Introduction

The global surface temperature has risen by around 1 °C since the pre-industrial period. It has been estimated that if greenhouse gas (GHG) emissions continue to rise, global warming will likely reach 1.5 °C between 2030 and 2052, and could reach over 4 °C by the end of the twenty-first century (IPCC 2022; IPCC 2021; Fawzy et al. 2020; Rogelj et al. 2018). Such projections prompted the United Nations Framework Convention on Climate Change (UNFCCC) to set targets in 2010 (Cancun Agreement) and 2015 (Paris Agreement) to keep global warming well below 2 °C above pre-industrial levels, preferably at 1.5 °C. A 2 °C rise in the planet’s average temperature is projected to produce devastating impacts, including, for example, the loss of nearly all of the world’s coral reefs and extreme and life-threatening heat waves that could affect over a third of the world’s population (IPCC 2018).

After the ratification of the agreement at the Conference of the Parties in Paris (COP-21) to limit warming to 1.5 °C, the number of studies assessing different global warming levels (GWLs) in regional climates has increased (Egbebiyi et al. 2020; Mentaschi et al. 2020; Arnell et al. 2019; Mbaye et al. 2019; Nikulin et al. 2018; Greve et al. 2018; Wartenburger et al. 2017; Dequé et al. 2016). Recently, a global commitment to accelerate climate action, and thus attempt to meet the Paris agreement target, was signed at COP-26 in Glasgow (UNEP 2021). Even more, participant countries were asked to update official government targets as soon as possible so that they could contain global warming to 1.5 °C. A package of decisions reaffirming countries’ commitment was delivered at COP-27 held in Egypt in 2022 (UNCC 2022). However, even if commitments are implemented on time and in full, global warming is expected to exceed 2 °C by the end of the century (UNEP 2022, 2021; IEA 2021). Nonetheless, there is still a lack of information on how GWLs affect smaller spatial scales, and whether a 0.5 °C difference between 2 and 1.5 °C would lead to significant regional changes. The main limitation of most studies is the use of Global Climate Models (GCMs) projections, which typically have coarse horizontal resolutions (100–200 km), to distinguish sub-grid processes at more refined spatial scales such as the mesoscale (Nikulin et al. 2018). Such processes are best represented in Regional Climate Models (RCMs), as they allow the use of a higher spatial resolution, on the order of tens of kilometers or more. RCMs have been adding value in depicting the effects of features such as topographic gradients, land–ocean contrasts, land use, urban areas, etc. (Ambrizzi et al. 2019; Chou et al. 2014a).

For instance, there is still a lack of information regarding the impact of different GWLs on Brazil’s water resources. The saturation vapor pressure of water in the air is known to be very sensitive to the surface temperature; therefore, disturbances in the global water cycle must accompany climate change, potentially producing more complex and uncertain regional effects than those directly linked to temperature (Milly et al. 2005). Brazil is highly dependent on hydroelectricity, which has a 65% (103 GW) share in the Brazilian energy matrix (EPE 2021; 2020). Changes in water availability can also impact the multiple uses of water in the Brazilian national territory. According to the Brazilian National Water Agency (ANA for its acronym in Portuguese), around 80% of water consumption is for irrigation, human supply (urban and rural), and industry (ANA 2019a). Despite the large total water supply in the country, which represents about 12% of the world’s water availability and 53% in South America (Veiga and Magrini 2013; Carvalho and Magrini 2006; Tucci et al. 2001), the distribution of water resources over the territory is spatially uneven and varies seasonally. The northern region of Brazil, home to the world’s largest hydrographic basin, the Amazon Basin, contains around 69% of Brazil’s available freshwater water and is inhabited by less than 8% of the population (Ghisi 2006). On the other hand, the Southeast and Northeast regions, which are home to 43% and 28% of the country’s population, have only 6% and 3% of the available water, respectively (Ghisi 2006). In some places, such as the semiarid region of the Northeast, the low water availability combined with droughts has serious consequences (Martins et al. 2018; Gutiérrez et al. 2014), making water a critical factor for the local population (Marengo 2008).

Previous studies have shown that hydrological processes in some Brazilian river basins are sensitive to global climate change (Brêda et al. 2020; Llopart et al. 2020; Rodrigues et al. 2020; Tiezzi et al. 2019; Neto et al. 2016; Alvarenga et al. 2016). However, studies using the GWLs approach to assess possible changes in water balance components and precipitation extremes are still scarce. Such assessments can aid decision-makers in seeking more adaptive water management systems and using proactive strategies to increase resilience to expected impacts.

