Introduction

Climate extremes and extreme weather are major threats to humans and ecosystems1,2. While dry extremes include warm temperatures, droughts, and enhanced fire weather, wet extremes comprise pluvials, heavy precipitation, and floods. Climate change is worsening the frequency, intensity, and impacts of climate extremes3. Relative to a 1.5 °C warming, the intensity of climate extremes is projected to double (quadruple) at 2 °C (3 °C) warming4.

As global temperatures rise, several regions across the globe face multiple climate extremes simultaneously. According to the United Nations Office for Disaster Risk Reduction (UNISDR), any disaster entails a potentially compounding process, whereby one event precipitates another5. The combination of multiple hazards is often referred to as compound extremes6,7. Climate extremes have serious environmental, economic, and social impacts on their own, but they can have devastating effects when they occur in combination8. The interaction of physical and societal mechanisms of compound extremes amplifies their impacts on vegetation health (e.g., tree mortality)8,9,10,11, economic growth12,13, and human health14,15.

Climate-driven warm temperatures and droughts can combine, thereby dangerously increasing the fire risk. Hot and dry conditions resulting from rising global temperatures have led to increases in the frequency and severity of fire weather16. Recent fierce wildfire seasons around the world (e.g.17,18,19,20,21) have been favored by extreme fire weather conditions (including high temperatures, dryness, and low humidity). Although fire activity is determined by several factors (including available fuel, land management, and ignition sources), under certain weather conditions, fires can spread easily out of control, threatening fragile ecosystems, human life, and property (e.g.22,23).

South America has also experienced severe impacts from concurring warm, dry, and high fire risk conditions (i.e., dry compounds). Although the rise of temperatures of the subcontinent closely follows the global path, the warming and drying have been more pronounced in some regions (Fig. S1). Likewise, heatwaves24,25 and droughts26,27,28 are surging. Concurrent warm and dry conditions often arise from atmospheric blockings. There is growing evidence suggesting that atmospheric blockings have been occurring more frequently in South America in recent decades29,30. The resulting heatwave-drought compounds have often led to enhanced fire weather conditions and catastrophic fire activity31,32,33.

Large-scale climate modes, such as El Niño-Southern Oscillation (ENSO), also play a major role in the South American climate regimes34,35. The phases of ENSO (i.e., El Niño and La Niña) are driven by the strength of trade winds36. During El Niño events, trade winds weaken, and warm water accumulates off the South American west coast. During La Niña events, trade winds strengthen, increasing upwelling and bringing cold, nutrient-rich water to the surface36. As discuss below, through atmospheric teleconnections, ENSO modulates the interannual variability of dry compounds in South America.

Although South America has been identified as one of the global hotspots of compound extremes7,37, dry compounds are still vastly understudied. Here, we use reanalysis datasets to analyze the progression of the concurring warm, dry, and high fire risk conditions over the period 1971–2022. We found that the frequency of the dry compound extremes has surged in key South American regions including the northern Amazon. We also found that the interannual variability of dry compounds in South America is influenced by ENSO. While El Niño enhances the fire risk in the northern Amazon, dry extremes in other regions appear to be more responsive to La Niña.

Methods

Regions of interest

South America has seen the largest relative increase in cropland in the world, with its cultivated land area nearly doubling since the early 2000s38. This large expansion was mainly due to rapid expansion often driven by an increasing demand of soy crops in southern Brazil, Argentina, Paraguay, Bolivia, and Uruguay38.

Most of the new crops are in savanna and dry forest regions including the Chaco basin, stretching across parts of Paraguay, Argentina, Bolivia, and the Brazilian states of “Mato Grosso do Sul” and “Mato Grosso”. The Chaco plain is the second largest forest in Latin America (after the Amazon Forest). The region also has one of the highest rates of deforestation in the world39. Satellite imagery indicates that since 1985 roughly 20% (142,000 square kilometers) of Chaco’s Forest has been converted into farmland or grazing land40.

