Signals associated with El Niño and warming of the tropical Atlantic waters can help identify, with several months of lead time, conditions favorable to episodes of extreme heat in the Amazon. A study published in the journal Earth’s Future found signals that, in some periods of the year, can emerge up to seven months before temperature extremes.
The investigation, led by scientists from the University of New South Wales (UNSW), in collaboration with Australian and Brazilian institutions, analyzed the relationship between large patterns of climate variability and temperature and precipitation extremes in the Amazon basin.
The results show that extreme heat episodes between March and May were associated with abnormally hot conditions in the tropical Pacific and in parts of the tropical Atlantic observed up to seven months earlier. The researchers also identified predictive signals for episodes in which high temperatures coincide with prolonged periods of lower precipitation, though these are more difficult to anticipate.
“It is quite remarkable that heat extremes are linked to conditions in the tropical Pacific and Atlantic oceans observed up to seven months earlier,” says Andrea Taschetto, associate professor at UNSW and one of the study’s authors.
A warning sign, not a forecast
The possibility of identifying conditions favorable to extreme heat in advance could be relevant for community preparedness and risk management. Simultaneous episodes of heat and drought can increase wildfire risk, affect ecosystems and pose threats to human health, food security, and water resources.
These phenomena can also reduce the Amazon’s capacity to absorb carbon dioxide, contributing to changes in the region’s climate balance.
However, the researchers make an important distinction: the identified relationship between ocean conditions and heat extremes does not mean that it is already possible to predict a heat and drought episode seven months in advance.
The period between March and May showed the clearest long-term signal for temperature extremes. The results related to the compounded phenomena of heat and drought were more limited. In some regions and seasons, models were able to identify useful signals one or two months earlier, but the best results also depended on the conditions recorded in the month of the event.
El Niño plays a central role
El Niño corresponds to the warm phase of the El Niño–Southern Oscillation (ENSO) phenomenon, associated with changes in sea surface temperatures and atmospheric circulation in the tropical Pacific.
During an El Niño event, warming of waters in the central and eastern Pacific can alter atmospheric circulation and reduce cloud formation and precipitation in some parts of the Amazon. With less rainfall, the amount of water available to cool the surface through evaporation also decreases.
El Niño can also influence temperatures in the tropical Atlantic. The combination of warmer conditions in the two oceans can favor periods of heat and drought in different parts of the basin, though the relationship varies throughout the year.
To assess these links, the researchers analyzed monthly data on temperature, precipitation, and large-scale climate conditions from 1950 to 2023, dividing the Amazon into six climatic regions.
The results showed that, between September and November, heat extremes were mainly associated with prior warming of the tropical North Atlantic. Between December and February, the relationship was more evident with recent warming of the tropical South Atlantic. Earlier signals associated with ENSO were also identified in southern Amazonia and in the transition zones between the Andes and the Amazon basin.
However, the relationships with precipitation extremes were weaker and less generalized. Rainfall is more variable than temperature and depends on regional factors, including meteorological systems, soil conditions, moisture transport, and interactions between the land surface and the atmosphere.
Artificial intelligence helps identify signals
The team used two complementary approaches. First, they assessed whether four major climate patterns — ENSO, sea surface temperature anomalies in the tropical North and South Atlantic, and the North Atlantic Oscillation — tended to precede temperature or precipitation extremes in different regions of the Amazon.
Then, they turned to machine learning models to determine whether different combinations of these signals could distinguish periods of extreme heat and drought from other conditions.
“Machine learning models helped us identify which climate signals are most promising for anticipating heat and drought extremes. ENSO generally provided the strongest predictive information, but Atlantic climate patterns were more informative in some regions and seasons. And, equally importantly, they showed us where these signals, on their own, were not sufficient,” explains Sanaa Hobeichi, the study’s lead author and a researcher at UNSW.
The models showed better results for heat and drought episodes between December and February in some northern areas of the Amazon and between March and May in the central-southern region of the basin. Predictability was lower between June and August and in some areas of the western Amazon, where local and regional processes appear to have greater influence.
From research to warning systems
The authors emphasize that the results demonstrate forecasting potential, not the performance of a real-time forecast system.
To translate these signals into operational alert tools, further studies will be needed, including climate models used in predictions and other variables, such as soil moisture, vegetation status, and regional atmospheric circulation.
Nevertheless, the study makes it possible to identify the regions and times of the year in which monitoring the Pacific and Atlantic conditions can provide advance information about an increased risk of extreme heat.
In a context of climate change and growing pressure on the Amazon rainforest, this information could contribute to more targeted forecasting systems and to better preparation for extreme episodes.
