Introduction

In order to satisfy future country’s energy needs, hydroelectric expansion plans in Brazil are concentrated in Amazonia, where projections of precipitation based on different models largely vary under climate change scenarios (Christensen et al. 2013). Therefore, it becomes crucial to assess how changing climate is likely to affect the energy security of the country.

In addition to the effects of climate change, the Amazon basin has been affected by the conversion of pristine forest to other types of land use since the 1970s due to agricultural expansion and resource extraction (Davidson et al. 2012; Walker et al. 2000). Singular policies for the region have been developed by the government (Growth Acceleration Program - PAC (BRASIL 2013)). These policies include the establishment of several protection areas and indigenous territories (Folhes et al. 2015). However, a substantial amount of deforestation occurs each year in the region (PRODES, INPE, http://www.obt.inpe.br/prodes/). The conversion of forest to pasture and crops promotes a decrease in evapotranspiration (Lathuillière et al. 2012) and soil water recharge and increases fast fluxes responses in the basin (Bruijnzeel 1991). Conversely, secondary forest growth in previously deforested and later abandoned areas has higher evaporative fractions (Giambelluca 2002) and higher evapotranspiration rates (Von Randow et al. 2017) than pristine forest areas. Thus, these areas can reduce the excess soil water and runoff.

The hydrological impacts due to climate change (CC) combined with land use and land cover change (LUCC) affect the availability of water resources and have important consequences for environmental and socioeconomic activities. Therefore, these effects should be carefully studied to improve water management and planning, even when considering that their impacts on water cycle are strongly dependent on scale and heterogeneity of landscape (D’Almeida et al. 2007).

To assess the effects of CC and LUCC on water resources, both climate and hydrological numerical models are frequently used (Demaria et al. 2013; Nóbrega et al. 2011; Cloke et al. 2013; Bravo et al. 2014, Siqueira Junior et al. 2015). This approach entails substantial uncertainty associated with the propagation of errors in the model chain (Jones 2000). Still, climate and hydrological projections provide important information for performing exploratory analyses of risk and uncertainties associated with decision-making processes related to CC adaptation (Mohor et al. 2015).

LUCC projections generally consider various public policies, biophysical properties, and socioeconomic assumptions (Prestele et al. 2016). In the context of the high level of uncertainty regarding future scenarios, scenario-based techniques can be used to explore and understand the consequences of alternative conditions in environmental studies. Aguiar et al. (2016) proposed three updated scenarios for the Brazilian Amazon through 2100. These scenarios combined exploratory and normative approaches, participatory methods, and qualitative and quantitative elements (Raskin et al. 2005).

Due to its abundant water resources, Brazil has become highly dependent on water availability for several economic activities, such as hydropower generation and agriculture (Nóbrega et al. 2011). Brazil currently gets three-quarters of its electricity from hydroelectricity (Horner et al. 2016). In this context, an increased need for energy to sustain economic growth has boosted governmental plans to expand hydropower in Amazonia. The new plants will increase the contribution of the Amazon region to Brazilian power generation from 10 to 24% (Empresa de Pesquisa Energética (EPE) 2012).

Located in the eastern divide of the Amazon Basin and on the transition between the Amazon forest and the Brazilian savanna (cerrado), the Tocantins River is considered one of the most important rivers for water supply, irrigation, transportation, and hydropower in the country. In terms of hydropower, the Tocantins Basin is the hub between the North/Northeast and South/Southeast power grids. Thus, the basin is a strategic point in Brazil (EPE 2006–2007). Previous studies have shown impacts of climate variability and land use conversion on the basin: Costa et al. 2003 analyzed 50 years of data and found 25% of increase on the discharge, but no statistically significant trend in the precipitation. Coe et al. 2009 modeled the same data set and attributed 2/3 of the increase in discharge to deforestation that occurred in that period. However, those studies did not consider the impact of secondary vegetation regrowth on the water cycle.

The objective of this work is to analyze the combined effects of CC and LUCC in water yield and evaluate the extent to which these effects could influence future energy production in the Tocantins Basin. In addition, we analyzed the effects of the growth of secondary forests on water and energy production, because secondary forest vegetation is generally not included on LUCC maps; its associated effects, in combination with climate and hydrological effects, are largely unknown.

