Navigating optimal solar-wind trade-offs under climate change Authors: Jingyun Li, Dan Tong, Dongsheng Zheng, Yuanyuan Lin, Yuan Xu, Qiang Zhang Published: 2026 ABSTRACT AbstractThe optimal solar-wind ratio (SWR) plays a critical role in shaping the cost and reliability of renewable electricity systems, yet its adaptability under climate change remains poorly understood. Here, we develop a climate-driven SWR optimization framework that couples multi-model climate projections with an integrated investment and dispatch model to quantify how future climate variability reshapes cost-optimal solar-wind configurations. We find that optimal SWRs exhibit pronounced latitudinal differences and show modest changes under climate change. Deployment pathways inherited from the historical preference scenario may diverge from optimal SWRs, leading to increases in system costs and capacity requirements. In many regions, these SWR mismatches amplify electricity supply costs far more than the cost impacts associated with climate-induced changes in renewable resources alone. Cost escalation is driven mainly by SWR mismatch rather than solely by climate-induced resource changes. These findings highlight the importance of climate-responsive and region-specific SWR optimization as a key element of resilient and economically efficient power system planning.Subject terms: Climate change, Energy supply and demand Introduction Accelerating the transition toward carbon neutrality has led to rapid deployment of wind and solar technologies, which are widely regarded as the backbone of low-carbon power systems due to their technological maturity and cost competitiveness1–3. Over the past decade, global installed capacity of wind and solar has grown exponentially, accounting for nearly one-quarter of total capacity by 20224. While wind capacity initially dominated the expansion, the recent surge in solar deployment has driven a shift toward a more balanced solar-wind mix, even culminating in a solar investment boom in some regions5. This evolving deployment pattern reflects not only cost and policy drivers but also the distinct temporal profiles of wind and solar resources, which often exhibit partial complementarity at daily and seasonal scales6. Such complementarity can mitigate supply fluctuations and reduce reliance on costly flexibility measures such as energy storage or dispatchable thermal generation7. However, the extent to which complementarity can be harnessed is constrained by the intrinsic variability and geographic dependence of these resources6, highlighting the importance of optimizing the solar-wind ratio (SWR), solar capacity’s share of total wind and solar capacity, in system design. Meanwhile, climate change is progressively reshaping the spatial and temporal characteristics of wind and solar resources, altering their short-term and long-term variability8,9, with warming projected to raise solar potential at some high latitudes and shift wind availability toward lower latitudes10,11, which in turn undermines the performance of historically optimal configurations. These shifts in resource patterns have direct implications for solar-wind deployment strategies that underpin power system planning, with growing climate uncertainty underscoring the need to identify configurations that remain robust under future conditions. Growing evidence has refined our understanding of solar-wind deployment by delineating cost-optimal SWR within long-term energy transition pathways12,13. Although integrated assessment models and long-term planning studies have provided valuable insights into solar-wind deployment strategies14,15, they often emphasize long-run cost optimization under simplified assumptions16, typically neglecting hourly variability and climate-induced extremes. These approaches commonly rely on low temporal resolution15 and fixed climate baselines17,18, limiting their ability to capture the operational challenges posed by hourly supply-demand imbalance and increasing climate variability under future climate change7. More recently, a growing body of research has begun to address this by optimizing SWR from operational or reliability-driven perspectives19–21, using metrics such as cost-effectiveness7, load mismatch6, unmet demand22, and storage adequacy23,24. These findings demonstrate that the effectiveness of SWR strategies depends on the alignment between deployed and climate-consistent optimal SWRs under future climate conditions, yet they are based on fixed climate baselines or single-year simulations, and consequently offer limited insight into whether historically preferred or institutionally projected SWR configurations remain robust under future climate variability. Despite growing attention to wind and solar complementarity7, few studies systematically assess how optimal solar-wind ratios evolve under future climate conditions across countries and scenarios. A unified framework that integrates multi-model climate projections, hourly generation variability, and high-resolution demand profiles within long-term planning horizons is still lacking. In particular, the cost and resilience implications of adaptive deployment strategies, as compared to existing institutional prediction pathways and historically preferred configurations, have not been comprehensively evaluated. As a result, important questions remain about how to align solar-wind deployment with climate-consistent optimal SWRs in a robust manner while minimizing the risks associated with SWR mismatch. The SWR is a key determinant of long-term system cost, reliability and capacity configuration in renewable power systems under climate change, yet its climate-driven evolution has not been systematically examined. This study contributes in two respects. First, we integrate multi-model, multi-decadal climate projections and climate-responsive demand into an hourly capacity-expansion model to examine how the cost-optimal SWR evolves under long-term climate change, thereby shifting the analytical focus from short-run dispatch effects to climate-adaptive planning. Second, we systematically compare historically observed deployment, IAM-based pathways and climate-consistent optimal configurations, and quantify the associated system-wide cost and capacity implications of SWR misalignment, thereby identifying the system-wide penalties associated with SWR mismatch. Here, we introduce a climate-responsive optimization framework that integrates evolving climate projections into long-term capacity planning to directly quantify the economic and structural penalties arising from SWR mismatch. Details are provided in the Methods section and the Supplementary Information (SI). In summary, our framework integrates climate data processing, electricity demand projections, integrated investment and dispatch optimization, and SWR strategy evaluation. We leverage hourly climate variables from seven CMIP6 global climate models (GCMs)25 to capture future changes in renewable resource availability and temperature-sensitive electricity demand. These data are used to compute hourly capacity factors and demand curves for different countries (Supplementary Fig. 1) that are selected based on data availability and regional representativeness (Supplementary Note 1) during two multi-year windows: 1981–2020 (representing the present climate) and 2061–2100 (representing the future climate). The ensemble enables an uncertainty assessment under two emissions pathways, SSP126 and SSP245, which represent contrasting levels of climate forcing and energy system transformation. We then use the average SWR derived from 2015 to 2022 to represent historical preference configurations and from the AR6 Scenario Database for 2060 to 2100 to represent future prediction configurations (Supplementary Fig. 3), and subsequently employ an integrated investment and dispatch model that minimizes total system costs while balancing hourly supply and demand across