Potential Impacts of Climate Change on South China Sea Wind Energy Resources Under CMIP6 Future Climate Projections

Abstract

Wind is an important renewable energy source, and even minor variations in wind speed will significantly impact wind power generation. The objective of this study was to systematically assess the impacts of climate change on wind energy resources in the South China Sea (SCS) under future climate projections. To achieve this, we employed a multi-model ensemble approach based on Coupled Model Intercomparison Project Phase 6 (CMIP6) data under three Shared Socioeconomic Pathways (SSP1-2.6, SSP2-4.5, and SSP5-8.5). The results demonstrated that, in comparison with scatterometer wind data, the CMIP6 historical results (1995–2014) showed good performance in capturing the spatiotemporal distribution of wind power density (WPD) in the SCS. There were regional discrepancies in the central SCS due to the complex monsoon-driven wind dynamics. Future projections revealed an overall increase in annual mean wind power density (WPD) across the entire SCS by the mid-21st century (2046–2065) and late 21st century (2080–2099). The seasonal analyses indicated significant WPD increases in summer, especially in the northern SCS and the region adjacent to the Kalimantan strait. The increase in summer (>40 × 10⁻⁴ m/s/year under SSP5-8.5) is about triple that in winter. In the late 21st century, an increase in WPD exceeding 10% can be generally anticipated under the SSP2-4.5 and SSP5-8.5 scenarios in all seasons. The extreme wind in the northern and central SCS will further increase by 5% under the three scenarios, which will add an extra extreme load to wind turbines and related marine facilities. These assessments are essential for wind farm planning and long-term energy production evaluations in the SCS. Based on the findings in this study, specific areas of concern can be targeted to conduct localized downscaling analyses and risk assessments.

1. Introduction

Climate change has affected the large-scale atmospheric circulation since the mid-20th century and has created several challenges in the 21st century. The energy sector is the primary driver of climate change, accounting for 75% of global CO₂ emissions [1]. Renewable energy is vital for reducing the emissions of greenhouse gases. In terms of installed capacity, growth rate, and technological maturity, wind energy ranks as the second largest renewable energy source globally [2]. As the global demand for clean energy intensifies, wind energy has emerged as a key renewable resource, with offshore wind farms increasingly deployed in coastal regions [3,4,5]. The potential of wind energy is proportional to the cube of wind speed; therefore, even minor variations in wind speed will significantly impact wind power generation [6,7].

The Intergovernmental Panel on Climate Change (IPCC) Assessment Report 6 has provided projections for winds under various greenhouse gas emission scenarios up to the year 2100 [8]. These projections are based on global climate models (GCMs) in the framework of the Coupled Model Inter-comparison Project Phase 6 (CMIP6), which show strong seasonal and regional dependence in response to climate change. The latest CMIP6 was started in 2015. The advanced multi-model datasets are helpful for elucidating the mechanisms of climate change and improving the reliability of future projections [9]. On the global scale, significant drops in wind energy density are projected for the mid-latitudes of the northern hemisphere while concentrated increases in wind resources are anticipated for tropical regions [10]. Analyses of Europe have predicted a continuous, generalized drop in wind resources by 2100 [6]. In North America, onshore wind power density is projected to drop by ~15%, reaching 40% in some regions under SSP5-8.5, whereas significant increases in wind power density are predicted for Hudson Bay [11]. In the United Kingdom, the CMIP6 data indicate that the monthly mean wind speed will rise from a baseline of 3.41 m/s (1950–2014) to new levels under various climate-change scenarios: 3.60 m/s (SSP1-2.6), 3.63 m/s (SSP2-4.5), 3.48 m/s (SSP3-7.0), 3.59 m/s (SSP4-6.0), and 3.61 m/s (SSP5-8.5) [12]. Based on the SSP2-4.2 and SSP5-8.5 scenarios, the European offshore wind resources will decline overall. However, much of the Atlantic coast of continental Europe will experience increasing extremes in the high-emission scenario [13]. The long-term production of offshore wind farms faces uncertainty due to future changes in wind resources. These changes could also affect the marine environment [14]. Wind turbines must be able to operate safely and reliably in extreme weather and marine environments for long periods of time [15].