Another region of great interest in South America is the Amazon basin, which spans about 6 million square kilometers and is home to Earth’s largest rainforest as well as the largest river41. The Amazon rainforest is not only an enormous carbon sink but also a huge source of water. Pumped into the air through evapotranspiration, Amazon moisture feeds vast atmospheric rivers that are an important source of rain for the region and elsewhere42.

Although both the Amazon and Chaco basins are rapidly changing, the Maracaibo basin is the South American region that has warmed and dried the most since 1971 (Fig. S1). The Maracaibo basin is also one of the major oil-producing areas of the world, located close to the spot where Columbia meets Venezuela. Although it supports a large population working and living nearby, including the second-largest Venezuelan city (Maracaibo), about 38% of the basin is still forested43. The exposure of relatively large pockets of population living in wildland-urban interfaces increases the risks of fires becoming disasters, as seen in those that resulted in dozens of fatalities during recent seasons in Central Chile33.

Here, we focused on three of the South American regions that have undergone major land-use changes39 and/or massive precipitation losses in recent decades (see Fig. S1):

  • Region 1 (6–12°N, 65–75°W, hereafter Maracaibo region) encompasses the Maracaibo basin and most of northern Venezuela.

  • Region 2 (3°N–8°S, 55–63°W, hereafter northern Amazon) covers part of the northern Amazon basin including the Brazilian state of “Roraima”, the western part of the Brazilian state of “Pará”, and the eastern part of the Brazilian state of “Amazonas”.

  • Region 3 (13–31°S, 55–65°W, hereafter Gran Chaco) encompasses most of the Chaco basin including the Brazilian state of “Mato Grosso do Sul”, the southern part of the Brazilian state of “Mato Grosso”, and most of the Brazilian Pantanal, the largest continuous tropical wetland in the world31.

Please note that we utilized raw latitude/longitude coordinates for defining the 3 regions mentioned above and did not employ cover/change maps to delineate their boundaries.

Compound and extreme analysis

In order to assess the changes in the frequency of dry compounds, we applied the methodology described by Sutanto et al.44 and by Feron et al.45. It involves building up binary maps assigning, to each grid cell, 1 if an extreme occurs or 0 if it does not. We created daily binary maps for the “warm” days, for the “dry” days, and for the “flammable” days, which allowed us to separately count the number of warm, dry, and flammable days per year. Adding up the binary maps also enabled us to compute the number of days per year with concurring warm, dry, and high fire risk conditions (i.e., dry compounds). Here, relative to the base period 1971–2000, we consider a day to be:

  • “warm” if the corresponding daily maximum temperature (TX) falls above the 90th percentile of the daily TX anomaly distribution;

  • “dry” if the corresponding 30-day running mean precipitation (P) falls below the 50th percentile of the daily P anomaly distribution;

  • “flammable” if the corresponding fire weather index (FWI) falls above the 90th percentile of the daily FWI anomaly distribution. The FWI (a metric of the fire danger derived solely from weather data) computed from daily noon estimates of the wind speed, 24-h accumulated precipitation, temperature, and the relative humidity46.

The anomaly distributions for TX, P, and FWI were defined as it follows. Over a base period of 30 years (1971–2000), we used a 15-day rolling window of the daily estimate of TX and FWI in order to form datasets of 450 values for each day (in the case of the precipitations, we used a 30-day rolling window of the daily estimate of P in order to form datasets of 900 values for each day). The mean of these datasets defined a daily base climatology from which anomalies (for TX, for P, and for FWI) were calculated. The histograms of the anomalies (the departure of the daily estimates from the base climatology) enabled us to recognize “warm”, “dry”, and “flammable” days. In the case of both TX and FWI, we used daily values to establish anomalies, as extreme temperatures or fire weather indexes can trigger climate disasters within a matter of hours or days. For P, we opted for 30-day running means to establish anomalies, as daily precipitation deficits typically have minimal impact. Instead, short-term droughts (i.e., 30-day deficits) can be consequential. Similar reasoning was applied in the case of the percentile. Temperatures and fire weather indexes generally have to reach very high or extreme levels (90th percentile) to affect human health, ecosystems, or properties, while even small but long enough deficits (50th percentile) can be consequential in the case of precipitation.

