Offshore wind farms in the northeastern United States play a vital role in the region's transition to renewable energy. By converting the kinetic energy of wind to electricity, these large-scale installations of wind turbines in coastal ocean environments are also known to alter near-surface wind patterns. The extent to which wind turbines modify near-surface meteorology, surface winds, and atmospheric boundary layer profiles has been the subject of extensive investigations in recent years (e.g., Akhtar et al., 2021Akhtar et al., , 2022;;Bodini et al., 2019Bodini et al., , 2021;;Fischereit, Brown, et al., 2022;Golbazi et al., 2022;Quint et al., 2024Quint et al., , 2025;;Raghukumar et al., 2022;Rosencrans et al., 2024). For example, a regional study by Quint et al. (2024) found that wind farms significantly reduce wind speeds both at hub height and at 10 m, and the magnitudes of these responses are strongly dependent on prevailing atmospheric stability conditions.
Because of the strongly coupled nature of the surface waves and surface winds, these changes in wind fields are expected to modify surface wave characteristics, including energy, mean and peak frequency, and direction (e.g., Bärfuss et al., 2021;Christensen et al., 2014;Fischereit, Brown, et al., 2022;Larsén et al., 2024;McCombs et al., 2014;L. Wu et al., 2020). In the North Sea, previous studies using in situ observations and coupled modeling experiments (Bärfuss et al., 2021;Larsén et al., 2024) have shown that wind wakes reduce significant wave height and wave energy downstream of the wind farm. Evidence exists that these changes also impact nearsurface ocean turbulent kinetic energy (TKE), upper-ocean mixing and stratification, and sea surface temperature, particularly in seasonally stratified shelf regions (Benkort et al., 2024;Broström, 2008;Christiansen, Daewel, Djath, & Schrum, 2022;Christiansen, Daewel, & Schrum, 2022;Daewel et al., 2022;Dorrell et al., 2022;Floeter et al., 2022;Paskyabi & Fer, 2012).
Offshore wind farms may also affect wave conditions through interaction involving turbine foundations and wake-induced ocean currents (e.g., Alari & Raudsepp, 2012;Carpenter et al., 2016;Christiansen et al., 2023;Hosseini et al., 2025;Ludewig, 2015;Ponce de León et al., 2011;Schultze et al., 2020;Zhao et al., 2024). However, such mechanisms are not considered in this study, which focuses specifically on wind wake-induced effects on surface waves.
Understanding the influence of offshore wind farms on ocean surface waves via the wind wake effect is important from the air-sea interaction perspective, as momentum exchange across the atmospheric and oceanic boundary layers is primarily supported by strongly coupled wind waves with wavelengths on the order of O(0.1-10 m) (Edson et al., 2013;Sullivan et al., 2014Sullivan et al., , 2018)). In numerical models, air-sea momentum exchanges are generally represented through the roughness length parameterization, which depends either solely on the surface wind speed or a combination of wave characteristics such as wave age, wave slope, and wave direction (e.g., Edson et al., 2013;Sauvage et al., 2023;Taylor & Yelland, 2001). Thus, by altering near-surface wind fields and ocean surface wave characteristics (Fischereit, Larsén, & Hahmann, 2022;Golbazi & Archer, 2019;Larsén et al., 2024), offshore wind farms may modulate the air-sea momentum flux. Furthermore, wave-modulated surface stress can influence the kinematic and thermodynamic structure of the marine atmospheric boundary layer, with implications for turbine wake development and power generated by the wind farms (AlSam et al., 2015;Deskos et al., 2021;Kalvig et al., 2014;Ning & Paskyabi, 2024;Ning et al., 2023;Porchetta, Muñoz-Esparza, et al., 2021;Porchetta, Temel, et al., 2021;L. Wu et al., 2020;C. Wu et al., 2022;Yang et al., 2022). However, the extent to which the wake-induced surface wave changes feed back to the atmospheric boundary layer and power output is not well understood.