Materials and methods

Study site

The Tocantins River Basin is the largest basin completely located within the Brazilian territory. The drainage area of the basin is more than 960,000 km2, or 11% of the Brazilian territory (ANTAQ 2013). The main rivers in the basin are Tocantins River and Araguaia River. The altitude of the Tocantins River Basin before it reaches the Araguaia River varies from 200 to 500 m in the majority of the area and is higher than 1000 m in the upstream portion of the basin. The average precipitation in the basin is 1869 mm per year, with a maximum of 2.565 mm on the Pará state coast. The precipitation regime occurs from south to north. Low amounts of precipitation are received along the border with Goiás and in the Northeast region, with an average of 1 mm/month between June and August (Tomasella et al. 2009). The precipitation decreases from 1850 mm in the west to 1000 mm/year in the east (MMA 2006). Additionally, evapotranspiration averages approximately 1.371 mm/year, and the long-term mean discharge is approximately 14,000 m3/s (MMA 2006). The main soils in the region are classified as Latossolos Vermelho-Amarelos and Vermelhos, with textures that range from average to clay. Although the land use has become more intense over the last two centuries, the Tocantins Basin still maintains some of the original phytophysiognomies. In the northern part of the basin, the original cover is tropical rain forest, and in the central and southern parts of the basin, known as Alto Tocantins, the original cover is mainly Brazilian savanna (cerrado) (Tomasella et al. 2009). However, this original cover is being replaced by other land use and land cover types. In Alto Tocantins, the cerrado that covered approximately 63% of the region in 1984 decreased to 47.9% in 2013 due to increases in anthropogenic cover (49.8%) and open water (2.3%) (Aquino Martins et al. 2015). The Tocantins-Araguaia region is one of the most important Brazilian hydrographic regions because of its location and hydropower potential which totals 26.285 MW. Additionally, the installed hydropower potential of the region is 6981 MW.

Distributed hydrological model—MHD-INPE

The Distributed Hydrological Model of the National Institute for Space Research (MHD-INPE) (Rodriguez and Tomasella 2016) is a regular grid-cell model that includes modules to solve the soil water budget, potential evaporation, transpiration, surface, sub-surface, and baseflows in each grid cell where several types of soil and land use are considered. Each grid cell is sub-divided into hydrological response units (HRUs), which are used to solve the water balance. The model uses the percentage of each type of land use and soil to determine the fluxes of a grid cell based on a weighted arithmetic mean. Further details about the MHD-INPE model are provided by Siqueira Junior et al. (2015) and Rodriguez and Tomasella (2016).

The model has been applied successfully in Amazonian basins such as the Ji-Paraná (Rodriguez and Tomasella 2016), Madeira (Siqueira Junior et al. 2015), Purus (Dalagnol et al. 2017), and Tapajos (Mohor et al. 2015) basins, as well as for flood forecasting in the Tocantins Basin (Falck et al. 2015).

Hydrological model calibration and validation

The Tocantins Basin was divided into nine sub-basins (Fig. 1). The MHD-INPE model was calibrated for each sub-basin using observed discharge data from 1970 to 1990 and the shuffled complex evolution method of automatic calibration developed at the University of Arizona (SCE-UA) by Duan et al. (1992, 1994). Observed meteorological data were used as input drivers for the model. Vegetation parameters were based on literature values for most vegetation types, of the same than those used by Rodriguez and Tomasella (2016). For the secondary forest parameters, we changed albedo and stomatal resistance based on flux tower data measured in a 21-year old secondary forest in Central Amazonia. Those parameters affect directly the evapotranspiration rates in the model. We also changed the vegetation height, zero displacement height, and roughness length, based on the vegetation characteristics of the site (Von Randow et al. 2017). Additionally, the calibration performance of MHD-INPE was qualitatively and quantitatively evaluated. The qualitative evaluation was based on a visual analysis of the hydrographs of observed data and simulated estimates of flow. In addition, for the quantitative evaluation, three different performance indices were used: the Nash-Sutcliffe efficiency (NSE), Nash-Sutcliffe efficiency of the logarithmic values of discharge (NSELog), and relative volume error (∆V) (Krause et al. 2005; Moriasi et al. 2007). Thus, the objective function was based on the average of NSE and NSELog efficiency coefficients and the convergence criteria was defined based on the minimal percentage of improvement in the objective function, which was fixed in 0.001.

