold for several months, earning it the 'very powerful' label. Others, integrating additional variables such as thermocline depth, surface wind intensity, and anomaly persistence, argue this event is 'unprecedented' due to its exceptional thermal amplitude, duration, and geographical extent since reliable records began.

This difference in criteria is also reflected in the databases used. The National Center for Atmospheric Research (NCAR) and the NOAA's El Niño-Southern Oscillation (ENSO) program prefer continuous time series since the 1950s, while the European Centre for Medium-Range Weather Forecasts (ECMWF) integrates more model-derived measurements, such as ocean-atmosphere heat flux calculated by neural networks. Depending on the dataset chosen, the same event may appear under two distinct aspects, contributing to the polarization of opinions.

How AI models evaluate El Niño: neural networks and satellite data

Modern forecasts rely on machine learning models capable of assimilating millions of data points in real-time. GraphCast, developed by the Met Office, and Pangu-Weather, from Hong Kong University's AI lab, use graph neural network architectures to connect surface temperature observations, zonal winds, and heat flux to predictions up to 30 days ahead. These models are fed by Copernicus satellite data, Argo fleet measurements, and ECMWF re-analyses, offering unprecedented spatial resolution that reveals fine structures of thermal anomalies.