El Niño and the Pacific signal that could help forecast some episodes up to a year in advance

Three climate models anticipated the transition from La Niña to El Niño with more than 70% accuracy at twelve months. (Illustrative Image Infobae)El Niño could be anticipated a year in advance in certain cases, according to a study that examined forecasts from nine climate models developed by centers in different countries. The key lies in the state of the Pacific at the time the projection is made: when it is going through a La Niña phase, the signals about the following year's event gain precision. Three of the nine models managed to foresee the passage from La Niña to El Niño with more than 70% accuracy at 12 months. In one of them, the accuracy reached 82.21%. The result could provide more time to prepare for droughts, floods, and other effects associated with changes in the tropical Pacific. The research, published in Science Advances, was led by scientist Nandini Ramesh of the Commonwealth Scientific and Industrial Research Organisation (CSIRO), together with specialists from the University of Melbourne. The authors warn that the method is not yet part of operational alert systems and that not all one-year forecasts have the same reliability. El Niño and La Niña are the two extreme phases of the phenomenon known as the El Niño-Southern Oscillation, or ENSO. They originate in the interactions between the ocean and the atmosphere of the tropical Pacific, but their effects reach much of the planet. The certainty of the projections falls below 50% when the models extend the calculation to the second year. El Niño is linked to an anomalous warming of the surface waters of the central and eastern equatorial Pacific. La Niña, on the other hand, is associated with a cooling of that same region. These variations can modify rainfall and temperature patterns thousands of kilometers away. For the study, the scientists analyzed retrospective forecasts from nine climate models developed by centers in different countries. The set included projections initiated between October and January and evaluated which ENSO phase was recorded during the period of greatest intensity of the following event, between October and February. The most relevant finding was that three of the nine models correctly predicted the transition from La Niña to El Niño more than 70% of the time with a year's advance notice. In one of the systems examined, the CESM1-1-CAM5-CMIP5, that accuracy reached 82.21%. Scientists from CSIRO and the University of Melbourne published the study on the ENSO phenomenon in the journal Science Advances. (Illustrative Image Infobae)The conclusion changes the way a long-term climate projection is read. Until now, the focus was usually on which model offered the best overall performance. The new work suggests that, before deciding how much to believe a warning, it is worth observing what state the Pacific is in at the beginning of the forecast. The advantage does not repeat itself when the ocean is in a neutral phase, that is, without clear signs of El Niño or La Niña. In those cases, the dynamic models showed much lower performance when trying to anticipate what will happen a year later. The researchers found that forecasts issued from a neutral condition had an accuracy of between 23% and 43%, depending on the model. As a reference, a basic projection built only from the historical behavior of the system achieved an accuracy of 61% in those situations. The accuracy of the dynamic models drops to values between 23% and 43% when the ocean starts from a neutral phase. This implies that, under certain neutral conditions, a decision-maker could obtain better results by following historical patterns than by taking an initialized climate model projection as definitive. The study detected that these systems tend to generate false alarms, especially when they anticipate La Niña or El Niño after a neutral period. The contrast was marked in the CESM1-1-CAM5-CMIP5. While its forecasts of a passage from La Niña to El Niño were correct in 82.21% of cases, when it anticipated a La Niña from a neutral phase it was right only 8.94% of the time. The difference helps to understand the contribution of the research. The new tool does not aim to replace existing forecasts with a single annual prediction. Instead, it proposes accompanying them with a measure of confidence: determining what type of transition the model is projecting and reviewing how well that kind of warning worked in the past. Predicting ENSO is not simple. Events usually peak between the end of the year and the beginning of the next, but the ability of models to anticipate them diminishes during the months of March to May in the northern hemisphere. That period is called the "spring predictability barrier." The persistence of cold anomalies in the ocean makes it possible to overcome the effects of the spring predictability barrier. (Illustrative Image Infobae)During that time, small alterations in the tropical Pacific can considerably change the evolution of the system. Therefore, although meteorological services usually issue seasonal projections with several months of lead time, extending that horizon to a year maintains a high level of uncertainty. The new analysis shows that this barrier does not affect all scenarios in the same way. Projections initiated during La Niña retained a predictive capacity comparable to that of operational forecasts issued with only three or four months' notice, according to the comparison made by the authors. A possible explanation lies in the very evolution of La Niña. Unlike El Niño, which often dissipates quickly toward the end of the boreal summer, cold anomalies can persist for longer. That persistence would offer a clearer signal about the ocean temperature and the energy accumulated in the tropical Pacific. The researchers also consider that during La Niña there may be less random atmospheric variability, a factor that would facilitate prediction. In any case, the physical causes of this greater anticipatory capacity still require further study. The database used brings together retrospective climate simulations of a decade, initiated each year between 1961 and around 2020. In total, the archive contains 125 ensemble forecast members among the nine models analyzed. The research evaluated more than 7,500 retrospective forecasts produced by nine international systems between 1961 and 2020. (DAVID MCNEW / AFP)To evaluate the one-year projection, the team examined more than 7,500 initial forecasts. If all the annual transitions within the ten-year simulations are counted, the set totals 67,500 cases. The sea surface temperature data came from records of the United States National Oceanic and Atmospheric Administration. The study classified each year into three categories —La Niña, neutral, or El Niño— based on the Niño 3.4 index, which measures temperature anomalies in a central zone of the equatorial Pacific. That definition largely coincided with the NOAA categorization for events recorded since 1960. The results showed that the starting phase weighs more than the general differences between the models. That is, a model considered more accurate on average can issue an unreliable projection if it starts from a neutral state, while another can give a useful signal if the Pacific was under La Niña. The practical application could reach sectors sensitive to climate variations, such as agriculture, water management, disaster prevention, and public health. A more reliable signal with a year's lead time would allow decisions to be anticipated in the face of the possibility of droughts, intense rains, or temperature changes. The method analyzed in the study is not yet part of the operational alert systems of meteorological agencies. (Illustrative Image Infobae)The economic impact of these warnings is also broad. The article points out that seasonal forecasts influence activities and assets valued at around USD 13 trillion globally, from agricultural markets to insurance systems and infrastructure planning. Even so, the work raises clear limits. The authors analyzed retrospective projections from climate models and not an alert system that is already in operation. In addition, they observed that the accuracy of the transitions falls below 50% in all cases when the horizon is extended to the second year. The only consistent two-year exception was the ability to foresee the absence of a new El Niño during the two years following an El Niño. For the scientists, that result could be useful in the face of the risk of prolonged periods with cold conditions in the Pacific, linked to multi-year droughts in subtropical regions. The next step will be to verify how this confidence reading can be incorporated into the prediction systems used by meteorological agencies and to evaluate whether statistical or machine learning methods improve the results when the Pacific starts from a neutral state.