This study employs a process-based life cycle assessment (LCA) approach to systematically evaluate the life cycle carbon emissions and net carbon mitigation benefits of global onshore wind power systems during the historical (2014-2023) and future (2025-2060) periods [Figure 1]. The system boundary encompasses the entire life cycle from material production, component production, international transportation, construction and installation, operation and maintenance, as well as decommissioning and recycling. The framework adheres to ISO 14040/14044 standards^[33,34]. The functional unit is defined as the life cycle carbon emissions per MW of installed onshore wind power capacity. Carbon mitigation benefits are measured by the greenhouse gas reductions achieved by substituting fossil-fuel-based electricity during the operational phase. The lifespan of wind turbines follows a Weibull distribution^[35], with a mean lifespan of 20 years, as an assumption widely recognized in the wind industry^[36]. This study quantifies four core indicators, namely life cycle carbon emissions, direct carbon mitigation benefits, net carbon mitigation benefits, and regional contributions. Life cycle carbon emissions refer to the emissions associated with wind turbine production, transportation, installation, and related life cycle stages. Direct carbon mitigation benefits represent the emission reductions from substituting non-renewable electricity generation with wind power. Net carbon mitigation benefits are derived by subtracting life cycle carbon emissions from carbon mitigation benefits. Regional contributions reflect the variations in mitigation benefits associated with regional deployment and production patterns.

The Life Cycle Inventory (LCI) is compiled by drawing on data from relevant academic literature, LCA reports published by wind turbine manufacturers, and datasets provided by the Chinese Wind Energy Association and wind energy enterprises^[5,37-40]. Multi-source data are integrated to calibrate the material composition of diverse turbine models. The manufacturing stage includes key material requirements for core components such as towers, nacelles, blades, generators, and gearboxes. These materials encompass steel, aluminum, copper, fiberglass, resin, concrete, and rare earth metals. For future turbine models, material demands are extrapolated to larger units (e.g., 8.5 MW) via power-based linear interpolation to reflect the trend of turbine upscaling. The construction and installation stage covers foundation works, fuel consumption of lifting equipment, and on-site energy consumption. The operation and maintenance stage focuses on lubricants, spare part replacements, and transportation-related energy consumption. The end-of-life stage involves the decommissioned capacity (Outflow), which is tracked in the material flow analysis. However, the greenhouse gas emissions associated with dismantling, transportation, and material recycling are excluded from the system boundary, given limited data availability and the absence of harmonized methods at the global scale. For the transportation stage, domestic transportation is assumed to cover a distance of 200 km and is modeled using road freight by trucks. All inventory parameters are quantified based on mass and energy consumption to ensure a consistent accounting framework across different regions.

Moreover, based on data availability and the Global Change Analysis Model (GCAM) regional classification, we cluster-weighted and averaged activity levels, dynamic emission factors, grid carbon intensities, and wind resource parameters across regions. This harmonization ensures consistent cross-regional comparison of the emission and mitigation estimates. The corresponding regions defined in the GCAM model are further aggregated in this study. The corresponding aggregated regions and their constituent GCAM regions are listed in Table 1^[41].

To reflect the spatiotemporal heterogeneity of carbon emission impacts, this study ensures that emission factors for materials in the LCI are spatiotemporally dynamic. Given the lack of high-precision spatiotemporal emission factors in existing LCA databases, the emission factors are processed and transformed. Region-specific emission factors are sourced from the GaBi Professional and Ecoinvent databases^[42,43]. For regions without exact matches, “Rest of World” (RoW) data are applied. For temporal dynamics, grid emission factors are prioritized, given their high spatiotemporal resolution and ready availability. For key materials with high emission contributions (steel, iron, aluminum, resin, and concrete), we disaggregated their emission factors into electricity-related and non-electricity-related components, as shown in Equation (1) below:

(1) $$ \mathrm{EF}_{p, i, t} = \mathrm{EF}_{\mathrm{p}}^{\mathrm{fix}}+\mathrm{EF}_{\mathrm{p}}^{\text {elec }} \times \mathrm{EF}_{i, t}^{\text {grid }} \\ $$

where EF_p,i,t denotes the emission factor of material p in region i during year t, $$ \mathrm{EF}_{p}^{\text {fix }} \\ $$, represents the fixed emission portion inherent to material p, $$ \mathrm{EF}_{p}^{\text {elec }} \\ $$, is the electricity consumption required to produce one unit of material p, and $$ \mathrm{EF}_{i, t}^{\text {grid }} \ $$, denotes the grid emission factor of region i in year t, calculated by integrating historical emission factor data with projected trends from relevant institutions.