In this work, a set of projections from the Eta Regional Climate Model (Eta-RCM) and the GWL approach were adopted with the following objectives: (1) to identify if there are significant changes in water availability and the occurrence of climate extremes over Brazil, under global warming levels of 1.5 °C and 2 °C in different RCP scenarios (Representative Concentration Pathways; Van Vuuren et al. 2011) and (2) if a difference of 0.5 °C between the GWLs could produce relevant changes in the Brazilian hydrographic regions. The following sections present details on the study area, the climate models and climate scenarios used, the method for determining the GWL approach, and the final considerations.

Data and methodology

Study area

The study area includes the major Hydrographic Regions (HRs) of Brazil, chosen for their socioeconomic importance to the country. The selected HRs cover approximately 81% of the national territory (ANA 2015), including most Brazilian states (Figure S1 in the supplementary material), and together contribute 91% (~ 98 GW) of Brazil’s operational hydropower potential (SIPOT 2018). The Paraná HR (PR) has the greatest contribution to the country’s hydroelectric potential, with almost 41% of the total potential (SIPOT 2018; ANA 2015). In addition, the PR is the one nearest the South-East region of Brazil, which has the highest regional demographic density and largest economic development in the country, thus increasing the importance of this HR in the national context (Soares et al. 2008). The Amazon HR (AMZ) has the largest hydrographic basin in the world; it occupies 45% of the Brazilian territory and holds a large part of the country’s freshwater water availability. This basin plays a significant role in the continental and global hydrological cycle (ANA 2019b); it has been estimated that the Amazonian flow is responsible for approximately one-fifth of the volume of freshwater affluent to the oceans (Chaudhari et al. 2019; Chen et al. 2010). Considering only the energy security point of view, the AMZ currently contributes approximately 22% of the hydroelectric potential in operation (SIPOT 2018). The Parnaíba (PRN) and São Francisco HR (SF) comprise a large part of the semiarid region of northeastern Brazil, characterized by critical drought events. For this reason, these two HRs play an essential role in the region for other uses of water as well, such as urban supply and irrigation. The Tocantins-Araguaia HR (TOA), besides its significant portion of the hydroelectric potential (~ 12% of the total), is also important in the national context since it is a region characterized by the expansion of the agricultural frontier, especially concerning the cultivation of grains (ANA 2015). The Uruguay HR (URU), besides its hydroelectric potential (~ 6% of the total), is relevant because of the agricultural and industrial activities present in the region.

Climate models and dynamical downscaling

In this study, it was used a data set generated from the dynamic downscaling of AR5 projections (Fifth Assessment Report of the IPCC- Intergovernmental Panel on Climate Change) through the Eta-RCM (Chou et al. 2014a, b) nested in four CMIP5 GCMs (Coupled Model Intercomparison Project—Phase 5; Taylor et al. 2012). Eta downscaling used 20 km of horizontal resolution and included South America, a part of Central America, and the Caribbean (Chou et al. 2014a, b). These simulations are available on the ProjEta platform (projeta.cptec.inpe.br). Downscaling simulations and their driven models were used in the evaluations, thus totaling eight members (Eta-RCM × GCMs).

The Eta-RCM was nested with the following GCMs: Canadian Earth System Model, version 2 (CanESM2; Arora et al. 2011; Chylek et al. 2011); Hadley Center Global Environmental Model, version 2 Earth System (HadGEM2-ES; Collins et al. 2011; Martin et al. 2011); Model for Interdisciplinary Research, version 5 (MIROC5; Watanabe et al. 2010); and Brazilian Earth System Model (BESM; Capistrano et al. 2020; Veiga et al. 2019; Nobre et al. 2013).

The climate projections regionalized by the Eta-RCM were used to support the Third Brazilian Communication of the United Nations Framework Convention on Climate Change (MCTI 2016) and have been extensively used in previous studies on the impacts of climate change, mainly in South America (Ferreira and Miranda 2021, 2020; Brito et al. 2019; Ferreira et al. 2019; Tavares et al. 2018; Arias 2018; Debortoli et al. 2017; Neto et al. 2016; etc.). In this research, we evaluate simulations of the present climate (1961–2005) and climate projections (2011–2100) of the precipitation and latent heat flux (used to estimate actual evapotranspiration; Allen et al. 1998). The difference between precipitation and evapotranspiration is considered a measure of the water availability of the terrestrial branch of the hydrologic cycle (Llopart et al. 2020; Marengo et al. 2016) and is considered here to represent water availability (hereinafter also called water resources -WR).

