Abstract
The increase in the greenhouse gas emissions leads to changes in the mechanisms connecting the two major river basins in South America, the Amazon and La Plata basins, at subcontinental scale. Studies very often neglect to address the impact of the model-component choices on the projected change in precipitation in the two river basins. Within that context, the present study investigates the probable causes of changes in the hydroclimate of the two river basins through projections from three global climate models—driven by the pathway with no stabilization of the emissions growth by 2100—with focus on the warming of regions in the equatorial Atlantic and Pacific Oceans. Because the annual cycle of the precipitation differs in the northern and southern portions of the two river basins, changes are then preferably assessed in subregions. The model-dependent results project the following changes in the physical and dynamic mechanisms toward the end of the twenty-first century: (i) intensification of the 850-hPa northerly moisture flux from the western tropical Atlantic in the eastern side of the central Andes; and (ii) increase in the magnitude of the 200-hPa wind core whose location largely coincides with the La Plata basin. Those changes may increase the precipitation in the northern Amazon and southern La Plata basins by the end of the century. In contrast, the decrease in precipitation in the northern La Plata basin may result from the decrease in length of the rainy season associated with South American Monsoon System.
Similar content being viewed by others
Explore related subjects
Discover the latest articles and news from researchers in related subjects, suggested using machine learning.Avoid common mistakes on your manuscript.
1 Introduction
Among the biggest unknowns in the global climate projections are the changes in precipitation led by adjustments in physical and dynamic mechanisms at subcontinental scales, more likely to impact the frequency and intensity of extreme hydro-events. Despite the uncertainty in the assessment of the future climate, dynamically consistent modeling and observational studies have projected an increase in the frequency and intensity of droughts and floods around the world (e.g., Milly et al. 2002; Hirabayashi et al. 2013; Spinoni et al. 2019).
In South America, the Amazon and La Plata basins (henceforth, AMB and LPB, respectively) occupy a large portion of the continent. They are geographically connected over an extended area in central and southeastern South America. The water cycle in the two river basins is strongly influenced by global and subcontinental patterns (e.g., low-frequency modes of climate variability and South American Low-Level Jet) that cause the hydrological connection between them (Zanin and Satyamurty 2020).
Zemp et al. (2014) discussed AMB importance to the LPB hydroclimate through moisture recycling during the rainy season in AMB, using modeling approaches and observation-based products. However, studies using in situ observations in AMB might prove challenging because rain gauges are sparse and have discontinuities in their historical records, mostly due to the extension of the tropical rainforest, which makes it difficult to install and to maintain observational networks.
To help bridge that gap, remote sensing-based products, global reanalysis, and climate models are often used in studies on AMB. For instance, Garcia et al. (2018) and Panisset et al. (2018) pointed out the scarcity of long-term records for climate studies in AMB. The former took advantage of the recently developed South American hydroclimate reconstruction (Nunes 2016) and the latter of satellite-derived products in the assessment of major AMB droughts in the twenty-first century.
Changes in the AMB and LPB hydrological cycles throughout the twenty-first century have been assessed by both Global Climate Models (GCMs) and downscaled projections as reported in many scientific studies (e.g., Seth et al. 2010; Kitoh et al. 2011; Llopart et al. 2014; Duffy et al. 2015; Cabré et al. 2016; Penalba and Rivera 2016; Sorribas et al. 2016; Montroull et al. 2018; Zaninelli et al. 2019).
Further than the scientific reports, changes in the AMB and LPB regions have repeatedly been recorded through the public media, such as the increased deforestation and fire outbreaks in the Amazon rainforest (Libonati et al. 2021), and the larger interannual variability in Southeast South America (SESA) precipitation that might ultimately affect water availability at Itaipu Hydroelectric Power Plant in the future (e.g., Rivarola Sosa et al. 2011). Therefore, we examine changes in the interconnection of the two basins that might potentially have implications on water supply and hydropower production in SESA, which is a densely populated area in South America.
South American Monsoon System (SAMS; Zhou and Lau 1998) influences the summer precipitation regime in most of the continent, particularly across an extended band from the southeastern Amazon to southeastern Brazil. During the rainy season in South America, it is possible to notice a precipitation band above 4 mm/day, extending from AMB to Southeast Brazil, which is characteristic of the SAMS active phase. SAMS onset and demise are defined in Gan et al. (2004) through the average starting and ending months of October and April, respectively. This period is marked by maximum warming in the Southern Hemisphere, which favors convection in South America.
Han et al. (2019) investigated changes in precipitation from the global monsoon systems, using the Global Precipitation Climatology Project (GPCP; Adler et al. 2003) and the European Centre for Medium-Range Weather Forecasts (ECMWF) ERA-Interim reanalysis (ERA-I; Dee et al. 2011) datasets from 1979 to 2016. They concluded that the summer monsoons had increased rainfall values for the period covered by the study, indicating a strengthening of the monsoon systems around the world, with the exception of an observed decrease in SAMS precipitation.
Bombardi and Carvalho (2009) investigated the ability of the Coupled Model Intercomparison Project, Phase 3 (CMIP3; Meehl et al. 2007) GCMs to reproduce the SAMS characteristics in the twentieth century (1981–2000) through comparisons with satellite- and gauge-based products. They also examined the variability of SAMS by the end of the twenty-first century (2081–2100), using ten CMIP3 models driven by the A1B scenario. Their results indicated that most of the analyzed GCMs projected a decrease in summer precipitation over the monsoon region by the end of the twenty-first century. Similarly, using nine CMIP3 models driven by the A2 scenarios, Seth et al. (2010) found a decrease in September–November precipitation in central-western Brazil and continental South Atlantic Convergence Zone (SACZ; Kodama 1992, 1993; Satyamurty et al. 1998), which is a component of SAMS (e.g., Carvalho et al. 2004; Jones and Carvalho 2002), therefore indicating an increase in the duration of the SAMS dry season. They suggested that the decrease in the continental SACZ precipitation might be associated with its shift southward. On the other hand, their results showed a projected increase in the spring precipitation in SESA due to changes in South American Low-Level Jet (SALLJ; Virji 1981) as well as in South Atlantic Subtropical High (SASH; Sun et al. 2017) with impact on SACZ (Bombardi and Carvalho 2011; Bombardi et al. 2014; Sun et al. 2017).
In particular, SALLJ plays an important role in connecting the two basins, and is characterized by a northerly low-level wind stream that follows the eastern Andes, with diurnal variations, and associated with heat and moisture transports from AMB to SESA (Marengo et al. 2004; Silva et al. 2009). According to Jones (2019), SALLJ might occur in separated periods or even simultaneously in the northern and central Andes at the lee side and presents significant variability at seasonal, annual and decadal timescales. Furthermore, the low-level moisture and heat transport from AMB has been shown to influence the LPB precipitation (e.g., Marengo et al. 2004; Vera et al. 2006; Soares and Marengo 2009) due to its role in the occurrence of mesoscale convective complexes (MCCs; Velasco and Fritsch 1987). Zipser et al. (2006) identified SESA among the regions with the strongest thunderstorms on Earth. These super-storms over land are more likely associated with extremely large MCCs, responsible for extreme events (e.g., floods and landslides), developed in dynamic and thermodynamic “favored zones” (Laing and Fritsch 1997), such as SESA.
Upper-level divergence and low-level moisture convergence are main environmental conditions that favor the development of MCCs (Laing and Fritsch 2000). At the upper troposphere, the westerly jet stream triggers divergence at its equatorward (poleward) entrance (exit) (e.g., Uccellini and Johnson 1979; Uccellini and Kocin 1987; Nunes and Roads 2007), regardless of the hemisphere considered. Therefore, the upper-level divergence at the equatorward (poleward) entrance (exit) of the Subtropical Jet (STJ; Krishnamurti 1961) core may interact with the moisture flux from SALLJ to form the MCCs in SESA through their relative positioning (Durkee et al. 2009). The two jets’ interaction would allow that deep-convection takes place above the exit of the low-level, northerly warmer flow that carries moisture from AMB to south-southeastern LPB, which would explain the highest occurrence of South American MCCs between 20° S and 30° S (Durkee and Mote 2009).
Cavalcanti (2012) described the two distinct precipitation regimes in LPB, one located in the northeastern and central areas associated with the active phase of SAMS, and the other in the southern region that includes northern Argentina, southern Paraguay and the western part of southern Brazil, and Uruguay. In the southern LPB, both frontal and mesoscale convective systems influence precipitation. However, the latter individually is responsible for large precipitation amounts in the region (Durkee et al. 2009) with the highest frequency of MCCs (Nesbitt et al. 2000).
Silva et al. (2009) analyzed the influence that different El Niño (La Niña)–Southern Oscillation (ENSO; Trenberth 1997) phases have on SALLJ in summertime, using composites of precipitation anomalies from rain gauges, together with 6-hourly global reanalysis outputs from 1977 to 2004. Their results indicated that during El Niño (La Niña), SALLJ increases (decreases) in frequency and intensity, as well as intensifies the southern (northern) LPB precipitation because of its preferential positioning toward the south (north) of the basin. They also concluded that STJ becomes stronger (weaker) during El Niño (La Niña) events. Furthermore, Andreoli et al. (2017) assessed the impact of El Niño on precipitation in South America, considering three different Niño regions in the equatorial Pacific, specifically: central, eastern and mixed. They concluded that the positive SST anomalies of the eastern-Pacific Niño increase precipitation in eastern Brazil and SESA in comparison to the two other Niño types, through a stronger SALLJ, for all seasons.
Using 16-year Tropical Rainfall Measuring Mission (TRMM) precipitation radar datasets and storm composites from ERA-I, Bruick et al. (2019) determined that storms can develop at the eastern side of the Andes in SESA, regardless the ENSO phases; however, they also concluded that the strengthening and relative positioning of SALLJ and the upper-level jet stream that define the storm structures are controlled by ENSO-type conditions. As a result, El Niño conditions favor deeper/taller storms, which are more prone to develop in areas beneath the equatorward entrance of a strong upper-level jet stream.
ENSO’s phases drive the behavior of precipitation, temperature and evapotranspiration in AMB as discussed in Moura et al. (2019), using satellite-based products from January 2000 to December 2016. Their results indicated decrease (increase) in precipitation, and increase (decrease) in temperature and evapotranspiration in El Niño (La Niña) years when compared to neutral years.
