El Niño and tropical Atlantic warming could offer early warning of extreme heat in the Amazon

El Niño and warming in the tropical Atlantic could help provide advance warning of extreme heat in the Amazon, with some signals appearing as much as seven months earlier, new research has found.

Researchers from the ARC Center of Excellence for the Weather of the 21st Century and the National Institute of Space Research in Brazil studied how major patterns of natural climate variability influence temperature and rainfall extremes across the Amazon Basin.

Published in the journal Earth's Future, the study found that extreme heat in the Amazon was more likely to follow unusually warm conditions associated with El Niño in the tropical Pacific and warming in parts of the tropical Atlantic.

The clearest long-range signal occurred from March to May. Across much of the Amazon, extreme heat during these months was associated with climate conditions observed as much as seven months earlier.

The researchers also investigated whether these large-scale climate patterns could help predict compound hot–dry events, when unusually high temperatures coincide with prolonged rainfall deficits.

The results revealed useful early-warning potential in some seasons and regions, particularly in the northern and lower central Amazon. However, predictability was considerably weaker elsewhere.

Associate professor Andrea Taschetto of UNSW, a contributing author to the paper, said, "The main mechanisms that affect the Amazon Basin and are related to the Pacific and Atlantic oceans are well established on the seasonal mean timescale.

"But in this study, machine learning techniques allowed us to examine the relationship with extreme compound events, which is much harder. It is quite surprising that heat extremes are linked to conditions in the Pacific and tropical Atlantic oceans observed up to seven months in advance."

Why advance warning matters

Compound hot–dry events can increase fire risk, damage ecosystems and threaten human health, food security and water supplies. They can also weaken the Amazon's ability to absorb carbon dioxide, increasing the risk of large-scale ecological and climatic destabilization.

Hot–dry events have become more frequent and intense in recent decades and are expected to become more common and severe as the climate continues to warm. Land-use change may intensify these hazards further, with previous research indicating that deforestation can prolong dry periods and increase temperatures.

Being able to identify heightened risks months or even weeks in advance could help communities, governments and environmental managers prepare for periods of dangerous heat and drought.

Climate connections across two oceans

El Niño is the warm phase of the El Niño–Southern Oscillation (ENSO), a recurring pattern of changing ocean temperatures and atmospheric circulation across the tropical Pacific.

During El Niño, unusually warm waters in the central and eastern tropical Pacific alter atmospheric circulation. These changes can suppress cloud formation and rainfall over parts of the Amazon, allowing more sunlight to reach the surface while reducing the amount of water available to cool the land through evaporation.

El Niño can also influence temperatures in the tropical Atlantic. Together, warming in the Pacific and Atlantic oceans can amplify hot and dry conditions across the Amazon. These relationships vary throughout the year and across the vast Amazon Basin, which contains regions with very different rainfall patterns, landscapes and seasonal cycles.

The researchers divided the Amazon into six climatic regions and examined monthly temperature, rainfall and large-scale climate conditions from 1950 to 2023. Heat and dryness were assessed using conditions accumulated over three-month periods.

The March–May period produced the clearest long-range heat signal, with unusually warm conditions associated with Pacific and Atlantic climate patterns observed up to seven months earlier.

During September–November, hot extremes were primarily associated with earlier warming in the tropical North Atlantic. During December–February, extreme heat was linked to more recent warming in the tropical South Atlantic across much of the basin. Earlier ENSO signals were also evident in the southern Amazon and the foothills between the Andes and the Amazon.

The connections between these ocean patterns and Amazon rainfall extremes were weaker and less widespread than those involving temperature.

Rainfall is inherently more variable than temperature and can be strongly shaped by regional weather systems, land conditions, moisture transport and interactions between the land and atmosphere. This makes rainfall extremes—and compound events involving both heat and low rainfall—more difficult to anticipate from large-scale climate patterns alone.

Testing the potential for early warning

The researchers used two complementary approaches to investigate these relationships.

First, they tested whether unusually strong phases of four major climate patterns—ENSO, tropical North Atlantic and tropical South Atlantic sea surface temperature anomalies, and the North Atlantic Oscillation—tended to occur before temperature or rainfall extremes in different parts of the Amazon.

This analysis used a statistical technique capable of focusing specifically on relationships at the extremes. Rather than simply asking whether two climate variables generally rise or fall together, it tested whether exceptionally warm ocean conditions were disproportionately associated with exceptionally hot conditions in the Amazon.

The team then used machine learning to determine whether different combinations of these earlier climate signals could distinguish compound hot–dry months from other conditions.

The machine-learning models performed best for December to February hot–dry events in parts of the northern Amazon and for March to May events in the lower central Amazon. ENSO was the most important source of predictive information across much of the basin, while tropical Atlantic temperatures also made a substantial contribution.

Predictability was generally weakest during June to August and in parts of the western Amazon, where local and regional processes not fully represented by the four climate patterns may exert a stronger influence.

In a small number of regions and seasons, the models identified useful signals one or two months before a compound hot–dry event. However, the strongest results usually also drew on climate conditions during the month being predicted.

This distinction is important—the seven-month relationship represents a long-range precursor signal for extreme heat, rather than evidence that compound hot–dry events can already be reliably forecast seven months in advance.

UNSW's Dr. Sanaa Hobeichi, the lead author of the paper, said, "Machine learning helped us identify which climate signals offer the most promise for anticipating hot–dry extremes. ENSO generally provided the strongest predictive information, but Atlantic climate modes were more informative in some regions and seasons. Just as importantly, it showed where these signals alone were not enough."

The results also demonstrate potential predictability rather than the performance of a real-time forecasting system. Further research using operational climate models and additional information, including soil moisture, vegetation conditions and regional atmospheric circulation, will be needed to translate these relationships into practical warning systems.

Nevertheless, the study identifies the regions and seasons in which monitoring Pacific and Atlantic climate conditions may provide valuable advance notice of heightened risk.

It also shows where ocean-based climate indicators are insufficient on their own, helping to guide the development of more targeted seasonal forecasting and climate-risk services for one of the world's most climate-sensitive regions.