Will climate change lead to spatiotemporal changes in the global distribution of wind resources in any region of the globe?

Many studies predict a decrease in mean wind speed in the mid-latitudes of the Northern Hemisphere and an increase in mean wind speed in the tropics. However, the average wind resource conditions only represent one aspect of the meteorological wind potential. The dynamics of the temporal variability and spatial complementarity of wind resources under climate change are not well understood. Therefore, in order to investigate the impact of climate change on the temporal variability of wind resources and the potential for complementary use of wind resources, we calculated time series capacity factors for a global dataset of wind farm sites using a set of 18 global climate models and statistical downscaling methods. The coefficient of variation was used to assess temporal variability and the Spearman rank correlation coefficient was used to assess complementarity. The results show that spatial complementarity of wind resources is rare in the historical period 1989–2014. In many regions of the Northern Hemisphere, the temporal variability of wind resources is increasing in the future period to 2099, while the potential for complementary wind resources remains low. In China and the United States, the average of the climate model ensembles predicts a significant increase in the coefficient of variation of almost 10% by the end of the 21st century compared to the historical period under high radiative forcing. When this is combined with a decrease in the average wind resources and a low potential for spatial complementarity, meteorological conditions for wind energy deployment become increasingly unfavorable. In contrast, conditions for wind energy expansion are improving in Brazil, as average capacity coefficients are increasing and coefficients of variation are decreasing. The findings of this study should be taken into account when integrating wind energy into the electricity system, as they are crucial for energy conversion and, in particular, for balancing electricity supply and demand. Transitioning to renewable energy sources is essential for mitigating climate change and reducing the reliance on fossil fuels in the power sector [1]. As of 2023, wind energy has become the most relevant renewable resource for combating global climate change, after solar photovoltaics, with an installed capacity exceeding 1 TW [2]. Notably, 92.8 % of this wind capacity is derived from onshore wind turbines. In 2022 alone, onshore wind turbines generated a remarkable total of 1,936 TWh of electricity [2]. The importance of wind energy is expected to grow even further in the coming years. Between 2024 and 2030, the rate of global wind capacity expansion is projected to double compared to the period from 2017 to 2023 [3]. However, the actual growth of wind capacity will hinge on the availability of wind potential [4]. Therefore, the wind resource, which is primarily determined by the wind speed (U), will be a critical factor in the future success of wind energy development [5]. The spatial and temporal variations in solar irradiation, air temperature, air pressure, orography, and surface roughness result in a complex and variable U pattern across the Earth [6]. Consequently, there are specific regions and time periods where the utilization of wind energy is more economically viable than in others [7]. Additionally, large-scale and small-scale circulation patterns are shifting due to the temperature change brought about by climate change altering the global spatiotemporal distribution of U [8]. Global climate models (GCM), that rely on different Shared Socioeconomic Pathways (SSP), project the future U development. They have a very coarse resolution and do not reflect the mesoscale properties of U. Thus, regional climate models are used to downscale U at the mesoscale level [9].Additionally, machine learning methods can be applied to consider local U properties [10]. The direction and magnitude of the projected future mean U changes from climate models vary depending on the region, the specific climate models utilized, and SSP scenarios [11]. Climate change-related changes in wind resources are often minor compared to the internal natural variability of U [12]. Typically, the magnitudes of mean U changes are highest under the most pessimistic SSP scenarios [13]. In general, wind resources tend to decrease in many mid-latitude regions of the Northern Hemisphere, while they increase near the equator [14]. For instance, in large parts of Europe, wind resources are projected to decrease significantly by 5–15 % under SSP245 and by 10–20 % under SSP585 by the end of the 21st century [15]. The declines in wind resources are particularly pronounced in northern Continental Europe and the Central Mediterranean [16]. In contrast, the majority of climate models expect an increase in wind resources in the tropics of South America [17]. The influence of climate change on U often varies with time, leading to potentially changing seasonal fluctuations in wind energy [18]. This variation in variability complicates the balancing of electricity supply and demand [19]. In fact, Liu et al. [20] assume a notable reduction in the alignment of supply and demand for 32 % of non-Antarctic land areas by the end of this century under SSP245. Additionally, as highlighted by Feng et al. [21], climate models show an increase in the frequency and duration of low-output wind-power events. Consequently, power systems are increasingly impacted by climate change [22]. To balance the power system during periods of low wind power