Over the period 1971–2022, we analyzed the progression of the number of warm, dry, and flammable days per year and per season. Following meteorological conventions, we utilized the seasons as follows: December-January-February (DJF), March-April-May (MAM), June-July-August (JJA), and September-October-November (SON), except when focusing on specific regions of interest. In these instances, we chose to examine the progression of compound warm, dry, and flammable days during the latter part of the dry season (which often coincides with peaks in wildfire activity). Accordingly, we utilized August, September, October (ASO) in the case of Northern Amazon and the Gran Chaco regions, and January, February, and March (JFM) in the case of the Maracaibo basin.

It should be noted that, although other 30-year reference periods such as 1981–2010 or 1991–2020 are viable options, changing the reference period does not impact trend calculations or absolute rankings shown below.

ENSO footprint

In order to assess the impact of ENSO on the simultaneous occurrence of warm, dry, and flammable conditions, we compared the changes in the frequency of dry events with variations in sea surface temperature (SST) in two El Niño regions: the Niño 3.4 region (5°N-5°S, 170°W–120°W) and the Niño 1 + 2 region (0–10°S, 90°W–80°W) (Fig. S2). While anomalies in the Niño 3.4 region are known to have global impacts, the Niño 1 + 2 region has a strong influence on the South American climate34.

We conducted Pearson Correlation Tests to determine whether there is a dependence between the frequency of dry compounds and the SST anomalies in the Niño Regions. A p value, below 0.05, indicates statistical significance, suggesting that the two time series are dependent. The tests involved comparing annual averages as well as seasonal averages. In this case, we used meteorological seasons: DJF, MAM, JJA, and SON.

Datasets

Here we used daily surface weather data (including near-surface (2-m) temperature, precipitations, and the FWI) come from the atmospheric reanalysis ERA5, produced by the European Center for Medium-range Weather Forecasts (ECMWF)47. ERA5 data are available at: https://www.ecmwf.int/en/forecasts/datasets/reanalysis-datasets/era5.

SST anomalies in the Niño regions come from the Climate Prediction Center (CPC)48, part of the National Oceanic and Atmospheric Administration (NOAA), available at https://www.cpc.ncep.noaa.gov/data/indices/wksst8110.for.

Results

Compound extremes have surged in key South American regions

Dry compounds with concurrent warm, dry, and flammable conditions have surged in recent decades in key South American regions (Fig. 1). The rise in dry extremes has been particularly sharp in the northern Amazon, Maracaibo, and the northeastern Gran Chaco regions (Fig. 1a). While warm, dry, and flammable conditions were generally present for less than 20 days per year over the period 1971–2000 (Fig. S3a), these conditions increased up to 70 days per year in the northern Amazon, Maracaibo, and the Brazilian Pantanal (in the northeastern Gran Chaco) regions in recent decades (Fig. S3b). Increases from 1971–2000 to 2001–2022 in the number of days per year with concurring warm, dry, and flammable conditions were less considerable in Ecuador, and in Patagonia (southern Chile and Argentina) (Fig. 1a).

Fig. 1: Compound extremes have surged in key South American regions.
figure 1

a Change from 1971–2000 to 2001–2022 in the number of days per year with concurring warm, dry, and flammable conditions (i.e., dry compound days). b Progression of the number of days per year with concurring warm, dry, and flammable conditions in the three key South American regions. The number of dry compound days per year were derived (see “Methods”) from daily estimates from the ERA5 dataset over the period 1971–2022. Plots were generated by using Python’s Matplotlib Library82.

