Reducing uncertainties of climate projections on solar energy resources in Brazil | Scientific Reports

Scientific Reports volume 14, Article number: 23586 (2024) Cite this article

Metrics details

The share of solar power in Brazil’s electrical grid has rapidly increased, relieving GHG emissions and diversifying energy sources for greater energy security. Besides that, solar resource is susceptible to climate change, adding uncertainty to electrical grid resilience. This study uses satellite and reanalysis data to evaluate the performance of CMIP6 models in replicating and predicting surface solar irradiance (SSR) in Brazil. The results from the most reliable models indicate an increase in SSR by 2% to 8% in most regions, with a decrease of around 3% in the South. These findings highlight the potential for increased photovoltaic (PV) yield if backed by supportive public policies while underlining the importance of uncertainty assessment of climate models.

Solar energy is a promising alternative to meet the growing electricity demand while reducing greenhouse gas emissions. However, they are weather-dependent and require careful planning to minimize the impact of the intrinsic intermittence on the energy distribution system1. Several worldwide studies investigated the time and spatial variability and tendency of weather and climate-driven renewable energy resources based on data provided by climate models2,3,4,5,6,7. Solar radiation assessments based on climate models found an average decrease of 0.1 to 0.4 W/m² per decade from 1979 to 2014 globally8. Several studies for Europe show that PV power generation will increase at the end of the century2,5,6 despite some evidence that the future climate scenarios can drive a more complex spatial change of surface solar radiation (SSR)9. Nevertheless, other studies adopting a small subset of models or even downscaled high-resolution climate projections point in the opposite direction, indicating a decrease in PV potential for Northern European countries by the end of the century10,11. In summary, previous studies suggested that the uncertainty endures showing distinct climate change signals depending on the methodological approach used in investigation2,10,12.

Since 2018, Brazil has been witnessing a significant surge in its installed PV capacity, which has now surpassed 30.7 GW in the second quarter of 202313,14,15. As PV power generation is set to play a more substantial role in Brazil’s future energy mix, it becomes imperative to delve into the impact of climate change on the spatial and temporal variability of solar energy.

It’s worth noting that more assessment studies need to be conducted that specifically address the impact of climate on solar radiation over Brazilian territory. For instance, historical records of surface solar irradiance at 129 automated weather stations have shown both positive (\(+40 Wh/m^2\) per year) and negative (\(-50Wh/m^2\) per year) trends for the northern and southern portions of the Brazilian Northeastern (NEB), respectively16. Recent studies based on global climate models have indicated an increase in solar energy potential for most of Brazil, with a high level of resilience for a 4-degree specific warming level scenario17,18. However, these results also present conflicting outcomes for Brazil’s Southeast and Midwest regions, suggesting high uncertainty. Zuluaga et al. showed that PV power generation in Brazilian territory would likely decrease by the end of the century in both SSP2-4.5 and SSP5-8.5 scenarios, except for the northern Amazon region19 based on an ensemble from the CMIP6 climate models.

The previous results revealed a high level of uncertainty in climate change impact assessments, partly due to the different methodologies and datasets adopted. A more rigorous selection of the climate models to be used in an ensemble analysis, focusing on selecting those with the best performance and ability to represent current climate patterns, is essential in improving the analysis of future climate scenarios. Bias-correction methods and statistical indicators to evaluate the model’s skill in reproducing spatial and seasonal patterns observed in historical reference datasets, like satellite-based or meteorological reanalysis, are fundamental to achieving more confidence in the climate change impact assessment.

This work aims to reduce uncertainties in the future surface solar irradiance \(SSR\) and PV yield for Brazilian territory by departing from an accurate historical dataset combining satellite and reanalysis and adopting reliable statistical methods to rank the CMIP6 climate models regarding their ability to represent spatiotemporal variability of \(SSR\). We used the ensemble with the highest-skill CMIP6 models to assess the impact of climate change on solar energy up to 2100 under SSP2-4.5 and SSP5-8.5 scenarios in order to support energy entrepreneurs, governmental and non-governmental organizations in planning and building the Brazilian power generation system resilient to future climate conditions. Case studies for nine particular interesting sites for PV power generation delivered geographical outcomes that can help design and develop public policies to promote environmental sustainability and social energy justice. All research results and Python codes developed for data analysis are available for public access20,21.

Figure 1 provides a comprehensive view of the performance of CMIP6 models in reproducing \(SSR\) spatial patterns, providing visual information on the alternation between positive and negative bias for Brazilian territory. Uncertainty in model estimates is noticeable due to the large spread of deviations. The 40-models’ ensemble (\(ENS\)) reproduces the \(SSR\)’s spatial pattern over Brazilian Northeast and Central regions with reduced bias. Nevertheless, the \(ENS\) overestimates (around \(50 W/m^2\)) the climatological \(SSR\) in the Amazon region. These results agree with findings showing a negative bias for precipitation outputs of CMIP6 models for the north of the Amazon22,23,24.

The panel presents the mapping of the BIAS deviation (in W/m2) shown by the \(SSR\) estimates provided by the ensemble (upper left corner) and by each of the forty climate models from CMIP6 used in the study. The model names are positioned above the corresponding map. The authors prepared maps using the available Python libraries.

Figure S1 in the Supplementary Material shows the bias-adjusted reanalysis (\(ERA5_{QM}\)) for the 1980-2014 period used as the reference dataset to evaluate CMIP6 models’ skill in representing spatial and seasonal patterns in three target regions (see Fig. 2): Brazilian Northeast (area A1), South (area A2) and Central (area A3). Regarding the target area A1, model M2 overestimates the \(SSR\) around \(40 W/m^2-50 W/m^2\)\((\approx 15\%-20\%)\) while M5, M18, M26, and M34 underestimate by the same amount. Most models overestimate around \(30 W/m^2\)\((\approx 12\%)\) in areas A2 and A3, while models M5, M6, and M31 underestimate up to \(20 W/m^2\)\((~8\%)\).

The results of the spatial correlation between CMIP6 outputs and \(ERA5_{QM}\) are available in the supplementary material (Table S1) that delivers information on the monthly spatial correlation presenting statistical significance (p-values \(< 0.05\)) achieved by climate models in the three target areas. The ensemble (\(ENS\)) and twenty-two CMIP6 models presented a significant positive correlation \((p < 0.05)\) in all twelve months for all three target areas. The \(ENS\) presented the best correlation index in the three target areas, supporting the benefit of working with model ensembles. The other eighteen models presented no significant correlation, mostly in transition months between wet and dry seasons (April and October), and were discarded for the following analysis.

