Thanet offshore wind farm is a fully-operational Round 2 offshore wind farm site operated by Vatenfall Vindkraft AB. The wind farm covers an area of 35 km² and is located 11 km from the nearest point of the Kent coast and at the margin of the high turbidity zone associated with the outer Thames region (Figure 2A). Construction of Thanet started in January 2008 with a cost of £900 million, and the farm has been operational since September 2010 consisting of 100 Vesta V90 wind turbines with a total capacity of 300 MW. Each of the turbines is supported by a 4.1–4.9 m diameter monopile driven into the seabed between water depths between 14 and 23 m (Figure 3), with no scour protection.

Individual monopile wake presence, direction and suspended particulate matter were analysed using a series of high-resolution satellite imagery from initial operation in 2010 through to 2022. This included 55 Landsat 8 (30 m resolution) scenes (2013–2022) and 58 Sentinel-2 (10 m) scenes (2015–2022). A further nine Landsat 5 (60 m) scenes (2010–2011) were included to extend the remotely sensed time series (i.e., pre-Landsat 8 and Sentinal-2 launch) to the start of operation at Thanet. Top of atmosphere Landsat scenes were downloaded from the USGS Earth Explorer (https://earthexplorer.usgs.gov) and Sentinel-2 from the Copernicus Open Access Hub (https://scihub.copernicus.eu/dhus). All images where the Thanet site was not obscured by cloud were included in the analysis.

Atmospheric effects were first corrected using the open access Acolite code (https://odnature. naturalsciences. be/remsem/software-and-data/acolite; Vanhellemont and Ruddick, 2014; Vanhellemont and Ruddick, 2018). This code uses an automated image-based “dark spectrum fitting” approach developed by Vanhellemont and Ruddick (2018) based on the red and near infrared channels to retrieve water-leaving radiance reflectance (ρ_w) in the red band. Suspended particulate matter estimates for each colour pixel (e.g., Figures 2B, C) were then calculated from ρ_w using following the algorithm: SPM = A ρ w 1 − ρ w / C (1) where A and C calibration coefficients are based on the wavelength of the red band of each satellite imager, as derived by Nechad et al. (2010). The A and C coefficients have been further calibrated specifically for Landsat 8 and Sentinel-2 in the Acolite code. Here, the Landsat 8 red band values for 654 nm were used: A = 346.32 g m⁻³ and C = 0.19905; and the Sentinel-2 red band values at 655 nm were used: A = 342.10 g m⁻³ and C = 0.19563. Estimates for suspended particulate matter in the wake of E01 (Figure 4) were obtained from satellite scenes by creating a perimeter at 1 km distance from the monopile and collecting data where it overlaps the optically distinct wake.

Temporal (seasonal to annual) variability in suspended sediment concentration within Thanet and the surrounding waters was assessed using coarser resolution (1–4 km pixel size) satellite imagery. Monthly mosaics of MERIS, MODIS/AQUA, VIIRS and SeaWiFS satellite imagery between 1997 and 2014 were acquired from the Copernicus Marine Environmental Monitoring Service (https://marine.copernicus.eu). Suspended particulate matter was then derived from the satellite reflectance using the algorithm of Gohin et al. (2005). To determine the potential impacts of wind farm construction and operation, average monthly suspended particulate matter estimates for the Thanet site prior to construction were compared to measurements post-construction. Four areas, each 10 by 10 km in size, to the northwest, northeast, southwest and southeast (Figure 3) were also analysed as control sites to account for any inter-annual natural variability. These sites were chosen to represent variability in water depth, proximity to sediment sources, and wave and tidal energy.

Sea-level data from the nearest tide gauge site at Sheerness (Figure 2A) were downloaded from the British Oceanographic Data Centre (https://www.bodc.ac.uk) and wave measurements from the South Knock (Figure 2A) Cefas WaveNet buoy (https://www.cefas.co.uk/data-and-publications/wavenet).