Fig. 1
figure 1

Tocantins River Basin, the hydropower plant locations (left) and the sub-basins considered for calibration of the MHD-INPE model (right)

Full size image

For validation purposes, the ability of the model to simulate the present climate conditions was assessed using two different runs: (i) hydrological simulations based on observed meteorological data and (ii) hydrological simulations based on meteorological conditions provided by a regional climate model. The latter validation is important for the assessment of uncertainties in analyses of future hydrological projections resulting from bias in the climate model. Both runs were evaluated using four different indices derived from flow duration curves (FDCs) (Yilmaz et al. 2008; Ley et al. 2011): a measure of the high flow volume, MWH, which represents the high part of an FDC; a measure of the low flow volume, MWL, which represents the low part of an FDC; SEASON, an index that represents the difference between dry and wet seasons; and QSM, the slope between the 0.2 and 0.8 quantiles of an FDC.

Input data

To drive the hydrological model, different input data are needed to represent current or future environmental conditions. Current vegetation cover was obtained from the Radar da Amazonia Project of the Instituto Brasileiro de Geografia e Estatística (RADAM-IBGE) (IBGE 1992) and the Projeto de Elaboração de Mapas de Vegetação (PROVEG) (Sestini et al. 2002). Notably, these sources provided a vegetation map of 15 vegetation classes that was previously used in the SiB land surface model (Sellers et al. 1986). Yearly information regarding LUCC was obtained from the historical reconstruction of Leite et al. (2011) based on historical census data and contemporary land use classification and considering cultivated areas, natural areas, and planted pastures in Amazonia. The soil type distribution was extracted from the SOTERLAC/ISRIC soil map (Dijkshoorn et al. 2005), and meteorological and hydrological data were assembled from ANA, INMET, and INPE. Geomorphologic information was derived from the Shuttle Radar Topography Mission (SRTM) (Farr et al. 2007).

In the analyses of the effects of CC, dynamically downscaled data from the regional atmospheric model Eta-INPE (Chou et al. 2012, Marengo et al. 2012) were used as meteorological drivers in MHD-INPE. Eta downscaling uses a range of members of a perturbed physical ensemble of the HadC-M3 model as boundary conditions (Gordon et al. 2000; Collins et al. 2001), with different sensitivities based on the emission scenario SRES A1B (Nakicenovic et al. 2000). This approach is used because it accounts for the uncertainties of different possible outcomes of climate projections, from low to high emissions (Collins et al. 2006, Chou et al. 2012). In this work, the set of HadCM3 boundary conditions passed to Eta-INPE, described by Chou et al. (2012) and Marengo et al. (2012), are the unperturbed member (hereafter referred to as M1), the low-sensitivity member (referred to as M2), and the high-sensitivity member (referred to as M4).

The use of meteorological data from climate models in hydrological applications has spatial and temporal resolution constraints, and biases in the distributions of variables should be addressed before they are used as drivers in the hydrological models (Wood et al. 2002, 2004). Similar to Mohor et al. (2015), we corrected biases in the precipitation data using a percentile-to-percentile approach based on Bárdossy and Pegram (2011), and linear scaling (Teutschbein and Seibert 2013; Lenderink et al. 2007) was applied to correct the other meteorological data.

The future LUCC scenarios used in this study are the results of a combination of the LUCC-ME Modeling Framework and INPE-EM Emission Modeling Framework, as described in Aguiar et al. (2016). The LUCC-ME modeling framework (Aguiar et al. 2012) was used to generate annual rainforest deforestation maps through 2050, and the INPE-EM modeling framework (Aguiar et al. 2012) was used to generate the subsequent secondary vegetation dynamics in deforested areas based on a spatially explicit grid of 25 × 25 km2. Because information about savanna deforestation was not available, LUCC-ME projections were restricted to the Amazon Forest and, consequently, changes in savanna vegetation were not considered in this work.

The Land Use maps combine qualitative and quantitative elements using the Story and Simulation (SAS) approach proposed by (Alcamo 2001). The qualitative scenarios were built using a participatory approach (Folhes et al. 2015; Aguiar et al. 2016), and the resulting storylines were aligned with the Intergovernmental Panel on Climate Change (IPCC) Shared Socioeconomic Pathways (SSP). Storylines were constructed considering two main scenarios. The first, sustainability scenario A, was an ideal/desired scenario with equilibrium between socioeconomic achievements and the environment. This scenario was aligned with SSP 1. The second, fragmentation scenario C, was the opposite of scenario A and considered the depletion of natural resources and significant social inequality. This scenario was aligned with SSP 3.