In addition, extreme climate indices were calculated using the ClimDex package, developed by the CCl/CLIVAR/JCOMM Expert Team (ET) on Climate Change Detection and Indices (ETCCDI) (Alexander et al. 2019; Alexander et al. 2006; Zhang et al. 2004). The extreme indices considered were the maximum number of consecutive dry days (CDD) and the total annual precipitation exceeding the 95th percentile (R95p). The CDD and R95p extreme indices function as proxies for meteorological drought (Ferreira et al. 2021; Valverde and Marengo 2014) and flooding events (Zhang et al. 2020). Therefore, they were included in the analyses to check the precipitation distribution on a daily scale. In contrast, the analysis of the water balance components projections is carried out based on the annual mean. Additionally, the annual trends for the present were calculated considering Sen’s slope test (Sen 1968) and the non-parametric Mann–Kendall test (Wilks 2011; Kendall 1975). Sen’s slope measures the magnitude of the trend, while the Mann–Kendall test indicates statistically significant trends. The combination of Sen’s and Mann–Kendall tests has been used in several studies of extreme climate indices trends (Marengo et al. 2022; Regoto et al. 2021; Dereczynski et al. 2020; Avila-Diaz et al. 2020; Bezerra et al. 2019; Ávila et al. 2016; Skansi et al. 2013). Finally, in order to highlight the projected mean changes, the statistical significance was evaluated using the Student t-test (Wilks 2011), and the future climate is depicted as the difference of the future minus the baseline period (1961–1990).

Global warming levels and representative concentration pathway scenarios

The GWL approach is considered in this study. In this approach, the periods in which the increase in the average global temperature becomes greater than the average global warming levels established in the various COP meetings are considered to evaluate projections (Nikulin et al. 2018). The timing of the GWLs is determined from the central year of 30-year moving averages when the global mean temperature reaches target warming levels in relation to the climate of the pre-industrial period. Therefore, the 30-year period where the GWLs are reached depends on the global warming level, the model simulation, and the climate forcing scenario. In this study, the target GLWs are 1.5° and 2 °C, warming limits proposed by the UNFCCC, and the period 1861–1890 was assumed to be pre-industrial, as adopted in N’Datchoh et al. (2022), Kumi and Abiodun (2018) and Nikulin et al. (2018). The period of the GWLs was established for the selected scenario considering the first member (r1i1p1) of each GCM. For the members of the Eta-RCM set, the same period that it takes to reach the GWL estimated by each GCM used in the nesting was adopted. The IPCC AR5 scenarios were not designed to address GWL concerns, and some assessments have demonstrated different regional impacts (in magnitude and direction) depending on GHG forcing (Kumi and Abiodun 2018; Montroull et al. 2018; Wartenburger et al. 2017). Therefore, two RCP scenarios were selected for analysis: RCP4.5 and RCP8.5 (Moss et al. 2010).

The RCP4.5 scenario is considered an intermediate between the most optimistic and the most pessimistic scenarios. The RCP8.5 scenario is the most pessimistic of the AR5 scenarios and can be considered the closest to reality, given the trajectory of GHG emissions in recent years. Recently, even with reduced emissions due to the COVID-19 pandemic (Foster et al. 2020; Le Quéré et al. 2020; Liu et al. 2020), the CO2 concentration has already exceeded 417 ppm, which represents an increase of 50% from the levels recorded since the beginning of the industrial period (AFP 2021).

The climate sensitivity is here defined as the period necessary for a GCM to reach the GWLs in each RCP scenario (Figure S2 in the supplementary material). Among the models considered, the CanESM2 model presented the highest climate sensitivity since it reached GWLs faster. BESM and MIROC models present lower climate sensitivity, regardless of the RCP scenario.