In the next step, the Taylor Skill Score (\(TSS\)) (described in \(eq. 2\)) was used to determine how accurately the twenty-two CMIP6 models depict the seasonal changes in \(SSR\) within the three target areas. Figure 3 displays the Taylor diagram and Table S2 (Supplementary Material) lists the time correlation index (\(r\)), unbiased root mean square deviation (\(uRMSD\)), and standard deviation (\(SD\)) ratio attained by the 22 models and the \(ENS\), assuming the \(ERA5_{QM}\) dataset as a reference. Time correlation was computed over the twelve-month climatological cycle. The blue markers denote the ten models presenting \(TSS\) greater than 0.9.

(left) Surface solar irradiance \(SSR\) reference map based on the \(ERA5_{QM}\). The rectangles indicate the target regions used in the performance evaluation of CMIP6 models: area A1 in the Northeast, A2 in the South, and A3 in Central Brazil. Standard deviation maps of \(SSR\) calculated among members of the \(ENS\) (central) and \(SME\) (right). The authors prepared maps using the available Python libraries.

A second ensemble (referred to as \(SME\)) was processed using the ten best-performing CMIP6 models. The statistical parameters achieved by \(SME\) in the three target areas are also listed in Table S2. The \(SME\) showed a little higher standard deviation ratio than the ensemble \(ENS\) with all 40 CMIP6 models; however, it overcomes the \(ENS\) in all other metrics, including the \(TSS\) (0.96 compared to 0.90). The maps shown in Fig. 2 indicate around 20% reduction in the standard deviation of the SSR data from the \(SME\) compared to the \(ENS\) in the whole Brazilian territory. The dispersion among members is related to the uncertainty in the ensemble mean, evidencing that the \(SME\) attained an expressive reduction in uncertainty while sustaining similar skill in reproducing \(SSR\) historical climatology.

Taylor’s diagram compares the CMIP6 model’s performance regarding the \(ERA5_{QM}\) database. Each marker represents the statistical metrics achieved by the CMIP6 climate models and the ENS ensemble. The blue (red) markers indicate the CMIP6 climate models with a Taylor index \(TSS\) above (below) 0.9. The black marker indicates the statistical metrics achieved by (\(ERA5_{QM}\)). Table S2 (Supplementary Material) lists the statistical metrics achieved by each model and ensemble.

Model M25 (HadGEM3-C31) is the top-performing model in terms of \(TSS\), with the highest time correlation and lowest \(uRMSD\). The \(ENS\) has the second lowest (\(SD\)) but performs poorly in other statistical indexes compared to the ten best-performing models.

The \(SSR\) changes for future scenarios obtained from \(SME\) are presented in three timeslices: near-future (2015-2040), mid-term future (2041-2070), and end-of-century (2071-2100). Complete plots and maps for the three timeslices and both climate scenarios (SSP2.45 and SSP5.85) are available at https://doi.org/10.6084/m9.figshare.25396612 for public access.

Figure 4 shows the seasonal variation of the climate change factor (\(CCF\)) and \(SSR\) in the three target areas. More details on seasonal variation of \(SSR\) derived from the \(ERA5_{QM}\) and CMIP6 Smart Ensemble (\(SME\)) are shown in Figs. S3 to S5. In Brazilian Northeast (Area A1), Fig. 4b shows that \(CCF\) is positive throughout the year, except in January and February for the end-of-century. In both climate scenarios, the highest \(CCF\) occurs in transition periods between dry and wet seasons, September to November and March to April. Due to the \(CCF\) seasonal variation, the monthly mean \(SSR\) increases around \(10 W/m^2\) in the austral autumn and spring seasons in both scenarios and all timeslices. Such an increase in \(SSR\) agrees with the reduction in precipitation, notably for the SSP5-8.5 scenario25. The seasonal \(CCF\) variation for the central region of Brazil (area A3, Fig. 4d) is similar to area A1. The \(CCF\) assumes positive values throughout the year, with the highest \(CCFs\) in the wet season from October to March (3-5% in SSP5-8.5 and 2-3% in SSP2-4.5) at the 2071-2100 timeslice.

(a) Seasonal variation of mean (\(SSR\)) based on the \(ERA5_{QM}\) database; (b) Seasonal variation of mean \(CCF\) obtained from \(SME\) for SSP2-4.5 and SSP5-8.5 pathways for target area A1; (c) for area A2 and; (d) for area A3.

Figure 4c shows an opposite seasonal pattern in the South of Brazil (area A2). The \(CCF\) shows negative values most of the year except for January and February, ranging from \(0.5\) to \(1.5\%\) in both scenarios and timeslices. The decrease in \(SSR\) is more severe during the Wet-Dry transition months when the predicted \(CCF\) is around\(-2.0\%\) (\(-4.5\%\)) in SSP2-4.5 (SSP5-8.5) at the end-of-century.

Figure 5 displays seasonal maps of the \(CCF\) for the Brazilian territory according to \(SME\). For the near-future timeslice (Fig. 5a and d), the impacts are quite similar in both scenarios (SSP2.45 and SSP5.85). The \(SSR\) is expected to increase by up to \(3\%\) in the Amazon and up to \(5\%\) in Central and Southeastern Brazil, though more pronounced during Spring and Summer. The 2070-2100 timeslice (Fig. 5c and f) shows a decrease of up to \(-2\%\) (\(-4\%\)) in \(SSR\) in the northern part of the Brazilian Northeastern region under SSP2-4.5 (SSP5-8.5) during Summer. An \(SSR\) decrease of up to \(5\%\) is also noticed in Southern Brazil for SSP5.85, lingering from Fall to Spring. On the other hand, \(SSR\) is likely to increase by up to \(6\%\) (\(10\%\)) in the Amazon, Central, and Southeastern Brazilian regions under SSP2-4.5 (SSP5-8.5).

The seasonal mean CCF predicted by the SME for the SSP2-4.5 in 2015-2040 (a), 2041-2070 (b), and 2071-2100 (c) timeslices; and for SSP5-8.5 in 2015-2040 (d), 2041-2070 (e) and 2071-2100 (f) timeslices. The columns are from left to right: summer, autumn, winter, spring, and annual. The gray dots over the maps represent the grid locations with statistical significance (p-value <0.05). The authors prepared maps using the available Python libraries.