generation, other renewable energy sources, such as solar photovoltaics [23] and hydropower [24], can be utilized. However, this relies on the availability of solar energy and hydropower as complementary sources to wind energy [25]. Complementarity assessments are typically conducted by evaluating the strength of the negative correlation between different renewable energy resources [26]. The temporal complementarity may vary depending on the time scale analyzed (hourly, daily, seasonally, or annually). For example, in Poland, temporal complementarity between solar and wind resources predominantly occurs on a seasonal scale [27]. Numerous studies have explored the temporal complementarity of solar and wind resources in various regions. For instance, the wind-solar complementarity was analyzed in China using the Pearson correlation coefficient [28]. Another study assessed the temporal complementarity between solar and wind resources on the Iberian Peninsula [29]. Significant potential for the complementary use of wind and solar energy was identified in Portugal [30]. Despite these regional studies, global scale complementarity studies are lacking in the literature [31]. In addition to temporal complementarity between different energy source, there is also the potential to utilize spatial complementarity of wind resources [32]. Spatial complementarity is based on the premise that wind speed varies between neighboring regions, influenced by factors such as diurnal wind patterns and time lags associated with the passage of frontal systems. However, there are fewer studies that examine the spatial complementarity of wind resources. One study found that wind power generation in most Central and Western European countries is strongly correlated, which limits the potential for leveraging spatial complementarity [33]. It is also uncertain how spatial complementarity may be affected by climate change. The degree of complementarity could shift due to varying climate change-related developments in wind resources, particularly in terms of direction and characteristics [34]. Currently, there are no global studies available that investigate the long-term development of spatial wind complementarity. lower seasonality of the wind resource near the equator. North America, Asia, and Oceania also show CVcon values below 50 %. In contrast, the CVcon values for South America and Europe are approximately 60 %. A continental perspective on CV has limited relevance due to the extensive spatial coverage and high spatial variability of wind resources [55]. Consequently, Fig. 1b presents CVnat values. Some countries, particularly in Africa, lack CVnat values because they do not have enough wind turbines for evaluation. All examined countries have CVnat values exceeding 50 %. The USA and China exhibit the lowest CVnat values globally, at just over 50 %. In other countries with large geographical areas, such as Canada, Australia, and Brazil, the CVnat levels are also comparatively low by global standards. A northwest-southeast gradient of CVnat is evident across Europe. The United Kingdom has a relatively low CVnat value of 75 %, while Germany’s CVnat value is higher at 95 %. In contrast, Italy has a CVnat value that reaches even 118 %. One reason may be the greater continentally in south-eastern Europe. Fig. 1c illustrates ρWT,con. Most sites (99.3 %) exhibit positive ρWT,con values, indicating a similarity between the temporal patterns of wind resources at the wind turbine sites and the continental mean. Consequently, the potential for using the spatial complementarity of wind resources on a continental scale is low. This is consistent with the results of a Europe-wide study which also showed only a low potential for spatial wind resource complementarity [33]. Variations in ρWT,con patterns occur across the continents. In Europe, similarity is particularly strong in England, Germany, and the Benelux countries because the wind turbines in these countries greatly influence the temporal patterns of CFcon. The ρWT,con values tend to decrease with increasing distance from the continental center of the wind turbine sites. Nonetheless, a weak similarity remains evident in the peripheral areas of Europe. A similar pattern of ρWT,con is observed across other continents. In North America, the highest ρWT,con values are in the Great Plains; in South America, they occur along the east coast of Brazil; in Asia, they are concentrated in northern China; and in Oceania, the highest ρWT,con values are seen in southeastern Australia. Africa has no major regions with high ρWT,con values due to the dispersion of wind turbines. Notably, in Asia, the wind turbines on China’s southeast coast exhibit weak complementarity with CFcon. Additionally, negative ρWT,con values are also common in island regions. The similarity between CFnat and CFWT sites is often strong to very strong in many countries (Fig. 1d). Thus, the availability of spatial complementarity of the wind resource at national level is even lower than at continental level. In total, 20 % of all wind turbine sites exhibit ρWT,nat values greater than 0.90. Particularly in Europe, where countries are generally smaller and wind turbine density is high, some regions demonstrate very strong similarity. Consequently, the potential to smooth CFWT,nat through the distribution of wind turbines is quite limited in Europe. In the United States, the similarity is weaker along the coasts compared to the Great Plains. In China, regions in the south and along the coasts also show weaker similarities. A certain potential for wind resource complementarity in China along a north–south gradient was also identified