Thus, this study investigates the surface wave response to a realistic scenario of large-scale offshore wind farm development in the northeastern U.S continental shelf from the viewpoint of air-sea momentum exchanges. Our analysis based on high-resolution fully coupled ocean-atmosphere-wave model simulations focuses on the boreal summer season. We focus on this season because it features relatively stable atmospheric conditions and pronounced turbine wake effects (e.g., Bodini et al., 2021;Quint et al., 2024), conditions that facilitate a clearer assessment of ocean wave response to the wind farm wake effect.
The high-resolution coupled modeling framework is described in Section 2. Section 3 presents the main results based on simulations conducted with and without wind farms. In Section 3.1, we examine how reduced nearsurface wind speeds associated with wind farms lead to decreases in wind-sea energy and wind stress. Section 3.2 demonstrates that suppressed wind-wave growth causes a spectral shift toward longer-period swell waves. In Section 3.3, we find that this shift to aerodynamically smoother, swell-dominated conditions reduces wave-supported roughness length and momentum flux. However, we also show that this effect is secondary to the reduction in wind stress directly caused by wind speed deficits in turbine wakes. A summary of these findings is provided in Section 4.
This study utilizes the Scripps Coupled Ocean-Atmosphere Regional (SCOAR) modeling system (Sauvage et al., 2023(Sauvage et al., , 2024;;Seo et al., 2007), which integrates the Weather Research and Forecasting (WRF) model (Skamarock et al., 2021) for the atmosphere with the Regional Ocean Modeling System (ROMS) (Haidvogel et al., 2000;Shchepetkin & McWilliams, 2005) for the ocean and the third-generation spectral wave model WaveWatch III (WW3) (Tolman et al., 2002) for surface waves.
In this study, WRF employs a one-way nesting approach, with horizontal grid spacing refined from 7.5 km in the outer domain to 1.5 km in the nested domain (Figure 1). The vertical grid consists of 50 levels, with 20 levels below 250 m. The lowest level of the model layer is centered at 5.5 m, and the second level at 12 m. The outer WRF domain dynamically downscales the ERA5 reanalysis (Hersbach et al., 2020) and is run with spectral nudging to retain the low-wave number characteristics of the downscaled circulation close to those in ERA5. In the inner domain, which is not nudged, atmospheric processes are resolved at 1.5 km resolution. WRF in the nested domain is coupled with ROMS and WW3. ROMS is driven by the daily 1/12°MERCATOR International global reanalysis (Lellouche et al., 2018) and WW3 by 14 spectral points obtained from the global 1/2°WW3 simulations (Rascle & Ardhuin, 2013). These three models share identical grids and land-sea masks and are coupled hourly. A detailed description of the physical parameterizations and schemes used in the model is provided in Table 1.
Air-sea momentum, heat, and freshwater fluxes are computed using the COARE wave-based bulk flux algorithm (Edson et al., 2013;Fairall et al., 1996Fairall et al., , 2003)), integrated within the Mellor-Yamada-Nakanishi-Niino (MYNN) surface layer scheme (Olson et al., 2021). In addition to standard meteorological outputs from WRF, sea surface temperature, and surface currents from ROMS, the COARE wave-based formulation utilizes peak wave phase speed (c p ) and significant wave height (H s ) from WW3 to estimate the surface roughness length (z 0 ) (Edson et al., 2013), given by
where u * is the friction velocity and ν the kinematic viscosity. At the wind speed ranges considered in this study (Figures 2 and 3a), z 0 is dominated by the first term, representing the roughness elements driven by the wind stress in the form of surface gravity waves. As wind farms modify sea state characteristics such as c p and H s by lowering wind speeds through wake effects, we examine the corresponding changes in the parameterized z 0 and consequently in the drag coefficient (C D ), wave-supported stress (τ aw ) and the total momentum flux (τ). In addition to Journal of Geophysical Research: Oceans 10.1029/2025JC023156
spectrally averaged wave outputs, full directional wave spectra are saved every 15 km across the domain. These outputs are used to illustrate the spatial distribution and fetch dependence of the spectral changes in wave fields.