Figure 5 shows a noticeable signal in the annual average maps with \(CCF\) ranging from \(2\%\) to \(6\%\) in the near future (SSP2-4.5) to \(-3\%\) to \(8\%\) by the end of the century in SSP5-8.5. Those amplified positive signals in \(CCF\) during spring over semi-arid and Central regions of Brazil play a crucial role in reducing the vulnerability of the Brazilian electrical system to climate change, especially where most of the utility-scale PV power plants are currently operating. The Brazilian Interconnected Electricity System (SIN) mainly relies on hydropower and faces critical operation from September to November due to the end of the dry season. During these months, hydropower reservoirs are usually depleted and exposed to climate variability, implying high risks. Past electricity supply crises are clear evidence of this situation26. In this sense, the higher solar resource levels during spring add resilience to the future of the national electricity system for both scenarios. However, impacts on other renewable resources (wind, hydro) are also expected, and an integrated evaluation should be conducted in the work in progress.

Solar PV technologies have rapidly grown in Brazilian metropolitan regions (MR) due to a sharp cost reduction and recent regulations encouraging distributed generation13. The SSR’s spatial distribution and future trends highlight the challenges in optimizing solar power benefits for Brazil’s energy mix while reducing risks and GHG emissions to fulfill international commitments. Based on recent works using data from PV power systems operating in Brazil27, we used the performance ratio (PR) around 0.8 to evaluate the impact of climate change on solar PV yield.

Figure 6 shows the annual PV yield from 1980 to 2100 assessed using the SSR outcomes of the SME for SSP2-4.5 and SSP5-8.5 pathways in seven MRs and two remote areas, covering different climate regimes. We assumed that technological advancements in PV technology will offset the losses in solar energy conversion due to the rise in ambient temperature. Table 1 lists the trend slope and p-value of the linear regression fitted for the nine locations and climate pathways. The statistically significant trends are highlighted in bold blue numbers.

Annual trends of \(SSR\) generated from \(SME\) outputs for seven metropolitan areas of Brazil - (a) Petrolina, (b) Fortaleza, (c) Brasília, (d) Belo Horizonte, (e) São Paulo, (f) Porto Alegre and (g) Manaus - and two remote areas (h) Boa Vista (located further north in the Brazilian Amazon), and (i) Colniza (located in the deforestation belt in the Southern Amazon). The geographical location of the nine spots is shown in Fig. 2.

Fortaleza and Petrolina are MRs located in the Northeastern region of Brazil, where \(SSR\) is at its highest. Fortaleza is on the coast near the Equator and holds more than three million inhabitants. It also has abundant wind energy resources throughout the year28,29, which allows hybrid wind-solar projects to take place, reducing power intermittence. On the other hand, Petrolina is located in the semiarid region close to the largest regional hydropower reservoir, Sobradinho (1050 GW), where floating PV power plants could improve water storage and management during extreme drought periods and meet water demands for other uses besides power generation30,31.

Based on Fig. 6, the \(SME\) predicts that PV yield will increase \(0.11 kWh/kW_p\) per year (\(0.18 kWh/kW_p\) per year) in Petrolina, until the end of the century for SSP2-4.5 (SSP5-8.5) pathway (with low statistical significance). However, \(SME\) predicts a significant negative trend (\(-0.25 kWh/kW_p.y^{-1}\) (\(-0.36 kWh/kW_p.y^{-1}\))) on annual PV yield for the same climate scenarios in Fortaleza. Although the results indicate a likely reduction in solar PV resources on the equatorial coast of the Brazilian Northeast, the portfolio is rather resilient since most large-scale PV utilities are being implemented in the semi-arid region. The impacts are limited to \(+/-2\%\) over the current PV yield (reference timeslice) and may not primarily affect the financial feasibility of the sector.

Brasilia and Belo Horizonte are important MRs where solar power rapidly expands due to the region’s abundant solar resources, and reduced seasonal variation. Recent studies indicate that investments in PV power distributed generation (PVDG) have the lowest payback period in Brazil32. For both climate scenarios, the \(SME\) projections show a positive trend of PV systems yield in both areas in the upcoming years. However, the trend slope is around twice as high in SSP5-8.5 (\(0.46\) to \(0.66 kWh/kW_p.y^{-1}\)). It means around a \(5\%\) increase in \(SSR\) in the more extreme scenario.

São Paulo is the largest Brazilian MR, with nearly 22 million inhabitants living in around \(8000 km^2\) in the Brazil Southeast. São Paulo is the country’s primary energy consumption center and has seen an exponential increase in PVDG since 202013,14. Figure 6 and Table 1 indicate that climate change will not particularly affect the annual PV yield as the trend slope is slight and has no statistical significance.

Porto Alegre is in the Brazilian Southern region, where \(SSR\) has the highest seasonal variability due to the solar geometry and climate dynamics in the region33,34,35. The SME predicts a robust decreasing trend up to \(-3\%\) around \(-40 kWh/kW_p.y^{-1}\) under SSP5-8.5 for Porto Alegre. The decreasing trend is also present under SSP2-4.5 but is not statistically significant. However, the South of Brazil also has a high wind energy potential29, and hybrid wind-solar power projects can be an alternative to reduce the impact of climate change in future solar PV generation.

The three remaining locations, Manaus, Boa Vista, and Colniza, are facing a pressing issue of energy access. Despite being far from the leading consumer centers, these regions urgently need to address their energy challenges. Manaus, the largest urban center in the Brazilian Amazon region, is a hub of economic activity. Boa Vista, the northernmost state capital in Brazil, is not served by the National Interconnected Electricity System (SIN). Still, diesel-powered thermal generation mainly meets its electricity demand, with a small fraction imported from neighboring countries. Colniza, a small town in the southern Amazon region, has an economy heavily based on agriculture, with a large portion of the population living in rural areas without access to electricity utilities.

Most of the Amazon region is not linked with the Brazilian Interconnected Electricity Distribution System (SIN). Instead, isolated power systems that rely on fossil fuels are spread throughout the region, and their costs are financed by compulsory taxes included in the energy tariff paid by all Brazilian electricity consumers. Solar PV systems are the primary alternative for isolated power systems to reduce greenhouse gas emissions and lower high taxes on electricity bills. According to the SME outcomes, the PV yield scenarios show the highest increasing trend in the Brazilian Amazon region for both SSP pathways. In the SSP5-8.5 scenario, the PV yield may increase up to \(4\%\), strengthening the solar power option to meet the power demand in the region.