in an earlier study by Liu et al. [65]. Here, the complementarities of the entire capacity factor time series were analyzed and it was found that they are mostly non-existent. If the time series were decomposed using wavelet spectrum analysis, there may be at least some complementarity over certain time scales [66]. These results reveal that the initial situation of wind resource availability, temporal variability, and complementarity is very diverse globally. Therefore, the climate change-induced changes shown in the following sections are of uneven relevance.3.2. Wind resource availability, temporal variability, and complementarity in the 21st century The properties of the CFcon time series change over time. PC compared to 2004 of the continental development of CFcon, ρcon, CVcon, and stdcon in the 21st century, assuming the SSP585 scenario, is shown in Fig. 2 by 30-year moving averages. This scenario is chosen for the presentation of the results as it represents the upper range of the climate change signal [67]. In North America, CFcon is steadily declining (Fig. 2a). At the end of the century (2099), CFcon is 8.0 % lower than at the beginning (2004). The remarkable CFcon decrease, in combination with only a slight stdcon decrease at 2.8 % in 2099, leads to a strong CVcon increase at 7.0 % in 2099. Complementarity hardly changes since PC of ρcon values fluctuate around 0 % for decades. However, in the last twenty years of the study period, ρcon increases by around 2.0 %. In South America, CFcon rapidly rises by 6.8 % until 2040 and then slightly decreases (Fig. 2b). Thus, a major part of the CFcon increase is already reflected in the currently very high CFcon values in South America. The stdcon values increase slightly and peak around 2060, while CVcon hardly changes. The CFcon similarity increases by PC of ρcon of about 4.0 % at the end of the century. In Europe, the ensemble of climate models projects decreasing wind resource availability during the 21st century (Fig. 2c). Accordingly, CFcon decreases by 5.7 % in 2099 compared to 2004. A relevant stdcon change is not simulated, leading to a sharp CVcon increase of almost 9 % by 2099. There is a slight tendency towards higher similarity between the wind turbine sites, as indicated by ρcon increasing about 2 % by the end of the 21st century. In Africa, the analyzed variables slightly decrease in the long term (Fig. 2d). The greatest decrease has stdcon, with 5 % at the end of the century. The ρcon values first slightly increase and then decrease. In Asia, the development of wind resource properties is very unfavorable for wind energy use (Fig. 2e). It has the greatest ρcon increase of all continents, with Δρcon at 8.6 % in 2099. While CVcon increases by about 5 %, stdcon decreases by around 5 %. In Oceania, CFcon is declining similarly to Europe and Asia (Fig. 2f). The CVcon value increases by around 5 %. The similarity also increases slightly until the end of the 21st century. The results suggest that in the majority of continents, an unfavorable development of wind resource characteristics under climate change must be anticipated. However, it is critical to note that this evaluation does not consider further improvements in wind turbine technology. Previous studies have shown that these could compensate for potential CF reductions [17] moderating the relevance of the unfavorable wind resource development in many continents. Fig. 3 illustrates the long-term development of CFnat, ρnat, CVnat, and stdnat, assuming the SSP585 scenario for the six countries with the highest installed wind capacity by PC. The ensemble of climate models projects a substantial decrease in CFnat for China under SSP585, with a decline of nearly 10 %, while stdnat remains close to constant (Fig. 3a). Consequently, CVnat also increases enormously by 10 %. The spatial distribution of wind resources in the country becomes slightly more uniform, as indicated by PC of ρnat of 2.7 %. In the USA, the projected development of CFnat and CVnat is similar to that of China (Fig. 3b). In contrast to China, stdnat decreases while ρnat remains almost constant. In Germany, CFnat decreases, while CVnat increases by approximately 5 % (Fig. 3c). Significant changes in ρnat and stdnat are not projected by the ensemble of climate models. In India, CFnat remains close to constant for several decades (Fig. 3d). After 2060, the ensemble of climate models projects increasing CFnat. Additionally, ρnat, CVnat, and stdnat also show favorable trends toward the end of the 21st century from a wind energy perspective, each exhibiting a percentage decrease of approximately 3 %. In Spain, CFnat and stdnat are continuously decreasing, with a decline of nearly 9 % (Fig. 3e). Consequently, CVnat is changing only slightly. Additionally, complementarity shows a modest improvement, with PA of ρnat being − 3.0 %. The development of CFnat in Brazil differs from that of the countries previously discussed (Fig. 3e). Initially, a significant increase of approximately 8 % in CFnat is projected by 2040. Following this period, CFnat decreases slightly. CVnat also declines until 2040 but then levels off due to a subsequent increase in stdnat. Throughout the investigation period, the PC of ρnat remains consistently positive. 