Observation datasets

Monthly and daily data were used to assess annual trends based on a 45-year time series (1961–2005) of water balance components and selected climatic extremes. The TerraClimate database provided precipitation and actual evapotranspiration monthly data. With a horizontal resolution of approximately 4 km, the TerraClimate data (Abatzoglou et al. 2018) was produced from the combination of high spatial resolution climatological normals from the WorldClim dataset with data from the Climate Research Unit version 4.0 (CRU; Harris. et al. 2017) and the Japanese 55-year reanalysis (JRA-55; Kobayashi et al. 2015). In addition, due to the great complexity of climate extremes, mainly at a higher spatial resolution, three sources of daily precipitation data were used in their analysis: the Global Meteorological Forcing Dataset (GMFD; Sheffield et al. 2006) with 25 km of horizontal resolution; meteorological stations precipitation dataset interpolated with a spatial resolution of 1° × 1°, made available by the National Oceanic and Atmospheric Administration of the United States (NOAA; Liebmann and Allured 2005); and, lastly, a dataset of daily precipitation data developed from meteorological stations data in Brazil obtained from various sources, such as INMET (National Institute of Meteorology), Sudene (Superintendence of Northeast Development), Cargill, ANEEL (National Electric Energy Agency), Cenibra (Celulose Nipo-Brasileira S/A), and others. The latter dataset was named “ProjEta,” for being data compiled and organized by the researchers of the Eta Model Project (INPE) (http://etamodel.cptec.inpe.br/). The ProjEta dataset was interpolated to a regular grid of 20 km × 20 km using the inverse square distance method. The domain considered for the interpolation covers Brazil. A summary of the main characteristics of the datasets used follows in Table S3 in the supplementary material.

Results and discussion

Historical trends

The accumulation of monthly anomalies of precipitation, evapotranspiration, and water availability from TerraClimate monthly data over six selected HRs was used to identify the historical trends of the water balance components. The anomalies were calculated based on the climatological period of 1961–1990 and accumulated from January 1991 to December 2005. As shown in Fig. 1, the Parnaíba and São Francisco HRs present systematic reduction patterns in the three water balance components. These HRs cover a large part of the Brazilian semiarid and are essential for regional subsistence. In the São Francisco HR, despite presenting similar anomalous patterns, the accumulation of rainfall and water availability deficits are smaller than in Parnaíba. In the HRs located in the Southeast and South regions of Brazil (i.e., Paraná and Uruguay HRs), the results indicate typically wetter conditions, with increasing rainfall and higher monthly water availability. In the other HRs, no systematic trends were identified for drier or wetter conditions (Figure S4 in the supplementary material). Thus, there is no indication of increasing accumulation of excess or deficit over the study period.

Fig. 1
figure 1

Accumulation of monthly precipitation anomalies (PREC; mm), first column, evapotranspiration (ET; mm), second column, and water resources (WR; mm), third column, from January 1991 to December 2005. The rows indicate the Brazilian hydrographic regions selected: Parnaíba, São Francisco, Paraná, and Uruguay

Full size image

In general, a direct relationship between the accumulated anomalies of precipitation and evapotranspiration was observed. The reduction in rainfall reduces the soil moisture and, consequently, the evapotranspiration rate. Likewise, decreasing evapotranspiration also represents a decrease in rainfall amounts. In summary, we conclude that trends in water availability in Brazilian hydrographic regions were associated with changes in rainfall. This is in agreement with previous results from Magrin et al. (2014) that found that in the La Plata River Basin (encompassing the Brazilian HR of Uruguay and Paraná), the increase in runoff since the second half of the twentieth century was associated with increases in precipitation.

In some HRs, the increase in total annual precipitation is followed by increasing trends in extreme rainfall events, such as those associated with the occurrence of floods, inundations, and landslides (R95p – Fig. 2). Consistent signs of increasing R95p in the Paraná and Uruguay HR were found. Similar results from other observed databases and different evaluation periods were found by Regoto et al. (2021), Dereczynski et al. (2020), Avila-Diaz et al. (2020), and Skansi et al. (2013). Thus, contributing to the robustness of the signal of extreme rainfall changes in the Paraná and Uruguay HR. In these regions, the pattern of increasing very wet days events may be associated with a more southerly displacement of the South Atlantic Convergence Zone, thus making these types of events more frequent (Zilli and Carvalho 2021; Dereczynski et al. 2020; Zilli et al. 2019; Liebman et al. 2004). A similar wetting pattern was observed for the Tocantins-Araguaia HR, where the ProjEta dataset showed an increase in the R95p index. The opposite pattern arises for the Parnaíba HR, with negative trends in total annual precipitation, although not statistically significant. In Amazon HR, significant trends of a 2 mm yr−1 reduction were identified when considering the ProjEta dataset. Opposite trends among the dataset were found in the São Francisco HR, although they were statistically non-significant. It is important to emphasize that for most HRs there is an observational uncertainty shown by statistical significance in only one of the databases or by sign inversion between series. Observational uncertainty is a common issue in Brazil due to the sparseness of meteorological stations, especially in the north region (as Amazon and Tocantins-Araguaia HRs) (Olmo et al. 2022).