This study shows that future climate projections present significant uncertainties, introduced not only by scenarios and assumptions but mainly by the spread of climate models’ performance. Results show that uncertainties are significantly reduced by choosing an appropriate subset of best-performing models while maintaining skill in simulating \(SSR\) patterns. In general, the CMIP6 models tend to overestimate \(SSR\) in the Amazon, agreeing with other assessments showing a decrease in precipitation in this region22,23 while performing better in Brazil’s Northeast region. Future projections show an increase in \(SSR\) by 2% to 8% in most Brazilian regions, with a decrease of around 3% in the South, particularly under the SSP5-8.5 scenario. The \(SSR\) change signal over the Brazilian Northeastern region matches with other studies18,19,24. However, for central and southeast Brazil, results vary depending on the methods and models adopted, evidencing the higher uncertainty. The low magnitude of changes compared to model ensemble spread remains as the main source of uncertainty and is an inherent limitation of this study. Nevertheless, this approach improves the confidence on climate change impact over \(SSR\) due to a broad and systematic assessment of a large sample of models.

Higher SSR levels were significant in the trend analysis for most Brazilian metropolitan areas. Moreover, the seasonal change depicts an increased \(SSR\) during Brazilian drier months (from September to November), reducing the vulnerability of Brazil’s electrical system, which relies heavily on hydropower. The results also indicate a trend of increased PV generation productivity for future climate scenarios across much of Brazilian territory, with a more pronounced impact in the Amazon region that is not served by the National Interconnected Electricity System (SIN). However, in some areas, climate models point to a decrease in productivity of less than -3% along the northeastern coast and the southern part of the country by the end of the century. From an energy planning perspective, these results may support regional development strategies for improving the resilience of the Brazilian power system to future climate conditions. More studies investigating changes in the frequency of \(SSR\) extremes could contribute to a more complete assessment, inviting further research and engagement.

The study area comprises continental Brazilian territory, the fifth-largest country in the world. From North to South, Brazil extends for almost 4400 km, with the Equator and the Tropic of Capricorn running through it. Most of the population lives near the Atlantic coast, and the largest cities are in the Southeastern region. Brazil encompasses diverse important biomes, including the Amazon Forest in the North region, Pantanal wetland in the Mid-west area, Caatinga (semiarid) in the Northeast, and Pampa in the South.

Despite the diverse climate and environmental features33,35, previous studies indicated that surface solar irradiation is relatively uniform in Brazilian territory36. The semi-arid area of the Northeastern Brazilian region receives the highest incoming solar irradiation, up to 6.27 \(kWh.m^{-2}.day^{-1})\). The \(SSR\) is around 6.0 \(kWh.m^{-2}.day^{-1}\) in Brazil’s Northern and Central regions during the dry season from July to September. The Southern and Southeastern coastal regions present higher annual variability due to the higher latitude, the onset of the monsoon system during summer, and the higher frequency of cold fronts during the fall and winter. South Brazil receives higher \(SSR\) than the Northern region during the austral summer due to the more extended day length and higher cloud coverage in the North linked to the southern shift of the Intertropical Convergence Zone (ITCZ).

Figure 7 illustrates the analytical steps for assessing the impacts of climate change on the solar energy resource in Brazilian territory. The investigation used \(SSR\) data from three repositories: the Coupled Model Intercomparison Project Phase 6 (CMIP6), the ERA5 reanalysis provided by ECMWF (European Centre for Medium-Range Weather Forecasts), and satellite-based data provided by INPE (Brazilian Institute for Space Research).

The flowchart shows the step sequence of the methodology used to investigate future solar energy resource scenarios based on CMIP6 climate models.

The CMIP6 provided the \(SSR\) data from 40 global climate models. Before using CMIP6 future projections, we assessed each model’s reliability by comparing them with a truth reference for spatiotemporal patterns. Several studies suggest that the ERA5 reanalysis database fulfills the required features (long and continuous time coverage, spatially homogeneous, and reliable ground data assimilation) to be the truth reference database22,34. However, the ERA5 overestimates the \(SSR\) throughout Brazilian territory due to limitations in its numerical radiative parameterization like aerosol optical depth (\(AOD\)) due to biomass burning events in the Amazon and Central area of Brazil during the dry season and sub-grid shallow clouds that prevails in the tropical humid atmosphere over the continent37,38,39.

The BSR database comes from the radiative transfer model BRASIL-SR, a semi-empirical model based on the two-stream approach to solve the radiative transfer equation in the atmosphere using cloudiness data obtained from GOES-East satellite imagery37. Previous studies comparing the BSR database and ground data revealed low uncertainties for monthly averages of the daily downward surface solar irradiation. The root mean squared error (RMSE) obtained for monthly averages were around 4% and 12% depending on the climate season (dry or wet) and environmental conditions like atmospheric aerosol load, land use, and others40. The BSR database is available at http://labren.ccst.inpe.br/atlas_2017-en.html or by contacting the research team. This study used daily averages of global downward surface solar irradiance (\(SSR\)) provided by the model BRASIL-SR from 2005 to 2017 at a spatial resolution of \(0.03^{\circ }\) x \(0.05^{\circ }\) (around 4 \(km\) x 4 \(km\)).

The ERA5 reanalysis dataset is based on the Integrated Forecasting System (IFS) Cy41r2, which uses a 4D-Var observational data assimilation scheme. The ERA5 data are available on the Copernicus Climate Change Service Data Store - CDS web platform https://cds.climate.copernicus.eu/cdsapp#!/dataset/41 in hourly time steps with a spatial resolution of \(0.25^{\circ }\) ( ≈27 \(km\)). This study used the daily integrated downward surface solar irradiation data (identified as \(SSRD\) in the output variable list) between 1980 and 2014 (34 years) for the entire Brazilian territory. Previous study42 demonstrated that ERA5 performed well for the Brazilian territory, particularly in Brasília, Petrolina, and Florianópolis ground measurement sites.