The South China Sea (SCS) is situated at the intersection of the Eurasian, Pacific, and Indo-Australian plates (Figure 1), rendering it particularly vulnerable to climate change [16,17]. The SCS possesses substantial wind energy resources due to the East Asian monsoon, which is characterized by strong seasonal variations [18,19]. Climate change-induced alterations in atmospheric circulation impact both wind resource availability and marine infrastructure safety [20,21]. Zhang et al. [22] utilized ERA5 reanalysis data to analyze the spatial distribution and trends in wind energy across various time scales over the SCS from 1979 to 2021. Their results indicated that the SCS possesses abundant wind energy resources, which show different trends in different areas and obvious seasonal changes. Hong and Zhang’s analysis [23] revealed that, over the past decades, the annual mean wind speed in the northern SCS has shown a decreasing trend in the coastal area and an increasing trend in the open sea. After analyzing the long-term spatiotemporal patterns of the annual, seasonal, and monthly wind fields and wind power density in the SCS, Wang et al. [24] found that maximum wind power generation occurs in December while the minimum occurs in May.

There are several studies that focused on the wind in mainland China. Jiang et al. [25] predicted that the maximum wind speed in China will decrease by 2046–2065 and 2080–2099. Based on CMIP6 data, the surface wind speed in most parts of China is projected to decrease in the middle and late 21st century under the three projection scenarios [26]. Moreover, the wind speed averaged over China will show a reduction under the SSP1-2.6, SSP2-4.5, and SSP5-8.5 scenarios, both annually and in winter, but it will increase in summer under SSP5-8.5. Studies on the future trends in wind resources in the SCS are still quite limited. Based on CMIP6 data and a deep learning-based downscaling method, Zhang et al. [27] conducted a survey on the offshore wind energy resources in China and the results of multi-model ensemble forecasting showed that the wind power density in the East China Sea will decrease slightly while in the South China Sea, it will increase in 2041–2060 and 2081–2100 under scenarios SSP2-4.5 and SSP5-8.5. Analyses of the wind power density in Australasia and South-East Asia showed that there will be marked changes at the end of the 21st century (exceeding 150%) in certain areas (e.g., Vietnam and Borneo) and the mean values and temporal variability changes will be greater in the high-emission scenario [28]. Extreme wind speeds (EWSs) can cause severe mechanical loads on the blades and other components of wind turbines [29]. Climate change can cause changes in extreme winds, which could impact the marine engineering design parameters of EWS loads with different return periods [30]. Therefore, to ensure the structural safety of wind turbines, it is necessary to examine the projected changes in EWSs in the SCS.

Although the impacts of climate change on the future sea surface wind energy resources have been the focus of ocean engineering and marine environment studies, the evolution of the wind energy resources in the SCS under future climate-change scenarios has not been systematically assessed. The purpose of this study was to assess the impacts of climate change on the wind energy resources in the SCS under the CMIP6 future climate projections. Firstly, the performance of the CMIP6 GCMs in the SCS was calibrated for the historical period (1995–2014) using Level 3 global daily gridded scatterometer observations. Secondly, the projected changes in wind energy resources were analyzed to determine their changes and trends under the SSP1-2.6, SSP2-4.5, and SSP5-8.5 scenarios in the medium term (2051–2060) and long term (2091–2100). Thirdly, the projected changes in extreme winds in the SCS were analyzed to illustrate the possible impact of future climatic scenarios on EWS loads with different return periods. Assessing the mid-term and long-term changes in wind energy is essential for supporting regional wind energy resource utilization and for planning and managing wind farms.

The remainder of this paper is organized as follows. Section 2 describes the data and methods. In Section 3, the results and discussion are presented. Finally, conclusions are drawn in Section 4.

