Compound changes in hydrological and climate variables. Regions suffering a long-term warming and/or drying trend (1981-2020) are identified through the analysis of spatial patterns of these trends for different radiative, atmospheric, and hydrological variables (Fig. 1). EAC experienced a widespread and significant warming trend (0.38±0.15ºC/decade, p<0.05) during the dry-to-wet transition season July-October (JASO) over last four decades (Fig. 1a). The observed actual evapotranspiration (EVP) reduction (Fig. 1b) tends to elevate temperature, which increases sensible flux to offset the net downward radiative flux. Generally, an increase of surface net radiation and consequent increase of temperature lead to an increase of EVP if there is sufficient moisture in plants and soil. The widespread increase of vapor pressure deficit (VPD, Fig. 1c) is in line with the spatial pattern of warming observed in air temperature. This also agrees with global increases of VPD leading to reductions in vegetation growth46.
Overall, precipitation trends during the dry-to-wet transition season do not show a statistically significant widespread spatial pattern, but negative trends predominate over southern and southeastern Amazonia (Fig. 1d). A delayed wet season onset (WSO) is also noticed over eastern Amazonia (Fig. 1h), associated with an increase in atmospheric subsidence, as suggested by the positive trend in vertical velocity (omega) at 500 hPa over this region (Fig. 1g). Accordingly, EAC is characterized by a significant increase of the frequency of dry days (DDF, Fig. 1e). While not statistically significant, observed rainfall decreases associated with increased subsidence over this region, are partially related to an intensification of the Hadley and Walker cells and to a poleward shift of the subtropical jet over South America32,47,48. Moreover, the increase in DDF over this region is related to a warming of the northern tropical Atlantic Ocean and a weakening of moisture transport from the tropical Atlantic Ocean29. This is consistent with previous findings demonstrating the increased dry season length2,47,49,51, also observed through the delayed WSO (Fig. 1h). There is a reduction in the atmospheric water vapor content (TCWV) in eastern Amazonia (Fig. 1f), and a northwest to southeast gradient, wet over the north and dry over the south.
In water-limited areas such as the eastern Amazon, however, an increase of temperature is unlikely to increase EVP, especially in the dry and dry-to-wet seasons. Variation in water availability governs EVP in the seasonally dry tropical forests in the south and southeast Amazon, towards the transition with the Cerrado biome22,51. Dry-adapted plants can control stomata opening or shed their leaves in response to water deficits, but unadapted plants cannot. If the stomata are closed for too long, an increase in plant mortality by carbon starvation is expected. On the other hand, if plants are unable to avoid water loss, mortality is likely to increase because of cavitation52. All these processes are fundamentally linked to canopy–atmosphere coupling, with complex interactions between climate and plant phenology.
The analysis of long-term trends (Figs. 1a-1h) evidenced that some of the hydrological and climate changes are already widespread in EAC, whereas other changes are focused on southern/southeastern Amazonia or even finer regional scales. By combining changes of all the variables into a single compound indicator (Fig. 1i), we show that EAC is suffering a combined dry and warming trend. The EAC sensitive region is mainly composed of Cerrado and encompasses roughly the MATOPIBA region (Extended Data Fig. 1). Therefore, the MATOPIBA region shows the strongest heating and drying trends observed across the whole of the Amazon and Tocantins basins and Cerrado biome (Figure 2). This agrees with the fire distribution focused across the southern boundary of the Amazon basin and in the EAC during the May-August dry season because the disturbed forests are more prone to burning in the dry-to-wet transition season than in the wet and dry seasons53.
Regional scale trends over the MATOPIBA. The EAC, including MATOPIBA has already been changing towards a drier and warmer climate (Fig. 2). The warming trend (Fig. 2a), is 0.3 ºC/decade (p<0.05); the total change from 1981 to 2020,1.2 ˚C. For the post-2000 period it increases to 0.45 ºC/decade (p<0.05), giving an increase of almost 1 ºC over the last two decades. The highest temperature anomalies (1.1 ºC) are observed in 1998 and 2015 (strong ENs), and 2017 (0.9 ºC). However, the MATOPIBA reached higher values of monthly-mean temperature anomalies, with a record 2.9 ºC in December 2015 (Extended Data Fig. 2). Rainfall trends show systematic reductions (-0.08 mm/day per decade, p<0.05) and an increase in the DDF (+1.5 days per decade, p<0.05) (Figs. 2b, 2c). The trend is slightly greater (around +2 days per decade) during the first half of the period (1981-1999), but this is not statistically significant.
These decreases in rainfall since 1980 are related to positive changes in the vertical velocity omega (Figure 1g), indicating increasing subsidence (less convective activity, thus dryness and warming) in the EAC region. This change since the late 90s has been identified in previous studies as part of changes in regional atmospheric circulation in JASO29. Warming and drying (temperature, precipitation and DDF) are stronger over Tocantins state (Extended Data Fig. 3) than anywhere else throughout the MATOPIBA region (Fig. 1i).