The CMIP6 database includes the climate models’ outputs for the historical timeframe (1850-2014) and future periods: 2025-2040 (“near future”), 2040-2070 (“mid of the century”), and 2070-2100 (“end of the century”) based on updated emissions scenarios43,44. The climate predictions are related to the RCPs (Representative Concentration Pathway) denoting a greenhouse gas concentration (not emissions) trajectory adopted by the IPCC, and the Shared Socioeconomic Pathways (SSPs), which describe projected socioeconomic global changes up to 2100 regarding the different mitigation and adaptation strategies to climate changes impacts.

This work encompasses two future scenarios, the SSP2-4.5 and SSP5-8.5. Both SSP2 and SSP5 scenarios assume a delay in global cooperation and were chosen due to the higher energy demands at the end of the century45. The SSP2 is the “middle of the road” pathway in which the global population growth is moderate, challenges to reducing vulnerability to societal and environmental changes remain, and energy use intensity declines after 2040. The SSP2-4.5 is the most probable scenario considering the current trajectory and the exhaustible character of non-renewable fuels, presenting intermediate challenges for mitigation46.

The SSP5 is the “fossil fuel development” scenario with increasing inequalities and stratification across and within countries, and the economic and social development is based on energy-intensive lifestyles linked to the abundant exploitation of fossil fuel resources around the world. In SSP5-8.5, emissions continue to rise throughout the 21st century, and it is generally considered the basis for worst-case climate change scenarios due to the higher mitigation challenges47. For Brazil, SSP2 implies GDP and population peaking around 2050, pressing for higher energy demand, while SSP5 relies mainly on fossil fuels48.

Table 2 lists the forty CMIP6 climate models (GCM) used in the study. The table also provides the spatial resolution of each model. The criteria for model inclusion was the data availability of monthly Surface Downwelling Shortwave Radiation (SDSR) with a spatial resolution of up to 250 km. The CMIP6 data was downloaded from https://esgf-node.llnl.gov/search/cmip6/. We used only the first member (realization1) of each GCM.

Satellite-based \(SSR\) products have shorter time series than meteorological reanalysis but present better accuracy42. Previous studies show that the BSR time-series presents low bias throughout the Brazilian territory40. Nevertheless, its time coverage is shorter than required to serve as reference data for the performance evaluation of CMIP6 historical simulations. In order to meet the time coverage and confidence required for the reference database, we used a bias correction method for the ERA5 database, assuming the BSR database as the reference truth. The Quantile Mapping (QM) method49,50 consists of a numerical transformation of the ERA5 surface solar irradiation (\(SSR_{ERA5}\)) database such that its cumulative probability distribution (CDF) equals the distribution of the reference database provided by the BSR. Each climatological month results in one specific transformation to segregate the influence of seasonal variability. All transformations were developed at daily resolution as Eq. (1) described.

where \(t\) and \(s\) represent the day and the grid point for each climatological month. Figure S1 shows the \(SSR\) mapping using ERA5 database and the bias-corrected ERA5 dataset \(SSR_{ERA5_{QM}}\)

CMIP6 models were evaluated concerning their ability to represent the \(SSR\)’s spatial variability and intra-annual seasonality during the historical timeframe. Figure 2 shows the three target areas used to investigate the models’ performance, comprising regions of interest for the Brazilian solar energy sector:

Area A1 includes the semi-arid region in the Brazilian Northeast with the highest surface solar irradiation and the lowest seasonal variability40;

Area A2 comprises the Southern region of Brazil, where the solar energy resource has the highest seasonal variability but presents a high demand for distributed PV systems and could take advantage of hybrid Wind-PV systems due to the high wind speed in the region28;

Area A3 covers the region that combines two advantages: a high surface solar irradiation and proximity to the main electricity consumer centers with fair coverage of the Brazilian interconnected electricity distribution system (SIN).

Sixteen of the forty CMIP6 models provide \(SSR\) with a 250 \(km\) horizontal resolution, while only one offers a 50 \(km\) resolution. Before evaluation, all CMIP6 models were downscaled (bilinear approach) to the same horizontal resolution of the \(ERA5_{QM}\) grid (around 27 \(km\)). The interpolation procedure was essential for a fair and unbiased comparison of the GCM models’ achievements.

The performance evaluation was accomplished at the intersection between the ERA5 and CMIP6 databases (1980-2014). As discussed by51, both temporal and spatial statistical performances were evaluated for the three target areas based on the following metrics:

Spatial correlation (\(Rs\)): the Pearson correlation was calculated using the monthly average surface solar irradiation data from each grid point of the climate model (\(SSR_{Mx}\)), where \(Mx\) is the ID used for CMIP6 models listed in Table 2 and \(ERA5_{QM}\) (\(SSR_{ERA5_{QM}}\));

Seasonal correlation (\(Rt\)): the Pearson correlation index was obtained from the \(SSR_{Mx}\) and \(SSR_{ERA5_{QM}}\) time series of the monthly climatological averages (Jan to Dec);

\(uRMSD\): the unbiased root of mean squared deviations between the \(SSR_{Mx}\) and \(SSR_{ERA5_{QM}}\) monthly climatological averages;

\(SD\) ratio: the average value of the ratio between \(SSR_{Mx}\) and \(SSR_{ERA5_{QM}}\) standard deviations in all three target areas.

The statistically significant spatial correlation \(Rs\) (p-value < 0.05) demonstrates the skill of CMIP6 models in representing the mean spatial patterns of the \(SSR\) in the three target areas. In this sense, we select a subset of CMIP6 models with a significant spatial correlation for all twelve months in the three target areas.

The Taylor diagram was constructed using three metrics (\(Rt\), \(uRMSD\), and \(SD\) ratio) to assess how well the SSR data from CMIP6 models align with the \(SSR\) climatology of the \(ERA5_{QM}\). By employing the Taylor Skill Score (\(TSS\)) approach52, we utilized Eq. ( 2) to combine these metrics and identify the CMIP6 models that best represent the \(SSR\) climatology in the three regions of interest.

where \(f\sigma _{Mx}\) is the ratio between the variance of the CMIP6 model (Mx) and the variance of \(ERA5_{QM}\), and \(R_{t0}\) is the maximum correlation attainable. According to this approach, a TSS value closer to unity indicates better performance of the model in representing the SSR climatology. Equation (2) yields a value of 1.0 as the climate model variance approaches the \(ERA_{QM}\) variance (i.e., \(f\sigma _{Mx}\) tends to 1) and \(R\) tends to \(R_{t0}\). On the contrary, TSS decreases towards zero as the correlation becomes increasingly negative or the model variance approaches zero or infinity. For fixed variance, skill increases linearly with correlation.