2. Materials and Methods

2.1. Data

2.1.1. Scatterometer Wind Data

The wind data used in this study are the bias-corrected monthly wind data produced by the Royal Netherlands Meteorological Institute (KNMI), which are based on Level 3 global daily scatterometer observations and European Centre for Medium-range Weather Forecasts (ECMWF) ERA5 reanalysis model winds (hereafter referred to as obs-ERA5) [31]. The Level 3 global daily gridded scatterometer observations rely on the products from the EUMETSAT Ocean and Sea Ice Satellite Application Facility (OSI SAF) Wind Centre at KNMI. Specifically, the observations are taken from the several scatterometers at a 0.25° horizontal resolution in different combinations depending on availability. Belmonte Rivas and Stoffelen [32] identified biases in the ERA5 model winds. The bias correction fields are calculated based on the monthly accumulation of local differences between available Level 3 scatterometer data and their collocated ERA5 model winds. In addition to the monthly averages, the product also includes the standard deviation of scatterometer-model differences and the total number of observations used for bias correction. The available data range from July 1994 to December 2024.

2.1.2. CMIP6 Wind Data

CMIP6 is a global collaborative framework designed to advance climate model development through multi-model comparisons, allowing users to understand past climate changes, make projections, and estimate future uncertainty. The CMIP6 dataset serves as a cornerstone for the IPCC Sixth Assessment Report, offering comprehensive, multi-model climate projections and enabling detailed evaluations of sensitivities to greenhouse gas emissions, aerosols, and natural variability [9]. It includes both historical data from 1850 to 2014 and projected data under three climate-change scenarios (SSP1-2.6, SSP2-4.5, and SSP5-8.5, which represent low-, medium-, and high-emission scenarios, respectively) from 2015 to 2100.

In this study, we used monthly near-surface wind speed (at a height of 10 m) data. The global climate models (GCMs), along with their institution ID, name, and horizontal resolutions (longitude × latitude), are summarized in

Table 1. List of 22 CMIP6 global climate models (GCMs) used in this study. The grid value (fourth column) is the number of grid points (1° × 1°) in the South China Sea where the bias values (CMIP6 model results relative to obs-ERA5) falls in the range of [−1 1]. Three low grid values are marked by triangles and the corresponding GCMs were excluded from the subsequent analyses. The GCMs with daily maximum wind speed data are marked by stars.

Institution_id	CMIP6 Model	CMIP6 Model Resolution (Longitude × Latitude)	Grid Value
CSIRO-ARCCSS	ACCESS-CM2	192 × 144	240
CSIRO	ACCESS-ESM1-5	192 × 145	257
AWI	AWI-CM-1-1-MR ✺	384 × 192	284
AWI	AWI-ESM-1-REcoM	192 × 96	256
BCC	BCC-CSM2-MR ✺	320 × 160	287
CAMS	CAMS-CSM1-0	320 × 160	175 ▲
CAS	CAS-ESM2-0	256 × 128	65 ▲
NCAR	CESM2-WACCM	288 × 192	281
CMCC	CMCC-CM2-SR5 ✺	288 × 192	262
CMCC	CMCC-ESM2 ✺	288 × 192	264
CCCma	CanESM5 ✺	128 × 64	255
CCCma	CanESM5-1 ✺	128 × 64	258
CAS	FGOALS-g3 ✺	180 × 80	275
FIO-QLNM	FIO-ESM-2-0	288 × 192	286
CCCR-IITM	IITM-ESM	192 × 94	252
MIROC	MIROC6	256 × 128	152 ▲
DKRZ	MPI-ESM1-2-HR ✺	384 × 192	259
MPI-M	MPI-ESM1-2-LR ✺	192 × 96	219
MRI	MRI-ESM2-0 ✺	320 × 160	296
NCC	NorESM2-LM	144 × 96	227
NCC	NorESM2-MM	288 × 192	252
AS-RCEC	TaiESM1	288 × 192	262

. Additionally, daily maximum near-surface wind speed (at a height of 10 m) data were used to analyze the extreme wind speeds under the projected scenarios. The GCMs with daily maximum wind data are marked in Table 1. To ensure that all the GCMs are weighted equally in the multi-model statistics, only one realization “r1i1p1f1”, which means the first realization (r1), first initialization (i1), first physics, and first forcing (p1f1), from the CMIP6 GCMs was used.