The effects of wind turbine wakes in the WRF nested domain are represented using the Fitch parameterization (Fitch et al., 2012) as implemented in the MYNN planetary boundary layer (PBL) scheme (see Section 2.2). A series of coupled and complementary uncoupled model simulations is conducted. First, two fully coupled oceanatmosphere-wave model simulations are performed: a control run without the wind farm parameterization (NOWF, unperturbed case) and a simulation incorporating the wind farm parameterization (WF). The lease areas (i.e., the areas leased by wind farm developers) are shown in Figure 1. In addition, a set of WRF-only simulations (WF_NWC) is carried out to isolate the effect of sea state modifications on momentum fluxes. All simulations span five boreal summer seasons (June, July, and August, JJA) for 2017-2021.
Table 1 Model Physics Model Parameter Physics details WRF Time step 20 s, 4 s Horizontally 7.5 km and 1.5 km Vertically 50 levels, with 20 below 250 m. 1 st model level is at 5.5 m and 2 nd is at 12 m Outer domain only Forcings Hourly 0.25• ERA5 (Hersbach et al., 2020) Spectral nudging Above the PBL to zonal and meridional wavelengths longer than 850 km and 730 km Deep convection Kain-Fritsch (Kain, 2004), outer domain only Cloud microphysics Thompson (Thompson et al., 2008) Land surface NOAH-LSM (Tewari et al., 2004) Radiations RRTM (Iacono et al., 2008) PBL MYNN 2.5 (Nakanishi & Niino, 2009) Surface layer MYNN (Olson et al., 2019(Olson et al., , 2021) ) Wind Source term ST4 package (Ardhuin et al., 2009(Ardhuin et al., , 2010) ) Wave reflection Enables reflection of shorelines (Ardhuin & Roland, 2012) Depth-induced breaking Battjes-Janssen model (Battjes & Janssen, 1978) Bottom friction SHOWEX bottom friction formulation (Ardhuin et al., 2003)
Wind turbine wake effects are represented using the Fitch parameterization (Fitch et al., 2012), incorporating the updates recommended by Archer et al. (2020). In this scheme, wind turbines act as momentum sinks and sources Journal of Geophysical Research: Oceans 10.1029/2025JC023156
of TKE at the model levels intersecting the rotor-swept area. The TKE generated by turbines is not computed directly within the Fitch scheme but is prescribed to the scheme as a parameter, α (Archer et al., 2020).
Previous studies have shown that simulated wake characteristics are sensitive to the choice of α, which determines the fraction of turbine-generated TKE added to the model (Archer et al., 2020;Bodini et al., 2021;Fischereit, Brown, et al., 2022;Larsén & Fischereit, 2021;Rosencrans et al., 2024). The default value α = 0.25 assumes that 25% of the turbine-produced TKE is injected into the atmosphere; however, several studies have demonstrated that model sensitivity varies across different environments and conditions (Ali et al., 2023;Larsén et al., 2024).
To evaluate the robustness of the simulated wave response to this parameter, two configurations are tested: α = 0.25 (WF) and α = 1.0 (WF_100).
Turbine characteristics, including dimensions, power and thrust coefficients, and layout configurations, follow those described by Rosencrans et al. (2024). A total of 1,418 turbines are distributed across the Mid-Atlantic Bight lease areas (Figure 1). Each turbine has a rated capacity of 12 MW, a hub height of 138 m, and a rotor diameter of 215 m, corresponding to blade tip heights of 30.5 m (bottom) and 245.5 m (top). In WRF, 15 vertical model levels intersect the rotor-swept region, with an additional 4 levels below the lowest blade tip. This fine vertical resolution is essential for accurately resolving turbine wake effects within the atmospheric boundary layer (Tomaszewski & Lundquist, 2020).