The CMIP6 models achieving \(TSS\) greater than 0.9 were combined to create a Smart Model Ensemble (\(SME\)) used to provide the \(SSR\) data for future climate scenarios. Similarly, the multi-model ensemble (ENS) was fetched by averaging the \(SSR_{Mx}\) data provided by each of the forty CMIP6 models. The individual model’s assessment for comparison included the \(SME\) and \(ENS\) performances.

The “climate change factor” approach proposed by Ref.53 was adopted to assess the impacts of future climate pathways SSP2-4.5 e SSP5-8.5. The climate change factor (\(CCF\)) represents the percentage change in \(SSR\) in future scenarios over the model’s prediction for the historical timeslice (1980-2014). This approach assumes that the model bias is preserved along the simulations, dismissing the need for bias corrections to evaluate the climate change signal. Equations (3) and (4) were used to estimate, respectively, the \(CCF\) and future \(SSR\) in each timeslice based on the truth reference (\(ERA5_{QM}\)). This approach was performed for each climatological month to produce future climatologies of CCF over SSR.

where the subscript \(Mx\) represents a specific CMIP6 model or ensemble, \(tslice\), and \(hist\) concerns the model´s outcomes for future and historical data frames. The \(CCF\) depends on the geographical location \(geo=[lat,lon]\), and the month represented by \(m\). The \(SSP\) is one of the future pathways for climate change (SSP2-4.5 or SSP5-8.5).

This manuscript includes supplementary tables and figures to complement the information and results presented in the submitted paper. Results, additional datasets, and codes used in data analysis are deposited for public access at https://figshare.com/s/bc57a485308cbf01ea29 and https://figshare.com/s/22cccad668c8e3872254.

Osman, A. I. et al. Cost, environmental impact, and resilience of renewable energy under a changing climate: a review. Environ. Chem. Lett. 21, 741–764. https://doi.org/10.1007/s10311-022-01532-8 (2023).

Article CAS Google Scholar

Gernaat, D. E. H. J. et al. Climate change impacts on renewable energy supply. Nat. Clim. Change 11, 119–125. https://doi.org/10.1038/s41558-020-00949-9 (2021).

Article ADS Google Scholar

Bloomfield, H. et al. Quantifying the sensitivity of European power systems to energy scenarios and climate change projections. Renew. Energy 164, 1062–1075. https://doi.org/10.1016/j.renene.2020.09.125 (2021).

Article Google Scholar

Solaun, K. & Cerdá, E. Climate change impacts on renewable energy generation. A review of quantitative projections. Renew. Sustain. Energy Rev. 116, 109415. https://doi.org/10.1016/j.rser.2019.109415 (2019).

Article Google Scholar

Huber, I. et al. Do climate models project changes in solar resources?. Sol. Energy 129, 65–84. https://doi.org/10.1016/j.solener.2015.12.016 (2016).

Article ADS Google Scholar

Wild, M., Folini, D., Henschel, F., Fischer, N. & Müller, B. Projections of long-term changes in solar radiation based on cmip5 climate models and their influence on energy yields of photovoltaic systems. Sol. Energy 116, 12–24. https://doi.org/10.1016/j.solener.2015.03.039 (2015).

Article ADS Google Scholar

Pereira, E. B., Martins, F. R., Pes, M. P., da Cruz Segundo, E. I. & de A. Lyra, A. The impacts of global climate changes on the wind power density in brazil. Renew. Energy 49, 107–110, https://doi.org/10.1016/j.renene.2012.01.053 (2013). Selected papers from World Renewable Energy Congress - XI.

Ha, S., Zhou, Z., Im, E.-S. & Lee, Y.-M. Comparative assessment of future solar power potential based on cmip5 and cmip6 multi-model ensembles. Renew. Energy 206, 324–335. https://doi.org/10.1016/j.renene.2023.02.039 (2023).

Article Google Scholar

Hou, X., Wild, M., Folini, D., Kazadzis, S. & Wohland, J. Climate change impacts on solar power generation and its spatial variability in Europe based on cmip6. Earth Syst. Dyn. 12, 1099–1113. https://doi.org/10.5194/esd-12-1099-2021 (2021).

Article ADS Google Scholar

Dutta, R., Chanda, K. & Maity, R. Future of solar energy potential in a changing climate across the world: A cmip6 multi-model ensemble analysis. Renew. Energy 188, 819–829. https://doi.org/10.1016/j.renene.2022.02.023 (2022).

Article Google Scholar

Jerez, S. et al. The impact of climate change on photovoltaic power generation in Europe. Nat. Commun. 6, 10014. https://doi.org/10.1038/ncomms10014 (2015).

Article ADS CAS PubMed Google Scholar

Monerie, P.-A., Wainwright, C. M., Sidibe, M. & Akinsanola, A. A. Model uncertainties in climate change impacts on Sahel precipitation in ensembles of cmip5 and cmip6 simulations. Clim. Dyn. 55, 1385–1401. https://doi.org/10.1007/s00382-020-05332-0 (2020).

Article Google Scholar

ABSOLAR. Panorama da energia solar fotovoltaica no brasil e no mundo. Tech. Rep., Associação Brasileira de Energia Solar Fotovoltaica (2023).

EPE. Brazilian Energy Balance - Sumary Report,. (Reference Year 2022) (Empresa de Pesquisas Energéticas 2023 (Brazil, Ministério de Minas e Energia, 2023).

IRENA. Brazil energy profile. Tech. Rep., International Renewable Energy Agency (2022).

de Lima, F. J. L. et al. The seasonal variability and trends for the surface solar irradiation in northeastern region of Brazil. Sustain. Energy Technol. Assess. 35, 335–346. https://doi.org/10.1016/j.seta.2019.08.006 (2019).

Article Google Scholar

Santos, A. J. L. & Lucena, A. F. Climate change impact on the technical-economic potential for solar photovoltaic energy in the residential sector: a case study for brazil. Energy Clim. Change 2, 100062. https://doi.org/10.1016/j.egycc.2021.100062 (2021).

Article CAS Google Scholar

de Jong, P. et al. Estimating the impact of climate change on wind and solar energy in brazil using a south American regional climate model. Renew. Energy 141, 390–401. https://doi.org/10.1016/j.renene.2019.03.086 (2019).