This study is based on historical country-level grid emission factor data provided by Our World in Data^[44]. These data are clustered and averaged using national electricity generation as weights, yielding historical grid emission factors for the regionally aggregated classifications adopted in this study. We applied the percentage reduction in grid emission factors by 2050 from the EnerBlue scenario (reflecting Nationally Determined Contributions (NDCs) and announced pledges), as provided by Enerdata. By integrating actual historical grid emission factors and applying the same reduction rates, we projected future grid emission factors for the corresponding regions, thereby deriving the total grid emission factors across regions^[45].

This study employs the Material Flow Analysis (MFA) method to establish a unified model covering the newly installed capacity, in-service stock, and decommissioned volume of wind power systems^[46]. Since future scenario projections for capacity growth, electricity demand, and generation are derived from the GCAM, the regional classification strictly follows its region system to ensure consistency between historical accounting and future scenario predictions. The in-service capacity S_i,t satisfies the stock conservation relationship in Equation (2):

(2) $$ S_{i, t}=S_{i, t-1}+I_{i, t}-O_{i, t} \\ $$

where S_i,t denotes the in-service capacity in region i during year t; I_i,t denotes the newly installed capacity in region i during year t; O_i,t denotes the decommissioned capacity in region i during year t.

For the future period, we assume that wind turbine lifetime follows a Weibull distribution, which characterizes the retirement probability of turbines installed in a given year in the t-th year of operation. Based on this distribution, the decommissioned capacity is then calculated^[35,47].

(3) $$ O_{i, t}=\sum_{x=1}^{t} I_{i, t-x} \cdot[F(x)-F(x-1)] \\ $$

where the cumulative distribution function F(x) denotes the decommissioning probability at lifespan x years, with shape parameter k = 2 (dimensionless) and scale parameter λ = 20 years (mean lifespan).

(4) $$ F(x)=1-\exp \left(-\left(\frac{x}{\lambda}\right)^{k}\right) \\ $$

Notably, the Weibull distribution models the temporal probability distribution of wind turbine retirement, instead of presuming a uniform fixed retirement age for all turbines. This approach enables the model to capture the variability inherent in turbine service lifetimes.

Due to the absence of systematic and harmonized global statistics on wind turbine production, this study reconstructs historical production activities across regions using a trade-flow-based inversion approach. By integrating import and export records of wind turbine nacelles, towers, generators, and other major components from national customs and regional trade databases, it establishes an interregional trade flow structure. This structure is then used to infer the spatial distribution of the global supply chains across different years^[48-51]. The core principle is that the correspondence between a region’s newly installed wind capacity and its import sources can serve to estimate its production output. Conversely, for regions with domestic production and consumption, the self-supply quantity can be estimated from trade balance identities.

At the turbine level, regional turbine flows are represented in the form of a trade matrix. Based on the mass balance between each region’s new installations (imports plus domestic production) and its outward flows (exports plus domestic production), the implied domestic production volume is resolved. This enables estimation of regional production shares during the historical period. This approach circumvents the need for complete production datasets, offering an analytical basis for reconstructing the global spatial pattern of production when statistical data are incomplete.

To further improve the spatial accuracy of carbon emission attribution for the manufacturing stage, a net-export adjustment is applied to key components such as towers and generators. For each component category, the regional net exports (exports minus imports) are utilized to reallocate production-related carbon emissions across regions. Regions with sustained net exports of a specific component assume a larger share of component-related carbon emissions, whereas net-importing regions receive correspondingly lower allocations. The reallocation process is conducted using component-specific unit emission factors, thereby aligning regional carbon emissions with the actual upstream structure of the supply chains. The final adjusted carbon emissions attributed to regional wind turbine production are expressed as Equation (5):

(5) $$ C_{i}^{\text {adj }}=C_{i}^{\text {unit }}+\sum_{k} \mathrm{NE}_{\mathrm{k}, i} \times E F_{k} \\ $$

where $$ C_{i}^{\text {unit }} \\ $$ denotes the preliminary carbon emissions of region i assigned based on turbine-level production attribution; NE_k,i denotes the net export volume of component k in region i; EF_k denotes the unit emission factor of component k; and $$ C_{i}^{\text {adj }} \\ $$ represents the final carbon emission allocation after net-export adjustment.