The equatorial Atlantic Ocean also plays an important role in shaping precipitation in South America through teleconnections with the tropical Pacific Ocean. This coupled ocean–atmosphere mode—Atlantic Equatorial Mode (AEM) of variability or Atlantic Niño—was defined in Zebiak (1993) through SST anomalies from 1967 to 1988, area-averaged over 3° N–3° S and 20° W–0°, as the ATL3 index (hereafter, ATL3), which, despite their similarities, exhibits weaker SST anomalies and wind patterns than Niño 3.
Rodríguez-Fonseca et al. (2009) demonstrated through observational data and model outputs that ATL3 shows larger SST anomalies in the boreal summer; and it influences ENSO’s variability in the following winter. They also showed that a positive ATL3 phase would strengthen the ascending (descending) branch of the Walker circulation over the Atlantic (central Pacific), which might lead to La Niña-type response. Tokinaga et al. (2019) reexamined this Atlantic–Pacific Niño connection. They called it a two-way mode interaction, with positive ATL3 phases associated with multi-year La Niña events. Torralba et al. (2015) also stated that the ENSO and AEM are the two most important phenomena of ocean–atmosphere interaction in the tropical ocean, and that both influence precipitation variability in northeastern South America (NESA), which encompasses a large portion of AMB. Hounsou-Gbo et al. (2020) discussed the activity of AEM from October through December in the ATL3 region, and that secondary peak observed in the year might influence an earlier development of ENSO in the Niño 3 region.
Although many studies have addressed projected changes in the AMB and LPB precipitation, they have rarely associated those changes with the dynamic and physical processes at subcontinental scale, and the model-component choices. Therefore, to seek supporting processes that ultimately lead to an increasing in extreme hydro-events, we investigate changes in precipitation throughout the twenty-first century in the northern and southern portions of AMB and LPB, from December through April, corresponding to the rainy season over most of South America. To connect the changes in precipitation to the model-dependent responses, we use three GCMs from the Coupled Model Intercomparison Project, Phase 5 (CMIP5; Taylor et al. 2012) under the most extreme climate change forcing, the Representative Concentration Pathway 8.5 (RCP8.5; Moss et al. 2010).
We chose those three global models because of their complexity and similarities, which would explain any changes in their results could be traced back to controlled differences in their components. For that, changes in the surface temperature in the equatorial Atlantic and tropical Pacific Oceans are sought through the models’ ocean components, together with changes in the low-level meridional moisture flux and upper-level zonal wind resulting from their atmospheric components, both influenced by the ocean–atmosphere coupling responses from each of the global models that might be explained through a matrix of the model components.
2 Methodology
2.1 Models
This study uses monthly outputs from three climate models from the CMIP5 project, namely: Coupled Model 3 (CM3; Donner et al. 2011), and two Earth System Models (ESMs), ESM2G and ESM2M (Dunne et al. 2012, 2013). All three global models were developed at the National Oceanic and Atmospheric Administration (NOAA) Geophysical Fluid Dynamics Laboratory (GFDL).
According to Dunne et al. (2012, 2013), the three GFDL models were developed from the Coupled Model version 2.1 (CM2.1; Delworth et al. 2006), with two different views. The first focuses on the ESMs, with improved representation of biogeochemical processes, with the same atmospheric and ice components as in CM2.1. Therefore, ESM2G and ESM2M are similar, except for their different ocean components. Although other components remain as in its predecessor CM2.1, the second view aims for better representation of the atmospheric processes through CM3, which uses an updated version of the atmospheric model, with a new dynamic core grid and more atmospheric-based processes, including new aerosol–cloud interactions, and increased chemistry-climate and troposphere–stratosphere responses (Donner et al. 2011).
The two ESMs employ version 2 of the Atmospheric Model (AM2; Anderson et al. 2004), and CM3, version 3 (AM3; Donner et al. 2011). AM2 uses a 2.0225° latitude × 2.5° longitude horizontal grid, and a 24-level hybrid vertical coordinate with σ-coordinate at the lower levels of the atmosphere, where the lowest level is located 30 m above ground and the top-level at 40 km. In turn, AM3 uses a new dynamic core represented in a finite volume cubed-sphere grid, with higher horizontal and vertical resolutions than AM2, precisely twice the number of the vertical levels in AM2, and the model’s surface and top located at the 1013.25- and 1-hPa levels, respectively.
Both ESM2M and CM3 employ the Modular Ocean Model 4p1 (MOM4p1; Griffies et al. 2011), and ESM2G, the Generalized Ocean Layer Dynamics (GOLD; Adcroft and Hallberg 2006; White et al. 2009), both ocean models developed and implemented at GFDL. MOM4p1 uses a horizontal grid of about 1°, increasing from 30° S to maximum resolution of 1/3° at the equator; and 50 vertical levels in total, with 22 levels of 10-m thickness within the first 220 m of the ocean. GOLD uses a horizontal grid of 1° that varies meridionally up to 1/3° at the equator and 58 layers.
Table 1 provides a matrix of the atmospheric and ocean components of the GFDL global models used in this study, as well as available horizontal resolutions. We then consider two 30-year periods in the GCM analyses, specifically: current climate (1976–2005), and future climate or projection for the end-century (2071–2100).
An assessment of the models’ ability to reproduce seasonal precipitation features of the current climate is initially performed using observation-based, global precipitation products.
2.2 Observation-based products
Monthly means from observation-based precipitation analyses comprehend three widely used datasets: (1) the Climate Research Unit (CRU) (Harris et al. 2014) version 4.01 (CRU TS 4.01) from the University of East Anglia; (2) the Full Data Monthly Product V.2018 (V.8) from the Global Precipitation Climatology Centre (GPCC) (Schneider et al. 2017); and (3) only precipitation analyses from the University of Delaware Air and Precipitation (UDEL) (Willmott and Matsuura 2001), version 4.01. Overall, the three observation-based products exhibit small variations, according to how their quality-control and interpolation methods are applied to the available raw datasets, providing a range of admissible results that allow better assessing the reliability of the models’ outputs. Here, an arithmetic mean of the seasonal climatology of three observation-based products is computed and used as reference in the models’ comparisons. Table 2 summarizes the main characteristics of the three precipitation products, namely the spatial resolution and temporal coverage.
2.3 Region of interest
This study was carried out using the hydrologic delimitations of AMB and LPB. As depicted by Fig. 1, each basin was divided into its northern and southern portions. Differences between the elevation heights from the coarse-resolution models (Fig. 1, CM3’s grid cell orography in meters) and the actual heights could reach several hundred meters, particularly in the Andes, implying several shortcomings in the representation of the physical processes, resulting in large systematic errors in the precipitation amounts. Furthermore, problems with excessive precipitation in global and mesoscale models over steep-high mountains have been reported by previous studies, as discussed in Chao (2012). Therefore, we removed a narrow region encompassing the higher elevations of the Andes (Fig. 1) from our statistical analyses.
Model orography (m) based on CM3 degraded to the 2.5° × 2.5° resolution. Showing the four regions used in the area-averaged precipitation computations, for a the Amazon river basin (AMB) and b La Plata river basin (LPB). Black contour delimitates the river basins. The dashed lines divide AMB and LPB into their northern and southern portions at the latitudes of 6° S and 27° S, respectively. Color bar at 50 m intervals
2.4 Pre-processing and statistical analysis
All GCM outputs (CM3, ESM2M and ESM2G), and CRU and UDEL were area-weighted interpolated to match the GPCC 2.5° grid cell, using a latitude-weighted, interpolating function from the Grid Analysis and Display System (GrADS), which allows manipulation and visualization of meteorological and climate data, and currently available at http://cola.gmu.edu/grads/. Two distinct time periods were used as follows: the current (historical), from January 1976 through December 2005, and the end-century, from January 2071 through December 2100. The future climate outputs were driven by RCP8.5, which is the pathway with no stabilization of the emissions growth by 2100.
For each dataset, seasonal means corresponding to the rainy and dry seasons in the central part of South America, here represented by DJFMA (December, January, February, March, and April) and JASON (July, August, September, October, and November), respectively. The datasets were obtained by averaging the corresponding monthly values.
Moreover, seasonal means (DJFMA and JASON) for the current climate were obtained at each grid point belonging to AMB (69 points) and LPB (42 points), for each of the GFDL model outputs (CM3, ESM2M and ESM2G) and observation-based precipitation products (CRU, GPCC and UDEL), and its mean (OBS), which is used as the reference. Then, the standard deviation, the centered root-mean-squared error (RMSE) and the Pearson’s correlation coefficient values were separately computed and used in the Taylor Diagram (Taylor 2001) to compare the seasonal means of the GFDL models and observation-based products with OBS over AMB and LPB.
Additionally, the time series of precipitation monthly means of the GFDL model outputs for the current and end-century climates, and the observed-based product that best compares with OBS from the Taylor Diagram analysis were area-averaged over four distinct areas, as follows: (1) northern AMB (NAMB); (2) southern AMB (SAMB); (3) northern LPB (NLPB); and (4) southern LPB (SLPB). The four subregions analyzed here are influenced by global and subcontinental patterns and have distinct precipitation annual cycles. The 30-year subsets are then used in the analysis of the current and end-century annual cycles, and their interannual variability through boxplots. Boxplots were processed using an R package for statistical computing (R; R Core Team 2018).
To assess the changes in SST in specific regions of the equatorial Atlantic and Pacific, ATL3 (3° N–3° S, 20° W–0°; Zebiak 1993) and Niño 3.4 (5° N–5° S, 120°–170° W; Trenberth 1997), each of the three global models had their SSTs monthly means from the current climate subtracted from their corresponding values from January 2006 throughout December 2100. We used a centered moving average of 11 months to reduce the impact of the high-frequency modes on the resulting anomaly time series.