Full size image

Most of the surge in dry compounds across the northern Amazon, Maracaibo, and Gran Chaco regions occurred during the last two decades, which suggests accelerated changes (Fig. 1b). Regardless of the interannual variability, few changes in the frequency of dry compounds were apparent in the 1970s and 1980s (Fig. 1b). However, since the late 1990s, concurring warm, dry, and flammable conditions started to be increasingly more frequent, especially in the northern Amazon and Maracaibo regions (Fig. 1b). In these two regions, the decadal average of dry compound conditions has seen a threefold increase over the period 1971–2022.

The rapid rise in dry compounds across the northern Amazon, Maracaibo, and Gran Chaco regions resulted from the parallel increase in the frequency of warm, dry, and flammable days in these regions (Fig. 2). From 1971–2000 to 2001–2022, warm conditions climbed by about 60 days per year in the Amazon and Maracaibo Basin regions (Fig. 2a and Fig. S4). Although the rise in warm days was not similarly meaningful in the Gran Chaco region, the latter was one of the South American regions that exhibited the largest increase in the number of dry days per year (Fig. 2b). From 1971–2000 to 2001–2022, annual precipitations in the Gran Chaco and Maracaibo regions dropped by about 100 mm and 200 mm, respectively (Fig. S1b). Although precipitations in these regions remain high (more than 1000 mm per year; Fig. S5), from 1971–2000 to 2001–2022, dry conditions in the Gran Chaco and Maracaibo regions rose by more than 50 days per year (Fig. 2b, Fig. S6).

Fig. 2: Warm, dry, and flammable conditions have become more frequent in South America.
figure 2

Change from 1971–2000 to 2001–2022 in the number of: (a) warm days per year, (b) dry days per year, and (c) flammable days per year. The number of warm, dry, and flammable days per year were derived (see “Methods”) from daily estimates from the ERA5 dataset over the period 1971–2022. Plots were generated by using Python’s Matplotlib Library82.

Full size image

Climate-driven warm temperatures and droughts are contributing to an enhanced fire risk (i.e., flammable conditions). The rise in extreme fire weather conditions has been particularly steep in the northern Amazon, Maracaibo, and Gran Chaco regions (Fig. 2c). While flammable conditions were present generally less than 40 days per year over the period 1971–2000 (Fig. S7a), these conditions arose up to 120 days per year in the northern Amazon and Maracaibo regions over the last decade (Fig. S7b). Increases in the number of days per year with flammable conditions were less considerable in western Patagonia (Fig. 2c), one of the few South American regions where precipitations grew from 1971–2000 to 2001–2022 (Fig. S1b).

Within the Gran Chaco, the northeastern region experienced the largest increase in the number of flammable days (Fig. 2c). This trend aligns with the significant changes in the frequency, duration, and severity of compound drought-heatwaves observed in the Brazilian Pantanal in recent years31,32. Some regions in eastern Brazil, such as the Brazilian state of Tocantins, have also undergone considerable changes in the number of flammable days (Fig. 2c). However, the temperature in this region did not rise at the same pace over the same period as it did in the northern Amazon region, for instance (Fig. 2a). Therefore, as shown in Fig. 1a, the changes in the frequency of compound extremes in eastern Brazil are considerably smaller than those observed in the northern Amazon region. The latter is the region that experienced the most substantial changes in the frequency of dry compounds across the Brazilian territory.

The changes in the frequency of dry compound extremes exhibit different seasonal trends (Fig. 3 and Fig. S8). Although the occurrence of dry compounds in the Maracaibo region has increased throughout the whole year, the increment appears to be slightly sharper in the DJF season (Fig. 3a). In the northern Amazon region, the surge in dry compounds seems to be steeper in the JJA season (Fig. 3c), which is also the season that has seen large and simultaneous rises in temperatures (Fig. S8c) and drops in precipitation (Fig. S8g). In the Gran Chaco region, the changes in the frequency of dry compounds appear to be particularly strong in the austral winter (JJA) and the austral spring (SON) (Fig. 3c, d). Within the Gran Chaco, the Brazilian Pantanal experienced the largest increase in the number of dry compound days (Fig. 3d), likely driven by the simultaneous rise in spring temperatures (Fig. S8d) and the loss of spring precipitation (Fig. S8h).

