Increases in the temperature seasonal cycle indicate long-term drying trends in Amazonia

The Amazon rainforest, a historically significant carbon pool storing up to 200 petagrams of carbon (PgC), has functioned as a carbon sink, absorbing 0.42-0.65 PgC yr⁻¹ from 1990-2007. However, climate change and deforestation threaten to transform it into a carbon source. Further warming and drying could trigger a tipping point, potentially leading to widespread dieback. While the possibility of Amazon dieback has been debated, recent CMIP6 climate models show greater agreement, with several models projecting localized abrupt dieback due to warming and drying associated with elevated CO₂ levels. The dry season in southern Amazonia has lengthened in recent decades, accompanied by prolonged fire seasons and increased frequency of dry days. This observed drying, exceeding natural variability, is linked to elevated greenhouse gas levels and deforestation. Tropical tree growth is strongly tied to dry season rainfall, and over three-quarters of the rainforest has shown resilience loss since the early 2000s, indicating an approaching critical threshold. Direct measurement of evaporation via eddy covariance is limited by sparse data across South America. However, changes in evaporation affect near-surface meteorology; since over half of Amazonian precipitation originates from evapotranspiration, reduced evaporation leads to increased temperatures. This study hypothesizes that changes in the amplitude of the temperature seasonal cycle (difference between maximum and minimum monthly means) will reveal changes in water availability, particularly during the dry season. Previous CMIP5 models showed a link between increased temperature variability and decreased evaporative fraction, especially in the southern hemisphere. However, CMIP5 models exhibited significant biases. The improved representation of evapotranspiration in CMIP6 models motivates this study to re-examine these relationships using both CMIP6 models and observational data from ERA5 reanalysis, aiming to estimate Amazon drying since 1900.

Existing research highlights the Amazon's role as a major carbon sink and its vulnerability to climate change and deforestation. Studies have documented increased dry season length, more frequent dry days, and increased hot extremes in recent decades. These changes are linked to elevated greenhouse gas levels and deforestation, impacting tropical tree growth and rainforest resilience. While direct measurements of evaporation are limited, previous research using CMIP5 models established a connection between temperature variability and evaporative fraction, although with substantial model biases. The improved representation of evapotranspiration in the newer CMIP6 models provides a stronger basis for investigating these relationships.

This study utilizes data from ERA5 reanalysis, other reanalysis products (NCEP-DOE R2, MERRA-2, JRA-55), CMIP6 climate models, and HadCRUT5 observational temperature data. The Amazon basin is divided into four regions (NWS, NSA, NES, SAM) as defined by the IPCC AR6. The analysis focuses on the relationship between the annual mean evaporative fraction (EF) and the temperature seasonal cycle amplitude. EF is calculated as LE/(LE+H), where LE is latent heat and H is sensible heat. The temperature seasonal cycle amplitude is defined as the difference between the maximum and minimum monthly mean temperatures for each year. The study first examines correlations between EF and temperature seasonal cycle amplitude anomalies (relative to 1979) in ERA5 reanalysis data for 1979-2020, and extends this analysis to other reanalysis products to assess robustness. The same analysis is performed using CMIP6 model data (historical and SSP5-8.5 scenarios from 1900-2099), also using SSP2-4.5 and a 1% CO₂ increase per year scenario for sensitivity analysis. Linear regressions are fit to the data to quantify the relationship between EF and temperature seasonal cycle amplitude. Spatial maps of correlations are generated to investigate regional variations. Finally, using the relationships derived from CMIP6 models, historical changes in EF are reconstructed from HadCRUT5 temperature data back to 1900, and compared to CMIP6 model projections under future global warming scenarios. All data is interpolated onto a 1°x1° grid for consistent comparison.

Analysis of ERA5 reanalysis data (1979-2020) reveals decreasing trends in evaporative fraction and increasing trends in temperature seasonal cycle amplitude in three out of four Amazon regions, indicating drying. Strong negative correlations between EF and temperature seasonal cycle amplitude anomalies are found in the NSA region (r=-0.61), with similar correlations in other regions, although weaker in areas near coastlines. Analysis of three additional reanalysis products shows consistent negative correlations, except for JRA-55. CMIP6 model data (1900-2099) shows even stronger negative correlations (r=-0.75 in NSA region), with a slope approximately double that of ERA5. Most CMIP6 models (23 of 25) exhibit strong negative correlations (r<-0.5) in the NSA region. Spatial analysis reveals heterogeneous correlations in ERA5 data, with high negative correlations in central and eastern Amazonia, but positive correlations near coastlines. CMIP6 model correlations are more homogeneous, with predominantly negative correlations across most of the Amazon basin. Reconstructing historical EF changes using HadCRUT5 data and CMIP6 derived relationships indicates a continuous downward trend in EF since 1900, consistent with CMIP6 model ensemble mean EF. CMIP6 projections suggest continued drying under future global warming, with a potential 5% decrease in EF at 2°C of global warming. The NSA region shows remarkable agreement between reconstructed and CMIP6 ensemble mean EF, indicating substantial drying at 3°C of global warming. While reconstructions slightly underestimate drying in NES and SAM regions.

The findings provide strong evidence for a robust drying trend in Amazonia, consistent across CMIP6 models and ERA5 reanalysis data. The strong correlation between reduced EF and increased temperature seasonal cycle amplitude highlights the impact of reduced evaporative cooling during longer and more intense dry seasons. The consistency of this relationship across various scenarios (SSP5-8.5, SSP2-4.5, 1% CO₂ increase) underscores its robustness. The agreement between reconstructed historical EF from temperature data and CMIP6 model projections strengthens confidence in the drying trend and its projected continuation under future climate change. The potential for substantial drying (5% decrease in EF at 2°C warming) raises significant concerns about Amazon forest dieback and its implications for the global carbon cycle and climate system.

This study demonstrates a robust and consistent drying trend in Amazonia, supported by multiple datasets and climate models. The strong correlation between evaporative fraction and temperature seasonal cycle amplitude provides a powerful tool for assessing past and future drying trends. Projections indicate continued drying under future climate change, raising serious concerns about Amazon forest dieback and its cascading impacts. Future research could focus on improving the representation of specific processes that might affect the climate sensitivity of the region or investigate the non-linear dynamics of drying in the Amazon.

The study relies on reanalysis and climate model data, which may contain biases or uncertainties. The spatial resolution of the data may limit the ability to detect localized drying events. The analysis focuses primarily on the relationship between EF and temperature seasonal cycle amplitude, and other factors influencing Amazonian drying (e.g., deforestation, land use change) are not directly addressed.