The SST anomalies were also computed in the ATL3 and Niño 3 regions for the current and end-century climates of each GFDL model, using the averaged monthly means from October to December (OND) and December to February (DJF), respectively. To determine the OND-AEM and DJF-ENSO impacts on the DJF 850-hPa moisture flux (as well as on its meridional component), which is very often associated with extremes in NESA and SESA precipitation fields, composites were then built from the years of OND-AEM and DJF-ENSO. We used the thresholds of ± 0.35 °C (Hounsou-Gbo et al. 2020) and ± 0.5 °C, which is currently the NOAA Climate Prediction Center Oceanic Niño Index threshold (Glantz and Ramirez 2020), to separate the OND-AEM and DJF-ENSO years from those presenting neutral conditions; subsequently, the DJF 850-hPa moisture flux (vector) and the meridional component means from the neutral years were subtracted from their corresponding means obtaining from the OND-AEM- and DJF-ENSO-driven years.
Furthermore, to evaluate changes in the end-century climate from the current, we computed the DJFMA climatological values for the following: the 850-hPa moisture flux and its meridional component, and the 200-hPa wind vector and its zonal component for each model and the ensemble mean (EGCM), which is an arithmetic mean of the GFDL models’ seasonal (DJFMA) climatology.
The projected changes in the DJFMA precipitation (%) for the end-century climate were assessed for each of the GFDL models and EGCM, only over the region encompassed by the hydrologic delimitations of AMB and LPB.
Lastly, normal distributions were considered the best fit for the precipitation monthly means in the rainy months (DJFMA) of the current and end-century climates for each of the four-region subsets, being the null hypothesis accepted at the 5% level of significance, using the two-tailed critical values from a standard Gaussian (normal) distribution to test differences of the mean for paired samples (Wilks 2011). The probability of occurrence of extreme precipitation was also analyzed through empirical cumulative distribution functions (ECDFs) using the R packages built from Venables and Ripley (2002), Garrett (2013, 2018) and Marchini et al. (2019) for the current and end-century climates from, in this case, time series of the mean value of the DJFMA precipitation, areal-averaged as described earlier. In this analysis, the Kolmogorov–Smirnov test statistic at the significance level of 5% (K–S test; Wuertz et al. 2020) determined which sample was stochastically greater than the other and, by comparison, which ECDF is significantly leaning toward the higher precipitation values, depending upon the maximum distance (D+) between ECDFs to be superior to its critical value (Dc+), the latter obtained from the two samples of equal size, one-sided K–S test table found in Rohatgi and Ehsanes Saleh (2015).
3 Results and discussion
3.1 Precipitation assessment in the current climate
The ability of the GFDL models to reproduce precipitation over the two river basins delimited as in Fig. 1 is assessed through the Taylor Diagram (Fig. 2).
Taylor Diagram for precipitation in AMB and LPB, for a, b DJFMA and c, d JASON, respectively. The blue dotted curves indicate the centered RMSE (mm/day); the gray solid curves, the standard deviation (mm/day); the gray solid lines, the Pearson’s correlation coefficients. The numbers (in color) represent each one of the GFDL models (CM3, ESM2M and ESM2G), and the observations (CRU, GPCC and UDEL), according to the legend on the right. The black star represents the reference value, which corresponds to the observations’ mean (OBS)
In AMB for DJFMA (Fig. 2a), the ESMs correlated to some degree better than CM3 with OBS, despite the lowest RMSE of CM3. The correlation (RMSE) values were 0.694 (2.066 mm/day) for ESM2G, 0.663 (2.035 mm/day) for ESM2M, and 0.648 (1.818 mm/day) for CM3. Among the observations, GPCC and UDEL displayed a higher (lower) correlation coefficient (RMSE) in comparison with OBS than CRU, 0.982 (0.322 mm/day), 0.982 (0.348 mm/day) and 0.953 (0.506 mm/day), respectively. The spatial variability given by the standard deviation showed its lowest value for CRU (1.646 mm/day), followed by OBS (1.666 mm/day), GPCC (1.698 mm/day), UDEL (1.793 mm/day). The GFDL models overestimated the standard deviation of the DJFMA precipitation in comparison with the each of the observations and OBS, with CM3 displaying the lowest standard deviation of 2.383 mm/day, and then ESM2M and ESM2G with 2.712 and 2.840 mm/day, respectively.
Unlike DJFMA in the AMB region, in LPB (Fig. 2b), ESM2M was the model with the highest correlation coefficient (RMSE), 0.764 (1.446 mm/day), in comparison with the other two GFDL models, followed by ESM2G, 0.735 (1.424 mm/day); and CM3 showed the lowest correlation (RMSE), 0.723 (1.341 mm/day), among the models. GPCC and UDEL had the same and slightly higher correlation coefficients (0.995) than CRU (0.992), which had the highest RSME (0.170 mm/day) followed by GPCC and UDEL with 0.142 and 0.125 mm/day, respectively. In LPB, GPCC exhibited the lowest standard deviation, 1.082 mm/day, followed by OBS, UDEL and CRU, with values of 1.164, 1.192, 1.241 mm/day, respectively. An overestimation in the standard deviation of the DJFMA precipitation was also present in the GFDL models in comparison with each of the observation-based products and OBS, with 1.915, 2.040 and 2.126 mm/day for CM3, ESM2G and ESM2M, respectively.
In AMB for JASON (Fig. 2c), CM3 correlated at 0.851 with OBS and showed the lowest RSME (1.1 mm/day) in the dry season. Although significant at 95% confidence level, ESM2M (0.546) and ESM2G (0.359) showed lower correlation with OBS in comparison to their values in the rainy season, but lower RMSE (1.473 and 1.630 mm/day) due to smaller precipitation amounts in the dry season. Among the observations, GPCC was better correlated with OBS at 0.994, in comparison with UDEL (0.984) and CRU (0.976) and had the smallest RMSE (0.197 mm/day). ESM2G (0.622 mm/day), ESM2M (0.787 mm/day and CM3 (0.880 mm/day) displayed lower values of the standard deviation in comparison with UDEL (1.655 mm/day), GPCC and OBS (1.747 mm/day), and CRU (1.922 mm/day), which reveals an underestimation of the standard deviation of the JASON precipitation by the GFDL models in AMB, therefore, lower spatial variability than the observation sets.
In LPB for JASON (Fig. 2d), the observation sets were almost indistinguishable from OBS, with correlation coefficient (RSME) values of 0.998 (0.060 mm/day), 0.996 (0.085 mm/day) and 0.994 (0.110 mm/day), for UDEL, GPCC and CRU, respectively. ESM2M and ESM2G presented similar behavior and are not significantly correlated with OBS at 95% confidence level, with correlation coefficients of − 0.146 and − 0.164, and RSME of 1.159 and 1.175 mm/day, respectively. CM3 was remarkably correlated with OBS at 0.613 and had the lowest RSME of 0.808 mm/day, in comparison to the ESMs. Like AMB, the GFDL models displayed very lower standard deviation values in LPB, specifically, 0.352, 0.501 and 0.516 mm/day, for CM3, ESM2M and ESM2G, in comparison with the standard deviation values of the observation-based products, 0.972, 0.974, 0.975 and 0.990 for CRU, UDEL, OBS and GPCC, respectively.
Although exhibiting nearly the same correlation coefficients, the observation-based sets were more scattered in the Taylor Diagram for AMB (Fig. 2a, c), mostly due to their higher and more distinct RSMEs in comparison to LPB (Fig. 2b, d), possibly associated with the reduced number of available rain gauges in the Amazon region. All correlation coefficients displayed in the Taylor Diagram were significant at a 95% confidence level, except for those from the ESMs for JASON in LPB.
Among the observation-based products, GPCC emerged as the one with the best agreement with OBS and, therefore, used in the subsequent analyses.
3.2 Precipitation variability and change in the two river basins
The variability and change in the areal-averaged precipitation are now assessed over northern and southern AMB and LPB. Figure 3 exhibits boxplots and the annual cycles (solid blue line) for GPCC (Fig. 3a–d), and the GFDL models (Fig. 3e–p), considering time series of monthly mean precipitation (mm/day) for the current and end-century climates, for each of the four subregions as displayed in Fig. 1. From left to right, Fig. 3 shows NAMB, SAMB, NLPB and SLPB regions, respectively. The variability of the precipitation series for each month is given by the difference between the 1st quartile and the 3rd quartile, i.e., the interquartile range (IQR), and their distance to the median (2nd quartile) in each boxplot. Figure 3 also displays whiskers showing the series minima and maxima falling inside ± 1.5 times IQR.
Boxplots for the monthly mean precipitation (mm/day) for the current (Current) and end-century (End) climates. From left to right, displaying northern AMB (NAMB), southern AMB (SAMB), northern LPB (NLPB) and southern LPB (SLPB), for: a–d GPCC, e–h CM3, i–l ESM2M and m–p ESM2G, respectively. The blue (magenta) color denotes the current (end-century) climate. Solid lines exhibit the precipitation annual cycles. Whiskers display the minimum and maximum monthly values in the time series falling inside ± 1.5 times IQR
In NAMB, GPCC (Fig. 3a) shows through the annual cycle and median that the rainy season begins in December and extends to May (mean values above 6 mm/day), with maxima occurring in April and May (above 9 mm/day), while the dry season begins in June (abrupt decrease in the precipitation mean) and extends to November, with the minimum occurring in September. It is noteworthy that GPCC presents the lowest intra-seasonal variability, with the annual cycle ranging from a maximum above 9 mm/day in the rainy season and a minimum around 5 mm/day in the dry season. January has the largest IQR, with a reduction in the ascending slope between January and February, while August and November have the lowest IQRs. Torralba et al. (2015) discuss the bimodal or unimodal character of the precipitation annual cycle in NESA, and they attribute that to changes in the positioning of the Intertropical Convergence Zone (ITCZ), which is the tropical belt of convective clouds and rain, reaching its southernmost position in the austral autumn, and known to impact precipitation in Northeast Brazil (Utida et al. 2019). For the CM3 (Fig. 3e) and ESM2M (Fig. 3i) current climates, the annual cycles display bimodal behavior with maximum in December in both models, and in May and April, respectively. The ESMs present higher interannual variability (larger IQRs) than CM3 in the rainy season. In comparison to CM3 and ESM2M, ESM2G (Fig. 3m) has less marked secondary maximum, which is shifted to January. The ESMs show higher interannual variability in December, and maxima occurring in April and May (around 8 mm/day). All three models underestimate the mean monthly values in NAMB, with a reduction in values for the austral summer months with a relative minimum in February, showing higher intra-seasonal variability in comparison with GPCC. ESM2M presents the greatest IQR among models and GPCC in the rainy season, which might be attributed to an increased interannual variability from December through May.