This correction enhances the representativeness of regional production emissions, as it accounts for the impact of concentrated component production and large-scale component trade on the spatial distribution of upstream carbon emissions in the global wind power supply chains.

This study adopts the SSP1-2.6 scenario provided by the GCAM model to obtain projections of total electricity demand and wind power generation across global regions over the period 1995-2060. This scenario is consistent with the global target of achieving net-zero greenhouse gas emissions by 2050, offering a unified macro-level scenario framework for defining future wind power deployment scenarios. Based on these projections, regional capacity factors - estimated based on local wind resource endowments - are applied to back-calculate the operating wind power capacity (stock) and annual newly installed capacity (inflow). These variables serve as the core activity inputs for future wind power deployment in the life cycle assessment framework.

To verify the global-scale plausibility of the scenario settings adopted in this study, the projected wind power installation scale is compared with key milestone projections from major international institutions under net-zero-oriented scenarios^[52,53], as summarized in Table 2. The comparison indicates that both the magnitude and growth trajectory of global wind power capacity under the scenario adopted in this study are broadly consistent with those reported in mainstream international net-zero scenarios. This consistency demonstrates that the assumed wind power deployment scale is reasonable and comparable, thereby providing a robust foundation for subsequent scenario analysis.

This study constructs three deployment scenarios and three production scenarios to evaluate carbon emissions and carbon mitigation under alternative “production-deployment” scenarios [Table 3]. Among the deployment scenarios, C1 represents the baseline deployment scenario and is constructed based on the regional allocation provided by the GCAM model. This scenario implies a continuation of historical electricity mix inertia, under which developed economies - with historically high penetration levels - continue to dominate new wind power installations. Accordingly, the newly installed capacity in each region directly adopts the inflow values derived from the GCAM model. C2 represents a balanced deployment scenario, where annual installations are allocated across regions in proportion to their shares of global electricity demand. It represents a more balanced and equity-oriented deployment scenario that decouples future capacity expansion from historical installation patterns. C3 represents a climate-priority deployment scenario, which preferentially allocates new wind capacity to regions with carbon-intensive power systems, particularly those characterized by high grid emission factors. Unlike C1, which serves as a baseline, C2 and C3 are counterfactual scenarios representing different deployment principles. They are designed to explore the theoretical carbon mitigation potential of alternative regional allocation schemes, rather than to serve as realistic transition pathways.

The grid emission factor of region i in year t is expressed as Equation (6):

(6) $$ E F_{i, t}^{\text {grid }}=E F_{\text {tradition }, i, t}^{\text {grid }} \times S_{\text {tradition }, i, t}+E F_{\text {renew }, i, t}^{\text {grid }} \times S_{\text {renew }, i, t} \\ $$

In the baseline scenario, the projected temporal evolution of the total grid emission factor $$ \mathrm{EF}_{i, t}^{\text {grid }} \\ $$ for region i in year t is taken from authoritative scenario outlooks; S_renew,i,t denotes the share of renewable generation in total electricity generation of region i in year t, derived from GCAM-projected renewable generation and electricity demand; and $$ \mathrm{EF}_{\text {tradition }, i, t}^{\text {grid }} \\ $$ is set to zero. Assuming that the emission factor of non-renewable power generation evolves independently with technological progress, $$ \mathrm{EF}_{\text {renew }, i, t}^{\text {grid }} \\ $$ can be back-calculated from the baseline scenario and subsequently used to rank regions in Scenario C3 for climate-efficient allocation of new installations.

In addition, two feasibility constraints are introduced into the dynamic allocation process to ensure that the constructed deployment scenarios remain consistent with realistic power system conditions. (1) To maintain the stability of regional power systems, the share of wind power generation is capped at 70% of total electricity generation^[54,55]. Regional electricity demand and total generation are sourced from GCAM projections. (2) The annual wind power generation associated with each region’s installed capacity cannot exceed its maximum technically available wind resource potential, thereby preventing deployment beyond the physical limits of wind resource. Based on these constraints, this study derives the annual cumulative wind power operating capacity across regions from 2025 to 2060 for each of the three deployment scenarios [Figure 2].