Table 1 shows that the CMIP6 dataset includes GCMs with different horizontal resolutions. To facilitate the comparison and statistical analyses, data from each GCM were interpolated to a 1° × 1°grid by using bilinear interpolation. To evaluate the data from each GCM, we calculated the bias value for each model relative to obs-ERA5. Then, we counted the number of grid points (1° × 1°) in the SCS where the bias values fell within the range of [−1 1]. The resulting grid values are also listed in Table 1. The CAMS-CSM1-0, CAS-ESM2-0, and MIROC6 models have lower grid values in comparison with the other GCMs. Therefore, the data from these three GCMs were excluded. The remaining 19 out of 22 GCMs were used in the subsequent analyses.

2.2. Methods

The Taylor’s skill score was utilized to evaluate the performance of the CMIP6 GCMs relative to the scatterometer wind data. The Taylor’s skill score is a comprehensive metric used to evaluate the performance of model predictions, which has been widely applied in fields such as meteorology, climatology, and environmental science [33]. It is defined as

$$Skill={\displaystyle \frac{4{\left(1+R\right)}^4}{{\left(\mathrm{S}\mathrm{D}\mathrm{R}+\frac{1}{\mathrm{S}\mathrm{D}\mathrm{R}}\right)}^2\times {\left(1+R_0\right)}^4}},$$

(1)

where $R$ denotes the correlation coefficient, $R_0$ denotes the maximum correlation coefficient (set to 1 in this study), and SDR is the ratio of the standard deviations of each model to that of the observed values.

A Taylor diagram was used to provide a concise and intuitive way to visualize multiple statistical metrics simultaneously, including the correlation coefficient, standard deviation, and root mean square error (RMSE) between the CMIP6 GCM outputs and scatterometer wind data [34]. The diagram was constructed in a polar coordinate system, with the radial axis representing the standard deviation and the angular axis representing the correlation coefficient. Points closer to the reference point (typically the observational data) indicate better model performance, with higher correlation and lower RMSE values.

The wind power density (WPD) was employed as the metric for assessing wind resources [35,36], which can be obtained using the equation

$$\mathrm{W}\mathrm{P}\mathrm{D}={\displaystyle \frac{1}{2}}\rho v^3,$$

(2)

where $\rho$ is the air density (assumed to be 1.225 kg m⁻³) and v is the near-surface wind speed. WPD is proportional to $v^3$, so $v$ has a significant impact on WPD.

The percentage change [37] was estimated as

$$\mathrm{percentage}\mathrm{change}={\displaystyle \frac{{\mathrm{W}\mathrm{P}\mathrm{D}}_{\mathrm{f}\mathrm{u}\mathrm{t}\mathrm{u}\mathrm{r}\mathrm{e}}-\mathrm{W}\mathrm{P}{\mathrm{D}}_{\mathrm{h}\mathrm{i}\mathrm{s}\mathrm{t}\mathrm{o}\mathrm{r}\mathrm{i}\mathrm{c}\mathrm{a}\mathrm{l}}}{{\mathrm{W}\mathrm{P}\mathrm{D}}_{\mathrm{h}\mathrm{i}\mathrm{s}\mathrm{t}\mathrm{o}\mathrm{r}\mathrm{i}\mathrm{c}\mathrm{a}\mathrm{l}}}}\times 100\%,$$

(3)

The percentage change is a simple and intuitive measure. It expresses the changes in terms of percentages, which can better illustrate the variation in WPD in the future compared to the historical period.

The Mann–Kendall trend test with Sen’s slope estimator was used to estimate the trends in the seasonal and annual mean wind speeds [38]. The Mann–Kendall trend test is a non-parametric statistical method. It calculates a statistic by comparing all pairwise data points:

$$S={\sum }_{i=1}^{n-1}{\sum }_{j=i+1}^{n}sgn\left(x_j-x_i\right),$$

(4)

where $sgn\left(x\right)$ is the sign function (1 for $x_j$ > $x_i$, −1 for $x_j$ < $x_i$, and 0 for ties). The variance of $S$ is then used to derive the $Z$ statistic:

$$Z=\left\{\begin{array}{cc}{\displaystyle \frac{S-1}{\sqrt{var\left(s\right)}}}& forS>0\\ 0\hfill & forS=0\\ {\displaystyle \frac{S+1}{\sqrt{var\left(s\right)}}}& forS<0\end{array}\right.,$$

(5)