Fig. 3: Dry compound extremes exhibit different regional and seasonal trends.
figure 3

Changes from 1971–2000 to 2001–2022 in the number of days per season with concurring warm, dry, and flammable conditions (i.e., dry compound days). The following meteorological seasons were considered: (a) December-January-February (DJF), (b) March-April-May (MAM), (c) June-July-August (JJA), and (d) September-October- November (SON). The number of dry compound days per season were derived (see “Methods”) from daily estimates from the ERA5 dataset over the period 1971–2022. Plots were generated by using Python’s Matplotlib Library82.

Full size image

Dry extremes exhibit a considerable interannual variability

The number of dry compound days has set several records during the last two decades. For example, across the Gran Chaco region, fire weather conditions were extremely recurrent in the 2004, 2007, 2010, and 2020 seasons (Fig. 4a). Across the northern Amazon region, fire weather conditions were also extremely frequent in the 2005, 2010, 2015, 2017 and 2020 seasons (Fig. 4b). Across the Maracaibo region, fire weather conditions were frequent in the 2003, 2010, 2016, and 2020 seasons (Fig. 4c). While wildfires in the northern Gran Chaco and northern Amazon regions often occur from August to October (at the latter part of the dry season in those regions), the fire season in the Maracaibo region peaks between January and March. During the fire season and especially under dry compound conditions, naturally occurring fires (not uncommon in South American drier forests and grasslands) or prescribed fires (for brush clearing or deforestation) can easily get out of control and invade other areas.

Fig. 4: The interannual variability of dry extremes have increased in key South American regions.
figure 4

Progression of the number of compound, warm, dry, and flammable days per season, averaged across three key South American regions: (a) Gran Chaco, (b) Amazon, and (c) Maracaibo. Seasons in this case were selected to coincide with the latter part of the dry season (which often aligns with peaks in wildfire activity): August, September, October (ASO) in the Northern Amazon and the Gran Chaco regions; January, February, and March (JFM) in the Maracaibo basin. The number of warm, dry, and flammable days per year were derived (see Methods) from daily estimates from the ERA5 over the period 1971–2022. Taken as a measure of the variability, the standard deviation of the number of dry compound days has seen a twofold increase from 1971–2000 to 2013–2022 in all three regions. Plots were generated by using Python’s Matplotlib Library82.

Full size image

The interannual variability of fire activity (the number of fires, monitored by remote sensing over the last two decades) have generally mirrored fire weather conditions in South America. For example, satellite imagery49 have confirmed that the 2020 fire season was very active across the northern Gran Chaco region in the northern Amazon region, likely fanned by the nearly-record fire weather conditions shown in Fig. 4a, b for the year 2020. In April 2020, over 30% of central-east South America experienced negative soil moisture anomalies larger than two standard deviations50. The Brazilian Pantanal, in particular, witnessed in 2020 the most catastrophic fire season in the last two decades, attributed to the simultaneous occurrence of dry and hot spells31. Although prior efforts have shown that the actual fire activity is strongly determined by climate extremes (e.g.51), as discussed below, fire weather conditions are not the only driver of fire activity in South America.

The year-to-year variability of dry compounds has increased in key South American regions. As dry compound conditions became more frequent, their interannual variability also climbed across the northern Amazon, Maracaibo, and Gran Chaco regions (Fig. 4). Taken as a measure of the variability, the standard deviation of the number of dry compound days has seen a twofold increase from 1971–2000 to 2003–2022. Similar strong increases in the interannual variability are apparent in the case of warm conditions and in the case of flammable conditions. Although changes are less clear in the case of dry conditions, the interannual variability in the number of dry days remains strong across the northern Amazon, Maracaibo, and Gran Chaco regions.