six technologies, including wind, solar, hydro, thermal, nuclear and storage. To isolate climate change effects, only climate-related drivers are allowed to influence the SWR, while non-climatic factors remain unchanged. Accordingly, total annual wind and solar generation shares are held constant across simulations within each scenario26. Based on over 680,000 dispatch simulations under different solar-wind configurations and climate scenarios (Supplementary Table 4), we assess how climate change affects system cost and capacity. To represent alternative solar-wind deployment strategies, we construct three planning scenarios: Optimization SWR (OR), Historical preference (HP), and Future prediction (FP) (Supplementary Note 2), as defined in Supplementary Table 5. Finally, we compare these scenarios to quantify the additional cost and capacity burdens. Sensitivity analyses across seven CMIP6 climate models, multiple emission pathways, extended time periods, and modeling assumptions confirm the robustness of our findings. These results emphasize the importance of aligning deployed SWRs with climate-consistent optimal SWRs to avoid SWR mismatch and support an efficient and resilient power system transition. Results Wind and solar investment trends and projections Figure 1 illustrates historical trends and future projections of the SWR under climate change, showing a global shift toward solar dominance in solar-wind configurations, with substantial regional heterogeneity under future scenarios. High SWR areas are mainly in low-latitude regions, whereas high-latitude countries generally exhibit lower SWR levels (Fig. 1a). Countries with very high SWR values (≥90%), such as Malawi and Niue, rely almost entirely on solar energy due to abundant irradiation27. China and Brazil exhibit a relatively balanced deployment of wind and solar power (about 50%), supported by strong resource availability and diverse regional conditions28,29. Norway and Ireland present low SWR ( ≤ 10%), where abundant wind resources result in wind-dominated solar-wind systems30,31. Present situation and future forecast of SWR. Spatial distribution of solar-wind ratio (SWR) across countries and regions in 2022 (a), and historical solar and wind capacity together with SWR trends for the world and selected key regions (b). Historical values cover 2000–2022, while future projections extend from 2030 to 2100 under SSP126 and SSP245. Mean values denote regional or global results, with uncertainty capturing the 10th–90th percentile range across Integrated Assessment Models from the International Institute for Applied Systems Analysis (IIASA) AR6 Database. Global solar-wind capacity trends reveal a structural shift from wind-led to solar-driven growth. During the 2000s, capacity expansion was dominated by wind, with solar accounting for less than 10% of total wind and solar capacity (Fig. 1b). Since 2015, solar capacity has grown rapidly, exceeding 50% of total wind and solar capacity by 2022. This marks a global surge in solar investment, driven by declining costs and deployment scalability32. While wind energy remains important in high-demand countries such as the United States and China, solar capacity has expanded rapidly across most regions, triggering a global solar investment boom (Supplementary Figs. 4–8). Comparable trends are evident in other major economies, including Europe, India, Brazil, Australia, Japan, and South Africa. Future SWR exhibits substantial uncertainty under climate change. In most cases, SSP245 yields higher SWR than SSP126 (Fig. 1b), suggesting greater solar dominance under a more emission-intensive pathway. Globally, the mean SWR exceeds 50% under both SSP126 and SSP245 scenarios, indicating a general shift toward solar dominance. Similar trends are observed in countries such as India, Australia and the United States. In countries such as Japan and South Africa, wind development is more pronounced under SSP126, whereas solar deployment is more prominent under SSP245. Climate-driven shifts in the optimal SWR Figure 2 illustrates projected changes in the optimal SWR and solar-wind capacity under future climate scenarios. Low-latitude countries typically exhibit higher SWR, whereas high-latitude countries show markedly lower values. This contrast reflects latitudinal differences in atmospheric circulation and in the distribution of surface shortwave radiation (Supplementary Note 3), which, through changes in wind-speed distributions and irradiance intensity, drive systematic variations in wind and solar capacity factors under different climate scenarios (Supplementary Fig. 25) and thus influence their relative system cost in cost-optimized configurations. Under SSP126, the average SWR in the 30°S–30°N region is 30.5% higher than that in other regions, and 0.6% higher than under SSP245 within this latitude band (Supplementary Fig. 26). Additionally, countries such as Brazil, Colombia, Malaysia, and Indonesia reach a 100% SWR, whereas the United Kingdom (25%), Canada (15%), New Zealand (9%), and Russia (9%) exhibit significantly lower values under SSP126 (Supplementary Fig. 27). SWR change and solar-wind capacity change under future climate change. SWR and SWR change for 2061–2100 relative to 1981–2020 are shown under SSP126 (a) and SSP245 (c). Solar and wind capacity changes for 2061–2100 relative to 1981–2020 are shown under SSP126 (b) and SSP245 (d). Mean values denote 40-year country-averaged results, with uncertainty capturing variability across climate models. Black dots indicate total solar-wind capacity changes. Colored dots represent results from individual climate models. Optimal SWR responses to climate change vary with latitude, with high-latitude countries showing stronger structural shifts than low-latitude countries (Fig. 2a). This differential sensitivity may reflect projected circulation-driven changes in near-surface wind regimes that exhibit greater sensitivity in some higher-latitude regions, which can contribute to comparatively larger adjustments in wind capacity factors (Supplementary Note 3). In low-latitude countries such as Brazil, the SWR remains effectively unchanged across scenarios, indicating limited climate sensitivity in their optimal solar-wind configurations. In contrast, high-latitude regions, such as Sweden, show substantially larger adjustments, with the SWR change of 7.6% under SSP126 (Supplementary Fig. 28). The SWR change in other regions is 1.8% higher than that in the 30°S-30°N region (-0.4%) under SSP126 (Supplementary Fig. 29). Apparent stability in the SWR may mask substantial shifts in the underlying capacity structure under climate change. Climate-driven shifts in installed capacities are widespread, with more pronounced changes occurring outside low-latitude zones. This pattern may be associated with projected climate-driven changes in wind and solar capacity factors and in the temporal distribution of electricity demand, which modify supply-demand balance patterns and flexibility requirements, requiring adjustments in total installed capacity within least-cost configurations (Supplementary Note 3). Under SSP126, the United Kingdom (7.97%) and Russia (5.07%) see notable increases in total solar-wind capacity (Fig. 2b). The relative contributions of wind and solar to total capacity change vary across countries. Solar capacity changes dominate in Sweden (13.61%), the United Kingdom (7.39%), and France (6.35%), while wind capacity changes are the primary driver in Russia (4.28%) and Venezuela (2.26%) (Supplementary Fig. 30). Total solar-wind capacity increases under both scenarios, with a larger mean change under SSP245 (1.03%) than under SSP126 (0.65%) (Supplementary Fig. 31). Low-latitude regions show a mean reduction in total capacity under SSP126 (-0.92%) but a shift toward expansion under SSP245 (2.21%), whereas other regions experience sustained increases under both scenarios, particularly under SSP245 (2.52%) (Supplementary