$Z$ is converted into a p-value using the standard normal distribution:

$$p=2\times \left(1-\Phi \left(\left|Z\right|\right)\right),$$

(6)

where $\Phi$ is the cumulative distribution function. A $p$-value < 0.05 indicated that the trend was statistically significant. Complementing the Mann–Kendall trend test, Sen’s slope estimator quantifies the trend magnitude by calculating the median of all pairwise slopes:

$$\begin{array}{cc}\beta =Median\left({\displaystyle \frac{x_j-x_i}{j-i}}\right)& forj>i\end{array},$$

(7)

A positive value of $\beta$ denotes an upward trend whereas a negative value corresponds to a downward trend.

The extreme wind speed was estimated through the Gumbel distribution, which can be obtained from the following equation [39,40]:

$$V(T)=-{\displaystyle \frac{1}{a}}ln[ln({\displaystyle \frac{T}{T-1}})]+b,$$

(8)

where V(T) is the wind speed for the return period of T years, a is the scale parameter, and b is the location parameter. a and b are given by Equations (9) and (10), respectively, which use the average value $\mu$ and root mean square value $\sigma$ of the sample group [41].

$$a={\displaystyle \frac{1}{0.78\sigma }},$$

(9)

$$b=\mu -0.45\sigma ,$$

(10)

3. Results and Discussion

3.1. Evaluation of Sea Surface Wind Data from CMIP6 GCMs

Proper evaluation of CMIP6 wind data is an important pre-requisite for estimating the uncertainties associated with each GCM before creating a multi-model ensemble mean to produce reliable projections of future changes in marginal seas [42]. The evaluation was conducted by utilizing the obs-ERA5 scatterometer wind data in conjunction with historical outputs from the CMIP6 GCMs from 1995 to 2014 to reconcile the data coverage of these two datasets.

Figure 2 illustrates the Taylor’s skill scores for individual CMIP6 models and the multi-model ensemble (MME) of 19 GCM results relative to the obs-ERA5 scatterometer observation data. The MME was constructed using the unweighted average, which has been frequently used by previous research [26,34]. The skill scores were generally > 0.4. The spatial patterns of most models were generally similar, with the higher values (> 0.6) occurring in the northern SCS and relatively lower values in the southeastern SCS. In addition, the skill scores of the models with lower resolutions, such as ACCESS-CM2 and IITM-ESM, were lower than those with high resolutions (e.g., AWI-CM-1-1-MR and BCC-CSM2-MR). These results are consistent with the findings of Jiang et al. [43] and Kumar et al. [44].

The Taylor diagrams for the different regions of the SCS are shown in Figure 3 to illustrate the gaps between the different datasets. The SCS was divided into three part to conduct separate assessments. The obs-ERA5 data were used as the baseline to compare the 19 GCMs from CMIP6. From the scattered distribution of points, we found that some models, such as MPI-ESM1-2-LR, performed poorly in the northern SCS but showed relatively reasonable performance in the central and southern SCS. The correlation coefficients of most of the models ranged between 0.4 and 0.9, with the southern SCS showing a more concentrated distribution with a range of 0.7–0.9. The RMSEs were between approximately 0.5 and 1.5 while the standard deviations ranged from approximately 0.5 to 1.6. Overall, the standard deviations in the central region were smaller than those in the other areas.

To further evaluate the seasonal features of the CMIP6 wind data for the SCS, the MME of 19 GCMs was compared with obs-ERA5 in each season (Figure 4a–h). The WPD was calculated according to Equation (2). The spatial distribution of WPDs from CMIP6 was generally consistent with that of obs-ERA5 across seasons, with a lower WPD in spring and higher WPD in winter. High WPDs appeared in the central and northern SCS in summer and autumn, respectively. Despite their simplicity, bias maps are highly effective for depicting spatial patterns over large domains, offering clear insights into regional variations on a global scale [44]. Bias maps were generated by computing the difference at each grid point (Figure 4i–l). They revealed that, unlike the long-term annual mean, the winter WPD from the CMIP6 MME was somewhat higher than that from obs-ERA5 in the central SCS but lower near the Vietnamese coast and northeastern SCS adjacent to Taiwan Island. During the winter monsoon, the sea surface wind in these areas is largely influenced by the topography, e.g., the mountains in mainland China, Taiwan Island, the Philippine Islands, and Vietnam, and straits like the Taiwan Strait and Luzon Strait (see Figure 1 for the topography) [45,46,47]. Some CMIP6 GCM models generally had low resolutions (Table 1). Thus, the CMIP6 models simulating the wind fields in these topographically sensitive areas showed relatively larger biases in comparison with the observations. Figure 2 also shows that the low-resolution models had low Taylor’s skill scores, which is consistent with the findings of previous studies [40,41]. It should be noted that, for the future climate-change projections, such biases will not change the consistency in the CMIP6 MME data itself.