SAMB has an annual cycle that is characteristic of the SAMS regime, with well-defined rainy and dry seasons. The rainy season in GPCC (Fig. 3b) comprises the period from October (SAMS onset, around 5.5 mm/day) to April (SAMS demise, approximately 6 mm/day), with a maximum above 9 mm/day in January, while the dry season comprises the period from May to September, with a minimum occurring in July. In general, the GPCC monthly means do not have much interannual variability. The GFDL models reproduce the annual cycle of the current climate close to that observed. However, maxima in the GFDL models occur in different months. In CM3 (Fig. 3f), the maximum in the annual cycle is in December, while in ESM2M (Fig. 3j) and ESM2G (Fig. 3n), it occurs in March. In addition, CM3 and ESM2M underestimate, while ESM2G slightly overestimates precipitation in comparison to GPCC in the rainy season, and all three models underestimate precipitation in the dry season. Like in the northern region, the ESMs show higher interannual variability in the rainy season than CM3 in SAMB. The intra-seasonal variability of the three models follows SAMS.
Like SAMB, NLPB is also marked by SAMS regime, which is characterized in GPCC (Fig. 3c) by the rainy season from October (onset) through March (demise), with a maximum in the annual cycle of the monthly mean precipitation in January (above 6 mm/day); and the dry season defined from April to September, with minimum in July (approximately 1 mm/day). The interannual variability is higher for the current climate of the three models, as well as the intra-seasonal variability, in comparison with GPCC. The higher intra-seasonal variability indicates a shorter (longer) rainy (dry) season for the three models in comparison with GPCC. Precipitation in average below 4 mm/day in October suggests later monsoon onset. Both CM3 (Fig. 3g) and ESM2M (Fig. 3k) place their maxima in January, approximately 8 and 7 mm/day, respectively. ESM2G (Fig. 3o) has its maximum shifted to February (around 7–8 mm/day). ESM2M has the highest interannual variability, followed by ESM2G and CM3.
In SLPB, the annual cycle of GPCC (Fig. 3d) shows the lowest intra-seasonal variability in comparison to the other regions in agreement with Cavalcanti (2012), which also states that the precipitation in the southern region of LPB is highly influenced by the frontal systems passing through the region. GPCC exhibits a difference of 2.5 mm/day on average between rainy (October to April) and dry (May to September) seasons; however, with the highest observed interannual variability in the rainy season, which might be due to the tropical Pacific variability (Pacific Niño) that modulates the occurrence and intensity of extreme hydro-events in that region (Bruick et al. 2019). Although with maxima in March and April, the difference within the rainy months is very small. The precipitation does not exceed 5 mm/day. The models’ current climate show differences in their annual cycles, with well-marked maxima in February for CM3 (Fig. 3h) and ESM2M (Fig. 3l), and in January for ESM2G (Fig. 3p) that differs from GPCC. Therefore, the models’ intra-seasonal variability is higher than the observed, but with lower interannual variability in comparison to the other regions. All the three models underestimate precipitation throughout the year in comparison with GPCC.
In Fig. 3, the end-century projections, especially for CM3 in NAMB (Fig. 3e), show a decrease in precipitation from June to November and an increase in the length of the dry season. It is remarkable the increase in maxima from the current climate to the end-century for all models (Fig. 3e, i, m). The interannual variability remains higher for the end-century projections in the rainy months, in particular for the ESMs (Fig. 3i, m). The intra-seasonal variability increases for the three models, with the dry season becoming drier and extended, and the rainy season wetter, but shorter. The CM3 and ESM2M resemblances in their annual cycles, such as the enhanced double maxima in the NAMB rainy season, might be controlled by a similar ocean component (MOM4p1 from Table 1).
Climate projections from the three models indicate that similar precipitation patterns are observed in the SAMB and NLPB regions (Fig. 3f, g, j, k, n, o), where the models project decrease in October and November, with ESM2G (Fig. 3o) displaying the largest reduction compared to CM3 (Fig. 3g) and ESM2M (Fig. 3k). In NLPB, the results show that the dry season tends to get drier, and extends to the months of October and November, causing the delay of the SAMS onset by the end of the century. CM3, ESM2M and ESM2G show a reduction in annual-cycle maxima in the end-century climate in comparison to the models’ current climate. This indicates a reduction in the average precipitation over the active phase of SAMS by the end of the century.
The models’ projections in SLPB (Fig. 3h, l, p) show a decrease in precipitation, especially from July to September, in comparison to the current climate. They also display an increase in precipitation during most of the rainy-season months, especially CM3 (Fig. 3h), with higher interannual variability for CM3 (Fig. 3h) throughout the year. As a result of the shortening of the rainy season, CM3 reaches its maximum in December, with a secondary maximum in February. ESM2M and ESM2G (Fig. 3l, p) have comparable behaviors, however, ESM2M shows higher interannual variability than ESM2G, with a distinct maximum in February.
3.2.1 Changes in supporting mechanisms of extremes
The tropical SST influences precipitation in various parts of the globe, including South America (e.g., Pezzi and Cavalcanti 2001; Coelho et al. 2002). Figure 4 displays time series of the 11-month centered moving average of SST anomalies for the GFDL models driven by RCP8.5 and computed in the Niño 3.4 and ATL3 regions, using the current climate as the background period (base field). All three models show positive trends for both Niño 3.4 and ATL3 anomalies toward the end of the century. They also show that the ATL3 anomalies are out of phase with and embedded in the Niño 3.4 anomalies, which exhibit higher variability. CM3 (Fig. 4a) reaches the highest values of change above 4 °C by 2100, for Niño 3.4 and ATL3. Despite distinct ocean models, ESM2M (Fig. 4b) and ESM2G (Fig. 4c) project similar changes in the smoothed SST anomalies around 2-3 ºC by 2100, however, ESM2M displays the highest variability in Niño 3.4, followed by CM3. In comparison with their respective current climates, the three GFDL models project an increase in the smoothed SST anomalies by 2100, with the largest increase seen in CM3 as discussed in Roy (2017), while the ESMs project a more modest increase. Although not so evident for CM3, the ESMs mainly exhibit a weakening of the Niño 3.4 SST anomalies by the end of the century, which might be related to a decline reported by Jia et al. (2019) in the Atlantic–Pacific Niño connection under a warmer climate.
Time series of the 11-month centered moving average of the anomalies computed from the SST (°C) monthly means driven by RCP8.5, area-averaged in the ATL3 (blue curve) and Niño 3.4 (magenta curve) regions, for the GFDL models: a CM3, b ESM2M and c ESM2G. Displaying June 2006 through July 2100. The SST monthly means from January 1976 to December 2005 provided the base field used in the computations of the anomalies
Table 3 exhibits the number of events for each mode’s phase, where ATL3− and ATL3+ represent the negative and positive phases of OND-AEM, respectively. CM3 shows different behaviors in the Niño 3.4 and ATL3 regions, with an increase from 8 (10) to 13 (11) for the La Niña (El Niño) events, and a decrease from 8 (10) to 6 (9) for ATL3− (ATL3+). ESM2M exhibits a decrease in the number of events toward the end of the century for both climate modes, with La Niña (El Niño) going from 11 (11) to 8 (8), and, especially in the Atlantic equatorial basin, with ATL3− (ATL3+), from 9 (9) to 3 (5). ESM2G behaves opposite to CM3, with La Niña (El Niño) presenting substantial reduction from 10 (10) to 4 (5), and most of doubling ATL3− (ATL3+) events from 3 (4) to 7 (9).
Figures 5 and 6 display the composites of the DJF moisture flux (vector) and its meridional component at the 850-hPa level for DJF El Niño and La Niña (Fig. 5), and for OND-AEM negative and positive phases (Fig. 6). From Fig. 5, we identify two anomaly patterns of the 850-hPa moisture flux in South American region that encompasses both basins, specifically: El Niño-type (EN) and La Niña-type (LN). EN exhibits a southeasterly difference in the moisture flux over northern and western AMB, and from north in central-eastern AMB, the SALLJ region in the eastern side of Andes and central and eastern LPB, which might contribute to the (hydro)extremes in SESA, and an anticyclonic feature over central-northern LPB that might inhibit precipitation. LN displays an opposite pattern with a predominant northerly difference in the moisture flux in northern and western AMB (increasing moisture availability), and a southerly one in central-eastern AMB, the SALLJ region, and central and eastern LPB that might reduce moisture availability to precipitation extremes, and a cyclonic feature over northeastern LPB that might contribute to an increase in precipitation. EN and LN are well captured by the ESMs in the current climate. CM3 shows reinforced EN and LN in the end-century climate in comparison to the current climate (Fig. 5a–d), while ESM2M, the opposite (Fig. 5d–h). ESM2G (Fig. 5i–l) only displays EN and LN consistent patterns in the current climate (Fig. 5i, k). For the end-century climate, the ESM2G moisture flux exhibits a less-coherent EN (LN), displaying an enhanced southeasterly (northwesterly) difference from northeastern LPB (northwestern AMB) to northwestern AMB (northeastern LPB) (Fig. 5j, l). It is noticeable the reduction in the El Niño (La Niña) events to about a half in the ESM2G end-century climate (Table 3), which might highlight a feature from a singular event rather than an overall pattern. The different ocean component might also play a role in such distinct patterns, because ESM2G has the lowest SST interannual variability among the models in the Niño 3.4 region (Fig. 4c), considering the current climate as the background.
DJF-ENSO-driven composites for the moisture flux (vector) and meridional component (color shaded) at the 850-hPa level for the DJF mean (values in g kg−1 m s−1). From left to right, showing the current and end-century climates for the El Niño and La Niña anomaly composites, respectively. Displaying a–d CM3, e–h ESM2M and i–l ESM2G. Black contour delimitates AMB and LPB hydrological boundaries
Same as in Fig. 5, except for OND-AEM
Examining Fig. 6, OND-AEM negative (positive) is usually associated with an earlier El Niño (La Niña) development. Therefore, EN (LN) might be found in their associated composites of the DJF moisture flux at 850-hPa level. Except for OND-AEM negative in the current climate (Fig. 6a), CM3 displays LN patterns (Fig. 6b–d) that are enhanced toward the end of the century for the OND-AEM positive phase (Fig. 6c, d). ESM2M (Fig. 6e–h) shows LN behavior in the current climate regardless the OND-AEM phases (Fig. 6e, g); however, for the ESM2M current climate, LN is well-defined in its positive phase (Fig. 6g), which might reinforce La Niña conditions in the current climate. For the ESM2M end-century climate, LN weakens, especially in the negative phase (Fig. 6f). For the negative phase, although less defined, ESM2G shows a LN pattern (Fig. 6i, j), which improves in the end-century climate. For the positive phase, ESM2G consistently displays LN in both climates, but enhanced in the current climate (Fig. 6k, l), despite the increased number in the end-century climate (Table 3).