Wave spectrum partitioning is performed offline using an algorithm available through the "Wavespectra" open source Python library and is based on the method used in WW3 model (The WAVEWATCH III Development Group, 2019;Tracy et al., 2007). Wind-sea and swell components are defined using a topographic partitioning method, which groups spectral bins into distinct wave systems. These systems are then classified according to a wind-sea fraction cutoff (Hanson & Phillips, 2001;Hanson et al., 2009;Portilla et al., 2009;Vincent & Soille, 1991). First, the spectrum is divided into different partitions representing energy from subpeaks within the spectral distribution. Each partition is characterized by parameters such as wave height, peak period, peak wavelength, and wind-sea fraction. The influence of wind is then quantitatively determined from the wave-age relationship, defined as
where c ws is the phase speed of the wind sea. The wind-sea region is then defined as the component of the wind speed (U 10 ) in the wave direction, multiplied by the wave age factor χ fac , with θ ws being the angle between the wind direction and that of the wind sea. Thus, any wave components traveling at a phase speed slower than c ws are considered to be wind-forced and initially classified as wind sea. Then, topographic partitions in which the proportion of wind-sea energy exceeds a defined threshold (W cut ) are grouped and assigned to the wind-sea component. The value of W cut enables classification of each partition as pure wind-sea (W cut = 1), pure swell (W cut = 0), or mixed sea states. Remaining partitions are categorized as swell components in order of decreasing wave height.
In this study, we define only one wind-sea and one main swell partition, using the standard values of W cut = 0.33 and χ fac = 1.7. While the results of the partitioning process are sensitive to the selection of W cut and χ fac , the overall conclusions of this study remain robust under reasonable variations of these parameters.
A two-sided Student's t test was employed to evaluate the statistical significance of time-mean differences between simulations. The effective sample size, nʹ, accounting for temporal autocorrelation within each time series, is computed following Bretherton et al. (1999). Since the time series is segmented by year, nʹ is computed separately for each summer and then summed to obtain a final nʹ for the entire period. Statistical significance is assigned only when the final p value indicates significance at the 99% confidence level (p < 0.01) and is represented as stippling on the figures, plotted every 6th grid point.
10.1029/2025JC023156
In order to validate simulated surface wind and wave summer characteristics, we use several National Data Buoy Center (NDBC) buoys located off the New England coast (Figure 1). These buoys provide near-surface meteorological and oceanographic measurements, including wave energy spectra and near-surface wind speeds. The in situ observations are compared in Figure 2 with JJA 2017-2021 time-averaged results from the unperturbed simulation (NOWF). Observed surface wind speeds from NDBC buoys are adjusted to the standard height of 10 m using the COARE3.5 bulk flux algorithm (Edson et al., 2013).
During summer, the Northeast U.S. experiences predominantly southwesterly winds, with moderate near-surface wind speeds of 5-6 ms -1 (Figure 2). The long-fetch conditions in the southwest direction allow the wind waves to develop before reaching the Massachusetts/Rhode Island (MA/RI) and New Jersey (NJ) lease areas, resulting in peak periods of approximately 6-8 s (Figure 2). In contrast, waves from the north-northwest are almost nonexistent during summer due to the limited fetch in that direction.
The directional wave spectra in Figure 2 also indicate greater wave energy coming from the south to southeast direction. This is consistent with the longer fetch in that direction, allowing for waves to grow and carry a larger amount of energy. These southeasterly waves typically have longer periods (8-10 s) and are decoupled from the prevailing southwesterly winds. The unperturbed simulation (NOWF) reproduces these observed seasonal patterns reasonably well, capturing both the surface wind characteristics and the general sea state seen in moored observations (Figures 2,3a,and 3b).
With wind farms (WF) in place, near-surface wind speed (U 10 ) decreases by 10%, or about 0.5 m s -1 of the unperturbed values (NOWF), most strongly around the lease areas (Figure 4a). The time-mean surface wave response is spatially well aligned with the wake patterns (Figure 4). As U 10 weakens, H s decreases by 3% or about 3 cm of the unperturbed value, while τ aw drops by almost 30% (1.10 -2 N m -2 ) (Figures 4b and 4c). Furthermore, c p increases by 3% (0.3 m s -1 ) within the wind farm and wake regions (Figure 4d). This suggests that reduced surface winds suppress local wind-wave growth, which promotes the occurrences of longer-period waves. This phenomenon is explored further in subsequent sections.
It is noteworthy that the largest c p differences occur in Nantucket Sound, a shallow (10-15 m) region characterized by strong tidal forcing. The local wave field there is further modulated by island sheltering effect, particularly under prevailing southwesterly winds, while weaker modulation is observed under other wind conditions (see also Figure 6).