For production scenarios, M1 represents the baseline production scenario, which retains the current global manufacturing structure. It assumes that each region’s production share remains consistent with recent historical patterns, reflecting a continuation of existing industrial distribution and supply-chain inertia. M2 represents a spatially diversified production scenario, in which manufacturing activities expand beyond the existing five major production hubs through increased production shares in Southeast Asia and Africa. This forms a more geographically distributed and resilient global supply chain structure. M3 represents a clean-oriented production scenario with stronger regional concentration. It is designed to explore the potential emission-reduction effect of relocating production to regions with lower manufacturing emission intensities. Under this scenario, approximately 90% of global wind turbine production is allocated to Europe and North America, forming a highly concentrated “clean production” structure aimed at minimizing life cycle manufacturing-related carbon emissions. M2 and M3 are counterfactual scenarios designed to explore the theoretical carbon emission potential of alternative production allocation schemes. For both scenarios, the target production structures are assumed to be realized by 2035, and the dynamic transition from the historically derived 2025 production structure to the 2035 configuration is modeled using linear interpolation. This study derives the annual wind turbine production capacity across regions from 2025 to 2060 for the three production scenarios [Figure 3].

The transportation stage includes three types of logistics routes: maritime shipping, cross-border land transportation, and intra-regional transportation. For intra-regional transportation, we assume an average distance of 200 km from production plants to installation sites and apply standard road-transportation emission factors. Cross-border transportation is modeled using a combination of maritime and land routes. Maritime distances are calculated between major representative ports of each region using the shortest available shipping routes, while land distances represent the connection from the regional port to the representative installation hub^[56]. Emission factors for both maritime and land transportation are sourced from established life cycle databases.

Based on these inputs, we developed an optimization model to determine transportation routes under each combination of production and deployment scenarios. The model assigns annual flows from production regions to installation regions by minimizing transportation distance, thereby ensuring consistent and comparable spatial attribution of transportation carbon emissions. Finally, we calculated transportation emissions for all nine “deployment-production” scenario combinations by integrating the optimized transportation routes with transportation mode-specific emission factors.

Life cycle carbon emissions for each region i in year t are calculated in a matrix framework to capture the temporal and spatial heterogeneity of the global wind power system. The manufacturing-stage emission matrix is expressed as Equation (7):

(7) $$ E_{i, t}^{\text {pro }}=\sum_{p}\left(A_{p, i, t} \times E F_{p, i, t}\right) \\ $$

where A_p,i,t denotes the activity level matrix of material p in region i and year t, and EF_p,i,t denotes the corresponding dynamic emission factor for material p. Construction, operation, maintenance, and decommissioning emissions are formulated in an analogous manner.

Transportation carbon emissions are calculated as Equation (8):

(8) $$ E_{t}^{\text {trans }}=\sum_{i} \sum_{j} D_{i, j, t} \times E F^{\text {trans }} \\ $$

where D_i,j,t represents the optimized transportation distance from production region i to installation region j in year t, and EF^trans is the transportation mode emission factor.

The total life cycle carbon emissions of the wind power system in region i and year t are the sum of all stages as given in

(9) $$ E_{i, t}^{\text {total }}=E_{i, t}^{\text {pro }}+E_{i, t}^{\text {trans }}+E_{i, t}^{\text {inst }}+E_{i, t}^{o m}+E_{i, t}^{d i s} \\ $$

Where $$ E_{i, t}^{\text {inst }} \\ $$, $$ E_{i, t}^{\text {om }} \\ $$, $$ E_{i, t}^{\text {dis }} \\ $$, denote emissions from the installation, operation, and maintenance, and decommissioning or recycling stages, respectively.

Carbon mitigation during the operation stage is estimated as the product of wind power generation and the regional grid emission factors. We assume that grid emission factors are not internally updated with manufacturing relocation and that wind power is fully grid-integrated. The estimated results thus represent the theoretical mitigation potentials. Historical wind generation is obtained from empirical datasets, whereas future generation is derived from projected operating capacity multiplied by region-specific capacity factors. For future wind installations in the period 2025-2060, annual carbon mitigation is calculated as the product of wind power generation from newly installed capacity and the non-renewable grid emission factor of each region. The direct carbon mitigation benefits are outlined in Equation (10):

(10) $$ M_{t}=\sum_{i=1}^{a_{i}}\left(E F_{\text {tradition }, i, t}^{\text {grid }} \times P_{\text {wind }, i, t}\right) $$

where M_t denotes the total mitigation in year t, $$ \mathrm{EF}_{\text {tradition }, i, t}^{\text {grid }} \\ $$ denotes the non-renewable grid emission factor for region i, and P_wind,i,t denotes the annual wind power generation from newly installed capacity in region i.