One might conclude that the warming of the equatorial Atlantic and Pacific basins produced different coupled responses in the GFDL models by the end of the century. Nevertheless, the three models have coherently shown EN and LN patterns associated with their corresponding 850-hPa moisture fluxes in the DJF-ENSO-driven composites, and, to some extent, have also replicated them in those driven by OND-AEM.
End-century changes in the 850-hPa meridional moisture flux are shown in Fig. 7 for DJFMA. Over most of South America, the projected changes might be associated with changes in SALLJ and in the moisture availability from the western tropical Atlantic. All models project an increase in the northerly 850-hPa moisture flux between 15° S and 35° S, with a maximum located near 20° S (Bolivia–Paraguay border), which is described in the literature as the region of maximum occurrence/activity of SALLJ (Marengo et al. 2004; Oliveira et al. 2018).
End-century projected changes from the current climate in the moisture flux (vector) and meridional component (color shaded) at the 850-hPa level (values in g kg−1 m s−1) (upper panels), and in the wind (vector) and zonal component (color shaded) at the 200-hPa level (values in m s−1) (lower panels). Displaying a CM3, b ESM2M, c ESM2G and d EGCM. Black contour delimitates AMB and LPB hydrological boundaries
CM3 (Fig. 7a) has projected changes in the low-level moisture flux higher than both ESMs (Fig. 7b, c), with ESM2G (Fig. 7c) being the model with the smallest. Although weakened, EGCM (Fig. 7d) shows comparable features to CM3 in AMB and LPB. CM3 projects an increase in the northerly moisture flux from the eastern part of the central Andes that splits into two branches, one leading to the southeastern coast of Brazil and another to northern Argentina. In CM3, the northeastern coast of Brazil presents higher positive values, as well as the central and eastern South Pacific, which might be related to intensification and positioning of SASH and the South Pacific Subtropical High (SPSH) toward the end of the century. Both highs are currently associated with variability of the seasonal precipitation over Argentina regions (Garbarini et al. 2019, 2021).
ESM2M (Fig. 7b) and ESM2G (Fig. 7c) also display an increase in the northerly moisture flux at the east side of the central Andes to southwestern LPB, as well as around the southeastern coast of Brazil that might be associated with the SASH reinforcement, also seen in CM3. ESM2G displays lower northerly values than CM3 and ESM2M. In all models, the eastern part of NAMB shows an increase in the northerly moisture flux that might be associated with an increase in the moisture flux from the warmer tropical Atlantic, which is known to supply moisture for AMB (Zanin and Satyamurty 2020).
An increase in the low-level northerly moisture flux might be associated with a strengthening of SALLJ toward the end of the century, which provides the thermodynamic support to the formation of storms and, consequently, an increase in precipitation in SLPB and adjacent regions, regardless of the ENSO phases (Bruick et al. 2019).
According to Marengo et al. (2012), the SACZ enhanced (suppressed) events are associated with the weakening (strengthening) of the southward moisture transport from AMB. Because the GFDL models projected an increase in the 850-hPa moisture flux toward SLPB in the rainy season, one expects the declining of the SACZ events by the end of the twenty-first century and, consequently, the weakening of SAMS active phase.
As seen in Fig. 7 for all models and the ensemble mean, the orientation of the vector difference and the strengthening of the southerly component of the 850-hPa moisture flux near the southwestern coast of South America and the eastern South Pacific Ocean might result from the climate change impact on SPSH, as discussed in Flores-Aqueveque et al. (2020). For further details on changes in both southern subtropical highs throughout the twenty-first century, see Cherchi et al. (2018).
Over SESA, the end-century projected changes in the DJFMA 200-hPa wind vector show an increase in magnitude of the westerly component (in shade) by the end of the century (Fig. 7, upper panels), which provides the dynamic support that favors an increase in the hydro-extremes in SLPB. CM3 (Fig. 7a) projects the westerly wind to be stronger by the end of the century, followed by ESM2M (Fig. 7b) and ESM2G (Fig. 7c), which displays the smallest changes; EGCM (Fig. 7d) shows changes in the westerly wind following between CM3 and ESM2G. Although the upper-level wind, in general, becomes stronger in both seasons, regardless of the phases of both equatorial modes (not shown), the model-dependent response unveils the impact of the choice of the atmospheric component on the strengthening of the upper-level jet by the end of the century over LPB, and especially over the South Pacific.
3.2.2 Changes in the precipitation spatial variability
For both river basins, the three models overestimated (underestimated) the precipitation spatial variability in DJFMA (JASON) in comparison with the observation datasets (Fig. 2), although they better correlated with OBS in DJFMA, which concentrates most of the higher monthly means in AMB and LPB (Fig. 3a–d). Besides, December and January are the months with the highest frequency of South American MCCs (Durkee and Mote 2009). Therefore, from now on, we analyze the changes in precipitation for DJFMA. Figure 8 shows the percentage in the DJFMA precipitation change projected for the end-century in comparison with the current climate.
In AMB, CM3 displays a slight increase in precipitation over the northern part, while in the southernmost region of AMB, a decrease is projected (Fig. 8a). Although higher in north- and northwestern AMB, ESM2M (Fig. 8b) projects an increase in precipitation over large areas of the river basin, except for a small area in its southeastern portion. ESM2G (Fig. 8c) shows an increase in precipitation in NAMB, with a decrease in precipitation over the southeastern area of SAMB. Changes in the precipitation from ESM2G and EGCM (Fig. 8d) better compare to Duffy et al. (2015) and Sorribas et al. (2016), showing an increase in precipitation confined to the northern and western portions of AMB.
The GFDL models project an increase in precipitation over SLPB (Fig. 8), also seen in downscaled projections (Cabré et al. 2016; Zaninelli et al. 2019). CM3 presents the largest area of positive values, with a maximum change between 30° S and 37° S toward the end of the century. The ESMs also project an increase in precipitation, however, constrained to the southernmost LPB region. Nevertheless, the GFDL models project a decrease in precipitation in NLPB that might be associated with a reduction in SAMS precipitation by the end of the century.
Figure 8 also show regional projected changes (especially CM3) compatible with the observational study by Spinoni et al. (2019) on regions around the world with changes in drought trends. They observed trends through comparisons of drought indices, considering two periods, specifically: 1951–1980 and 1981–2016. Constraining this discussion to only the overlapped regions, they found the most relevant increase (decrease) in drought frequency in the Amazon (northern Argentina and Uruguay).
For each of the models and the ensemble mean, one might consider an overall EN pattern associated with the projected change in precipitation by the end of the century (Fig. 8), which might be associated with similar pattern in the projected change in the 850-hPa moisture flux and its meridional component (Fig. 7), showing a reduction in precipitation in the central part of the continent, and an increase in the southeastern LPB. However, an increase in precipitation in the northern, central, and western AMB portions (the two latter seen only in the ESMs and EGCM) associates better with the LN pattern found in the low-level moisture flux.
As seen through the Gaussian (normal) Probability Density Functions (PDFs) (Fig. 9), the distribution of the precipitation monthly means for the rainy months (DJFMA) over 30 years (sample size equal to 150) vary more markedly among each of the four subregions than under the RCP8.5 scenario (Fig. 9 and Table 4).
Gaussian Probability Density Function (PDF) of the precipitation monthly means in the rainy months (mm/day) for: a–d CM3, b, c ESM2M and c–f ESM2G, over AMB (top row) and LPB (bottom row). The blue (magenta) lines indicate the northern (southern) region of each river basin. The solid (dashed) line represents the current (end-century) climate
For NAMB (blue curves in Fig. 9a–c), all models project an increase in the rainy months (CM3 from 4.68 to 4.86 mm/day; ESM2M from 4.96 to 6.07 mm/day; ESM2G from 5.35 to 6.58 mm/day) by the end of the century (Fig. 9 and Table 4). Significant increase is observed in ESM2M and ESM2G at the 95% confidence interval (Table 4). Under RCP8.5, all models project that the precipitation regime will become more variable, as reveled by the increase of the standard deviation toward the end-century (from 1.69 to 2.10 mm/day for CM3, from 2.76 to 3.05 mm/day for ESM2M, and from 2.24 to 2.90 mm/day for ESM2G). The changes in mean are significant at 95% confidence level for the ESMs. This increase in the standard deviation of precipitation also characterizes the rise in the probability of occurrence of extreme events throughout the twenty-first century what is corroborated by the end-century higher values of the 90th percentile (from 7.05 to 7.60 mm/day for CM3, from 8.55 to 10.84 mm/day for ESM2M, and from 8.40 to 10.91 mm/day for ESM2G). The increase in the precipitation mean and 90th percentile might be related to the warmer equatorial Atlantic that might favor evaporation, which causes the strengthening of deep convection associated with the ITCZ. A steep rise in the precipitation rates from February through April–May is also seen in Fig. 3, which is enhanced in all three models by the end of the twenty-first century. From 25-year time-slice (uncoupled) experiments, Kitoh et al. (2011) also projected an increase in precipitation from March through May at the end of century, markedly over ITCZ and the northwestern Amazon near the equator, using a 20- and 60-km resolution atmospheric general circulation model, with the lower resolution experiments driven by a combination of different SSTs and initial conditions.
SAMB (in magenta) shows changes in the mean (standard deviation) precipitation from 8.01 to 7.88 mm/day (from 1.17 to 1.22 mm/day) for CM3, from 8.16 to 8.42 mm/day (from 2.01 to 1.62 mm/day) for ESM2M, and from 8.58 to 8.66 mm/day (from 1.68 to 1.92 mm/day) for ESM2G (Table 4). Except for CM3 (from 9.51 to 9.28 mm/day), the ESMs exhibit an increase in the 90th percentile (10.50 to 10.51 mm/day for ESM2M, and 10.56 to 11.03 mm/day for ESM2G).