Fig. 2
figure 2

Annual precipitation of days when rain exceeded the 95th percentile (R95p; mm) and magnitude of annual trends (Q/Sen’s estimate) for 1961 to 2005. The R95p index was calculated from observed data from the ProjEta, GMFD, and NOAA databases and extracted for the average of the Brazilian hydrographic regions selected: Amazon, Tocantins-Araguaia, Parnaíba, São Francisco, Paraná, and Uruguay. The + symbol indicates significant trends at the 5% level

Full size image

The trends found in the CDD index, used as an indicator for meteorological drought occurrence, were weak and not significant for all HRs (Figure S5 in the supplementary material). In general, the trends found in the CDD index agree with the results by Skansi et al. (2013) and Bezerra et al. (2019). However, the results disagree with those found by Avila-Diaz et al. (2020), most likely due to dataset and time frame differences. The time series used in this study (1961–2005) is more similar to the 1950–2010 period used by Skansi et al. (2013), compared to the period of 1980–2016 used by Avila-Diaz et al. (2020). Studies that include time series starting in the 1980s and 1990s and going to more recent years (2016–2018) have shown trends towards an increase in the CDD index, especially in Northeastern Brazil (a region encompassing the São Francisco and Parnaíba HRs) (Avila-Diaz et al. 2020; Regoto et al. 2021). Through analysis of the CDD anomaly series, Regoto et al. (2021) showed that the climate of the Northeast started to present consistently longer periods of drought from the 1990s onwards, which persist until recent years.

In average annual values from 1961 to 2005 (Fig. 3), the results indicate negative trends for the water balance components of Parnaíba and São Francisco HRs. However, statistical significance at the 5% level was found only for the Parnaíba HR. The decrease in precipitation and water availability were, respectively, 4 and 3 mm year−1 in Parnaíba; and 1.6 and 0.8 mm year−1 in São Francisco. The other selected HRs showed positive trends in precipitation, evapotranspiration, and water availability, with higher magnitudes in the Uruguay HR. However, the trends were not significant at the level of 5%. The increases in precipitation and water availability in Uruguay HR were 4.3 and 3.3 mm year−1, respectively. Previous studies found similar results, with trends of increasing annual rainfall in hydrographic basins located in the North, South, and Southeast regions of Brazil (Dereczynski et al. 2020; Xavier et al. 2020; Zandonadi et al. 2016; Ávila et al. 2016; Skansi et al. 2013; Ely and Debruil 2017) and reduction in the Northeast (Avila-Diaz et al. 2020; Bezerra et al. 2019; Melo et al. 2018; Skansi et al. 2013). The obtained increase in rainfall of 1.6 mm year−1 in the Amazon HR was very close to the results found by Skansi et al. (2013) for the Amazon region for a long historical series (60 years).

Fig. 3
figure 3

Annual trends for the period 1961 to 2005 of precipitation (PREC; mm), evapotranspiration (ET; mm), water resources (WR; mm), precipitation of days when rain exceeded the 95th percentile (R95p; mm), and maximum number of consecutive dry days (CDD; days) calculated from the data observed (Obs.), from the Global Climate Models (GCMs): BESM, CanESM2 (CanE.), HadGEM2-ES (HadG.), MIROC5 (MIR.), and from the Eta Regional Climate Model (Eta-RCM) nested to the respective GCMs: BESM (EtaB), CanESM2 (EtaC), HadGEM2-ES (EtaH), and MIROC5 (EtaM). Shading in darker colors indicates significant trends at the 5% level. Shades of brown represent a negative trend, and shades of blue represent a positive trend. Average trends of the Brazilian hydrographic regions selected: Amazon (AMZ), Tocantins-Araguaia (TOA), Parnaíba (PRN), São Francisco (SF), Paraná (PR), and Uruguay (URU)

Full size image

The performance of the selected climate models in reproducing the signs of historical trends in water balance components and climate extremes was also investigated. Figure 3 shows the average annual trends of each HR extracted for the observed data, the nested Eta-RCM to the GCMs, and the GCMs. First, it is important to emphasize that there is a signal inversion between the trends of the RCMs and GCMs in some of the analyzed HRs. This could have been caused by the different physical and surface parameterizations used in the Eta-RCM, resulting in different trends in precipitation and evapotranspiration rates.

In general, most climate models were able to capture the trend signal of reduced precipitation and water availability in the Parnaíba and São Francisco HRs, and the increase in the Uruguay HR. For the latter, the entire set of models represented with good skill the trends of increasing evapotranspiration. In Parnaíba HR, despite the good performance, only BESM and Eta-MIROC simulations reproduced the statistically significant reductions in rainfall and water availability. In the Amazon and Paraná HR, half of the set (2 GCMs and 2 Eta-RCM) captured the signs of increasing rainfall trends, with a slightly better performance in the simulations of increasing trends in water availability in Paraná (2 GCMs and 3 Eta-RCMs). At Tocantins-Araguaia HR, most models failed to reproduce the observed signs of increasing trends in the water balance components.