Moreover, we conducted an uncertainty analysis that explicitly accounts for grid emission factors ( ± 10% baseline), scenario-specific coefficients (reflecting increasing uncertainty in counterfactual scenarios), a temporal factor (capturing growing uncertainty in long-term projections), and a regional data-quality factor. This framework is applied to both carbon emissions and net carbon mitigation benefits, with cumulative uncertainty derived by aggregating annual uncertainties through standard error propagation, where the total variance equals the sum of the annual variances. All reported uncertainties are reported as mean ± 1 standard deviation (σ).

The key assumptions adopted in the modeling framework of this study are categorized and summarized in Table 4. The table encompasses technical parameters, supply chain parameters, system boundary, model settings, and scenario settings, accompanied by corresponding descriptions for each assumption.

During 2014-2023, wind power systems generated cumulative global net carbon mitigation benefits of 3,690 Mt CO₂e, accounting for approximately 1.01% of total global carbon emissions over the same period [Figure 4]. Specifically, the cumulative direct carbon mitigation benefits amounted to 4,035 Mt CO₂e, while the life cycle carbon emissions associated with the wind power industry totaled 345 Mt CO₂e. Thus, total life cycle carbon emissions were equivalent to merely 8.55% of the direct carbon mitigation benefits. This indicates that although the wind power supply chains generate a certain amount of greenhouse gas emissions, these emissions are far smaller than the mitigation benefits achieved during the operational phase. These benefits stem from the substitution of fossil-fuel-based electricity generation. Overall, wind power offers substantial carbon mitigation benefits. Its associated emissions are negligible when compared with conventional generation technologies such as coal- and gas-fired power. Over the past decade, the equipment production stage dominated life cycle carbon emissions, contributing nearly 270 Mt CO₂e and accounting for 78% of total emissions. The construction stage followed, accounting for 18%. In contrast, operation and maintenance as well as transportation contributed only about 2% and 1%, respectively, indicating relatively limited contributions [Figure 4C].

Over the ten-year period, global net carbon mitigation benefits remained positive, showing a steady upward trajectory. Despite regional fluctuations in net mitigation levels, an overall upward trend was evident. Global direct carbon mitigation benefits increased from 220.78 Mt CO₂e in 2014 to 565.76 Mt CO₂e in 2023, a 2.56-fold rise that represents an 11.0% compound annual growth rate. During the same period, direct carbon mitigation benefits increased from 247 Mt CO₂e to 617 Mt CO₂e, a 2.50-fold increase that outpaced the 1.96-fold rise in carbon emissions, which grew from 26 Mt CO₂e in 2014 to 51 Mt CO₂e in 2023. This pattern reflects the rapid expansion of wind power capacity in recent years. As the cumulative stock of operating turbines increased, the growth rate of carbon mitigation outpaced that of emissions generated during turbine production, thereby further amplifying net mitigation effects. Taking China as an example, which has extensive wind power installed capacity and exhibits the most pronounced growth rate, carbon emissions from wind power activities increased by 2.91 times between 2014 and 2023. In comparison, net carbon mitigation benefits increased by 4.02 times. This robust growth propelled China’s share of global net carbon mitigation benefits from 31.9% in 2014 to 50.0% in 2023. Conversely, other countries and regions exhibited more modest growth in scale and installation rates, resulting in either fluctuating or declining shares of global net mitigation. For example, the European Union’s share dropped from 30.5% to 18.4%, and the United States’ share fell from 25.63% to 18.01%.