In situ measurements within the turbid wakes of monopiles and surrounding waters at the Thanet offshore wind farm were collected in August 2016 on the RV Meriel D. Data were collected over two surveys each consisting of three consecutive days (3rd–5th and 17th–19th August 2016) during spring tides (within a typical spring tidal range for the site) when current velocity would be greatest to ensure the highest chance of measuring within an optically-distinct monopile wake. Measurements were made either through the upper (0–10 m water depth) water column or as continual transects during which the engines of the research vessel were disengaged and allowed to drift in tidal currents. A greatly increased signal-to-noise ratio was obtained when the ship was drifting passively as compared to underway with power, most notably for acoustic Doppler current profiler data (see Section 2.4.3). We describe the results of two periods of passive drifting through the wake of monopile E01 at Thanet on 19th August 2016 (Figure 4). Defining the exact timing of entering and leaving the visible wake was difficult due to sun glint and wave chop meaning the optically-distinct wake was not always observable from the vessel at sea level. Drift A recorded in situ measurements within the wake of E01 for an estimated 34 min while drifting down-current after initially being tethered to the monopile (Figure 4B). Due to a change in wind effects the vessel took a different trajectory to that of the tidal current in Drift B. Here the vessel initially drifted through the water surrounding the wake of E01, before passing through the wake for an estimated 7 min, before leaving again (Figure 4C). Positional data for all other instruments were linked to the primary GPS using time-stamps, with the clocks of all synchronised at the start of surveys. Detailed descriptions of data collection methods are listed below.

Regional monthly mosaics of MERIS, MODIS/AQUA, VIIRS and SeaWiFS satellite images between 1997 and 2014 displayed an annual cycle of suspended particulate matter in the Thanet site and surrounding waters (Figure 5A). Similar patterns of variation were observed in significant wave height (Figure 5B). For example, sediment concentrations were highest in January through March when the sea state is roughest with estimates of up to 100 mg L⁻¹ in the outer Thames, with Thanet located on the edge of this high turbidity region. Sediment concentrations and wave height decreased greatly during Spring; by May most of the Thames regions showed values of suspended matter below 20 mg L⁻¹ with the Thanet site indistinguishable from its surrounding waters. Sediment concentrations remained low throughout the summer, until September where zones of higher turbidity were apparent off the Kent and Essex coastlines, with values increasing through the remainder of the year. The annual pattern was observed at Thanet and all the control sites (Figure 3), with higher suspended particulate matter observed in the more proximal southwest and northwest control sites, with a much greater amplitude of suspended matter during the winter than control sites to the northeast and southeast (Figure 5A). The Thanet site does show inter annual variation in suspended particulate matter. Regional satellite images showed annual averaged concentration ranged from 17 (2001) to 42 mg L⁻¹ (2010). This variation though is coupled with the neighboring northwest control zone, which most closely resembled the annual cycle of suspended matter to Thanet (Figure 6).

Optically distinct monopile wakes were observed in >90% of all satellite images. When wakes were absent, more than half of the images were of poor quality (e.g., partially covered by clouds) or displayed rough sea conditions. The direction of visible monopile wakes was consistent with the tidal ellipse within the Thanet wind farm (Figure 7B). At low water, when tidal flow is aligned from north to south, monopile wakes are also aligned in this direction (Figure 8A). Monopile wake direction rotates clockwise during the rising flood tide, such that the visible wake trends from north-east to south-west by mid-flood (Figure 8B). Clockwise rotation continues with monopile wakes trending south to north at high water (Figure 8C) and then west to east aligning with ebbing currents leaving the Thames Estuary (Figure 8D).

In situ sampling showed total suspended material in the surface waters (∼1 m water depth) was elevated in the wakes of both monopiles E01 and C15 in comparison to the surrounding water upstream, showing a mean increase of 50% (Figures 9A, C). The wake of C15 was enriched in suspended matter compared to the zone upstream in the mid (10–13 m) and lower (∼20 m) water depths (Figure 9C). However, the corresponding measurements for E01 showed the wake area was depleted in suspended matter in comparison to the surrounding water (measurements outside the wake had 11% higher volume of suspended particulate matter; Figure 9A). Samples recovered at high tide in waters adjacent to the wake of C15 were depleted in suspended particulate matter compared to corresponding samples collected during ebb tide (Figure 9C). The percentage of organic matter in samples showed an overall increase towards the surface (increases of 6% at E01; 4.5% C15); no obvious trend was observed between samples collected within and outside the monopile wakes (Figures 9B, D). Satellite-derived measurements of suspended particular matter (using Eq. 1) showed the direction of the monopile wake (depending on tidal flow) at E01 had no obvious influence on concentration (Figure 7A).