Regarding very wet days (R95p), most models of the set reproduced the observed increasing trends in the HRs located in the south and the southeastern of Brazil (Uruguay and Paraná HRs). However, the statistically significant trends observed in Paraná HR were not reproduced. In the Uruguay HR, only two Eta-RCM simulations (BESM and MIROC) captured significant trends. At Tocantins-Araguaia HR, only half of the set performed well. In the Amazonas, São Francisco, and Parnaíba HRs, most simulations showed signs opposite to those observed. The good performance of the models in simulating the signs of annual trends for HRs in the south and southeast of Brazil was also identified in the CDD index. Although the trends of the CDD index were not significant, the simulations compared well with the observations of an increase in Paraná and a decrease in the Uruguay HR.

Future climates under 1.5 °C and 2 °C of global warming

Changes of precipitation, evapotranspiration, and water resources

In this section, the spatial patterns of future changes in water balance components under RCP4.5 and RCP8.5 scenarios at the GWLs of 1.5 °C and 2 °C are presented (Fig. 4). The projected changes considered the average values of the set of 8 Eta-RCM models and their driving GCMs. In general, a 2 °C GWL under the RCP8.5 scenario is projected to have a higher impact on the water balance components, with amplification of drier conditions for a large part of Brazil (North, Northeast, Central-West, and Southeast). Reductions in water availability occur mainly due to reduced precipitation. The consequences of a 0.5 °C difference between the GWLs were more pronounced in the RCP8.5 scenario, indicating more intense reductions in both precipitation and water availability, especially in the extreme north and east coast of Brazil. The projected changes in evapotranspiration followed the pattern of projected changes in precipitation. Torres et al. (2021) and Santos et al. (2020) found a similar pattern of the expansion of areas with reduced precipitation towards the center of the country, which was strengthened and increased in higher GWLs.

Fig. 4
figure 4

Difference between the projections and the baseline period (1961–1990) of precipitation – PREC (mm/year), evapotranspiration -ET (mm/year), and water resources -WR (mm/year) for Brazil. Mean values of the climate models set (Global Climate Models + Eta Regional Climate Model) under Representative Concentration Pathway scenarios, RCP4.5 and RCP8.5, at the 1.5 °C and 2 °C global warming levels (GWLs), and the spatial patterns of the differences between the two GWLs in each RCP scenario. The hatched regions show statistically significant changes

Full size image

In the Southern region of Brazil, projections indicate trends toward wetter conditions. For this region, the signs of change are more robust and statistically significant, with consistently positive standard deviation spatial patterns, decreasing the uncertainties (Figure S6 in the supplementary material). The impacts were more accentuated at the 2 °C GWL. However, unlike other regions of Brazil, they presented a higher magnitude in the intermediate concentration scenario (RCP4.5). Similar results were reported by Montroull et al. (2018) for the La Plata Basin. Also, the differences between the GWLs within the RCP4.5 scenario shows that the reduction in precipitation is higher at the 1.5 °C GWL in some areas of the Southeast and Northeast regions.

A large reduction under the RCP4.5-GWL1.5 °C combination is more evident for the São Francisco Basin (Fig. 5). However, the most remarkable changes in the three water balance components (PREC, ET, and WR) take place in the RCP8.5 scenario at the 2.0 °C GWL (Fig. 5). Under this combination (RCP8.5-GWL 2.0 °C), a 0.5 °C difference in mean global warming is projected to produce additional decreases in water availability of around 4 to 7% per year in half of the selected HR. The projected reductions in water availability, although not statistically significant, were approximately 20% in Tocantins-Araguaia, Parnaíba, and Amazon HRs. In Amazon HR, similar reductions in water availability were projected in both RCP scenarios and GWLs. In São Francisco and Paraná HRs, smaller reductions are to be expected, with values of 9% and 3%, respectively. The Paraná HR is located in the transition area, between the South and Southeast regions of Brazil. In this area, drier conditions are noted in the north and wetter in the south of the hydrographic region. However, the average percentages indicate conditions slightly closer to the baseline period conditions. Only for the Uruguay HR, percentages of positive change were obtained. In this HR, projected wetting conditions were more intense under higher global warming; however, the highest percentages of changes occurred at the intermediate concentration scenario. Projections (RCP4.5-GWL2.0 °C) indicate significant mean annual increases in the three water balance components (PREC, ET, and WR). The precipitation and water availability increased by 15% and 30%, respectively. Regardless of the RCP scenario, the 0.5 °C difference between GWLs (2–1.5 °C) contributes to increasing water availability, in the order of 5 to 6% per year.