In response to the decrease in τ aw , z 0 is also decreased up to 20%
(1 × 10 -2 mm) (Figure 4e). Finally, the near-surface ocean TKE decreases by more than 20% (0.2 × 10 -3 m 2 s -2 ), suggesting further changes in breaking waves, turbulence, and mixing in the near-surface ocean (Figure 4f).
The reduction in near-surface wind speeds due to wind farms is consistent with previous studies (i.e., Golbazi et al., 2022;Larsén et al., 2024;Quint et al., 2024;Rosencrans et al., 2024). These studies also suggested that the magnitude of wind speed reductions varies with atmospheric stability and the α parameter in the Fitch scheme (i.e., Archer et al., 2020;Bodini et al., 2021).
To assess the sensitivity of our results to these two factors, we compare simulations with different α and classify the prevailing atmospheric stability conditions using the hourly Monin-Obukhov length (Monin & Obukhov, 1954), following the approach of Rosencrans et al. (2024).
Our analysis shows that stably stratified conditions occur most frequently (59.3%) in the unperturbed case (NOWF) (Figure 5), typically associated with winds having a westerly component. Conversely, unstably stratified conditions occur 35.6% of the time in association with easterly winds. These results are similar in the MA/RI and NJ lease areas.
Figures 6 and 7 compare the near-surface wind speed and wave responses under stable and unstable atmospheric conditions. Under stable conditions, wake-induced wind reductions extend farther downstream (Quint et al., 2024), leading to decreases in H s and τ aw and an increase in c p . In contrast, under unstable conditions, wind deficits are largely confined within the lease areas, with minimal or no downstream wake effects. Similar patterns of changes are found with H s , τ aw , and c p .
To quantitatively assess the changes in wave wakes under different atmospheric stability conditions, we define the wave wake extent as the area where H s deficit exceeds 1.5%. This threshold is a tentatively chosen value to produce a wave wake scale comparable to that obtained using the commonly used 0.5 m s -1 wind speed deficit criterion used in wind wake studies. The relative differences remain essentially unchanged with different thresholds. At the New Jersey site, the wave wake during the stable conditions covers an area of approximately Journal of Geophysical Research: Oceans 10.1029/2025JC023156 SAUVAGE ET AL. 9 of 18 21699291, 2025, 12, Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2025JC023156 by Cesar Sauvage -Naval School , Wiley Online Library on [23/12/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
6,874 km 2 , which decreases by a factor of 2.3-2,959 km 2 under unstable conditions (Figure 7). The larger extent of the wake during stable conditions is mainly associated with southwesterly conditions, whereas under unstable conditions, easterly to northeasterly winds dominate, and the coastline confines the wake extension (Figure 7).
At the MA/RI site, the opposite pattern occurs: the wave wake area increases from 7,148 km 2 in stable conditions to 9,988 km 2 in unstable conditions, about 1.4 times larger (Figure 6). This contrast arises because, under stable conditions, the southwesterly waves are partially blocked by Martha's Vineyard and Nantucket islands, limiting wake extension and development (Figure 6).
Atmospheric stability influences the spatial patterns of wave and sea state responses. However, the signs of these responses remain consistent regardless of stability. Both the all-stability composites (Figure 4) and the stabilitybased conditional composites (Figures 6 and 7) consistently show decreased H s and τ aw and increased c p near the wind farms.
Regarding the effect of α value, results from WF_100 (α = 1.0) (Figure 8) show considerably weaker responses in near-surface wind and wave variables than WF (α = 0.25). In WF_100, changes in U 10 , H s , and τ aw are minimal, or in some cases, slight increases are observed within the lease areas, especially under stable conditions. These findings are consistent with prior modeling studies of near-surface atmospheric and wave fields (Bodini et al., 2019;Larsén et al., 2024;Quint et al., 2024). Overall, the comparison between simulations across different α values shows consistent results. Therefore, for the remainder of this paper, we use the all-stability composites with α = 0.25 as the baseline for analyzing changes in the spectral characteristics of surface waves.