In quantitative scenarios, the main goals are to choose elements from storylines to use in quantitative models and to define the adequate model parameters. The elements selected from the storylines were related to natural resources and land use dynamics. In the model of scenario A, the elements are as follows: a decrease in old-growth forest degradation; clear-cut deforestation reaching “zero (non-authorized) deforestation” by 2025; the regeneration of all illegally deforested areas on private properties; and high-value secondary vegetation. The final element will allow areas to become secondary forests by not disrupting regrowth in these areas. The quantification of scenario C considers deforestation rates at levels prior to those in 2004 in response less environmental protection due to agricultural expansion.

Analysis of the impacts on hydrology and on hydroelectricity

Changes in the discharge response as a result of different CC and LUCC scenarios were assessed through the comparison between projected long-term average (LTA) discharge and LTA discharge simulated for the historical period.

The Brazilian interconnect generation system includes more than a 100 hydropower plants, in addition to solar and wind mills. Operational rules are defined to provide synergy of the system considering constrains such as power transmission lines, ecological and navigation requirements, hydrological seasonal regimes, cascade optimization, market regulations, and power plant profitability (Correa da Silva et al. 2016). It is clear that, under climate change scenarios, those operational rules are likely to change drastically in response to climate and market future conditions. Because of the large source of uncertainties, and considering this study is restricted to the Tocantins Basin only, the operational rule adopted in this study maximizes the generation of each reservoir regardless the rest of the cascade system. Although this simplification does not reflect the future operational conditions, it allows comparisons between current and potential future scenarios on an equal basis in each dam.

The impact on energy production was accessed separately on large reservoirs and run-of-river (ROR) hydropower plants (HPP) because of the different management rules. While large hydropower plants have seasonal and multi-annual management rules and continuous energy production, ROR have limited management capacity (usually between days and weeks) and long non-production period during dry season. For large reservoirs, we calculated the potential energy generation as recommended by Paish (2002) using Eq. 1

$$ \mathrm{Ep}\left[\mathrm{MWm}\right]=0.0088\times \mathrm{Hm}\times \mathrm{Qm} $$
(1)

where Ep is the average energy potential [MW]; Hm is the average net head [m], which is assumed constant in all scenarios; and Qm is the mean net flow [m3 s−1]. The coefficient 0.0088 results from the product of the specific weight of water (1000 kg m−3), the efficiency factors of the turbine (0.93) and generators (0.97), the force of gravity (9.81 m s−2), and the coefficient 1.10−6, which is used to the average energy to units of MW.

The effects of CC and LUCC on the potential energy generation of large reservoirs were studied at four HPP: Serra da Mesa HPP, which has been operational since 1998 and has an energy capacity of 1275 MW (MMA 2006) and a net head of 129 m (FURNAS 1996 apud Oliveira Mesquita 2006); Peixe Angical HPP, which has been operational since 2006 and has an energy capacity of 452 MW and a net head of 26 m (Martins Filho et al. 2009); Luiz Eduardo Magalhães-Lajeado HPP, which has been operational since 2001 and has an energy capacity of 903 MW (MMA 2006) and a net head of 29 m (Hahner 2009); and Tucuruí, which has been operational since 1984 and has an energy capacity of 8370 MW and a net head of 60.8 m (Cidade Tucuruí 2016).

For ROR HPPs, we used the method proposed by Vogel and Fenessey (1995) to assess the impacts of CC and LUCC on the hydrological regime. The method analyzes the viability of an ROR HPP through the use of power duration curves (PDCs) derived from annual FDCs. First, we sorted the values of daily discharge in each simulation year, which resulted in an FDC for each year. Second, for each day of a year, we calculated the median of all years, which led to one FDC for a typical (hypothetical) year. Third, we calculated the energy production associated with each discharge to obtain a PDC for a typical (hypothetical) year. Fourth, the area under the PDC was integrated to determine the typical annual energy production (Vogel and Fenessey 1994).