Drought and land cover changes. Drought indices evidence a drying trend over EAC for the period 1981-2020 (Extended Data Fig. 4). There is a strong interannual change in SPI, SPEI and scPDSI, with consistently negative trends in all indices, consistent with the decreasing rainfall. SPI shows the smallest negative trend (p<0.1). SPEI and scPDSI indices, which also include the impact of evapotranspiration, show a strong negative trend (p<0.0001). While SPI and SPEI show the largest negative values in 2015 and 2017, scPDSI provides the largest negative value in 2016, probably because of the different response to soil moisture memory between indices. scPDSI also evidences a strong negative trend from year 2000. SPEI and scPDSI agree on the negative index value in 2020. In sum, all of them show a significant drying trend during the last four decades
The exposure of affected areas to drought and its intensity can be assessed by the Integrated Drought Index (IDI), which combines the lack of precipitation and the vegetation response to water stress (Fig. 3): 1992, 1994, 1998, 2005, 2007, 2010, 2012, 2015 and 2017 were years when the area affected by severe to exceptional drought increased. Some of these years are recognized as strong El Niño years (1998, 2015) or having an anomalously warm Tropical North Atlantic (2005, 2010). Note that the area impacted by drought was at least as large as that in other years (e.g., 2007, 2012, 2017). This suggests that the impacts of the exceptional droughts of 2005 and 2010 may be surpassed by the exacerbated dry condition already established in the region in recent years, and not necessarily attributable to El Niño or warm Tropical North Atlantic conditions. Consequently, during the period 2007-2020, more than 25% of the MATOPIBA region was affected by severe to exceptional drought in four years (2007, 2012, 2015 and 2017) (Fig. 3).
This trend towards an increase in drought frequency also occurs in a scenario of land cover changes in the EAC, characterized by conversions to pastures and croplands since the mid 1980s54. Global land cover satellite MODIS MCD12C1 products from 2001 to 2019 show that area of forest has diminished gradually at a rate of 2.8% (p<0.01) per decade in the last 20 years (Extended Data Figs. 5 and 6). In contrast, the combination of savanna, grasslands, and croplands increased significantly. Between 2002 and 2011, deforestation rates in the Cerrado (1% per year) were 2.5 times higher than in the Amazon basin (EMBRAPA-www.embrapa.br). This overall drying and warming trend over MATOPIBA depend on land cover (Extended Data Fig. 7). The warming trend over forests during the period 1981-2020 is 0.4 ºC/decade (p<0.05), increasing to more than 0.5 ºC/decade (p<0.05) for 2000-2020. This is consistent with a decrease in evaporation (-0.26 mm/day per decade, p<0.05) over forests during this last period, combined with an increase in the frequency of dry days, albeit non-significant (1.8 per decade). The warming trend, however, is not necessarily in line with precipitation deficits. Trends in precipitation over the different classes are not statistically significant (except for the savannah class and the period 1981-2020), and the trend towards dryness is stronger in the period 1981-1999. Trends in air temperature and evaporation over forests resemble trends over areas converted from forests to savannah and grasslands (Extended Data Fig. 8). Over these converted areas, specifically, trends in precipitation and dry-day frequency are slightly greater, but not statistically significant.
Long-term trends of air temperature and evaporation variables documented in this study show that dry conditions have intensified during the last two decades, suggesting increasing water stress on vegetation. Dry and dry-to-wet seasons in eastern Amazonia are becoming longer, warmer and dryer. While deforestation in eastern Amazonia along the deforestation arc has increased in recent years, intensified fire seasons have occurred when drought conditions affect this region, and the impacts of compounded drought-heat events in eastern Amazonia extend to the Cerrado vegetation along the Tocantins River basin.
Analysis of evaporation and dry-day frequency agrees with previous studies that show a relationship between a delay of the wet season onset and area deforested and could further reduce evaporation and exacerbate the dryness over Amazonia5,53.
Socioeconomic and ecological implications. In 2015 the Brazilian government declared MATOPIBA to be the country’s “last agricultural frontier”55. The expansion of the MATOPIBA agricultural frontier, driven by agribusiness, is imposing a new functional reorganization on the use and occupation of the territory. In recent years this region has sacrificed a large part of native vegetation for soybean and cattle production. The expansion of agribusiness, and especially of soybean monocultures, began in this region since the 1980s, and accelerated in the early 2000s53. In the last two decades, the area cultivated with soybeans in South America more than doubled, with most of the soybean-driven deforestation observed in the Cerrado56 (Extended Data Fig. 9). In 2012, there were about 2.5 million hectares cultivated with soy in the MATOPIBA, producing more than 7 million tons and generating revenues exceeding R$5.5 billion (USD 1.05 billon).
Currently, almost a quarter of the Cerrado’s soybean area is in MATOPIBA57, where most of the agricultural expansion in occurred over native vegetation - 68% (780,000 hectares) between 2000 and 2007, and 62% (1.3 million hectares) in the following period, between 2007 and 2014, especially in Maranhão and Piaui, the current agricultural frontier of the Cerrado.