Fig. 32). SWR mismatch amplifies costs from present to future climate Figure 3 illustrates cost changes under the Optimization SWR, Historical preference and Future prediction between present climate and future climate (different scenarios can be obtained from Supplementary Table 5 in detail). The Optimization SWR comparison captures the cost impact of climate change, whereas the Historical preference and Future prediction comparisons reflect the combined effects of climate change and SWR mismatch. An optimal SWR enables some countries to benefit from cost reductions driven by climate change. These cost impacts are unevenly distributed across latitudes, and optimal solar-wind configurations reduce average costs by 0.6% in low-latitude countries under SSP126, while other regions experience an average increase of 0.6% (Supplementary Fig. 33). The magnitude of these cost impacts is further amplified under SSP245. In low-latitude regions, favorable shifts in solar radiation and optimized resource mixes help offset adverse climate effects7. By contrast, high-latitude countries face greater demand variability and weaker solar-wind complementarity, driving up investment needs and overall system costs7. Additionally, electricity costs decrease in Vietnam (-2.66%), Thailand (-2.56%), and South Korea (-1.98%), while increases occur in countries such as Russia (2.88%), Sweden (2.68%), and the United Kingdom (2.65%) (Supplementary Fig. 34). Cost changes across the Future prediction, Historical preference, and Optimization SWR scenarios under climate change. Cost changes of the Future prediction, Historical preference, and Optimization SWR scenarios for 2061–2100 relative to 1981–2020 under SSP126 (a) and SSP245 (b). The impact of climate change is evaluated by comparing Optimization SWR between the 2061–2100 and 1981–2020 climate periods, whereas the combined impact of climate change and SWR mismatch is assessed by comparing Historical preference and Future prediction in 2061–2100 with Optimization SWR in 1981–2020. Mean values denote 40-year country-averaged results, with uncertainty capturing variability across climate models. Error bars indicate the ranges across climate models. SWR mismatch can negate the cost-saving benefits of optimized deployment and markedly amplify electricity cost risks under climate change. For instance, under SSP126 with Historical preference, electricity costs rise by 80% in Brazil, 33% in Germany, 29% in India, and 23% in Algeria (Fig. 3), whereas climate-driven impacts alone account for only -0.4%, -0.1%, -0.5%, and 0.2%, respectively. This stark contrast indicates that SWR mismatch, rather than climate-driven changes in capacity factors, is the primary driver of the cost increases. Meanwhile, mismatched solar-wind configurations lead to larger cost increases, particularly in low-latitude regions, with an average increase of 28.5% under the Historical preference scenario (Supplementary Fig. 35). Moreover, a growing number of countries are exposed to dual risks arising from both climate change and misaligned solar-wind deployment (Supplementary Fig. 36). Under climate change, the modest cost reductions in wind and solar power are far from sufficient to offset the pronounced cost increases induced by solar-wind mismatches. Under the optimization SWR, South Korea exhibits a 2.9% reduction in wind power costs driven by climate change, yet incurs a 14.3% increase in solar power costs due to solar-wind mismatch under SSP126 in the Historical preference scenario (Supplementary Fig. 37). Similarly, climate change leads to a 0.4% decrease in solar costs for Brazil, whereas solar-wind mismatch drives a substantial 106.1% increase in wind costs. For the United Kingdom, climate change results in modest increases in both solar (2.5%) and wind (0.4%) costs, which are further compounded by solar-wind mismatch, adding 1.8% and 1.0%, respectively. Mismatched SWR strategies raise costs under future climate Figure 4 illustrates cost changes for the Optimization SWR, Historical preference and Future prediction scenarios under future climate conditions. Both Historical preference and Future prediction scenarios result in higher electricity supply costs relative to the Optimization SWR. Notably, in some countries, including Brazil, costs under Historical preference exceed those under Future prediction (Fig. 4a, b, d, e). Average supply costs in Venezuela, Brazil, and Mozambique rise by 96%, 81% and 74% under SSP126 in the Historical preference scenario (Supplementary Fig. 38), respectively, partly attributable to wind capacity additions in regions where solar deployment would have delivered lower system costs. Similar patterns are observed in Indonesia (69%) and Ghana (47%). By contrast, higher-latitude countries exhibit modest cost increases. In France (5.2%) (Fig. 4b), the limited increase is attributable to insufficient solar investment, whereas in New Zealand (1.2%), the United Kingdom (2.4%), and Russia (3.0%), cost increases are driven by excessive solar deployment in regions with less favorable solar conditions under different SWR strategies (Supplementary Fig. 39). Different cost comparison under the Future prediction, Historical preference, and Optimization SWR scenarios under future climate scenarios. Cost changes of the Future prediction and Historical preference scenarios relative to the Optimization SWR scenario under SSP126 (a, b) and SSP245 (d, e). Relationship between the cost change of Future prediction and the cost change of Historical preference under SSP126 (c) and SSP245 (f). Mean values denote 40-year country-averaged results, with uncertainty capturing variability across climate models. Error bars indicate the ranges across climate models. Historical preference leads to more widespread and pronounced increases in electricity supply costs than Future prediction (Fig. 4c, f). Under both SSP126 and SSP245, 71% of regions experience larger supply cost increases under the Historical preference scenario than under the Future prediction scenario. While supply cost increases in most regions are concentrated within the 0–30% range (Supplementary Fig. 40), certain countries exhibit markedly higher increases under the Historical preference scenario compared with the Future prediction scenario, with differences of up to 96% in Vietnam and 74% in Mozambique. Changes in wind and solar costs are the primary drivers of electricity supply cost increases. Wind power cost increases dominate in many low-latitude countries such as Venezuela (135%), Mozambique (121%), and Brazil (106%) under SSP126 in the Historical preference scenario (Supplementary Fig. 41), reflecting insufficient solar investment and the resulting increase in overall electricity supply costs. In contrast, solar power is the primary contributor to rising costs in several countries, including Libya (49%), Japan (47%), Algeria (45%), and Russia (27%), reflecting continued solar expansion in relatively less favorable conditions. Storage costs also increase substantially in countries such as Libya (14%) and Algeria (12%), following a pattern consistent with solar-driven cost increases. These cost patterns may be related to deviations between historical and optimal SWRs. Where observed SWRs, reflecting Historical preference, exceed optimal levels, solar may be overrepresented, as PV is generally cheaper and easier to deploy than wind. Conversely, where observed SWRs remain below optimal levels, earlier commercialization of wind power and accumulated grid-integration experience may have slowed subsequent solar adjustment toward the optimum (Supplementary Note 4). Under SWR mismatch, some countries rely more heavily on wind power to smooth solar generation, which can reduce renewable curtailment and storage throughput. Nevertheless, compared with configurations that more fully exploit low-cost solar resources in combination with energy storage, such strategies tend to result