Among all regions globally, China, the European Union, and the United States emerged as the dominant contributors. Their cumulative net carbon mitigation benefits totaled 1,587 Mt CO₂e, 822 Mt CO₂e, and 764 Mt CO₂e, respectively, accounting for 43.0%, 22.3%, and 20.7% of the global total. Collectively, these three regions contributed 86% of global net mitigation. Their high mitigation levels are closely proportional to their substantial cumulative installed capacities. Wind power development in these regions started earlier and reached more advanced stages. Over the decade, their cumulative installed capacities accounted for 43%, 26%, and 15% of global installed capacities, respectively, amounting to a combined share of 84.0%. Moreover, turbine production for these installations was predominantly undertaken domestically. This made these regions both major production hubs and the primary sources of carbon emissions within the wind power supply chains. During the ten-year period, their turbine production output accounted for 53.2%, 21.4%, and 11.1% of the global total, respectively, with a combined share of 85.7%. As a result, cumulative carbon emissions associated with wind power supply chain activities in these regions accounted for 83.8% of global life cycle carbon emissions. Among the remaining regions, India accounted for 4.72% of global carbon emissions, followed by Australia with 2.49% and Canada with 1.17%. In contrast, regions such as Brazil, Southeast Asia, Central America, Russia, and Eastern Europe exhibited the lowest net mitigation contributions, accounting for less than 1% of the global total. For Southeast Asia, Central America, Russia, and Eastern Europe, their low net mitigation contributions were primarily due to their limited installed capacities. By 2023, these regions accounted for merely 0.88% of global cumulative installed capacities. Brazil’s cumulative installed capacity accounted for 2.87% of the global total, far exceeding its net mitigation share of 0.06%. This extremely low net mitigation benefit stems primarily from the country’s relatively clean power mix. Consequently, wind power yields lower marginal mitigation benefits here than in other regions.

From a refined temporal and regional perspective, the net carbon mitigation benefits in most countries and regions remained positive across all years. Only in the early years did a few countries or regions record negative values. This indicates that, during those periods, the life cycle carbon emissions associated with the local wind power industry exceeded the mitigation benefits from wind turbine operation. Regions such as Brazil, the European Free Trade Association, Indonesia, and parts of southern South America fell into this category. The primary driver of these negative values was that a large proportion of turbines produced in the early period were exported, and their marginal carbon mitigation benefits were not significant. The previously discussed expansion of mitigation effects driven by increased installed capacity has improved the situation in these regions that initially had negative net carbon mitigation benefits. By 2022, all regions worldwide achieved positive annual net carbon mitigation. For instance, Brazil experienced negative net carbon mitigation between 2014 and 2016, with the magnitude gradually decreasing from 0.89 Mt CO₂e to 0.28 Mt CO₂e. It subsequently turned positive, reaching cumulative net carbon mitigation benefits of 2.23 Mt CO₂e by 2023. Moreover, southern South America exhibited the net carbon mitigation benefits of -0.02 Mt CO₂e in 2014. In the following years, the values remained positive and showed an overall increasing trend, reaching 1.39 Mt CO₂e in 2023.

The carbon emissions stemming from wind power systems, along with their carbon mitigation benefits, are not solely determined by installed capacity. Instead, they are shaped by multiple spatiotemporal factors^[5,60]. To capture the impact of these factors on wind power performance across different regions, this study constructs two metrics: carbon mitigation intensity and carbon emission intensity. Carbon mitigation intensity is defined as the amount of carbon mitigation achieved per unit of installed capacity within one year. It is expressed in kg CO₂e/(kW·a) and quantifies the actual mitigation effectiveness of wind power in substituting conventional fossil energy, especially coal-fired power. The spatiotemporal variations of this intensity are mainly determined by regional wind resource endowment and the carbon emission factors of the local grid. As grids gradually decarbonize (e.g., a declining share of coal power and an increased share of other renewables), the carbon mitigation intensity of wind power within the same region dynamically changes. Therefore, its calculation must account for the characteristics of the grid in each year.

In contrast, carbon emission intensity focuses on greenhouse gas emissions along the wind power supply chains. It denotes the cumulative carbon emissions per unit of installed wind capacity (kg CO₂e/kW) across the full life cycle, including turbine production, component transportation, on-site operation and maintenance, and decommissioning. Its magnitude is primarily shaped by production technologies, the electricity mix, and regional industrial features.

Carbon mitigation intensity is primarily determined by regional wind resource endowment and grid cleanliness. Spatially, over the past decade, the highest carbon mitigation intensity was observed in northern Africa in 2014, reaching 1,381.35 kg CO₂e/(kW·a) [Figure 5A]. This is mainly attributed to its abundant wind resources, vast deserts, and open terrain, which deliver stable and strong wind speeds conducive to large-scale wind power development. Additionally, coal power constituted a high share of the electricity mix in this region, and its emission factor was significant. Substituting wind for coal power avoided substantial carbon emissions, thereby greatly enhancing carbon mitigation intensity. Beyond northern Africa, regions such as Australia, South Africa, and the Middle East also exhibited relatively high overall carbon mitigation intensity. In contrast, the European Free Trade Association had relatively low carbon mitigation intensity, with a global minimum of 50.40 kg CO₂e/(kW·a) in 2021. This is because the region’s grid is highly decarbonized, so the carbon reductions from wind substituting coal are relatively limited. Moreover, the average regional wind speed that year might have been suboptimal, resulting in lower overall turbine output and, in turn, reduced carbon mitigation intensity. Similarly, South American countries like Brazil and Colombia had grids dominated by hydropower, leading to high grid cleanliness and relatively low carbon mitigation intensity, with mean values of 105.25 and 90.75 kg CO₂e/(kW·a), respectively.