Article Google Scholar

Zuluaga, C. F., Avila-Diaz, A., Justino, F. B., Martins, F. R. & Ceron, W. L. The climate change perspective of photovoltaic power potential in Brazil. Renew. Energy 193, 1019–1031. https://doi.org/10.1016/j.renene.2022.05.029 (2022).

Article Google Scholar

Martins, F. R., Lima, F. J. L., Gonçalves, A. R., Costa, R. S. & Pereira, E. B. Python Scripts for Data Analysis.[SPACE]https://doi.org/10.6084/m9.figshare.25396210.v1 (2024).

Martins, F. R., Lima, F. J. L., Pereira, E. B., Costa, R. S. & Gonçalves, A. R. Research results - Climate change influence on surface solar irradiance. figshare[SPACE]https://doi.org/10.6084/m9.figshare.25396612.v1 (2024).

Firpo, M. A. F. et al. Assessment of cmip6 models’ performance in simulating present-day climate in Brazil. Front. Clim.[SPACE]https://doi.org/10.3389/fclim.2022.948499 (2022).

Article Google Scholar

Almazroui, M. et al. Assessment of cmip6 performance and projected temperature and precipitation changes over South America. Earth Syst. Environ. 5, 155–183. https://doi.org/10.1007/s41748-021-00233-6 (2021).

Article ADS Google Scholar

Ortega, G., Arias, P. A., Villegas, J. C., Marquet, P. A. & Nobre, P. Present-day and future climate over central and South America according to cmip5/cmip6 models. Int. J. Climatol. 41, 6713–6735. https://doi.org/10.1002/joc.7221 (2021).

Article Google Scholar

Dantas, L. G., dos Santos, C. A. C., Santos, C. A. G., Martins, E. S. P. R. & Alves, L. M. Future changes in temperature and precipitation over Northeastern Brazil by cmip6 model. Water[SPACE]https://doi.org/10.3390/w14244118 (2022).

Article Google Scholar

Hunt, J. . D., Stilpen, D. & de Freitas, M. . A. . V. A review of the causes, impacts and solutions for electricity supply crises in Brazil. Renew. Sustain. Energy Rev. 88, 208–222. https://doi.org/10.1016/j.rser.2018.02.030 (2018).

Article Google Scholar

Gonzalez, J. O. & Martins, F. R. Performance study of a photovoltaic system operating on the southeastern coast of Brazil. IEEE Lat. Am. Trans. 22, 410–417. https://doi.org/10.5194/gmd-9-3461-2016 (2024).

Article Google Scholar

Davis, N. N. et al. The global wind atlas: A high-resolution dataset of climatologies and associated web-based application. Bull. Am. Meteor. Soc. 104, E1507–E1525. https://doi.org/10.1175/BAMS-D-21-0075.1 (2023).

Article Google Scholar

Amarante, O. A. C., Brower, M., Zack, J. & Sá, A. L. Atlas do Potencial Eólico Brasileiro, 2nd edn (Centro de Pesquisas de Energia Elétrica, Ministério de Minas e Energia, Brazil, 2017).

Velloso, M. F. A., Martins, F. R. & Pereira, E. B. Case study for hybrid power generation combining hydro- and photovoltaic energy resources in the Brazilian semiarid region. Clean Technol. Environ. Policy 21, 941–952. https://doi.org/10.1007/s10098-019-01685-1 (2019).

Article Google Scholar

Ferraz de Campos, E. et al. Hybrid power generation for increasing water and energy securities during drought: Exploring local and regional effects in a semi-arid basin. J. Environ. Manag. 294, 112989. https://doi.org/10.1016/j.jenvman.2021.112989 (2021).

Article Google Scholar

Filippo Antoniolli, A., Naspolini, H. F., de Abreu, J. F. & Rüther, R. The role and benefits of residential rooftop photovoltaic prosumers in Brazil. Renew. Energy 187, 204–222. https://doi.org/10.1016/j.renene.2022.01.072 (2022).

Article CAS Google Scholar

Beck, H. E. et al. Publisher correction: Present and future köppen-geiger climate classification maps at 1-km resolution. Sci. Data 7, 274. https://doi.org/10.1038/s41597-020-00616-w (2020).

Article PubMed Google Scholar

Avila-Diaz, A., Benezoli, V., Justino, F., Torres, R. & Wilson, A. Assessing current and future trends of climate extremes across Brazil based on reanalyses and earth system model projections. Clim. Dyn. 55, 1403–1426. https://doi.org/10.1007/s00382-020-05333-z (2020).

Article Google Scholar

Alvares, C. A., Stape, J. L., Sentelhas, P. C., de Moraes Gonçalves, J. L. & Sparovek, G. Koppen’s climate classification map for Brazil. Meteorol. Z. 22, 711–728. https://doi.org/10.1127/0941/2948/2013/0507 (2013).

Article Google Scholar

Martins, F., Pereira, E., Silva, S., Abreu, S. & Colle, S. Solar energy scenarios in Brazil, part one: Resource assessment. Energy Policy 36, 2853–2864. https://doi.org/10.1016/j.enpol.2008.02.014 (2008).

Article Google Scholar

Casagrande, M. S. G. et al. Numerical assessment of downward incoming solar irradiance in smoke influenced regions’a case study in Brazilian amazon and cerrado. Remote Sens.[SPACE]https://doi.org/10.3390/rs13224527 (2021).

Article Google Scholar

Sianturi, Y. & Marjuki Sartika, K. Evaluation of ERA5 and MERRA2 reanalyses to estimate solar irradiance using ground observations over Indonesia region. AIP Conf. Proc. 2223, 020002. https://doi.org/10.1063/5.0000854 (2020).

Article Google Scholar

Boilley, A. & Wald, L. Comparison between meteorological re-analyses from era-interim and merra and measurements of daily solar irradiation at surface. Renew. Energy 75, 135–143. https://doi.org/10.1016/j.renene.2014.09.042 (2015).

Article Google Scholar

Pereira, E. B. et al.Brazilian Atlas for Solar Energy, 2nd edn (Instituto Nacional de Pesquisas Espaciais, Ministério de Ciência, Tecnologia e Inovação, Brazil, 2017).

Hersbach, H. et al. The era5 global reanalysis. Q. J. R. Meteorol. Soc. 146, 1999–2049. https://doi.org/10.1002/qj.3803 (2020).