NLPB (blue curves in Fig. 9d–f) displays a decrease in the mean precipitation (90th percentile) from 6.55 to 6.38 mm/day (9.0 to 8.94 mm/day) for CM3, from 6.24 to 5.68 mm/day (8.89 to 8.41 mm/day) for ESM2M, and from 6.15 to 5.88 mm/day (8.82 to 8.68 mm/day) for ESM2G, with significant change in mean at the 95% confidence interval for ESM2M, all consistent with the weakening of SAMS. The standard deviation increases from 2.16 to 2.20 mm/day for CM3, from 2.22 to 2.40 mm/day for ESM2G, and decreases from 2.32 to 2.16 mm/day for ESM2M.
In SLPB (magenta curves in Fig. 9d–f), changes in mean (standard deviation) toward end-century are from 3.38 to 3.98 mm/day (from 1.21 to 1.25 mm/day) for CM3, from 2.88 to 3.12 mm/day (from 1.14 to 1.16 mm/day) for ESM2M, and from 2.92 to 3.02 mm/day (from 1.17 to 1.09 mm/day) for ESM2G. The changes in the occurrence of higher precipitation amounts are from 5.04 to 5.42 mm/day for CM3, from 4.47 to 4.77 mm/day for ESM2M, and from 4.55 to 4.28 mm/day for ESM2G. Changes in mean are significant at the 95% confidence interval for CM3 and ESM2M. The CM3 and ESM2M results indicate an increase in the occurrence of higher values of precipitation consistent with their projected increase in the northerly moisture flux at low levels from AMB to LPB (Fig. 7a, b).
Figures 10 and 11, together with Table 5 (K–S test table), corroborate the results from PDFs and Table 4. In AMB, ECDFs of the three models exhibit seasonal precipitation values that are increasing toward the end of century in NAMB (Fig. 10a–f); however, for CM3 (Fig. 10a, b), the increase is below the confidence level of 95% (Table 5, D ≤ 0.30), and no significant decrease (D = 0.17) or increase (D = 0.07) in SAMB; while ESM2M (Fig. 10c, d) displays a significant increase in precipitation toward the end of century over NAMB (D = 0.37), but does not show any significant decrease in precipitation in SAMB (D = 0.20); ESM2G (Fig. 10e, f and Table 5) is similar to ESM2M, with a significant increase in precipitation in NAMB (D = 0.60), and exhibits no decrease (D = 0.10) or increase (D = 0.17) in the SAMB precipitation for the end-century climate. For LPB (Fig. 11a–f), all models (Fig. 11a, c, e) show a reduction in the NLPB precipitation for the end-century climate, but only significant at the 95% confidence level for ESM2M (Fig. 11c) (Table 5, D = 0.47). In SLPB (Fig. 11b, d, f and Table 5), only ESM2G (Fig. 12f and Table 5) shows no significant increase (D = 0.17) in precipitation for the end-century climate, with CM3 and ESM2M presenting the following D-values: 0.47 and 0.33, respectively. Note that ESM2G is the GFDL model with the smallest change in the northerly moisture flux in the central-southeastern Andes and the 200-hPa wind in SESA, in comparison with the other two GFDL models under RCP8.5.
Same as in Fig. 10, except for LPB
Projected changes in thermodynamic and dynamic mechanisms by the end of the twenty-first century that might support increase (decrease) in precipitation in DJFMA, in northern AMB and southern (northern) LPB, based on GFDL models’ projections. In the gray boxes, up (down) arrows indicate an increase (decrease) in components in comparison to the current climate
Changes in thermodynamic and dynamic mechanisms might explain the projected changes in the rainy season. Figure 12 schematically exhibits some of the main mechanisms responsible for precipitation changes in the rainy season in South America that are compatible with the DJFMA precipitation changes investigated through the matrix of the GFDL models.
Although apparently constrained by the model’s choice of atmospheric and ocean components, the projected increase in moisture availability from the warming of the tropical Atlantic Ocean would provide an explanation to the significant change in the precipitation mean over NAMB in the GFDL ESMs, accompanied by an increase in extremes. Nevertheless, SAMS-modulated precipitation experienced reduction in SAMB in CM3, and in all three GFDL models in northern LPB, which might be associated with the strengthening of the northerly moisture flux at the central part of the eastern side of the Andes, which might, in turn, increase precipitation in the southern part of LPB (increase in mean for the three models). Furthermore, the increase in the 850-hPa moisture flux at the lee side of the Andes, together with the strengthening of STJ, might be attributed to the warming of the equatorial Atlantic and Pacific waters and might play a key role in controlling the intensity and frequency of occurrences of extreme hydro-events in southern LPB throughout the twenty-first century (increase in the 90th percentile values seen in CM3 and ESM2M).
Here, the hypothesis raised is that although the coarse resolution of GCMs makes them unable to directly simulate convective storms, intensification of the two mechanisms seen in the GFDL models’ projections—driven by CMIP5 RCP8.5—might cause an increase in the frequency and intensity of precipitation extremes in SLPB, as well as depletion of SAMS active phase, both by the end of the twenty-first century.
4 Concluding remarks
We initially performed an assessment of the ability of the GFDL’s global climate models, specifically, CM3, ESM2M and ESM2G, to reproduce general precipitation features in the rainy and dry seasons in the current climate over the two largest river basins in South America, the Amazon and La Plata basins. Through comparisons with observations, an improved atmospheric model turned out to be an essential component to CM3’s ability to reproduce precipitation spatial characteristics over the two river basins during both seasons, while the ESMs only showed a better skill in the rainy season. We then investigated probable causes for changes in the physical mechanisms capable of impacting the rainy season in the subregions of the two river basins, using projections from the same three models, under the assumption of the worst global warming scenario—RCP8.5.
We conclude from the GFDL models’ results that the significant warming in the equatorial Atlantic Ocean injects more moisture into the northern AMB atmosphere, which would increase precipitation in the region toward the end of the twenty-first century. Afterward, the increased moisture flux would cross western AMB, connecting to the low-level flow in the east side of the central Andes to enhance precipitation in southeastern South America. This would provide the thermodynamic support needed to favor the increase in the frequency of the extreme hydro-events in southern LPB. Furthermore, a stronger STJ would grant the dynamic mechanism to support an increase in precipitation over southern LPB. The northerly heat and moisture fluxes would then strengthen toward southern LPB, decreasing the frequency and intensity of SACZ events and, therefore, weakening the active phase of SAMS, and, ultimately, causing the precipitation (mostly) in northern LPB to reduce.
However, under an extreme climate scenario, specifically during the rainy season in South America, the GFDL models’ projections show more conservative changes in the precipitation mean over the AMB and LPB subregions in comparison to changes in the physical and dynamic mechanisms. This behavior might be associated with the inability of the global models’ coarser resolution to reproduce adequately convective systems, despite their ability to depict the subcontinental changes caused by the large-scale phenomena that would trigger them.
Data availability
The data used in this work are available to download from their developers’ or authorized websites listed in the Acknowledgements section.
References
Adcroft A, Hallberg R (2006) On methods for solving the oceanic equations of motion in generalized vertical coordinates. Ocean Model 11:224–233. https://doi.org/10.1016/j.ocemod.2004.12.007
Adler RF, Huffman GJ, Chang A et al (2003) The version-2 global precipitation climatology project (GPCP) monthly precipitation analysis (1979–present). J Hydrometeorol 4:1147–1167. https://doi.org/10.1175/1525-7541(2003)004%3c1147:TVGPCP%3e2.0.CO;2
Anderson J, Balaji V, Broccoli A et al (2004) The new GFDL global atmosphere and land model AM2–LM2: evaluation with prescribed SST simulations. J Clim 17:4641–4673. https://doi.org/10.1175/JCLI-3223.1
Andreoli RV, de Oliveira SS, Kayano MT et al (2017) The influence of different El Niño types on the South American rainfall. Int J Climatol 37:1374–1390. https://doi.org/10.1002/joc.4783
Bombardi RJ, Carvalho LMV (2009) IPCC global coupled model simulations of the South America monsoon system. Clim Dyn 33:893–916. https://doi.org/10.1007/s00382-008-0488-1
Bombardi RJ, Carvalho LMV (2011) The South Atlantic dipole and variations in the characteristics of the South American Monsoon in the WCRP-CMIP3 multi-model simulations. Clim Dyn 36:2091–2102. https://doi.org/10.1007/s00382-010-0836-9
Bombardi RJ, Carvalho LMV, Jones C, Reboita MS (2014) Precipitation over eastern South America and the South Atlantic Sea surface temperature during neutral ENSO periods. Clim Dyn 42:1553–1568. https://doi.org/10.1007/s00382-013-1832-7
Bruick ZS, Rasmussen KL, Rowe AK, McMurdie LA (2019) Characteristics of intense convection in subtropical South America as influenced by El Niño-Southern Oscillation. Mon Weather Rev 147:1947–1966. https://doi.org/10.1175/MWR-D-18-0342.1
Cabré MF, Solman S, Núñez M (2016) Regional climate change scenarios over southern South America for future climate (2080–2099) using the MM5 Model. Mean, interannual variability and uncertainties. Atmósfera 29:35–60. https://doi.org/10.20937/ATM.2016.29.01.04