ENSO plays a major role in the interannual variability

The influence of El Niño and La Niña on the South American climate makes the SST in the tropical Pacific of paramount importance. Driven by the strength of trade winds36, the SST in the Niño regions determines the phases of ENSO (i.e., El Niño and La Niña) (Fig. S9a), also affecting the frequency of dry compounds in the regions of interest. For example, the all-time record of dry compounds in the Gran Chaco region (Fig. 4b) coincided with the onset of the triple-dip La Niña from 2020 to 2022 (Fig. S9a). Similarly, the all-time record of dry compounds in the northern Amazon region (Fig. 4b) coincided with the El Niño event of 2015 (Fig. S9a). The latter is often referred to as El Niño Godzilla due to its record intensity, especially in the Niño Region 3.4 (Fig. S9b).

The interannual variability of dry compounds in South America is driven by the year-to-year variability of the SST in the Niño regions (Fig. 5). In the northeastern Amazon (in the central Gran Chaco), there is a strong correlation (anticorrelation) between the detrended occurrence of dry compounds and the SST anomalies in the Nino 1+2 Region (Fig. 5a). We found connections of similar statistical significance (above the 95% level) when comparing SST anomalies in the Nino 1+2 Region and the occurrence of individual extremes (warm, dry, or flammable conditions) in the northeastern Amazon and the central Gran Chaco (Fig. S10). Dry extremes appear in the Gran Chaco region (in the western Maracaibo region) to be less (more) connected with the interannual variability of the sea surface temperature in the Niño 3.4 region. In the Gran Chaco region (in the western Maracaibo region), the correlations became weaker (stronger) when comparing concurring warm, dry, and flammable conditions with SST anomalies in the Niño 3.4 region (Fig. 5b, Fig. S11).

Fig. 5: El Niño (La Niña) makes dry extremes more likely in the northern Amazon (in the Gran Chaco).
figure 5

Pearson correlation between the number of days per year with concurring warm, dry, and flammable conditions (i.e., dry compound extremes) and the corresponding annual SST anomalies in the: (a) Niño 1+2 Region, and (b) Niño 3.4 Region. Stippling indicates statistical significance (p value: 0.05). The number of warm, dry, and flammable days per year were derived (see “Methods”) from daily estimates from the ERA5 dataset over the period 1997–2022. SST anomalies in the Niño Regions come from the Climate Prediction Center (CPC), part of the National Oceanic and Atmospheric Administration (NOAA), available at https://www.cpc.ncep.noaa.gov/data/indices/wksst8110.for. Plots were generated by using Python’s Matplotlib Library82.

Full size image

The seasonal frequency of dry compounds in South America is also connected with the SST in the Niño regions (Fig. 6). In the northeastern Amazon and the Maracaibo regions, the correlation between the occurrence of dry compounds and the SST anomalies in the Nino regions is particularly strong in the austral summer (DJF) (Fig. 6a, b). In the Gran Chaco region, the anticorrelation between dry compounds and the SST anomalies in the Nino region in particularly strong in austral winter (JJA) (Fig. 6e–g), which is also the driest season in this region. We generally found connections of similar statistical significance when comparing the occurrence of dry compounds with SST anomalies in either the Nino 1+2 Region (Fig. 6, upper row) or the Nino 3.4 Region (Fig. 6, lower row). However, in the dry season (JJA), dry compounds in Paraguay (in eastern Bolivia) appear to be more connected with the SST in the Niño 1+2 region (in the Niño 3.4 region) (Fig. 6e–g).