Fig. 5
figure 5

Projected changes (in %) of precipitation – PREC, evapotranspiration – ET and water resources –WR between representative concentration pathway scenarios, RCP4.5 and RCP8.5, considering the period when the models reached global warming levels (GWLs) of 1.5 °C and 2 °C. Values were obtained from the climate models (Global Climate Models + Eta Regional Climate Model) extracted for the Brazilian hydrographic regions selected: Amazon, Tocantins-Araguaia, Parnaíba, São Francisco, Paraná, and Uruguay. The diamonds and squares indicate the mean values for the 1.5 °C and 2 °C GWLs, respectively. The circles and dashes indicate, respectively, the maximum/minimum values and the standard deviations. Squares and diamonds marked with “ + ” and “X” indicate statistically significant values at the 10% and 5% levels. The green and red colors represent the RCP4.5 and RCP8.5 scenarios

Full size image

Changes of extreme precipitation

Figures 6 and 7 show the changes in precipitation indices (R95p and CDD) for each hydrographic region. Projections indicate significant increases in extreme daily rainfall in the South region and parts of the Southeast region of Brazil, with higher intensities at the 2.0 °C GWL under the RCP4.5 scenario (Fig. 6). In Paraná and Uruguay HRs, projections indicate increases of 17% and 37% (15% and 34%) under the condition RCP4.5-GW2°C (RCP8.5-GWL2°C), respectively (Fig. 7). For the Uruguay HR, in addition to the significant mean increases, the obtained standard deviations were always positive, indicating less uncertainty about the projected changes.

Fig. 6
figure 6

Difference between the projections and the baseline period (1961–1990) of the annual precipitation of days when rain exceeded the 95th percentile (R95p; mm) and of the maximum number of consecutive dry days in the year (CDD; days) for Brazil. Mean values of the climate models set (Global Climate Models + Eta Regional Climate Model) under Representative Concentration Pathway scenarios, RCP4.5 and RCP8.5, at the 1.5 °C and 2 °C global warming levels (GWLs), and the spatial patterns of the differences between the two GWLs in each RCP scenario. The hatched regions show statistically significant change

Full size image
Fig. 7
figure 7

Projected changes (in %) of precipitation of days when rain exceeded the 95th percentile (R95p) and maximum number of consecutive dry days (CDD) for the Representative Concentration Pathway scenarios, RCP4.5 and RCP8.5, considering the period when the models reached global warming levels (GWLs) of 1.5 °C and 2 °C. Values were obtained from the climate models (Global Climate Models + Eta Regional Climate Model) extracted for the Brazilian hydrographic regions selected: Amazon, Tocantins-Araguaia, Parnaíba, São Francisco, Paraná, and Uruguay. The diamonds and squares indicate the mean values for the 1.5 °C and 2 °C GWLs, respectively. The circles and dashes indicate, respectively, the maximum/minimum values and the standard deviations. Squares and diamonds marked with “ + ” and “X” indicate statistically significant values at the 10% and 5% levels. The green and red colors represent the RCP4.5 and RCP8.5 scenarios

Full size image

Regarding the CDD index, projections showed that the frequency tends to increase in many parts of Brazil, especially in the North and Northeast, with higher coverage of areas of statistical significance under higher global warming (Fig. 6). On average, increases can be expected in all selected HRs. However, the changes are not statistically significant. The largest increases for consecutive dry days (9 to 12% per year) were achieved in the Tocantins-Araguaia, Parnaíba, and São Francisco HRs in both RCP scenarios, under higher GWL (Fig. 7). The 0.5 °C difference between the GWLs contributed to average increases of the order of 3–4% per year in some HR. In Paraná HR, an increase of more than 10% per year is expected at the 2.0 °C GWL. However, it is projected to occur only in the RCP8.5 scenario.

Conclusions and final considerations

This work aimed to evaluate the projected changes under 1.5° and 2 °C global warming levels in Brazil and its main hydrographic regions. To that aim, two greenhouse gas concentration scenarios and their radiative forcing on the climate of the IPCC AR5 were used: RCP4.5 and RCP8.5. Precipitation, evapotranspiration, water availability, and precipitation extremes were evaluated using a set of eight simulations from the Eta-RCM and their driving GCMs (BESM, CanESM2, HadGEM2-ES, and MIROC5).