Temporally, carbon mitigation intensity across most regions worldwide showed a clear decreasing trend due to grid decarbonization from 2014 to 2023 [Figure 5B]. For example, the carbon mitigation intensity of China decreased from 854.72 kg CO₂e/(kW·a) in 2014 to 721.77 kg CO₂e/(kW·a) in 2023, representing a reduction of 16%. This decline primarily stems from China’s active promotion of electricity mix adjustment over recent years. The country has gradually reduced coal’s share while expanding the capacity of other clean energy sources, such as hydro and solar, thereby lowering the grid carbon emission factor. As grid decarbonization accelerates, the relative carbon reductions achieved by substituting wind for coal decrease, leading to a decline in carbon mitigation intensity. However, in certain regions, the downward trend was not significant, with periodic fluctuations or even upward shifts observed. For instance, in eastern Africa, carbon mitigation intensity rose from 242.81 kg CO₂e/(kW·a) in 2014 to 261.50 kg CO₂e/(kW·a) in 2017, then declined to 209.69 kg CO₂e/(kW·a) by 2023. In western Africa, carbon mitigation intensity over the decade showed no clear downward trend, instead increasing from 376 kg CO₂e/(kW·a) in 2014 to 392.31 kg CO₂e/(kW·a) in 2023. This pattern may be attributed to lagging wind power development in these regions. Installed capacity had not increased significantly from 2014 to 2023, whereas electricity demand continued to grow alongside economic development. Coal-fired power remained the dominant source, and grid cleanliness fluctuated. Consequently, the mitigation effect from substituting wind for coal became more pronounced, resulting in a rise in carbon mitigation intensity. The observed decline in carbon mitigation intensity serves as a warning. With continued grid decarbonization, the marginal mitigation benefit per unit of installed capacity will continue to decline. In the long term, relying solely on capacity expansion may not sustainably amplify global mitigation effects.

The spatial distribution of carbon emission intensity exhibited significant regional differences. In global regions, Europe showed relatively low carbon emission intensity overall. For example, in 2014, carbon emission intensities of EU-12, EU-15, and the European Free Trade Association were 304.19, 302.30, and 294.91 kg CO₂e/kW, respectively, well below the global average [Figure 5C]. This is mainly due to Europe’s relatively clean electricity mix, where low-carbon power constitutes a high share of the electricity system, leading to lower emission intensity in turbine production. Furthermore, mature production technology and strict environmental regulations further promote energy efficiency and emission control during the production process. In contrast, emerging production centers such as China and India exhibit relatively high carbon emission intensity. In 2014, China and India reached 481.44 and 499.94 kg CO₂e/kW, respectively, significantly higher than European levels. This disparity mainly stems from the high coal share in the two countries’ electricity systems, which means turbine production relies heavily on high-carbon electricity, thereby elevating carbon emissions per unit of installed capacity.

Temporally, the global average carbon emission intensity exhibited a decreasing trend from 2014 to 2023, declining from approximately 414.19 kg CO₂e/kW in 2014 to 318.51 kg CO₂e/kW in 2023 [Figure 5D]. This indicates that technological progress and electricity mix optimization in turbine production have achieved notable results. In China, carbon emission intensity decreased from 481.44 kg CO₂e/kW in 2014 to 367.64 kg CO₂e/kW in 2023, a reduction of about 24%. The trend toward cleaner production has, to a certain extent, offset the emission pressures from capacity expansion. A brief decomposition shows that turbine size increase was the primary driver behind carbon emission intensity decline in China from 2014 to 2023, contributing 100.94 kg CO₂e/kW, which accounted for approximately 88.70% of the total decline. Meanwhile, production decarbonization contributes about 6.38%, indicating that improvements in the electricity mix during the production stage also play a key role in reducing emission intensity.