Fig. 6: The effect of ENSO on the frequency of dry compounds is season-dependent.
figure 6

Pearson correlation between the seasonal SST anomalies in the Niño Regions and the corresponding number of dry compounds. Upper row: Niño 1+2 Region; Lower row: Niño 3.4 Region. The following season were considered: (a, b) December-January-February (DJF), (c, d) March-April-May (MAM), (e, f) June-July-August (JJA), and (g, h) September-October- November (SON). Stippling indicates statistical significance (p value: 0.05). The number of dry compounds per season was derived from daily estimates from the ERA5 dataset over the period 1997–2022. SST anomalies in the Niño Regions come from the Climate Prediction Center (CPC), part of the National Oceanic and Atmospheric Administration (NOAA), available at https://www.cpc.ncep.noaa.gov/data/indices/wksst8110.for. Plots were generated by using Python’s Matplotlib Library82.

Full size image

The connections between the tropical Pacific SST and the occurrence of dry compounds shown in Fig. 5 suggest that El Niño (La Niña) makes dry extremes more (less) likely in the northeastern Amazon. These results are consistent with prior efforts that have shown ENSO as one of the main drivers of interannual climate extremes in the Amazon basin51. During an El Niño event, the Walker Circulation is greatly weakened, which suppresses convection and rainfall in the eastern Amazon52. Contrary, during an La Niña event, the resulting strengthening of the Walker circulation has often led to floodings in the region53. Dry and wet extremes in the precipitation reconstructions for the eastern Amazon also co-coincided with precipitation (and tree growth extremes) of opposite sign in the mid latitudes of South America54.

The correlation between the tropical Pacific SST and the occurrence of dry compounds shown in Fig. 5 also suggest that La Niña boosts concurring warm, dry, and flammable conditions in the Gran Chaco region. These results are also consistent with prior efforts that have shown that La Niña events disturb the South American monsoon, which transports moisture to Southern South America including the Chaco region55. The South American monsoon is characterized by strong year-long trade winds, channeled southward into the South American Low-Level Jet (SALLJ) after crossing the Amazon basin56. Likely influenced by disruptions in the monsoon in the Gran Chaco region, droughts have generally been found concurrent with La Niña events57. The strong connection between the fire weather conditions in the Gran Chaco region and the SST in the tropical Pacific is likely one of the reasons why the very active fire seasons in 2004, 2007, 2010, and 2020 in the Gran Chaco region (i.e., “Mato Grosso” Brazilian state) concurred with La Niña events (Niño 1+2 region) in 2004, 2007, 2010, and 2020 (Fig. S9a).

Although the interannual variability of the fire activity generally mirror fire weather conditions, the long-term trends are determined by regulation enforcement and policies (e.g.58,59). In the Brazilian states, burning of forests to clear land for agriculture and grazing was extremely high early this century (2001–2004) because it was not until 2004 that Brazil enacted a series of environmental regulations that reduced fires and the rate of deforestation. The adoption of these environmental regulations explains why we have not seen a dramatic increase in the severity of the fire season across key regions in South America over the last two decades60.

Discussion

We have shown that, despite significant interannual variability, the frequency of compound extremes (warm, dry, and high fire risk conditions) has climbed in key South American regions including the Amazon region, which have seen a threefold increase in the number of days per year with extreme fire weather conditions. These results are consistent with prior efforts showing that climate-related heatwaves24,25 and droughts26,27,28 are surging in South America.

Our results also suggest that, while anthropogenic warming likely drives this increasing trend, ENSO modulates the interannual variability of compound extremes in South America. The influence of ENSO on the increasing trend of dry conditions in South America is likely minor. The absence of a significant trend in SST in the El Niño regions suggests that ENSO is unlikely to play a major role in this trend. However, quantitatively confirming this would require an attribution analysis, which would involve a dedicated study.