3.2. Projected Change in Wind Energy Resources Under Future Climatic Scenarios

From a statistical standpoint, the MME demonstrated superiority over any individual model. The projected change in WPD in the SCS was examined using the MME data from 19 GCMs under the climate-change scenarios of SSP1-2.6, SSP2-4.5, and SSP5-8.5. Figure 5 presents the projected change in annual mean WPD. The MME data during the historical period (1995–2014) were used as the baseline. The spatial distribution of WPD shows that the SCS possesses substantial wind energy resources (Figure 5a). The percentage of change showed a strong spatial dependency. By the mid-21st century (2046–2065), the northern SCS is predicted to show an offshore increase in WPD but decrease in coastal areas (Figure 5b–d). A WPD decline is predicted in the southern SCS (3–14 °N) under SSP5-8.5. In the Gulf of Thailand, the WPD will decrease by nearly 10% (Figure 5d). By the late 21st century (2080–2099), the WPD will generally increase in most areas (Figure 5e–g). A large increase in WPD (>15%) is predicted to occur in the area south of 5° N under SSP5-8.5, especially near the Kalimantan Strait (>20%). An increase (>10%) will also occur near the central SCS west of the Philippine Islands. Moreover, the coastal area along the western coast of SCS also show obvious WPD increase under SSP2-4.5 (>15%). The WPD change under scenario SSP1-2.6 is relatively weaker than other scenarios. These features are consistent with previous findings from CMIP6 data either at global scale [10] or offshore in China [27].

The coefficient of variation (CV) can be used to evaluate the standard deviation. Here, we used the CV to measure the variability in WPD in the SCS. Interestingly, the CV of WPD was higher in the northwestern SCS and lower in the southeastern SCS (Figure 6a). Under future scenarios, a decrease in CV is expected in the western and southern SCS by the mid- and late 21st century (Figure 6b–g). The variability in wind generally showed a consistent increase (>10%) along the coastal sea of China, Malaysia, and the Philippines, where wind farms are usually deployed. The increase by the late 21st century will be higher than the increase by the mid-21st century.

The SCS wind energy exhibits strong seasonal variations due to the East Asian monsoon [22,24]. Generally, the WPD reaches its maximum in winter and minimum in spring (Figure 7a–d). In the northern SCS, a strong WPD appeared in both winter and autumn, with a magnitude three times higher than those in spring and summer. In the central SCS, the WPD was strengthened in summer due to the summer monsoon, especially off the Vietnamese coast. The WPD was generally weaker in the southern SCS than in other regions. Figure 7e–p depict the percentage change in WPD in each season by the mid-21st century. Similar to Martinez and Iglesias [10], concentrated increases in wind resources are anticipated in the tropical region. A consistent increase in WPD can be observed in each season, especially under scenarios SSP2-4.5 and SSP5-8.5. The WPD increase in summer will be greater than in the other seasons, especially in the northern SCS and the region adjacent to the Kalimantan strait (south of 6 °N). Both the Gulf of Thailand and region offshore of Malaysia were predicted to show prominent WPD decreases (more than 10%) in summer and autumn.

The observed increase in WPD during winter, spring, and summer by the mid-21st century under the SSP2-4.5 and SSP5-8.5 scenarios is also predicted to occur in the late 21st century (Figure 8). The WPD is expected to decrease over a broader area during winter under the SSP1-2.6 scenario, and during autumn in the SSP5-8.5 scenario. In summary, an increase in WPD exceeding 10% can be generally be anticipated across most regions of the SCS under both the SSP2-4.5 and SSP5-8.5 scenarios in all seasons. However, uncertainties arise within the SSP1-2.6 scenario, where increases and decreases in WPD coexist in each season throughout the century.