We analyzed the viability of three ROR HPPs under different scenarios. The first is Cana Brava, which has been operational since 2002 and is permitted thorough 2033. It is an ROR HPP with an installed capacity of 450 MW, a head fall of 43.1 m, a minimum flow discharge (MFD) of 150 m3 s−1 (ANA 2006), and a reservoir with a 139-km2 surface area. Cana Brava is located in sub-basin “Carolina” (4) (MMA 2006; ENGIE 2016). The second is São Salvador, which has been operational since 2009 (Tractebel Energia 2010) and is permitted to operate through 2023. It is an ROR HPP with an installed capacity of 243 MW, a head fall of 22.84 m, an MFD of 102 m3 s−1, and a reservoir with a 104-km2 surface area. The São Salvador plant is located in sub-basin Carolina (4) (MMA 2006; ENGIE 2016). The third is Estreito, which has been operational since 2011. It is an ROR HPP with an installed capacity of 1087 MW, a head fall of 18.9 m, an MFD of 585 m3 s−1, and a reservoir with a 555-km2 surface area. The Estreito plant is located between the outfall of sub-basin Carolina (4) and sub-basin “Descarreto” (5) (MMA 2006; ENGIE 2016). Figure 1 shows the locations of the ROR HPPs and large reservoir dams in the Tocantins Basin. For energy production purposes, we estimated the operational time of the plant as the time period when the river discharges were above the MFD.

Results and discussion

Hydrological model calibration and validation

Model calibration and validation show that MHD-INPE performed well in simulating historic stream flow at gauging stations in the Tocantins Basin. The NSE and NSELog indices were considered satisfactory at one of the gauging stations (Tesouro, superior to 0.5) and good and very good at the others (varying from 0.66 to 0.94) according to the range proposed by Moriasi et al. (2007). The volumetric errors were lower than 10% in most of the basin, except in São Félix do Araguaia, where a volumetric error of 13% was observed. The performance indices of the calibration of MHD-INPE model in the Tocantins Basin are shown in Table 1.

Table 1 Performance index results obtained for the calibration of nine gauging stations in the Tocantins Basin
Full size table

For validation purposes, we calculated the regression lines between the indices derived from FDCs of hydrological simulations based on observed meteorological data and of modeled hydrographs based on meteorological conditions provided by a bias-corrected regional climate model, both for historical period. The modeled hydrographs showed good agreement with simulations for the nine sub-basins, compared to the 1:1 line. For MWH, SEASON, and QSM indices, the slopes were between 1.01 and 1.07. A small deviation is observed for MWL in some sub-basins (slope = 0.84), indicating a possible underestimate of base flow. The coefficient of determination was higher than 0.89 for all indices, except for QSM (R2 is 0.73). Fig. S1 details the validation results.

Impacts on hydrology

To evaluate the effects of CC on water production, the land cover was held constant based on 2008 conditions, and only the effect of CC was analyzed. Then, LUCC was included in the projection to compare the effects of CC and the effects of interactions between CC and LUCC (CC + LUCC).

In the Tocantins Basin, the three CC scenarios considerably affected the LTA discharge compared to the historical period (Fig. 2). For simplicity, the baseline used as a reference corresponds to the average of the simulations in the historical period using all Eta climate scenarios. The highest impact occurred in the M1 scenario with a decrease of approximately 20% from 2011 to 2040 and 32% decrease from 2041 to 2070 in precipitation, resulting in corresponding decreased of 39 and 57% in discharge. The second highest impact occurred in the M4 scenario, which also exhibited reductions in precipitation of 12% from 2011 to 2040 and 24% from 2041 to 2070, resulting in corresponding decreases of 30 and 50% in discharge. The M2 scenario exhibited precipitation reductions of 11% from 2011 to 2040 and 13% from 2041 to 2070, resulting in discharge reductions of 26 and 28%.

Fig. 2
figure 2

Projections of monthly long-term discharge at Tucuruí gauging station (representing the entire Tocantins Basin) for the periods between a 2011 and 2040, b 2041 and 2070, and c 2071 and 2099. The baseline is the average of simulations of MHD-INPE in the historical period fed with all Eta climate variants