Carvalho LMV, Jones C, Liebmann B (2004) The South Atlantic Convergence Zone: intensity, form, persistence, and relationships with intraseasonal to interannual activity and extreme rainfall. J Clim 17:88–108. https://doi.org/10.1175/1520-0442(2004)017%3c0088:TSACZI%3e2.0.CO;2
Cavalcanti IFA (2012) Large scale and synoptic features associated with extreme precipitation over South America: a review and case studies for the first decade of the 21st century. Atmos Res 118:27–40. https://doi.org/10.1016/j.atmosres.2012.06.012
Cherchi A, Ambrizzi T, Behera S et al (2018) The response of subtropical highs to climate change. Curr Clim Change Rep 4:371–382. https://doi.org/10.1007/s40641-018-0114-1
Coelho CAS, Uvo CB, Ambrizzi T (2002) Exploring the impacts of the tropical Pacific SST on the precipitation patterns over South America during ENSO periods. Theor Appl Climatol 71:185–197. https://doi.org/10.1007/s007040200004
Dee DP, Uppala SM, Simmons AJ et al (2011) The ERA-Interim reanalysis: configuration and performance of the data assimilation system. Q J R Meteorol Soc 137:553–597. https://doi.org/10.1002/qj.828
Delworth TL, Broccoli AJ, Rosati A et al (2006) GFDL’s CM2 global coupled climate models. Part I: formulation and simulation characteristics. J Clim 19:643–674. https://doi.org/10.1175/JCLI3629.1
Donner LJ, Wyman BL, Hemler RS et al (2011) The dynamical core, physical parameterizations, and basic simulation characteristics of the atmospheric component AM3 of the GFDL global coupled model CM3. J Clim 24:3484–3519. https://doi.org/10.1175/2011JCLI3955.1
Duffy PB, Brando P, Asner GP, Field CB (2015) Projections of future meteorological drought and wet periods in the Amazon. Proc Natl Acad Sci 112:13172–13177. https://doi.org/10.1073/pnas.1421010112
Dunne JP, John JG, Adcroft AJ et al (2012) GFDL’s ESM2 global coupled climate-carbon earth system models. Part I: physical formulation and baseline simulation characteristics. J Clim 25:6646–6665. https://doi.org/10.1175/JCLI-D-11-00560.1
Dunne JP, John JG, Shevliakova E et al (2013) GFDL’s ESM2 global coupled climate-carbon earth system models. Part II: carbon system formulation and baseline simulation characteristics. J Clim 26:2247–2267. https://doi.org/10.1175/JCLI-D-12-00150.1
Durkee JD, Mote TL (2009) A climatology of warm-season mesoscale convective complexes in subtropical South America. Int J Climatol. https://doi.org/10.1002/joc.1893
Durkee JD, Mote TL, Shepherd JM (2009) The contribution of mesoscale convective complexes to rainfall across subtropical South America. J Clim 22:4590–4605. https://doi.org/10.1175/2009JCLI2858.1
Flores-Aqueveque V, Rojas M, Aguirre C et al (2020) South Pacific subtropical high from the late Holocene to the end of the 21st century: insights from climate proxies and general circulation models. Clim past 16:79–99. https://doi.org/10.5194/cp-16-79-2020
Gan MA, Kousky VE, Ropelewski CF (2004) The South America monsoon circulation and its relationship to rainfall over west-central Brazil. J Clim 17:47–66. https://doi.org/10.1175/1520-0442(2004)017%3c0047:TSAMCA%3e2.0.CO;2
Garbarini EM, González MH, Rolla AL (2019) The influence of Atlantic High on seasonal rainfall in Argentina. Int J Climatol 39:4688–4702. https://doi.org/10.1002/joc.6098
Garbarini EM, González MH, Rolla AL (2021) Modulation of seasonal precipitation in Argentina by the South Pacific high. Int J Climatol. https://doi.org/10.1002/joc.6924
Garcia BN, Libonati R, Nunes AMB (2018) Extreme drought events over the Amazon Basin: The perspective from the reconstruction of South American Hydroclimate. Water Switz 10:1594. https://doi.org/10.3390/w10111594
Garrett RG (2013) The ‘rgr’ package for the R Open Source statistical computing and graphics environment—a tool to support geochemical data interpretation. Geochem Explor Environ Anal 13:355–378. https://doi.org/10.1144/geochem2011-106
Garret RG (2018) rgr: Applied Geochemistry EDA. R package ver. 1.1.15. https://cran.r-project.org/web/packages/rgr/rgr.pdf
Glantz MH, Ramirez IJ (2020) Reviewing the Oceanic Niño Index (ONI) to enhance societal readiness for El Niño’s impacts. Int J Disaster Risk Sci 11:394–403. https://doi.org/10.1007/s13753-020-00275-w
Griffies SM, Winton M, Donner LJ et al (2011) The GFDL CM3 coupled climate model: characteristics of the ocean and sea ice simulations. J Clim 24:3520–3544. https://doi.org/10.1175/2011JCLI3964.1
Han Z, Su T, Huang B et al (2019) Changes in global monsoon precipitation and the related dynamic and thermodynamic mechanisms in recent decades. Int J Climatol 39:1490–1503. https://doi.org/10.1002/joc.5896
Harris I, Jones PD, Osborn TJ, Lister DH (2014) Updated high-resolution grids of monthly climatic observations—the CRU TS3.10 dataset. Int J Climatol 34:623–642. https://doi.org/10.1002/joc.3711
Hirabayashi Y, Mahendran R, Koirala S et al (2013) Global flood risk under climate change. Nat Clim Change 3:816–821. https://doi.org/10.1038/nclimate1911
Hounsou-Gbo A, Servain J, Vasconcelos Junior F das C et al (2020) Summer and winter Atlantic Niño: connections with ENSO and implications. Clim Dyn 55:2939–2956. https://doi.org/10.1007/s00382-020-05424-x
Jia F, Cai W, Wu L et al (2019) Weakening Atlantic Niño–Pacific connection under greenhouse warming. Sci Adv 5:eaax4111. https://doi.org/10.1126/sciadv.aax4111
Jones C (2019) Recent changes in the South America low-level jet. Npj Clim Atmos Sci 2:20. https://doi.org/10.1038/s41612-019-0077-5
Jones C, Carvalho LMV (2002) Active and break phases in the South American monsoon system. J Clim 15:905–914. https://doi.org/10.1175/1520-0442(2002)015%3c0905:AABPIT%3e2.0.CO;2
Kitoh A, Kusunoki S, Nakaegawa T (2011) Climate change projections over South America in the late 21st century with the 20 and 60 km mesh Meteorological Research Institute atmospheric general circulation model (MRI-AGCM). J Geophys Res 116:D06105. https://doi.org/10.1029/2010JD014920
Kodama Y-M (1992) Large-scale common features of subtropical precipitation zones (the Baiu Frontal Zone, the SPCZ, and the SACZ) Part I: characteristics of subtropical frontal zones. J Meteorol Soc Jpn Ser II 70:813–836. https://doi.org/10.2151/jmsj1965.70.4_813
Kodama Y-M (1993) Large-scale common features of sub-tropical convergence zones (the Baiu Frontal Zone, the SPCZ, and the SACZ) Part II: conditions of the circulations for generating the STCZs. J Meteorol Soc Jpn Ser II 71:581–610. https://doi.org/10.2151/jmsj1965.71.5_581
Krishnamurti TN (1961) The subtropical jet stream of winter. J Meteorol 18:172–191. https://doi.org/10.1175/1520-0469(1961)018%3c0172:TSJSOW%3e2.0.CO;2
Laing AG, Fritsch JM (1997) The global population of mesoscale convective complexes. Q J R Meteorol Soc 123:389–405. https://doi.org/10.1002/qj.49712353807
Laing AG, Fritsch JM (2000) The large-scale environments of the global populations of mesoscale convective complexes. Mon Weather Rev 128:2756–2776. https://doi.org/10.1175/1520-0493(2000)128%3c2756:TLSEOT%3e2.0.CO;2
Libonati R, Pereira JMC, Da Camara CC et al (2021) Twenty-first century droughts have not increasingly exacerbated fire season severity in the Brazilian Amazon. Sci Rep 11:4400. https://doi.org/10.1038/s41598-021-82158-8
Llopart M, Coppola E, Giorgi F et al (2014) Climate change impact on precipitation for the Amazon and La Plata basins. Clim Change 125:111–125. https://doi.org/10.1007/s10584-014-1140-1
Marchini JL, Heaton C, Ripley BD (2019) fastICA: FastICA algorithms to perform ICA and projection pursuit. R package ver. 1.2-2. https://cran.r-project.org/web/packages/fastICA/fastICA.pdf
Marengo JA, Soares WR, Saulo C, Nicolini M (2004) Climatology of the low-level jet east of the Andes as derived from the NCEP–NCAR reanalyses: characteristics and temporal variability. J Clim 17:2261–2280. https://doi.org/10.1175/1520-0442(2004)017%3c2261:COTLJE%3e2.0.CO;2
Marengo JA, Liebmann B, Grimm AM et al (2012) Recent developments on the South American monsoon system. Int J Climatol 32:1–21. https://doi.org/10.1002/joc.2254
Meehl GA, Covey C, Delworth T et al (2007) THE WCRP CMIP3 multimodel dataset: a new era in climate change research. Bull Am Meteorol Soc 88:1383–1394. https://doi.org/10.1175/BAMS-88-9-1383
Milly PCD, Wetherald RT, Dunne KA, Delworth TL (2002) Increasing risk of great floods in a changing climate. Nature 415:514–517. https://doi.org/10.1038/415514a
Montroull NB, Saurral RI, Camilloni IA (2018) Hydrological impacts in La Plata basin under 1.5, 2 and 3 °C global warming above the pre-industrial level. Int J Climatol 38:3355–3368. https://doi.org/10.1002/joc.5505
Moss RH, Edmonds JA, Hibbard KA et al (2010) The next generation of scenarios for climate change research and assessment. Nature 463:747–756. https://doi.org/10.1038/nature08823
Moura MM, dos Santos AR, Pezzopane JEM et al (2019) Relation of El Niño and La Niña phenomena to precipitation, evapotranspiration and temperature in the Amazon basin. Sci Total Environ 651:1639–1651. https://doi.org/10.1016/j.scitotenv.2018.09.242
Nesbitt SW, Zipser EJ, Cecil DJ (2000) A census of precipitation features in the tropics using TRMM: radar, ice scattering, and lightning observations. J Clim 13:4087–4106. https://doi.org/10.1175/1520-0442(2000)013%3c4087:ACOPFI%3e2.0.CO;2
Nunes AMB (2016) Satellite-enhanced dynamical downscaling for the analysis of extreme events. J Geophys Res Atmos 121:10617–10636. https://doi.org/10.1002/2016JD024927