3.3. Long-Term Trends in Sea Surface Wind: Seasonal and Regional Dependency

Given the significant spatial variations in the predicted changes in WPD, the data for the SCS was divided into three groups, i.e., northern, central, and southern SCS (see Figure 1), to elucidate the differences in each sub-region’s responses to future climate-change scenarios. The temporal evolution of the sea surface wind speed averaged over the three sub-regions are presented in Figure 9, Figure 10 and Figure 11, respectively. Vigorous interannual variations can be observed in the annual, summer, and winter time series (Figure 9a,c,e, Figure 10a,c,e and Figure 11a,c,e), which is similar to the findings of Wu et al. [26]. The time series for summer wind speed generally showed a consistent increase under three scenarios, especially in the northern and central SCS (Figure 9c and Figure 10c). In the northern and central SCS, the interquartile range is larger during summer (Figure 9b,d and Figure 10b,d) whereas in the southern SCS, the interquartile range is larger during winter (Figure 11b,d). The increasing trend for annual variations in sea surface wind speed is generally small, with the central and southern SCS showing larger increases under SSP5-8.5 and SSP2-4.5, respectively (Figure 9f, Figure 10f and Figure 11f). In the northern SCS (Figure 9f), the increase in summer is about three times higher than that in winter. On the contrary, the increase in the southern SCS is larger in winter than the summer and annual increases under SSP2-4.5 and SSP5-8.5 (Figure 11f). It should be noted that the winter wind speed shows a decreasing trend under SSP1-2.6. Since the trend was calculated using data from 2015 to 2099, the results under SSP1-2.6 show alternating increasing and decreasing trends on the decadal timescale (see Figure 9e, Figure 10e and Figure 11e). Such inconsistencies in the decadal variations resulted a long-term decreasing trend. Despite this, the results under SSP2-4.5 and SSP5-8.5 show consistent increases in winter. The specific values of these increases are listed in

Table 2. Trends in wind speed (×10⁻⁴ m/s/year) in northern, central, and southern SCS by late 21st century under SSP1-2.6, SSP2-4.5, and SSP5-8.5 scenarios. Trends for northern, central, and southern SCS were calculated from the data in Figure 9, Figure 10 and Figure 11.

		NSCS	CSCS	SSCS
	SSP1-2.6	5.33	5.01	1.20
Annual	SSP2-4.5	5.04	7.42	5.01
	SSP5-8.5	5.50	14.09	2.89
	SSP1-2.6	17.14	18.50	8.68
Summer	SSP2-4.5	19.78	22.93	7.21
	SSP5-8.5	44.06	50.86	2.85
	SSP1-2.6	−6.41	−9.10	−4.59
Winter	SSP2-4.5	3.87	7.55	9.03
	SSP5-8.5	15.74	24.51	14.23

. The increasing trends in wind speed in the northern and central SCS are generally larger than those in the southern SCS. The largest trend appears in the northern and central SCS in summer (>40 × 10⁻⁴ m/s/year under SSP5-8.5 and > 17 × 10⁻⁴ m/s/year under SSP1-2.6). The increases in winter are relatively larger under SSP5-8.5 (>24 × 10⁻⁴ m/s/year in central SCS). The annual increase are usually small, except in the central SCS, under SSP5-8.5 (>14 × 10⁻⁴ m/s/year).

3.4. Projected Changes in Extreme Wind Speed with Different Return Periods

Extreme winds threaten the safety of ocean constructions like wind farms. The design parameters that account for EWS loads with different return periods must also consider marine engineering strategies for extreme environmental conditions [48]. After obtaining the annual maximum wind speed, the EWS with a certain return period can be estimated using the Gumbel distribution, which has been used in previous study on the SCS [49,50]. Since the primary focus of this study was the relative changes between the historical period and the future, the EWS during the historical period, mid-21st century, and late 21st century were calculated to examine the percentage change.