in higher overall system costs. For example, in Brazil and Mexico, insufficient investment in solar capacity leads to pronounced reductions in renewable curtailment (by 88% and 46%) (Supplementary Fig. 42) and storage throughput (by 87% and 37%) (Supplementary Fig. 43), but the associated increase in wind-related costs substantially outweighs the cost savings from reduced solar and storage deployment (Fig. 4). Similar patterns are observed in China, India, and France. By contrast, in regions such as Russia and Japan, excessive investment in solar power results in substantial increases in renewable curtailment (by 598% and 332%) and storage throughput (by 158% and 64%). SWR mismatch intensifies solar-wind capacity requirements Figure 5 illustrates solar-wind capacity comparisons under different SWR strategies. Historical preference and Future prediction result in widespread solar underinvestment across most regions, with the shortfall being particularly pronounced in low-latitude countries, while some higher-latitude regions exhibit signs of overinvestment, with underinvestment driving larger losses in low latitudes and overinvestment amplifying losses in higher-latitude regions. Historical preference results in larger solar-wind capacity increases, particularly in low latitudes (Fig. 5). Countries with substantial additional solar-wind capacity requirements include Venezuela (66.9%), Brazil (58.6%), Mozambique (44.8%), and Ghana (43%), underscoring pronounced solar underinvestment under SSP126 in the Historical preference scenario. By contrast, higher-latitude countries such as Japan (105.8%) and Russia (46.3%) exhibit a different adjustment pattern, driven primarily by solar overinvestment rather than underinvestment. Additionally, low-latitude countries exhibit greater average increases in solar-wind capacity (24.85%) than other regions (13.3%) under SSP126 in the Historical preference scenario (Supplementary Fig. 44). In a few countries, solar underinvestment leads to a partial substitution of solar by wind capacity that results in a slight net decline in total solar-wind capacity and, given the higher cost of wind, an increase in the unit cost of electricity supply (Supplementary Fig. 45). Capacity comparisons under the Future prediction, Historical preference, and Optimization SWR scenarios. Relationship between latitude and SWR change for Future prediction relative to Optimization SWR under SSP126 (a) and SSP245 (c), and for Historical preference relative to Optimization SWR under SSP126 (b) and SSP245 (d). Mean values denote 40-year country-averaged results, with uncertainty capturing variability across climate models. Error bars indicate the ranges across climate models. Solar overinvestment results in substantially higher solar-wind capacity requirements than underinvestment, especially in higher-latitude regions. Under SSP126 in the Historical preference scenario, the average solar-wind capacity increase reaches 28.09% in cases of solar overinvestment, compared with 14.07% under underinvestment (Supplementary Fig. 46). This contrast is particularly pronounced in higher-latitude regions, where solar overinvestment is associated with a substantially larger capacity increase (31.81%) than underinvestment (1.92%) (Supplementary Fig. 47). Accordingly, Morocco exhibits solar underinvestment with a 13.9% solar-wind capacity increase, whereas Japan shows solar overinvestment and requires a capacity increase 91.9% higher than Morocco’s. By contrast, in low-latitude regions, Nigeria shows solar overinvestment with a 14.9% increase in the capacity, while Venezuela, under solar underinvestment, requires a capacity increase that is 52% higher than that of Nigeria (Supplementary Fig. 48). The increase in total solar-wind capacity is primarily driven by additional wind deployment in low-latitude regions. Average wind capacity increases by 30.74%, while solar capacity declines by 5.9% under SSP126 in the Historical preference scenario (Supplementary Fig. 49). In regions characterized by solar underinvestment, such as Venezuela, solar capacity requirements decline by approximately 80%, accompanied by a substantial increase in wind capacity of about 147% (Supplementary Fig. 50). Similar adjustment patterns are observed in Mozambique, Brazil, and Peru, where solar capacity reductions of 84%, 50%, and 31% are offset by corresponding wind capacity increases of 129%, 108%, and 70%, respectively. By contrast, in a limited number of regions exhibiting solar overinvestment, such as Algeria, solar capacity requirements increase markedly (by 99%), while wind capacity decreases by about 45%. Discussion This study develops an integrated assessment framework that couples long-term climate projections with the SWR index and power system planning to identify climate-adaptive solar-wind ratios under future climate change while evaluating system costs and capacity requirements under different solar-wind configurations. By capturing regional differences in wind and solar generation profiles (Supplementary Figs. 51–53), our analysis reveals that under future climate conditions, low-latitude regions tend to favor higher solar shares, whereas high-latitude regions shift toward greater reliance on wind power (Supplementary Fig. 2). Solar-wind configurations under the Historical preference and Future prediction scenarios often fail to ensure cost-effective system operation compared with the Optimization SWR scenario. Moreover, the SWR mismatch frequently results in system cost increases that exceed the direct impacts of climate change itself, amplifying both economic burdens and climate-related vulnerabilities. These findings highlight the importance of aligning deployed SWRs with climate-consistent optimal SWRs, as rapid expansion of solar or wind capacity under mismatched SWRs may substantially increase electricity system costs. Such mismatches may reflect not only climate-driven changes in resource complementarity, but also existing deployment pathways shaped by technology costs, project complexity and other contextual factors33–36. Where observed SWRs, represented by the Historical preference scenario, are higher than optimal SWRs, deployment appears to have tilted towards solar PV. This imbalance is consistent with the comparatively lower cost and easier deployment of PV, including shorter commissioning times and fewer project-specific constraints than wind power, whose development is typically more site-dependent and subject to longer permitting and construction processes33,34. Conversely, where observed SWRs fall below optimal SWRs, solar PV deployment may not have expanded fast enough to match the climate-consistent optimum. Such cases may reflect path dependence from earlier wind deployment, in which prior commercial maturity and accumulated grid-integration experience helped entrench wind-oriented investment, infrastructure and planning structures35,36. Further details are provided in Supplementary Note 4. The mechanisms underlying changes in optimal SWRs under future climate conditions are primarily associated with shifts in renewable resource availability and electricity demand. Under future climate change, shifts in near-surface wind-speed distributions, surface shortwave radiation, and temperature may modify the hourly structure of renewable supply and electricity demand37. Changes in large-scale atmospheric circulation alter wind regimes38,39 and wind capacity factors, while variations in irradiance and temperature-dependent photovoltaic performance may modify solar capacity factors40,41. Concurrent warming reshapes heating and cooling demand and thus load timing37. These climate-driven changes in renewable availability, variability, and demand timing may modify residual-load profiles and flexibility requirements7,22. As least-cost systems are determined by hourly supply-demand balance, shifts in capacity factors, temporal alignment, and variability directly alter