Full size image

To address the combined impacts of CC and LUCC on the water production, Table 2 presents the change in discharge resulting from CC and CC + LUCC compared to the baseline. Additionally, Table 2 illustrates the importance of considering the presence of regrowth (secondary forest) in the analysis. The decrease in discharge caused by CC was slightly minimized by deforestation effects. However, considering the presence of secondary forest growth in abandoned pastures reduced the attenuation caused by deforestation on the discharge in scenario C1 and enhanced the effects of CC in scenario A by further decreasing the discharge compared to CC only simulations. Scenario C1 is a less sustainable scenario, in which the deforestation rate increases over time. This increase in deforestation can lead to a discharge increase due to the replacement of forest by pasture (Roberts 2009). In scenario A, the decrease in deforestation and the abandonment of deforested areas lead to an increase in secondary forests (Fig. S2). Subsequently, the discharge decreases due to the increased evapotranspiration rate (Giambelluca 2002; Von Randow et al. 2017), in spite of the low percentage of secondary forest. Table 2 shows the results for CC scenario M1, which had a major effect on discharge. Note that the model does not consider the potential effects of soil erosion caused by deforestation, which may diminish the soil moisture input, affect soil moisture storage, and consequently influence runoff production (Wang and Shao 2013).

Table 2 Results of the discharge simulations considering CC (Eta M1) and both CC and LUCC scenarios. Next to simulated annual discharges, percentage changes in relation to the baseline (1970–1990 average) are presented in parentheses
Full size table

Impacts on energy production

We analyzed the effects of CC and CC + LUCC on energy production in three sub-basins: Carolina, which contains the majority of the HPPs in the Tocantins Basin, Descarreto, and Tucuruí. Carolina and Descarreto did not exhibit a significant difference between the A and C1 LUCC scenarios because most of these areas are covered by savanna (Fig. S2a and S2b), for which deforestation is not considered in the LUCC-ME scenarios. The Tucuruí basin includes areas drained by the Tocantins River where forest cover is expansive and differences exist between LUCC scenarios (Fig. S2c). Therefore, for simplicity, analyses of the Carolina and Descarreto sub-basins were only performed for scenario C1, but both scenarios were analyzed in the Tucuruí sub-basin.

Impacts on the energy production of large dams

Changes in potential energy production in the future are assessed using historical simulations as a baseline. The energy potential of all three large hydropower dams decreases (Fig. 3 and Fig. S3) because the average reductions in precipitation were similar in all three sub-basins. However, the projected LUCC-ME scenario only induces significant changes to the vegetation cover in the Tucuruí dam basin (Fig. S2c).

Fig. 3
figure 3

Projected impacts on the energy potential at Tucuruí dam caused by climate change only and climate change plus land use change scenarios

Full size image

As expected, changes in the potential energy production are similar to the changes on discharge. CC only causes a reduction of 27 to 38% in the first time-slice, and 29 to 56% in the second time-slice, in comparison with the energy production estimated with baseline discharge. When the C1 LUCC-ME scenario is considered, the reduction is slightly lower (reduction of 26 to 37% in the first time-slice, and 27 to 56% in the second time-slice), mainly because deforestation increases under this scenario, which increases discharge in each basin. Conversely, the decrease in the energy potential becomes more pronounced when scenario A is considered (reduction of 29 to 40% in the first time-slice and 32 to 58% in the second time-slice). Under this scenario, the deceleration of deforestation increases secondary forest growth (Fig. S2c.1) and promotes higher evapotranspiration rates (see Table S1) and lower water excess and runoff. Thus, the combined effects of CC and LUCC are enhanced.

The offset of CC due to LUCC can be seen in scenario C1. In this scenario, primary forest cover decreases and pasture land increases, while the areas of secondary forest remain nearly unchanged. When the LUCC scenario C1 is considered, the energy production decrease is lower than when only CC is taken into account.

In scenario A, even with only approximately 10% of secondary vegetation, in addition to the decrease on energy potential caused by CC, evapotranspiration further decreases the energy potential from 2011 to 2040 by 2% and that from 2041 to 2070 by 3%.

In scenario C1, when deforestation increases over time and the area of secondary forest is small and unchanged, the energy potential increases only 1% from 2011 to 2040 and 2% from 2041 to 2070 compared to only considering CC.

In the Carolina and Descarreto sub-basins, the decrease occurs mainly as a result of CC in both the 2011–2040 and 2041–2070 periods (Fig. S3) because these basins do not undergo significant LUCC (Fig. S2a and S2b). Although the precipitation reduction varied from 10 to 38% in the CC scenarios, reductions in the discharges were higher resulting in more severe impacts on the energy potential, which varied from 30 to 80% reductions based on the current potential. As expected, the highest impact occurred in the M1 scenario, which was associated with the most significant decrease in precipitation in comparison to that in the historical period.