Using the output of directional wave spectra across the model domain, this section examines the changes in spectral characteristics and spatial distributions of the surface waves. Figures 9a and 9d show maps of the total wave energy density spectra for the unperturbed case (NOWF) and difference (WF-NOWF), respectively. These data are sampled over the shelf region, with water depths between 15 and 500 m. Wave energy spectra averaged over the MA/RI lease area are also displayed as enlarged directional wave spectra in the lower right corner of each panel. As in Figure 4, the stippling highlights the statistically significant differences.
This figure focuses on southwesterly wind conditions, which generate the bulk of the wind-wave energy in the southerly to southwesterly sectors (Figure 9a). Weak southeasterly waves are also present, but since these waves Journal of Geophysical Research: Oceans In response to wake-induced reductions in southwesterly U 10 , wave energy decreases are most concentrated in the southwesterly sector, particularly at wave periods of around 5 s. In contrast, the southeasterly swell remains unaffected by the reduction in southwesterly U 10 (Figure 9d).
To assess wave energy spectra responses as a function of fetch, three coast-parallel areas are selected and colorcoded: the NJ lease area (green), the immediate wake region (blue), and the downstream and longer fetch region just southwest of the MA/RI lease areas (red). For each area, 1-D energy spectra in frequency space are calculated for NOWF (solid lines) and WF (dashed lines) in Figure 10a, and their differences (WF-NOWF) are shown in Figure 10d.
In all three regions, the total wave energy spectra peak at the swell frequency (7 s, Figure 10a), while the largest differences in wave energy between WF and NOWF occur at shorter periods. In the NJ lease area (green) and the immediate wake region (blue), wave energy reductions peak at 4.9-5.3 s, corresponding to wave energy decreases of approximately 2.1 × 10 -2 m 2 Hz -1 (7%) and 2.6 × 10 -2 m 2 Hz -1 (6%), respectively (Figures 9d and 10d). As the distance from the wind farm increases northeastward, in the red region with longer fetch, the spectral differences shift toward lower frequencies, peaking near 6 s and representing a smaller reduction of about 1.4 × 10 -2 m 2 Hz -1 (2%) in wave energy (Figures 9d and 10d).
Using topographic partitions and a wave age cutoff (see Section 2.3), the total and difference in directional wave spectra are separated into wind-sea (Figures 9b and 9e) and swell (Figures 9c and 9f) components. Corresponding 1-D energy spectra are shown in Figures 10b, 10c, 10e, and 10f. Under dominant southwesterly wind conditions (Figures 9b and 10b), wind-sea energy peaks just below 6 s for all regions. Wind-sea energy is the lowest at the NJ lease area (green) compared to the downstream locations. This suggests the presence of fully developed waves in that region, as wave growth likely occurs upstream of the NJ lease area under southwesterly winds.
Under northeasterly winds, wind-sea energy and period increase with fetch, extending from the MA/RI lease area downstream (Figures S1 and S2 in Supporting Information S1). This behavior is more consistent with the wave age-dependent fetch laws, which describe how wind-seas grow in the direction of the prevailing wind (i.e., Hasselmann et al., 1973;Kahma & Calkoen, 1992;Young, 1997). Journal of Geophysical Research: Oceans 10.1029/2025JC023156
The wind-sea component, typically consisting of shorter waves aligned with the prevailing wind, shows the most significant reduction in energy density within the wind farm region (green) with a 3.1 × 10 -2 m 2 Hz -1 (13%) decrease (Figures 9e and 10e). Smaller reductions are observed in the immediate wake region (blue), with 2.4 × 10 -2 m 2 Hz -1 (8%) and 1x10 -2 m 2 Hz -1 (3%) for region with longer fetch (red). These results are consistent with previous findings that wind farms reduce wave energy and that the reduction is fetch-dependent (Bärfuss et al., 2021;Larsén et al., 2024). Similar wind-sea energy decrease is found in the case of northeasterly winds (Figures S1 and S2 in Supporting Information S1).
The swell component exhibits broader directional spreading, ranging from southwest to southeast (Figure 9c). Swell wave energy originating from directions not aligned with prevailing wind remains unchanged (Figure 9f), as these waves are likely generated outside the model domain and are unaffected by wind speed reductions induced by the wind farms.