the optimal solar-wind ratio and total installed capacity7. The latitude-dependent pattern in optimal SWR reflects the differences in atmospheric circulation and surface shortwave radiation37. Mid- to high-latitude regions generally experience stronger and more seasonal wind resources37, whereas low latitudes experience more stable solar irradiance42, helping explain lower optimal SWRs in high latitudes and higher optimal SWRs in many low-latitude regions. Under continued warming, arctic amplification may increasingly influence atmospheric circulation43. As wind power responds nonlinearly to wind speed44, even modest changes may translate into amplified variations in power output45 and thus larger SWR shifts in some high-latitude countries. Even when SWR remains stable, climate-driven changes in variability and complementarity modify residual-load magnitude22,46, requiring adjustments in total installed wind, solar, storage, and backup capacity to satisfy hourly balance constraints. Further details are provided in Supplementary Note 3. Our study has several important limitations and uncertainties. First, socio-economic and institutional factors influencing regional energy transitions are not explicitly represented in this study, including country-specific government policies, commercial activity, domestic supply chains, and public preferences. Many of these factors ultimately influence energy system development through techno-economic parameters such as investment costs and financing conditions. Cost sensitivity tests and country-specific WACC tests, with WACC assumptions provided in Supplementary Table 3, indicate that the spatial patterns of optimal SWR and cost results remain robust (Supplementary Figs. 9–13). Second, wind and solar shares are fixed exogenously in the model to control for non-climatic influences. Despite different renewable shares under SSP126 and SSP245, the main conclusions remain robust. Third, capacity factors are derived from meteorological variables and bias-corrected against reanalysis data rather than directly validated against observed generation data, which may introduce additional uncertainty related to potential discrepancies between modeled and realized performance47 (Supplementary Note 5). Separately, we test alternative hourly downscaling methods48 to evaluate whether different treatments of climate variability affect the results. These tests show only minor impacts on flexibility requirements (Supplementary Fig. 23). Fourth, the model does not explicitly consider the impacts of climate change on hydropower availability. Hydropower tests show that variations in hydropower availability do not materially affect the optimal SWR (Supplementary Figs. 14–16). Finally, the estimated optimal solar-wind ratios and associated system costs are derived from a set of modeling assumptions regarding input data, system optimization, and scenario design. We perform extensive sensitivity analyses to assess the robustness of our results (Supplementary Figs. 2 and 9–23), including PV technology49, storage duration23,50 (Supplementary Table 2), cost test, load test, and grid interconnection test. Further details are provided in Supplementary Note 6. Despite these limitations, our results provide important implications for climate-adaptive renewable energy planning. Climate change reshapes the optimal balance between wind and solar deployment, and SWR mismatch may substantially increase system costs and flexibility requirements. This underscores the importance of aligning deployed SWRs with climate-consistent optimal SWRs through region-specific planning strategies to minimize the electricity system costs associated with SWR mismatch. In addition, although optimal capacity configurations vary across weather years and climate models, these differences remain limited (Supplementary Fig. 24), indicating that long-term mean configurations provide a robust basis for system planning rather than relying on any single weather realization (Supplementary Note 7). Flexible resources, including energy storage, dispatchable low-carbon generation, demand-side management, and enhanced transmission infrastructure, may remain essential for mitigating the cost penalties associated with renewable variability7,23 and SWR misalignment. Technological progress, such as reductions in renewable and storage costs and improvements in storage technologies examined in the sensitivity analysis, may further alleviate these system costs (Supplementary Fig. 11). Strategic planning of regionally optimal SWR, together with expanded regional grid interconnections and cross-border electricity trade, may further enhance system resilience by enabling spatial balancing of renewable resources (Supplementary Note 6). However, the effectiveness of regional balancing may vary under climate change due to changes in renewable resource variability, particularly when solar-wind deployment deviates from regionally optimal ratios (Supplementary Figs. 18 and 19). This further highlights the importance of integrated planning approaches that minimize SWR mismatch by accounting for regional resource characteristics and long-term climate conditions when designing future renewable power systems. Methods This study constructs an integrated climate-responsive framework to assess how SWR configurations influence power system performance under climate change. While the integrated investment and dispatch model (8,760 hours/year) is adapted from the structure developed by Zheng et al.⁷, our framework places particular emphasis on the integration of multi-source, high-resolution input data to quantify system costs, capacity requirements, and the implications of mismatched deployment strategies under evolving climate conditions. Using this framework, we conduct over 680,000 simulations covering combinations of SWR configurations and climate models. Climatological data collection and processing methods The climatological data used to compute hourly wind and solar capacity factors in this study are sourced from the CMIP6, incorporating outputs from seven GCMs. These models—BCC-CSM2-MR, CanESM5, MIROC6, MIROC-ES2H, MPI-ESM1-2-HR, MRI-ESM2-0, and NESM3—were developed by institutions across multiple countries, representing a wide range of expertise, methodologies, and parameterizations. The integration of these models could enhance the robustness and reliability of the power dispatch simulations. Supplementary Tables 6 and 7 provide detailed information on the models’ sources and parameters, along with scenarios for SSP126 and SSP245. For each GCM, we extract the meteorological inputs required for estimating wind and solar capacity factors, including shortwave radiation, clear-sky shortwave flux, wind components, and near-surface air temperature. To reduce systematic biases in climate model simulations, meteorological variables from CMIP6 were bias-corrected using the ERA5 reanalysis dataset as the reference. The correction was applied independently at each grid cell using a quantile mapping (QM) approach51,52. For the historical period, the classical QM method was used to align the cumulative distribution function (CDF) of CMIP6 simulations with that of ERA551,52. where Fm and Fo−1 denote the CDF of the CMIP6 historical simulations and the inverse CDF of the corresponding ERA5 historical data, respectively. xm represents the simulated variable and xm,adj is the bias-corrected value. To account for non-stationarity under future climate change, a quantile-based calibration method was applied for the future period, which preserves the projected changes in mean, variability, and distributional shape51,52. The corrected future value at the i th quantile is expressed as: where Oi is the i th quantile of the ERA5 historical variable; △¯ represents the mean difference between CMIP6 future and historical simulations; and Δai denotes the quantile-specific deviation from the mean difference. The scaling factors k