Nunes AMB, Roads JO (2007) Dynamical influences of precipitation assimilation on regional downscaling. Geophys Res Lett. https://doi.org/10.1029/2007GL030247
Oliveira MI, Nascimento EL, Kannenberg C (2018) A new look at the identification of low-level jets in South America. Mon Weather Rev 146:2315–2334. https://doi.org/10.1175/MWR-D-17-0237.1
Panisset JS, Libonati R, Gouveia CMP et al (2018) Contrasting patterns of the extreme drought episodes of 2005, 2010 and 2015 in the Amazon Basin. Int J Climatol 38:1096–1104. https://doi.org/10.1002/joc.5224
Penalba OC, Rivera JA (2016) Regional aspects of future precipitation and meteorological drought characteristics over Southern South America projected by a CMIP5 multi-model ensemble. Int J Climatol 36:974–986. https://doi.org/10.1002/joc.4398
Pezzi LP, Cavalcanti IFA (2001) The relative importance of ENSO and tropical Atlantic sea surface temperature anomalies for seasonal precipitation over South America: a numerical study. Clim Dyn 17:205–212. https://doi.org/10.1007/s003820000104
R Core Team (2018) R: a language and environment for statistical computing. The R Project for Statistical Computing, Vienna, Austria
Rivarola Sosa JM, Brandani G, Dibari C et al (2011) Climate change impact on the hydrological balance of the Itaipu Basin. Meteorol Appl 18:163–170. https://doi.org/10.1002/met.213
Rodríguez-Fonseca B, Polo I, García-Serrano J et al (2009) Are Atlantic Niños enhancing Pacific ENSO events in recent decades? Geophys Res Lett 36:L20705. https://doi.org/10.1029/2009GL040048
Rohatgi VK, Ehsanes Saleh AKM (2015) An introduction to probability and statistics, 3rd edn. Wiley, New York
Roy I (2017) Indian summer monsoon and El Niño Southern Oscillation in CMIP5 models: a few areas of agreement and disagreement. Atmosphere (basel) 8:154. https://doi.org/10.3390/atmos8080154
Satyamurty P, Nobre CA, Silva Dias PL (1998) South America. Meteorology of the southern hemisphere. American Meteorological Society, Boston, pp 119–139
Schneider U, Finger P, Meyer-Christoffer A et al (2017) Evaluating the hydrological cycle over land using the newly-corrected precipitation climatology from the Global Precipitation Climatology Centre (GPCC). Atmosphere (basel) 8:52. https://doi.org/10.3390/atmos8030052
Seth A, Rojas M, Rauscher SA (2010) CMIP3 projected changes in the annual cycle of the South American Monsoon. Clim Change 98:331–357. https://doi.org/10.1007/s10584-009-9736-6
Silva GAM, Ambrizzi T, Marengo JA (2009) Observational evidences on the modulation of the South American Low Level Jet east of the Andes according the ENSO variability. Ann Geophys 27:645–657. https://doi.org/10.5194/angeo-27-645-2009
Soares WR, Marengo JA (2009) Assessments of moisture fluxes east of the Andes in South America in a global warming scenario. Int J Climatol 29:1395–1414. https://doi.org/10.1002/joc.1800
Sorribas MV, Paiva RCD, Melack JM et al (2016) Projections of climate change effects on discharge and inundation in the Amazon basin. Clim Change 136:555–570. https://doi.org/10.1007/s10584-016-1640-2
Spinoni J, Barbosa P, De Jager A et al (2019) A new global database of meteorological drought events from 1951 to 2016. J Hydrol Reg Stud 22:100593. https://doi.org/10.1016/j.ejrh.2019.100593
Sun X, Cook KH, Vizy EK (2017) The South Atlantic subtropical high: climatology and interannual variability. J Clim 30:3279–3296. https://doi.org/10.1175/JCLI-D-16-0705.1
Taylor KE (2001) Summarizing multiple aspects of model performance in a single diagram. J Geophys Res Atmos 106:7183–7192. https://doi.org/10.1029/2000JD900719
Taylor KE, Stouffer RJ, Meehl GA (2012) An overview of CMIP5 and the experiment design. Bull Am Meteorol Soc 93:485–498. https://doi.org/10.1175/BAMS-D-11-00094.1
Tokinaga H, Richter I, Kosaka Y (2019) ENSO Influence on the Atlantic Niño, revisited: multi-year versus single-year ENSO events. J Clim 32:4585–4600. https://doi.org/10.1175/JCLI-D-18-0683.1
Torralba V, Rodríguez-Fonseca B, Mohino E, Losada T (2015) The non-stationary influence of the Atlantic and Pacific Niños on North Eastern South American rainfall. Front Earth Sci. https://doi.org/10.3389/feart.2015.00055
Trenberth KE (1997) The definition of El Niño. Bull Am Meteorol Soc 78:2771–2777. https://doi.org/10.1175/1520-0477(1997)078<2771:TDOENO>2.0.CO;2
Uccellini LW, Johnson DR (1979) The coupling of upper and lower tropospheric jet streaks and implications for the development of severe convective storms. Mon Weather Rev 107:682–703. https://doi.org/10.1175/1520-0493(1979)107%3c0682:TCOUAL%3e2.0.CO;2
Uccellini LW, Kocin PJ (1987) The interaction of jet streak circulations during heavy snow events along the east coast of the United States. Weather Forecast 2:289–308. https://doi.org/10.1175/1520-0434(1987)002%3c0289:TIOJSC%3e2.0.CO;2
Utida G, Cruz FW, Etourneau J et al (2019) Tropical South Atlantic influence on Northeastern Brazil precipitation and ITCZ displacement during the past 2300 years. Sci Rep 9:1698. https://doi.org/10.1038/s41598-018-38003-6
Velasco I, Fritsch JM (1987) Mesoscale convective complexes in the Americas. J Geophys Res 92:9591. https://doi.org/10.1029/JD092iD08p09591
Venables WN, Ripley BD (2002) Modern applied statistics with S, 4th edn. Springer, New York
Vera C, Baez J, Douglas M et al (2006) The South American low-level jet experiment. Bull Am Meteorol Soc 87:63–78. https://doi.org/10.1175/BAMS-87-1-63
Virji H (1981) A preliminary study of summertime tropospheric circulation patterns over South America estimated from cloud winds. Mon Weather Rev 109:599–610. https://doi.org/10.1175/1520-0493(1981)109%3C0599:APSOST%3E2.0.CO;2
White L, Adcroft A, Hallberg R (2009) High-order regridding–remapping schemes for continuous isopycnal and generalized coordinates in ocean models. J Comput Phys 228:8665–8692. https://doi.org/10.1016/j.jcp.2009.08.016
Wilks DS (2011) Statistical methods in the atmospheric sciences, 3rd edn, vol 100. International geophysics series. Academic Press, Oxford
Willmott CJ, Matsuura K (2001) Terrestrial Air Temperature and Precipitation: Monthly and Annual Time Series (1950-1999) (Version 1.02). In: Cent. Clim. Res. Dep. Geogr. Univ. Delaware. http://climate.geog.udel.edu/~climate/html_pages/README.ghcn_ts2.html. Accessed 22 May 2018
Wuertz D, Setz T, Chalabi Y, Maechler M (2020) fBasics: Rmetrics - markets and basic statistics. R package ver. 3042.89.1. https://cran.r-project.org/web/packages/fBasics/fBasics.pdf
Zanin PR, Satyamurty P (2020) Hydrological processes interconnecting the two largest watersheds of South America from multi-decadal to inter-annual time scales: a critical review. Int J Climatol. https://doi.org/10.1002/joc.6442
Zaninelli PG, Menéndez CG, Falco M et al (2019) Future hydroclimatological changes in South America based on an ensemble of regional climate models. Clim Dyn 52:819–830. https://doi.org/10.1007/s00382-018-4225-0
Zebiak SE (1993) Air–sea interaction in the equatorial Atlantic region. J Clim 6:1567–1586. https://doi.org/10.1175/1520-0442(1993)006%3c1567:AIITEA%3e2.0.CO;2
Zemp DC, Schleussner C-F, Barbosa HMJ et al (2014) On the importance of cascading moisture recycling in South America. Atmos Chem Phys 14:13337–13359. https://doi.org/10.5194/acp-14-13337-2014
Zhou J, Lau K-M (1998) Does a monsoon climate exist over South America? J Clim 11:1020–1040. https://doi.org/10.1175/1520-0442(1998)011%3c1020:DAMCEO%3e2.0.CO;2
Zipser EJ, Cecil DJ, Liu C et al (2006) Where are the most intense thunderstorms on Earth? Bull Am Meteorol Soc 87:1057–1072. https://doi.org/10.1175/BAMS-87-8-1057
Acknowledgements
We would like to thank the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES) for supporting the Graduate Program in Meteorology at the Federal University of Rio de Janeiro, Brazil, and therefore making this work possible. We would also like to thank the anonymous reviewers for their helpful comments, and Climate Dynamics’ Editor and Editorial Office for all support. R. Libonati was supported by CNPQ: Conselho Nacional de Desenvolvimento Científico e Tecnológico (grant 305159/2018–6), and by FAPERJ: Fundação Carlos Chagas Filho de Amparo à Pesquisa do Estado do Rio de Janeiro (grant E26/202.714/2019). T. Ambrizzi would like to acknowledge the partial support from CNPq, FAPESP and INCT Climate Change–Phase 2. NOAA GFDL’s data portal provided the global models’ outputs available at https://data1.gfdl.noaa.gov. The University of East Anglia Climate Research Unit (CRU) provided the CRU TS 4.01 precipitation data. The University of Delaware (UDEL) and the Global Precipitation Climatology Centre (GPCC) precipitation data products were provided by the NOAA/OAR/ESRL PSL, Boulder, Colorado, USA, from their website at https://psl.noaa.gov/data/gridded/data.UDel_AirT_Precip.html and https://psl.noaa.gov/data/gridded/data.gpcc.html, respectively. The Taylor Diagram script running under Python programming language was adapted from Yannick Copin’s version 2018-12-06, which has been placed in the public domain.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The authors declare no knowledge of any conflicts of interest that could influence the publication of this work.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
About this article
Cite this article
Gomes, G.D., Nunes, A.M.B., Libonati, R. et al. Projections of subcontinental changes in seasonal precipitation over the two major river basins in South America under an extreme climate scenario. Clim Dyn 58, 1147–1169 (2022). https://doi.org/10.1007/s00382-021-05955-x
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00382-021-05955-x