In contrast, swell wave energy aligned with the prevailing wind increases within the NJ and MA/RI lease areas (Figures 9f and 10f). This increase corresponds to a local reduction in wind-sea energy (green box, Figures 9e, 9f, 10e, and 10f). The shift in wave energy to lower frequencies occurs because wind waves generated upstream of the NJ and MA/RI lease areas become "swell-like," as they enter regions where near-surface wind speeds decrease due to the presence of wind farm wakes (Figure 4). This results in a transition in sea state from a fully developed wind-sea regime to a swell-dominated regime. The transition is most pronounced under southerly wind conditions, where longer fetch allows wind seas to develop upstream of the wind farms. It is less evident under fetch-limited conditions, such as northerly to westerly winds (not shown). The shift toward longer-period waves is consistent with the observed increase in peak phase speed (c p ) (Figure 4d). Journal of Geophysical Research: Oceans 10.1029/2025JC023156
Overall, wind farms induce broad spectral modifications to sea states across varying fetches. Most of the total wave energy reductions occur in the wind-sea component, with a partially compensating increase in swell energy near the lease areas, particularly when the swell is in the direction of the prevailing wind. As fetch increases, short wind waves continue to weaken within wind wakes but without a corresponding increase in swell energy. This leads to the maximum wave energy reductions in the immediate wake region of the NJ lease area (blue, Figure 10e).
In WW3, we used the source term 4 (ST4, Ardhuin et al., 2010) package, which represents energy transfer from wind to waves, leading to wave growth. Wind input in ST4 depends on the wind-wave growth rate, γ, which, according to the quasi-linear theory of wind-wave generation (Janssen, 1991), scales with the squared inverse wave age, expressed as
Here, ϵ denotes the air-sea density ratio, ω the angular frequency, c the phase speed, θ the wave propagation angle, and the Miles parameter β, which also depends on the inverse wave age.
Since wind wakes reduce τ aw and consequently u * while increasing c p , these changes collectively result in a lower inverse wave age, thereby reducing wind input and the wind-wave growth rate.
It is worth noting that this wind-wave growth parameterization assumes a logarithmic velocity profile in the atmospheric boundary layer. While wind farms can disrupt this profile, the parameterization may remain applicable if the boundary layer quickly adjusts to turbine wakes or if deviations occur above the critical layer heights. A complete assessment of these effects on wind-wave growth mechanisms falls beyond the focus of this study but may be valuable for future investigation.
The decrease in inverse wave age and H s has implications for air-sea momentum exchange and surface drag. This is because many modern bulk air-sea flux parameterizations define roughness length as a function of inverse wave age and/or wave slope (e.g., Drennan et al., 2005;Edson et al., 2013;Sauvage et al., 2023;Taylor & Yelland, 2001). In this study, we employ the wave age and wave slope-dependent formulation of Edson et al. (2013), as described in Equation 1.
This section quantifies the changes in z 0 , C d , and τ caused by the reductions in inverse wave age and H s reported in the previous section using additional WRF-only simulations (WF_NWC). In WF_NWC, the wind farm parameterization is switched on, but sea state is prescribed from NOWF simulations; hence, the sea state input to Equation 1 has been unaffected by wind wakes and corresponds to unperturbed states. Sea surface temperature and ocean surface currents are identical between WF_NWC and WF. Therefore, any difference in the atmospheric outputs can be attributed solely to sea state changes via c p and H s (Figures 4b and 4d), independent of bulk atmospheric variables.
Figure 11 shows the average percentage differences between WF and WF_NWC in z 0 , C D , τ, and U 10 . Compared to the 20% decrease in z 0 shown previously in Figure 4e, Figure 11a shows a 5 %-8% decrease in z 0 . This suggests that roughly one-third of the z 0 reductions shown in Figure 4e is attributable to changes in c p and H s , while the remaining two-thirds result from the decrease in friction velocity (u * ) . These wave-induced changes in z 0 lead to modest reductions in C D and τ, on the order of 1%-2% (Figures 11b and 11c). A collocated, but even smaller increase in U 10 is also observed, averaging less than 0.5% (Figure 11d).