and f correspond to the ratios of the mean and interquartile range between ERA5 and CMIP6 historical variables, respectively. As a result, the future CF data are corrected using Quantile Mapping, ensuring that they closely align with the statistical characteristics of the reanalysis data while maintaining consistency between historical and future periods (Supplementary Figs. 54 and 55). Wind and solar capacity factors Wind capacity factor is calculated using the power curve of a standard wind turbine, which shows how the actual electricity output varies with wind speed. For onshore wind, capacity factors are based on the V136–3.45 MW turbine53, while offshore wind capacity factors are derived from the IEA 15 MW reference turbine54 (V236–15.0 MW). Their power curves are shown in Supplementary Fig. 56. The capacity factor is given by the following equation: where CFW represents the wind capacity factor, and ws denotes the power curve of the standard wind turbine. VHhh is the actual wind speed at hub height. To estimate VHhh, the following model is used. where α and VH10 are the surface friction coefficient (1/7) and the hourly VHhh at 10 m, respectively. where NWC and EWC are the northward and eastward wind components from the GCMs, respectively. The relevant parameters are referenced from Zheng et al. 7. The solar capacity factor (CFS) serves as a key metric for assessing the gap between a photovoltaic system’s actual generation and its theoretical maximum potential55. TG represents the output power per unit area of a photovoltaic module. TW denotes the rated power per unit area under standard test conditions. TG is expressed in Eq. (7): TIΣ represents the total irradiance received by the photovoltaic panel, while PCE denotes the photoelectric conversion efficiency (set at 16.19%). TC and SPC correspond to the temperature coefficient and system performance coefficient (set at 80.56%), respectively. The detailed calculation of TIΣ refers to relevant research55. In this calculation of the photovoltaic capacity factor, the most advanced two-axis tracking system is chosen in this study. The TIΣ is divided into direct irradiance (DTIΣ), diffuse irradiance (DEIΣ), and reflected irradiance (RIΣ)55,56. where ∑ is the tilt angle of the photovoltaic panel. DTIH, DEIH, and TIH denote the direct irradiance, diffuse irradiance, and total irradiance on the horizontal plane, respectively. CCDI denotes the conversion coefficient of diffuse irradiance57. aoc represents the albedo constant (with a value of 0.2 for grassland and ground). α0 denotes the incident angle of direct irradiance7. where γ and φ denote the solar zenith and declination angles, respectively; ξ and ν indicate the azimuth angles of the solar and photovoltaic panel. LSP represents the panel’s geographic latitude, while SA refers to the solar incidence angle. The variable d corresponds to the day number within the year. d0=10 and D=365.25 are constants. The calculation of CCDI uses the method proposed by Perez et al. 57, which corrects CCDI based on the anisotropic sky hypothesis: where F1 and F2 are the sky scattering coefficients for the outer solar region and the horizon region, respectively. a/b represents the angular position of the outer region. TC is expressed in Eqs. (16) and (17)58: o represents the cell temperature (25 °C), while κ denotes the temperature coefficient, set at 0.005 °C⁻¹. ω denotes the operational temperature. z1, z2, z3 and z4 represent the equation constant (4.3 °C), temperature coefficient (0.943), irradiation coefficient (0.028 °C W⁻¹m²), and cooling coefficient (1.528 °C m⁻¹s⁻¹), respectively. ς and υ represent surface solar radiation and surface wind speed. Aggregated capacity factors Using the hourly grid-scale CF, we determined the national-scale hourly CF by excluding regions unsuitable for wind and solar farm construction due to geographical constraints (Supplementary Table 8). We identified the suitable areas for wind and solar energy separately, applying distinct criteria for each. Solar deployment suitability is screened using land-use constraints, terrain slope, and protected-area boundaries. We identify unsuitable grid cells using land-cover information from the European Space Agency Climate Change Initiative (ESACCI)59, excluding water bodies, wetlands, permanent snow and ice, forests, and croplands. Additionally, regions with slopes greater than 5% and those within protected areas were excluded. Slope data was obtained from the Global Multi-resolution Terrain Elevation Data60 (GMTED), a widely used dataset for global and continental-scale applications. Protected area information was sourced from the World Database on Protected Areas61 (WDPA). For wind energy suitability, we utilized the ESACCI dataset to exclude urban areas, water bodies, permanent snow and ice, forests, regions with slopes exceeding 20%, areas above 3000 m in elevation, and protected areas. High-altitude regions were excluded because lower air pressure and reduced wind speeds at such elevations limit the potential for effective wind energy generation. After excluding unsuitable regions at the grid scale, the suitable land area for wind and solar energy within each 1°× 1° grid cell was calculated. National-scale hourly CFs were then derived by aggregating grid-level CFs using suitable-area weighting within each country, with CF fields linearly interpolated to 0.1° × 0.1° during aggregation to better resolve national boundaries. Electricity demand projections To assess the impact of future climate change on electricity demand, this study builds on previous research7,22 (the data source is provided in Supplementary Table 9), focusing primarily on seasonal heating and cooling demands influenced by temperature changes. Hourly power demands for both historical and future periods are first calculated based on surface temperature. The difference between future and historical power demand is then added to the historical power demand to estimate future hourly power demand. Specifically, we determined the daily average temperature thresholds that influence cooling and heating demands, namely, Tcool as the cooling temperature threshold and Theat as the heating temperature threshold. Electricity demand is adjusted when daily mean temperature is below the heating threshold (Theat) or above the cooling threshold (Tcool). To quantify these changes, we assess the sensitivity of demand variations to cooling and heating using βcool and βheat, respectively. These coefficients represent the percentage change in demand relative to the historical demand (Dhis) for each 1 °C increase or decrease in temperature. It should be noted that the nonlinear changes in cooling and heating demand are not considered here. where Dfut,d, Dhis,d, Dfut,cool,d, Dfut,heat,d, Dhis,cool,d and Dhis,heat,d represent the daily electricity demand at time d for the future, historical demand, future cooling demand, future heating demand, historical cooling demand, and historical heating demand, respectively. Dcool,d and Dheat,d refer to the cooling and heating electricity demand, either historical or future. Td is the daily average temperature at time d. Tcool and Theat are the temperature thresholds for cooling and heating. βcool and βheat represent the sensitivity of demand changes to cooling and heating. These parameters are specified on a country-specific basis (Supplementary Table 10). The resulting daily demand changes are converted to an hourly resolution using representative 24-hour cooling and heating demand allocation profiles7,62, where the sum of the allocation indices for each day totals 24, and each index is dimensionless. where Dfut,h, Dhis,h, Dcool,h and Dheat,h represent the future hourly electricity demand, historical hourly electricity demand, daily variation in cooling demand, and daily variation in heating demand, respectively. Due to spatial and temporal heterogeneity, these daily changes may also be negative. αcool,h