Our results are comparable to the findings of other studies in Southern Amazon Basin (Mohor et al. 2015, Siqueira Junior et al. 2015, Stickler et al. 2013), which showed that the increase in deforestation rates may partly counterbalance effects of precipitation reduction. However, and unless previous studies, our analysis showed that secondary forest regrowth in abandoned areas might play an important impact on water availability.

Energy production of run-of-river dams

The typical PDCs for hydroelectric generation at the Cana Brava, São Salvador, and Estreito HPPs are presented in Fig. 4, considering the median values from 2041 to 2070, which is assumed as the HPP lifetime. For comparison, the baseline represents the PDC of the historical period simulations (1970 to 1990) obtained using observed data. We can see that in historical period simulation, although there was a reduction on the energy production during part of the time, the discharge was above the minimum flow of the plant, and the plant remained operational 100% of the time.

Fig. 4
figure 4

Annual processed power duration curves based on median daily values at Tocantins run-of-river HPPs for the period between 1970 and 1990 (baseline, black lines) and 2041 and 2070 (future simulations, colored lines). The baselines are obtained using observed data as input

Full size image

As previously discussed for large dams, in sub-basins 4 and 5, the impact on runoff is due to CC; therefore, only CC is examined in this section. The most critical projection scenario (M1) indicates that Estreito is the plant that is affected the most based on the time of operation. Notably, the plant is non-operational 77% of the time. The second most affected plant is São Salvador, which is non-operable 65% of the time from 2041 to 2070. Additionally, Cana Brava does not operate 59% of the time in this period (Fig. 4). In terms of the effect on energy production per year based between 2041 and 2070, the Estreito HPP production is 5.2 × 105 MWh/year, which is 87% lower than the hypothetical production of 3.96 × 106 MWh/year simulated for the historical period for this HPP. Cana Brava exhibited the second highest reduction in production per year of 85%, which represents a decrease of 2.16 × 106 MWh/year from the hypothetical production in the historical period to 3.2 × 105 MWh/year. Finally, São Salvador exhibited a production of 2 × 105 MWh/year, which represents a reduction of 83% of the hypothetical production of 1.2 × 106 MWh/year in the same period. Therefore, we noted that a reduction of 12 to 38% in precipitation causes a drastic reduction of up to 90% in the operating time of the HPPs.

Conclusion

CC projections of reduced rainfall over Tocantins basin induced decreases in discharge simulated by the MHD-INPE hydrological model in the entire basin. However, this reduction was weakened when land use change scenarios provided by the LUCC-ME model were included in the simulations. These changes increased the discharge when only deforestation, without regrowth, was considered. The weakening occurred mainly in scenario C1, which considers significant social inequality and the depletion of natural resources, i.e., high deforestation. However, when the regrowth of secondary vegetation in deforested area is considered in LUCC scenario C1, the effects are of similar magnitude to the CC only impacts, and even enhanced in sustainable scenario A as a result of the high evapotranspiration rate of secondary forests in comparison with the evapotranspiration of pastures and primary forests.

The energy potential of large dams and the expected operating time of ROR HPPs decrease significantly as a result of the decrease in discharge due to CC. Additionally, LUCC moderately counteracts these effects. Nevertheless, the region of the Tucuruí HPP includes an area of secondary forest and the decrease in its energy potential may again become pronounced when scenario A is considered due to the deceleration of deforestation and the conversion of pasture into secondary forest.

Due to the uncertainties about the rules of operation of the Brazilian interconnected system under climate change scenarios, the operational rule adopted in this study maximizes the generation of each power plant. Future studies should consider the large varieties of conditions that would affect the generation of the reservoir cascade of the basin. Also, our simulations did not include scenarios involving water consumption. In these cases, the reduction in potential energy production may be further magnified.

Despite the high variability among the climate model integrations, the annual energy production is likely to decrease based on the analyzed projections and the current HPP design. However, MHD-INPE does not consider the effects of erosion and consequent silting caused by deforestation, and these factors would affect water production.

Our results highlight the importance of including the effects of CC and LUCC, such as deforestation and secondary forest growth, in assessments of the water resource availability, energy production, and the feasibility of HPP deployment.