The results shown in Figure 11 are consistent with the sea state changes shown in Figures 4b and 4d. Overall, the magnitude of changes is small, except for z 0 . These results indicate that the influence of sea state changes on the drag coefficient and wind stress is secondary to the dominant effect of reduced friction velocity due to lower wind speeds.
Journal of Geophysical Research: Oceans 10.1029/2025JC023156
This study investigates the response of sea state to offshore wind farms in the coastal ocean of the northeast U.S. using fully coupled ocean-atmosphere-wave model simulations with and without wind farm wake parameterization. Consistent with previous studies (i.e., Bodini et al., 2021;Larsén et al., 2024;Quint et al., 2024), our results show that turbine-induced wind wakes cause a significant reduction in near-surface wind speeds near the wind farms (approximately 10%). This wind speed reduction leads to a corresponding decrease in significant wave height (by about 3%) and wave-supported stress (by approximately 30%).
Consistent with Quint et al. (2024), we find that wake-induced wind reductions extend further downstream under stable atmospheric conditions, whereas under unstable conditions, wind deficits remain largely confined to the lease areas, with minimal downstream propagation. Although the magnitude, and in some cases even the sign, of the response may vary under specific conditions (e.g., stable and α = 1.0), the overall comparison across simulations with different α values yields consistent results. This suggests that the simulated changes in near-surface wind and surface wave fields (Figure 4) are robust to the choice of α.
Spectral analysis reveals a reduction in the wind-sea component of wave energy within the wake regions, indicating suppressed local wind-wave growth. This reduction in short wind waves leads to lower surface roughness length, wind stress, and ocean surface turbulent kinetic energy, likely diminishing near-surface ocean mixing.
Additionally, the reduced surface wind speed and wind-to-wave energy result in an average increase of approximately 3% in the peak wave phase speed near the wind farms, indicating a shift toward higher wave age and an older sea state. This shift occurs because fully developed waves generated upstream of the wind farms begin to outrun the locally weakened winds within the lease areas. The transition toward a swell-dominated regime is most pronounced under southwesterly winds, which support longer fetch and sustained wave growth, and is less evident under northerly to westerly winds, where fetch is limited.
We demonstrate that wind farms reduce wind speed, and consequently, wind stress and friction velocity, while increasing peak wave speed, resulting in a lower inverse wave age. At the same time, the reduction in significant wave height lowers the wave slope. Together, these changes in wind and wave characteristics contribute to a reduction in surface roughness length (Figure 4e).
To isolate the role of sea state feedback in modifying surface drag, we performed additional WRF-only simulations, in which sea state is prescribed from unperturbed simulations (NOWF). Results show that the primary driver of wind stress reduction in the presence of wind farms is the wake-induced wind speed deficit, rather than changes in sea states.
The sea state changes examined in this study ultimately result in a reduction of mean wave energy flux or wave power, which depends on both significant wave height and the mean wave period. On average, during summer, wave power may decrease by about 3% (0.12 kW m -1 ) (not shown). Such systematic reductions in coastal wave power could potentially have implications for nearshore dynamics, including changes in coastal erosion patterns and sediment transport processes. While these coastal impacts lie beyond the scope of the present study, they represent important directions for future research.
Further research is needed to assess the long-term effects of offshore wind farms on regional wind and wave climates, as well as on upper-ocean processes. These findings are important not only for advancing our understanding of the complex interactions occurring within turbine wakes but also for informing robust environmental assessments near wind farm installations. The responses identified in this study are likely modulated by multiple factors not considered in this study, including variations in turbine locations, layouts, and sizes, as well as potential interactions among neighboring wind farms. Comprehensive, high-resolution coupled modeling studies that systematically assess these factors are crucial for guiding the effective and sustainable design, siting, and operation of offshore wind energy projects in a manner that mitigates unintended impacts on the coastal environment.
21699291, 2025, 12, Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2025JC023156 by Cesar Sauvage -Naval Postgraduate School , Wiley Online Library on [23/12/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License