and αheat,h are the hourly allocation indices for cooling and heating, respectively. The hourly allocation curve can be found in Supplementary Fig. 57. Future electricity demand is constructed in two steps to isolate climate-driven temperature effects from long-term socioeconomic growth. We first derive a fixed future demand baseline by applying scenario-consistent growth rates from Integrated Assessment Models (IAMs) provided by the International Institute for Applied Systems Analysis (IIASA) to observed electricity consumption in 2020. We then adjust this baseline for heating and cooling demand using temperature sensitivities and daily mean temperature changes relative to predefined thresholds, before disaggregating daily variations to hourly resolution with representative load profiles. This approach preserves scenario-consistent total demand while allowing temporal demand patterns to respond to climate change. Integrated investment and dispatch model To assess how climate change affects optimal solar-wind deployment, we employ an integrated investment and dispatch model that represents the electricity system through a set of externally determined inputs and endogenously optimized investment and operational decisions. The model covers major power generation technologies, including wind, solar, coal, oil, gas, nuclear, and hydropower, as well as long- and short-duration energy storage and interregional alternating current (AC) line and direct current (DC) line. The objective function is to minimize the overall electricity system cost (EC). where ECfix represents the overall capital and fixed O&M cost of electricity systems, and ECvar represents variable O&M cost. Relevant costs are shown in Supplementary Table 11, drawing on the latest available cost data63–65. The detailed calculation equation of ECfix is given as Eq. (25). where Gi and Gs denote the installed capacity for i category power generation technology and energy capacity for s category storage system, respectively. n includes nuclear power, hydropower, oil power, gas power, coal power, wind power, and solar power. m includes long-term energy storage and short-term energy storage. ωifix is the unit annualized fixed cost per installed capacity. We assume that long-term energy storage has a continuous operation duration of 16 hours, while short-term energy storage has a continuous operation duration of 4 hours7,66. where ECivar denotes the variable O&M cost of the i-th power generation technology and ECsvar represents the variable O&M cost of the s-th storage system. The detailed equation of ECivar and ECsvar is expressed as Eqs. (27) and (28). where EEi,t and ωivar represent power generation at hour t for technology i and variable O&M for i category power generation technology, respectively. ESs,tcha and ESs,tdis denote the charging and discharging power for s category of energy storage system, respectively. ωsvar denotes the variable O&M cost for s category of energy storage system. The model is restricted by a series of constraints in terms of supply-demand balance, energy output constraints, power generation, and energy storages. With regard to supply-demand balance, the formula is expressed as Eq. (29). where Pt denotes the electricity demand at hour t. CURtsolar and CURtwind represent the curtailed solar power and curtailed wind power, respectively. The data sources for electricity demand are provided in Supplementary Table 9. For the generation output at hour t (except wind and solar), the formula is expressed as Eqs. (30) and (31). where EEi,tmin represents the minimum output of different energy. ρ denotes the coefficient of minimum output. The corresponding coefficients are adopted from previous research7. For nuclear power, we follow previous studies to assume a stable output over the full year67,68. Hydropower is not simulated using climate-driven inflows or reservoir operations. Its annual output is fixed at its historical share of the non-wind-and-solar generation mix to isolate the effects of wind and solar variability on the optimal SWR and avoid uncertainty from hydrological projections. For wind power and solar power constraints, the formulas are expressed as Eqs. (32) and (33). where EEwind,t and EEsolar,t denote power generation of wind and solar power at hour t, respectively. Gwind and Gsolar are the installed capacities of wind and solar power, respectively. Fwind,t and Fsolar,t are the wind capacity factor and solar capacity factor at hour t, respectively. For energy storage constraints, the formulas are expressed as Eqs. (34)–(38). where ηs denotes the storage duration of storage technology s. SOCs,t represents state for s category of storage energy at hour t. ϱscha and ϱsdis denote the charging and discharging efficiency for s category of storage system, referring to relevant research69. ζs expresses the decay rate for s category of storage system, referring to relevant research7. Meanwhile, SOC at t = 1 is equal to SOC at t = 8760. To isolate climate effects on the optimal SWR, we fix the shares of all non-wind-and-solar technologies, including hydropower, at their 2020 proportions. This allows changes in the optimal SWR to reflect climate-driven wind and solar resource differences rather than shifts in the broader generation mix7. For electricity generation constraints, wind and solar power generation are constrained by IAM-predicted shares, and the shares of other generation technologies are adjusted based on the current energy mix and future wind and solar shares7. where ηj represents the generation share of j category of power technologies except wind power and solar power. l represents wind power and solar power. σ represents the proportion of wind and solar generation in the total electricity generation. For power transmission costs, there is the following constraint. where Capac and Capdc denote the installed capacity for AC line and DC line, respectively. Cacfix and Cdcfix represent the overall capital and fixed O&M cost of AC and DC, respectively. Cacvar and Cdcvar denote the variable O&M cost for AC and DC, respectively. Gact and Gdct represent the amount of AC and DC power, respectively. For power transmission constraints, the following restrictions apply. With regard to supply-demand balance, the electricity demand at hour t must equal the sum of electricity generated by all local power generation technologies, energy storage discharging/charging, and net power transmission. The net power transmission includes electricity transferred out of the region and electricity imported from other regions. For the power balance in region g at hour t, Eq. (45) holds: where Dg,t denotes the electricity demand in region g at hour t. Geng,i,t represents power generation in region g at hour t. Storg,s,tdis and Storg,s,tcha denote the discharging and charging power in region g at hour t for s category of energy storage system, respectively. w and r represent the sets of electricity input channels and electricity output channels in region g, respectively. Gacg,s,tu and Gdcg,s,tu represent the amount of AC and DC power transmitted from region u to region g at time t, respectively. Lacg,tu and Ldcg,tu represent the line losses of AC and DC from region u to region g, respectively. Gacu,s,tg and Gdcu,s,tg represent the amount of AC and DC power transmitted from region g to region u at time t, respectively. Lacu,tg and Ldcu,tg represent the line losses of AC and DC lines from region g to region u, respectively. For each climate model, scenario, and period, the optimization is solved independently for each weather year rather than for an averaged climate. The resulting configurations are evaluated over the full 40-year period, and costs and performance metrics are averaged to represent long-term conditions. Interannual stability is assessed by comparing annual optimal solar-wind capacity with 40-year means, showing generally small median deviations within -7.5% to 7.5% (Supplementary