The results indicate that most regions of Brazil are likely to face a decrease in mean annual precipitation and consequently a reduction in both evapotranspiration and water availability, regardless of the RCP scenarios and GWL targets. As expected, the impacts are more accentuated under higher global warming and GHG concentration. The largest reductions in freshwater availability, in the order of 20% a year, are projected for the Tocantins-Araguaia, Parnaíba, and Amazon hydrographic regions. Increases in dry spells are also expected. Such projections can have a direct impact on the two largest Brazilian biomes (Amazon and Cerrado – Brazilian tropical savannah) and one of the country’s most important agricultural frontiers, called MATOPIBA (an acronym formed from the names of the states of Maranhão, Tocantins, Piauí, and Bahia). The MATOPIBA region is strategic for the Brazilian economy, mainly due to the quantities of agricultural commodities produced there (Silva et al. 2021), such as soybeans (Marengo et al. 2022; Rattis et al. 2021; Bragança, 2018). In recent decades, soybean cultivation in the region has expanded, in part, due to irrigation (Lopes et al. 2021; Silva et al. 2021). Irrigation is the highest water-using activity in the country, accounting for around 70% of total use relative to other uses (ANA 2019a). Despite its high consumption, irrigation has become one of the most efficient ways to increase food production and has helped cement Brazil in the ranking of the world’s largest producers (CONAB 2019). The country is still one of the few countries that can sustainably increase its agricultural production, mainly through better use of its large areas of degraded pastures (Wendt et al. 2015). However, the prospect of increasing agricultural production and growing demand for irrigation can further threaten water security and the multiple water uses. Social conflicts over water use in MATOPIBA have already intensified (Silva et al. 2021; Pousa 2019). In addition, observations show that the region has suffered from increased combined drought-heat extremes over the last 20 years (2000–2020), with the dry rainy season becoming warmer, drier, and longer (Marengo et al. 2022). Therefore, the prospect of an even drier climate in the region could intensify water conflicts, impact the Brazilian economy, and jeopardize global food security since Brazil contributes approximately one-third of the world’s soybeans (Rattis et al. 2021). It is important to emphasize that the historical observational series in Parnaíba HR indicated significant drying conditions (statistical significance at the 5% level), in other words, a reduction in precipitation and water availability, since 1961–2005.

A drier climate is also projected for São Francisco HR, located in the semiarid region of the country. The decline in water availability varies by about 9% per year, which is also accompanied by an increase of about 12% per year during drought periods. Such conditions can increase the current level of exposure and socioeconomic vulnerability in this region due to water shortages caused by frequent droughts (Marengo et al. 2017). Several economic sectors are expected to be affected, such as hydroelectric power generation, agriculture, livestock, and industry. Unlike the large-scale agribusiness in the MATOPIBA region (Silva et al. 2021), agriculture in São Francisco is mainly subsistence farming, rainfed (Vieira et al. 2021), and has great social importance as it contributes to the maintenance of rural communities in the countryside (Marengo et al. 2020). Therefore, climate change may further exacerbate the region’s socioeconomic vulnerability.

On the other hand, in Uruguay HR, located in southern Brazil, observed trends already indicate a statistically significant wetting pattern, increasing water availability. The increase in annual totals is associated with extreme daily rainfall. This pattern was also found in projected changes in the RCP scenarios and GWL with high consistency, as all individual simulations (ensemble standard deviation) agree with the sign of change. However, under the global warming level of 2 °C, there is a higher order of magnitude in the intermediate concentration scenario (RCP4.5). The intensification of extreme rainfall events can increase the environmental susceptibility to hydrological disasters that commonly occur in the region, such as floods and landslides (Regoto et al. 2021; Avila-Diaz et al. 2020).

In summary, the higher concentration scenario (RCP8.5) at the 2 °C GWL will likely have a larger impact on the water balance components, amplifying drier conditions for a large part of Brazil. Conversely, in the South region, projections indicate wetting trends with more pronounced values in the RCP4.5 scenario. The 0.5 °C difference between the GWLs intensifies the reductions from 4 to 7% in water availability, mainly in the Tocantins-Araguaia, Parnaíba, and São Francisco HRs. Finally, the results suggest that some regions of Brazil may suffer impacts on water resources and extremes of precipitation, even limiting global warming to 1.5 °C (Paris Agreement). It is expected that the information presented here on the potential consequences of these global warming scenarios can contribute to more adaptive and resilient planning strategies to guarantee water security.