3.3. Temporal variability under climate change Fig. 4 displays the climate change-related CVcon changes across the continents. Most continents exhibit significant changes (at the 5 % level) in CVcon under various climate change scenarios. The only continent with no significant changes in CVcon is Africa. When significant CVcon values are present, they are predominantly positive. An increase in CVcon indicates more temporal variability. This is due to the fact that in many regions the mean wind resource decreases but the standard deviation of the wind resource remains at a high level. Consequently, the dispersion around the mean increases. As climate change intensifies, the continental CVcon values frequently increase, particularly in Europe. Under the SSP245-NF scenario (Fig. 4a), CVcon increases by only 1.65 %, whereas under SSP585-NF (Fig. 4b), ΔCVcon is 2.08 %. The most substantial impact of climate change is observed in the SSP585-FF (Fig. 4f) scenario, where ΔCVcon in Europe reaches 5.02 %. This pattern of CVcon development also occurs in Asia, North America, and Oceania; however, the extent of ΔCVcon is lower than in Europe. Under the SSP585-FF scenario, CVcon increases by 2.97 % in North America, 2.79 % in Oceania, and 1.69 % in Asia. South America is the only continent where CVcon shows a slight decline (Fig. 4a-d). A linear relationship between the intensity of climate change and CVcon changes is not discernible in South America. The highest CVcon decrease occurs under the SSP245-NF and SSP585-MF (Fig. 4d) scenarios. In the SSP245-FF (Fig. 4e) and SSP585-FF scenarios, no significant change in CVcon happens in South America.This also affects two countries with high installed wind capacity, Germany and India. There are significant positive CVnat changes in 38.5 % of all countries. The CVnat increase in Italy and Poland exceeds 3.0 %. Decreasing CVnat values are considerably rarer at 7.3 %. In Brazil and Uruguay, ΔCVnat is around − 1.8 %. The SSP585-NF (Fig. 5b) and SSP245-MF (Fig. 5c) scenarios have a similar pattern of non-significant, positive-significant, and negative significant CVnat shares. One of the few relevant differences to SSP245- NF is that Germany has a significant positive CVnat increase under SSP585-NF and SSP245-MF. Under the SSP585-MF scenario, the share of countries with significantly positive ΔCVnat values dominates at 56.3 % (Fig. 5d). Italy, Canada, and Poland stand out in particular with ΔCVnat values of almost 5 %. The share of significantly negative CVnat values also increases to 11.5 %. In Brazil and South Africa, ΔCVnat under SSP585-MF is around − 2.8 %. The SSP245-FF scenario has 52.1 % countries with significantly positive ΔCVnat values (Fig. 5e). The CVnat increase in Mexico is especially notable at 8.02 %. Poland and the UK also have ΔCVnat values exceeding 6 %. Only 9.4 % of all countries have a significant CVnat decrease, including Brazil (ΔCVnat = -2.2 %) and South Africa (ΔCVnat = -1.2 %). The SSP585-FF scenario reveals the most striking CVnat changes (Fig. 5f). Overall, 77.1 % of all countries have significant CVnat changes. ΔCVnat is significantly positive in 61.5 % of all countries and negative in 15.6 %. In 35 % of all countries, the CVnat increase exceeds 4 %, including countries with high installed wind capacity, such as China, the USA, the UK, and France. The Central American countries Costa Rica (− 19.3 %) and Nicaragua (− 10.1 %) have extremely low ΔCVnat values.Germany, the United Kingdom, and Sweden. In countries with larger geographical areas, such as the USA and China, wind turbine sites exhibiting a significant increase in ρWT,nat are confined to specific regions. Additionally, the proportion of wind turbine sites with significantly negative ρWT,nat values rises slightly, with such sites occurring frequently in Spain. In the far-future, only minor changes occur under SSP245 compared to the mid-future scenario (Fig. 7e). The SSP245-FF scenario reveals that 12.2 % of wind turbines exhibit significantly negative ρ values, representing the highest proportion. Meanwhile, under SSP585-FF, the share of wind turbine sites with significantly positive ρWT,nat values continues to rise, reaching 51.0 % (Fig. 7f). Additionally, at many wind turbine sites that show significantly positive ρWT,nat under other scenarios, ρWT, nat is increasing. 4. Conclusions This study investigated the impact of climate change on the temporal variability of wind resources and the potential for complementary wind resource utilization on a global scale. The results reveal that the decrease of the mean wind resource in the mid-latitudes of the Northern Hemisphere is often linked to increasing temporal variability as the standard deviation of the wind resource remains high. Besides it was found that the potential for complementary wind resource utilization remains low or often even reduces in most of these regions. This combination of declining average wind resources, increasing temporal variability, and limited complementarity creates deteriorating meteorological conditions for wind energy deployment in many regions with high installed wind capacities. However, as the wind turbine fleet compensates for potential capacity factor reductions, the relevance of unfavorable wind resource developments is softened. The results of this study show spatial diversity. There is a small number of regions where the wind resource properties develop favorably for wind energy expansion. One example is Brazil, where average capacity factors are projected to increase and the temporal variability declines. The findings of this study are critical for integrating wind energy into the power system and how to balance electricity supply and demand. The study underscores the relevance of climate protection by wind energy expansion as the negative developments are primarily associated with the highest emission scenarios. This means that the faster the expansion of wind energy progresses, the more favorable the radiative forcing and the better the conditions for the use of wind energy in many regions.