Fig. 24). Solar-wind ratio index The optimal SWR index is defined as the proportion of solar capacity to the total capacity of both wind and solar power installations after the optimization of the power system, referring to relevant studies6,7. where SWRi,jOR denotes optimal SWR in the jth mode of GCMs for the ith year. capsolar,i,jOR and capwind,i,jOR denote solar capacity and wind capacities in the jth mode of GCMs for the ith year after the optimization of the power system, respectively. The average optimization SWR is calculated as Eq. (47). where SWRmOR denotes the average optimal SWR over n periods across m GCMs. Change in the SWR in the Optimization SWR scenario under climate change is calculated as Eq. (48). where CSWROR denotes the change in SWR from period t−1 to period t. SWRtOR and SWRt−1OR represent SWR in period t and t−1, respectively. To characterize SWR deviations under different scenarios, we define the following expressions. where CDSWROR denotes the deviation of the SWR in the Historical preference or Future prediction scenario from that in the Optimization scenario. A positive value of CDSWROR indicates overinvestment in solar relative to the optimal SWR, whereas a negative value indicates underinvestment. The Future prediction SWR across different countries is projected under different climate scenarios, based on IAMs from the International Institute for Applied Systems Analysis (IIASA; https://data.ece.iiasa.ac.at/ar6/). The solar-wind ratio index can be directly calculated. Specifically, the Future prediction SWR, corresponding to Capacity|Electricity|Wind and Capacity|Electricity|Solar, is calculated by dividing solar capacity/(solar capacity+wind capacity). If data for a specific country is missing, we identify the regional projections (i.e., R10regions) to which the country belongs and use the regional solar-wind ratio. Regarding the Historical preference SWR, this study uses the average SWR from 2015 to 2022 to represent the historical preference configuration. Uncertainty analyses We test a range of modeling assumptions to verify the robustness of our results (Supplementary Note 6). First, we address climate uncertainty by coupling multiple GCM projections, evaluating independently optimized configurations across years rather than optimizing against an averaged climate, and expanding the analysis to 178 countries across three time periods (1981–2020, 2021–2060 and 2061–2100) (Supplementary Fig. 2). This design ensures that the identified SWR adjustments reflect robust directional shifts under climate variability across both spatial and temporal scales. Second, we assess sensitivity to key technological and cost assumptions (Supplementary Figs. 9−16), including PV technology choice, storage duration, ±10% variations in renewable and storage costs, hydro sensitivity, load sensitivity, and country-specific adjustments to the weighted average cost of capital, reflecting heterogeneity in financing conditions across regions. Across these scenarios, the spatial pattern and directional shifts in optimal SWR remain largely unchanged, and deviations from climate-consistent SWRs consistently lead to elevated system costs. In particular, low-latitude regions consistently exhibit larger cost penalties from SWR misalignment than higher-latitude regions across all sensitivity analyses. In most regions, the cost impacts associated with SWR misalignment exceed those arising from climate change. Third, we examine the impact of cross-border electricity interconnection through a European trade sensitivity analysis (Supplementary Figs. 17–21), finding that the transmission test results in relatively small changes and does not fundamentally alter the optimal SWR or the comparative cost penalties associated with SWR misalignment. Fourth, to evaluate potential distortions arising from temporal downscaling, we test alternative treatments of hourly variability to assess whether interpolating 3-hourly GCM outputs affects main results (Supplementary Figs. 9−13 and 22–23). Although it leads to relatively small changes in flexibility energy requirements, it does not alter the relative magnitude of system cost impacts associated with SWR misalignment. Further details are provided in Supplementary Note 6. Supplementary information Supplementary Information Transparent Peer Review file Source data Source Data_Figure 1-5 Acknowledgments The authors gratefully acknowledge technical and data-related support from colleagues and staff at Tsinghua University. Author contributions D.T. conceived and designed the study. J.L. led the model development and conducted the simulations. D.T. and J.L. performed the analyses, with support from Q.Z. and Y.X. on analytical approaches and from D.Z. and Y.L. on data compilation and model implementation. D.T. and J.L. wrote the paper with input from all co-authors. Peer review Peer review information Nature Communications thanks Johannes Schmidt and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available. Funding This work was supported by the National Natural Science Foundation of China (grant nos. 42130708, 42505164, and 72504158), Beijing Natural Science Foundation (grant no. 9254028), General Project of China Postdoctoral Science Foundation (grant no. 2024M761738), and China Meteorological Administration “Research on Value realization of climate ecological products” Youth Innovation Team Project (No. CMA2024QN15). We acknowledge the High Performance Computing Center of Tsinghua University for providing the facility support. Data availability All climate datasets used in this study are publicly available. Historical reanalysis data are derived from the ERA5 dataset (https://cds.climate.copernicus.eu/), while future climate projections are sourced from the CMIP6 archive hosted by the Earth System Grid Federation (ESGF; https://esgf-node.llnl.gov/projects/cmip6/). Country-level projections of electricity generation technology shares under different climate scenarios are obtained from the International Institute for Applied Systems Analysis (IIASA; https://data.ece.iiasa.ac.at/ar6/). Geospatial maps are generated using open-access shapefiles from the Database of Global Administrative Boundaries (GADM; https://gadm.org/), with data processing and visualization carried out using Python v3.11, GeoPandas v1.1 (https://geopandas.org/en/stable/), and Matplotlib v3.7.1 (https://matplotlib.org/). Source data are provided with this paper. Code availability The integrated investment and dispatch model and the corresponding datasets are accessible at https://zenodo.org/records/18265901. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Supplementary information The online version contains supplementary material available at 10.1038/s41467-026-75879-9. References Associated Data Supplementary Materials Supplementary Information Transparent Peer Review file Source Data_Figure 1-5 Data Availability Statement All climate datasets used in this study are publicly available. Historical reanalysis data are derived from the ERA5 dataset (https://cds.climate.copernicus.eu/), while future climate projections are sourced from the CMIP6 archive hosted by the Earth System Grid Federation (ESGF; https://esgf-node.llnl.gov/projects/cmip6/). Country-level projections of electricity generation technology shares under different climate scenarios are obtained from the International Institute for Applied Systems Analysis (IIASA; https://data.ece.iiasa.ac.at/ar6/). Geospatial maps are generated using open-access shapefiles from the Database of Global Administrative Boundaries (GADM; https://gadm.org/), with data processing and visualization carried out using Python v3.11, GeoPandas v1.1 (https://geopandas.org/en/stable/), and Matplotlib v3.7.1 (https://matplotlib.org/). Source data are provided with this paper. The integrated investment and dispatch model and the corresponding datasets are accessible at https://zenodo.org/records/18265901.