Figure 12 presents the changes in EWS by the mid-21st century. The spatial distribution of EWS during the historical period (Figure 12a,e,i) indicate that EWSs mainly occurred in the northern SCS and decreased towards the southwest, matching the occurrence of tropical cyclones [30]. The maximum EWS values were 23.6 for a 25-year return period, 25.5 for a 50-year return period, and 27 for a 100-year return period. By the mid-21st century, the EWS will increase in the northern and central SCS but decrease in the southern SCS under three scenarios. The EWS increase under SSP1-2.6 (Figure 12b,f,j) and SSP5-8.5 (Figure 12d,h,l) is stronger than that under SSP2-4.5 (Figure 12c,g,k). The EWS decrease near Malaysia, Vietnam, and Thailand is 10% for the 100-year return period under three scenarios. The EWS in the inner Gulf of Thailand is expected to increase under the SSP5-8.5 scenario (Figure 12d,h,l).

The changes in EMS by the late 21st century (Figure 13 are similar to those by the mid-21st century, i.e., increase in the northern and central SCS and decrease in the southern SCS under three scenarios. Specifically, the region showing an increase in EMS by the late 21st century is larger under SSP2-4.5 (Figure 13c,g,k) and shows a more consistent spatial pattern under three scenarios. This information reinforces the conclusion that the extreme wind in the northern and central SCS will further increase by 5% under future climate-change scenarios, adding to the extreme load on wind turbines and related marine facilities.

4. Summary and Conclusions

4.1. Summary of Major Findings

This study provides a comprehensive assessment of the wind energy resources in the SCS under different climate-change scenarios, utilizing 19 GCMs from the CMIP6 and three Shared Socioeconomic Pathways (SSP1-2.6, SSP2-4.5, and SSP5-8.5). The key findings of this study can be summarized as follows:

Comparison of the CMIP6 historical data (1995–2014) and scatterometer observations (obs-ERA5) demonstrated that the MME of the CMIP6 GCMs effectively captured the spatiotemporal distribution of WPD in the SCS. There are regional discrepancies in the central SCS, which had lower model skill scores and a higher RMSE, indicating challenges in simulating complex monsoon-driven wind dynamics using a global model.

At the mid-21st century, the WPD increase in summer will be greater than in other seasons, especially in the northern SCS and the region adjacent to the Kalimantan strait. Both the Gulf of Thailand and area offshore of Malaysia will show prominent WPD decreases (more than 10%) in summer and autumn. By the late 21st century, an increase in WPD exceeding 10% can be generally anticipated across most regions of the SCS under both the SSP2-4.5 and SSP5-8.5 scenarios in all seasons.

The time series of summer wind speed shows a consistent increase under three scenarios, especially in the northern and central SCS. The increasing trend in summer (>40 × 10⁻⁴ m/s/year under SSP5-8.5 and >17 × 10⁻⁴ m/s/year under SSP1-2.6) is about three times higher than that in winter. The variability in wind generally shows a consistent increase (>10%) along the coastal sea of China and Malaysia, and this increase will be greater in the late 21st century than the mid-21st century.

The extreme wind in the northern and central SCS will further increase by 5% under future climate-change scenarios, which will add extra extreme loads to wind turbines and related marine facilities.

4.2. Conclusions

The findings of this study provide the general large-scale trends in WPD in the SCS resulting from projected climate-change scenarios, which might be of great value for future wind farm planning and investment. The 5% rise in extreme winds in the northern and central SCS necessitates immediate revision of marine engineering standards to address the predicted heightened structural loads on turbines and support facilities. These changes directly challenge the current design paradigms, requiring reinforced foundations and dynamic operational protocols to mitigate production losses during intensified monsoon events. As offshore wind energy farms are increasingly being built in deeper waters and further offshore, the scientific analysis of the changes in marine wind resources can be used to improve the adaptability and sustainability of wind farm technologies. The identified regions in which the WPD, CV, and EWS are projected to change significantly can support industry risk assessments and offshore engineering financing. Based on the highlighted future wind energy changes in this study, specific areas of concern (e.g., regions with significant increases or decreases and areas affected by stronger extremes) can be targeted to conduct localized downscaling analyses and risk assessments to support detailed engineering and environmental planning.