Article ADS Google Scholar

Urraca, R. et al. Evaluation of global horizontal irradiance estimates from era5 and cosmo-rea6 reanalyses using ground and satellite-based data. Sol. Energy 164, 339–354. https://doi.org/10.1016/j.solener.2018.02.059 (2018).

Article ADS Google Scholar

Eyring, V. et al. Overview of the coupled model intercomparison project phase 6 (cmip6) experimental design and organization. Geosci. Model Dev. 9, 1937–1958. https://doi.org/10.5194/gmd-9-1937-2016 (2016).

Article ADS Google Scholar

O’Neill, B. C. et al. The scenario model intercomparison project (scenariomip) for cmip6. Geosci. Model Dev. 9, 3461–3482. https://doi.org/10.5194/gmd-9-3461-2016 (2016).

Article ADS Google Scholar

Riahi, K. et al. The shared socioeconomic pathways and their energy, land use, and greenhouse gas emissions implications: An overview. Glob. Environ. Change 42, 153–168. https://doi.org/10.1016/j.gloenvcha.2016.05.009 (2017).

Article Google Scholar

Jr, R. P., Burgess, M. G. & Ritchie, J. Plausible 2005–2050 emissions scenarios project between \(2^{\circ }\)c and \(3^{\circ }\)c of warming by 2100. Environ. Res. Lett. 17, 024027. https://doi.org/10.1088/1748-9326/ac4ebf (2022).

Article ADS Google Scholar

Gidden, M. J. et al. Global emissions pathways under different socioeconomic scenarios for use in cmip6: a dataset of harmonized emissions trajectories through the end of the century. Geosci. Model Dev. 12, 1443–1475. https://doi.org/10.5194/gmd-12-1443-2019 (2019).

Article ADS CAS Google Scholar

Dellink, R., Chateau, J., Lanzi, E. & Magné, B. Long-term economic growth projections in the shared socioeconomic pathways. Glob. Environ. Change 42, 200–214. https://doi.org/10.1016/j.gloenvcha.2015.06.004 (2017).

Article Google Scholar

Rajczak, J., Kotlarski, S. & Schär, C. Does quantile mapping of simulated precipitation correct for biases in transition probabilities and spell lengths?. J. Clim. 29, 1605–1615. https://doi.org/10.1175/JCLI-D-15-0162.1 (2016).

Article ADS Google Scholar

Boé, J., Terray, L., Habets, F. & Martin, E. Statistical and dynamical downscaling of the seine basin climate for hydro-meteorological studies. Int. J. Climatol. 27, 1643–1655. https://doi.org/10.1002/joc.1602 (2007).

Article Google Scholar

Lauer, A. et al. Benchmarking cmip5 models with a subset of esa cci phase 2 data using the esmvaltool. Remote Sensing of Environment 203, 9–39, https://doi.org/10.1016/j.rse.2017.01.007 (2017). Earth Observation of Essential Climate Variables.

Taylor, K. E. Summarizing multiple aspects of model performance in a single diagram. J. Geophys. Res. Atmos. 106, 7183–7192. https://doi.org/10.1029/2000JD900719 (2001).

Article ADS Google Scholar

Navarro-Racines, C. E., Tarapues-Montenegro, J. E. & Ramírez-Villegas, J. A. Bias-correction in the ccafs-climate portal: A description of methodologies. decision and policy analysis (dapa) research area. Tech. Rep., International Center for Tropical Agriculture (CIAT), Cali, Colômbia (2000).

Download references

The authors thank DIIAV/INPE and Unifesp for their institutional support. Thanks to CNPq for the research scholarships to Fernando Martins, José Orsini and Enio Pereira. The authors also thanks to National Institute of Science and Technology for Climate Change - INCT-MC Project Phase 2 (Grants FAPESP 2014/50848-9, CNPq 465501/2014-1, and CAPES/FAPS No 16/2014).

Division of Impacts, Adaptation and Vulnerabilities, Brazilian Institute for Space Research, São José dos Campos, SP, 12227-010, Brazil

Francisco José Lopes de Lima, André Rodrigues Gonçalves, Rodrigo Santos Costa, Marcelo Pizzuti Pes, Ana Paula Paes dos Santos & Enio Bueno Pereira

National Center for Monitoring and Early Warning of Natural Disasters, São José dos Campos, 12247-016, Brazil

Jose Antonio Marengo Orsini

Institute of Marine Science, Federal University of São Paulo - campus Baixada Santista, Santos, SP, 11070-102, Brazil

Francisco José Lopes de Lima, Marcelo Pizzuti Pes & Fernando Ramos Martins

Graduate Program in Natural Disasters, UNESP/CEMADEN, São José dos Campos, Brazil

Jose Antonio Marengo Orsini

Graduate School of International Studies, Korea University, Seoul, South Korea

Jose Antonio Marengo Orsini

You can also search for this author in PubMed Google Scholar

You can also search for this author in PubMed Google Scholar

You can also search for this author in PubMed Google Scholar

You can also search for this author in PubMed Google Scholar

You can also search for this author in PubMed Google Scholar

You can also search for this author in PubMed Google Scholar

You can also search for this author in PubMed Google Scholar

You can also search for this author in PubMed Google Scholar

Conceptualization and investigation: F.J.L.L., A.R.G., R.S.C., and F.R.M.; methodology: F.J.L.L, A.R.G., R.S.C., and F.R.M.; code developments: F.J.L.L, M.P.P.; data analysis: F.J.L.L., R.S.C., A.R.G., and F.R.M.; writing-original draft preparation: F.J.L.L., A.R.G., R.S.C., A.P.P.S. and F.R.M.; writing-review and editing: F.J.L.L., A.R.G., R.S.C. and F.R.M.; funding acquisition: J.A.M.O., E.B.P., and F.R.M. All authors reviewed the manuscript.

Correspondence to Fernando Ramos Martins.

The authors declare no competing interests.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

de Lima, F.J.L., Gonçalves, A.R., Costa, R.S. et al. Reducing uncertainties of climate projections on solar energy resources in Brazil. Sci Rep 14, 23586 (2024). https://doi.org/10.1038/s41598-024-73769-y

Download citation

Received: 10 April 2024

Accepted: 20 September 2024

Published: 21 October 2024

DOI: https://doi.org/10.1038/s41598-024-73769-y

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative