El Niño–Southern Oscillation (ENSO) remains the main source of climate variability on the planet, causing droughts, floods, and abnormal heat in different regions. Traditional forecasting models, including complex global climate models and statistical approaches, hit the so-called "spring barrier": their reliability drops sharply when trying to look beyond six months. This work stands among studies that seek precursors of El Niño in observational data, but it differs from neighboring approaches by relying not on ocean surface temperature but on the cooperative behavior of atmospheric temperatures across a vast network of points in the Pacific. Nearby in the archive are works on early signals in climate data and neutrino oscillations, but this study stands out for its long-term validation on real forecasts since 2012.
The authors built a climate network of 14 nodes within the "El Niño basin" (the equatorial Pacific corridor) and 193 nodes outside it. For each pair of nodes, they computed the time-lagged cross-correlation of daily surface air temperature anomalies over the preceding 365 days. Link strength was defined as the excess of the maximum correlation over the mean, normalized by the standard deviation, yielding a signal-to-noise ratio. All links were then averaged into a single index S(t). The alarm threshold Θ was fixed during the training period (1950–1980) and remained unchanged during the forecast phase. When S(t) crossed the threshold from below, and the latest available ONI index was below 0.5°C, the algorithm issued a warning of an El Niño onset in the following calendar year.
From 2012 to October 2019, the algorithm generated 10 alarms, of which 8 were correct, giving a success probability of 80%. It correctly predicted the absence of El Niño in 2012, 2013, and 2017, as well as the occurrence of events in 2014 and 2018. In September 2019, the index S(t) crossed all threshold values, and the latest ONI (JJA 2019) was 0.3°C, below the critical 0.5°C. This points to the return of El Niño in 2020 with a probability of 4 in 5. For comparison, the probability of randomly guessing such a sequence of forecasts from 2012 to 2018 is about 1.5%, and for the entire period from 1982 to 2018, approximately 2.2×10⁻⁵.
The work demonstrates that network analysis of atmospheric temperatures can overcome the "spring barrier" and provide a reliable El Niño forecast nearly a year in advance. Unlike neighboring studies that often rely on statistics of oceanic indices or machine learning, this one uses a physically motivated measure of cooperativity in atmospheric links that requires no parameter tuning in the forecast phase. This is the first test of such an approach over a long period of real forecasts, strengthening confidence in the method as an operational early-warning tool.
The method could be extended to forecast the strength and duration of El Niño, which the current version does not do. The authors point to the prospect of using information entropy to estimate the magnitude of future events. Furthermore, the network approach can be adapted to other climate phenomena, such as La Niña or monsoon systems, where long-range correlations are also important.
Early warning of El Niño would allow agriculture, water management, and emergency response services to prepare for droughts, floods, and extreme heat months before their peak. This would reduce economic damage and human risks in vulnerable regions.
The immediate next step is to wait for the actual development of events in 2020 and compare the forecast with reality, providing another independent test of the method. In the future, they plan to integrate an estimate of El Niño magnitude based on information entropy.
For climatology, the problem of predicting the strength and duration of El Niño remains unsolved, as does understanding the mechanisms that trigger the cooperative regime in the atmosphere. This work contributes to the first part but leaves open the question of the physical causes of the observed synchronization.
An average El Niño raises the global temperature anomaly by about 0.1°C. If a strong event develops in late 2020, 2021 could become a new record year for temperature, since the atmospheric response lags by three months.
The idea of predicting global catastrophes from barely perceptible signs echoes Kim Stanley Robinson's novel "Forty Signs of Rain," where scientists try to convince politicians of an impending climate shift based on early indicators.