Historical results indicate that the net carbon mitigation benefits of global wind power have primarily relied on the direct carbon mitigation benefits of wind power. These effects are influenced not only by installed capacity and electricity generation but also by the dynamically changing regional carbon mitigation intensity. In practice, they highly depend on the spatial alignment between production and deployment stages. Against the backdrop of continued global wind power expansion, different production-deployment scenario choices may further amplify or alleviate this mismatch between emissions and benefits^[60]. This study, grounded in a uniform assumption regarding global total installed capacity, systematically compares the carbon mitigation and emission performance of newly installed wind power systems under different deployment scenarios (C1-C3) and production scenarios (M1-M3). The objective is to reveal the relative contributions of each stage to future mitigation potential and their interaction mechanisms.

To assess the robustness of the relative differences in manufacturing-related carbon emissions across production scenarios, we conduct a one-at-a-time (OAT) sensitivity analysis for non-grid emission factors of key materials^[62]. For key materials with relatively high emission contributions that have undergone electricity-input decoupling, including steel, iron, aluminum, and resin, we impose perturbations to their non-grid emission factors. We preserve the dynamic evolution of the electricity-related component throughout this process. The resulting changes in future manufacturing-related carbon emissions are then quantified. As the responses exhibit an approximately linear relationship across different perturbation magnitudes, we primarily present the benefits for a 10% upward adjustment in the emission factors [Figure 9].

Under the 10% upward setting, steel is identified as the most sensitive material affecting future manufacturing-related carbon emissions, with increases of 5.52%, 5.54%, and 5.81% under Scenarios M1, M2, and M3, respectively. Resin ranks second, with corresponding increases of 1.93%, 2.00%, and 1.98%. The impacts of aluminum and iron are comparatively marginal, with all increases remaining below 1%. Temporally, the sensitivity of each key material exhibits slight variations over the study period, yet remains generally stable. For instance, steel shows a change from 5.08% in 2025 to 5.68%-5.96% in 2060 across the three scenarios. Resin changes from 1.72% to 2.01%-2.10%, while iron and aluminum display even narrower fluctuations. In summary, the effect of the non-grid emission factors of key materials on future manufacturing-related carbon emissions is subject to multiple factors, namely annual production scale, regional production distribution, and the structural evolution of turbine sizes^[18]. Nevertheless, the direction and relative ranking of these effects remain generally consistent. Therefore, the conclusions regarding the relative differences among the various production pathways in this study are reasonably robust.

In addition, to assess the robustness of the deployment scenario rankings among C1, C2, and C3, we analyzed the potential effects of key parameters based on study assumptions and existing literature [Table 5]. The parameters encompass resource conditions (wind capacity factors), grid characteristics (emission factors, decarbonization rates, and displacement factors), technical parameters (turbine lifetime and manufacturing intensity), demand (electricity demand and demand-side feedback), and other factors (grid access, social acceptance, and policy frameworks)^[63-70].

The ranking is primarily driven by the spatial pattern of deployment and the associated grid emission factors. Parameters such as wind capacity factors, regional grid emission factors, grid decarbonization rate, turbine lifetime, and manufacturing carbon intensity affect only the absolute magnitude of mitigation benefits, without changing scenario rankings. Wind capacity factors exhibit region-specific uncertainty due to natural climate variability^[64]. Their spatial distribution is also subject to long-term shifts driven by climate change. However, in this study, capacity factor scales all scenarios proportionally. Even a ± 10% deviation in capacity factors would not alter the relative ranking of the three deployment scenarios, as identical datasets and assumptions are applied consistently across all scenarios. Our uncertainty analysis confirms that the relative ranking of the three scenarios is stable even if regional grid emission factor disparities narrow, with cumulative mitigation estimates (mean ± 1σ) by 2060 of 23,198 ± 167 Mt CO₂e (C1), 25,198 ± 239 Mt CO₂e (C2), and 53,711 ± 935 Mt CO₂e (C3). Turbine lifetime and manufacturing emission intensity also affect the absolute magnitude of life cycle carbon emissions and carbon mitigation benefits, but do not alter the relative ranking of the three deployment scenarios. Nevertheless, carbon mitigation benefits are sensitive to electricity demand changes. It would also benefit from a positive demand-side feedback loop from capacity expansion. In this loop, greater installed capacity lowers electricity prices, which in turn stimulates demand and supports further capacity expansion^[57]. Despite these sensitivities, the comparison remains meaningful because all deployment scenarios are evaluated within a consistent framework. The relative ranking offers relatively robust insights into the carbon mitigation benefits of alternative deployment pathways.