While El Niño enhances the fire risk in the northern Amazon and Maracaibo regions, dry extremes in the Gran Chaco region appear to be more responsive to La Niña. Earlier studies have shown61,62 that El Niño events appear to have weakened in recent decades and their SST anomalies shifted westward towards the central Pacific. Such a trend would make warm events stronger in the Niño 3.4 region and weaker in the Niño 1+2 region. However, this trend (not apparent in Fig. 5a) can unlikely explain the surge of compound extremes observed in recent decades in key South American regions (Fig. 1).

Although the inter-annual variability of dry compounds in South America appears to have increased (Fig. 4), it is uncertain how it will evolve because it is unknown how ENSO may respond to future greenhouse gas emissions63. Yet, climate models show that under a likely emission scenario, extreme El Niño frequency increases linearly with the global mean temperature towards a doubling at 1.5°C warming64. Those projections suggest further increases in the inter-annual variability of dry compounds.

Despite the paramount importance of the tropical Pacific SST anomalies, prior efforts have shown that SSTs in the Atlantic Ocean also play a role in regulating drought and fire occurrence in South America. For example, the Atlantic Multidecadal Oscillation (AMO) index has been found to be closely linked with fire activity in the southwestern Amazon65, while SST anomalies in the North Tropical Atlantic (NTA) have been found to be connected with precipitation variability in the western Amazon region66.

Even though the climate-fueled dry extremes and SST anomalies drive the fire risk, climate is not the only driver of fire activity in South America. In fact, droughts in the twenty-first century have not increasingly exacerbated the severity of fire seasons in the Brazilian Amazon60. While major droughts do coincide with high-extreme fire years, anthropogenic activities play a significant role in the distribution of extreme fires in the Brazilian Amazon67. Poor management and lax laws are making the Brazilian Pantanal, the largest continuous tropical wetland in the world and a World Heritage Site, prone to fierce fires68. Fires in the subcontinent are often started intentionally to clear large swaths of forest and convert it to agricultural land69. When fires are unusually large, persistent, and along the forest edge, they are likely deforestation fires. Deforestation and slash-and-burn agriculture are still a major influence on the fire activity in South America70.

In the Amazon region, dry compounds act as a feedback loop disrupting forest carbon dynamics71 and amplifying the predominant drying trends. Enhanced fire activity associated with dry compounds releases into the atmosphere massive amounts of black carbon, which absorbs heat from the sun, causing the atmosphere to warm and also interfering with cloud formation and, consequently, with rainfall (e.g.72). This is affecting the water supply of trees whose water demand is increasing due to atmospheric warming. Such a trend may have already pushed the Amazon close to a critical threshold of rainforest dieback73.

Dry compounds are also public health hazard in South America32,74. The signal of excess deaths associated with extreme temperatures has already emerged in major South American cities14,15,74. However, few countries in South America (other than Chile and Argentina) have developed Heat/Health warning system (HHWS) according to the recommendations of the World Health Organization (WHO) and the World Meteorological Organization (WMO)75. Moreover, although black carbon particles from biomass burning are known to be small enough to pass through the lungs into the bloodstream, spreading sickness and death far beyond local sources76, too many people in South America still breathe unhealthy air.

As with other climate-related impacts, increasing compound extremes disproportionally affect vulnerable rural populations and minorities. In particular, Amazon fires are seriously threatening indigenous territories in which largely isolated indigenous groups build their homes77. Moreover, deposition of black carbon from biomass burning impacts the Andean snowpack, which is an important source of water for many small rural communities78. Black carbon has been found to darken the Andean snow, reducing the albedo, and accelerating melting79,80,81.

In contrast to the dominant framing of climate change in terms of global average temperature change, it is urgent to address regional variations, since changes vary between regions and their significance is also highly locally modulated. In addition, it is the manifestation of extremes -rather than averages- that must be brought into view in order to provide relevant and meaningful insights on how climate change affects environmental, human, infrastructural and ecological systems on the ground. These orientations to specific and unusual aspects of climate change should drive the research agenda and be used to develop common ground for scientists, policymakers, and civil society.