Physical and chemical water properties were broadly consistent between the upper 10 m of the water column in wakes of monopiles D03 (STN2A) and E01 (STN3A) and zones of water immediately adjacent (STN2B and STN3B; Table 1). Temperature and salinity profiles within and outside the monopile wakes were within the range of variability observed at the surface. Of the optically-active water column constituents, chlorophyll and cDOM were higher (depth-averaged increases of 10% and 18%, respectively) within the wake of D03 but almost no change was observed at E01 (Table 1). However, local and horizontal variability in chlorophyll and cDOM can be observed with relatively high depth-averaged measurements of both parameters recorded at STN1 upstream of the wind farm (Table 1). Similarly, depth-averaged backscatter increases in the upper 10 m of the water column at D03 (by 35%) yet remains remarkably consistent at E01 (Table 1).

Surface turbidity and ADCP measurements were recorded when the vessel was passively drifting through the wake of monopile E01 to greatly increase signal-to-noise (Figure 10). Prior to Drift A (Figure 4B) the vessel approached the plume of turbine E01 from the south-west, entering the plume at 9:08. Turbidity measurements, both from the on-deck system and the instrument frame at 2 m water depth, showed sharp increases between 9:08 and 9:15 before being positioned and tethered ∼30 m down current in the turbulent wake of E01 where the engines were disengaged at 9:27 (Figure 4B, 10C). Continuous measurements of turbidity were then recorded from this point and remained constant when the vessel detached from the turbine at 10:01. A small decrease in turbidity was observed when drifting out of the visible wake at 10:04 (Figure 10C). Surface acoustic backscatter recorded by the ADCP similarly decreased (by ∼2%) at this time (Figures 10D, E). Conversely, near-bed acoustic backscatter readings were lower within the wake and increased (by ∼3%) as the vessel drifted out of the plume (Figures 10D, E). It is important to note however, that a caveat to the use of acoustics is that higher surface sediment load, or surface bubbles, will attenuate backscatter from the underlying water column.

Turbidity increases were also observed within the wake of E01 during Drift B (Figure 4C); with a sharp increase detected at 10:35 followed by a gradual decrease during transit across the wake from 10:36 to 10:41 (Figure 10C). A higher-resolution record (10 Hz versus 0.016 Hz) of this event was captured from measurements from the instrument frame deployed at 2 m water depths located in the moonpool. Movement of the vessel into the wake similarly resulted in a sharp 40% increase in scattering at 532 nm. Moving away from the center of the monopile wake resulted in a gradual decrease in backscatter (Figure 11). Similar results were also obtained with the multi-waveband optical instruments during this period. Unlike Drift A, entering the plume during Drift B did not show an obvious response in near-surface acoustic backscatter. However, near-bed readings did show a spike in decreased backscatter while within the monopile wake, but this was not observed across the entire drift (Figures 10D, E).

The velocity structure of the water column is highly variable, even during the period when the vessel was stationary with the engines disengaged while tethered to E01. However, elevated vertical velocity can be observed during both drifts within the wake of E01 in comparison to measurements recorded in the surrounding water column (Figures 10F, G). This is most clearly observed between water depths of 4–12 m in Drift A where vertical upwards velocity increases up to 18 cm s⁻¹, with Drift B much less pronounced (Figure 10F). Additionally, vertical velocity abruptly reversed to a downwards direction on leaving the visible wake during Drift A and prior to entering during Drift B (Figures 10F, G).

Cycles of annual suspended particulate matter variation were observed whereby increases aligned with periods of higher waves heights. Similarly, the abundance of suspended matter varied spatially between the Thanet site and the surrounding waters. For example, a year-round lower concentration of suspended matter was observed in the northeast and southeast control zones due the position of the zones in deeper waters, further from the coast (Figures 3, 5A). While inter-annual variation in suspended particulate matter within Thanet was observed, no obvious trends (increases or decreases) exist during the construction or operational phases of the wind farm. Furthermore, post-construction measurements (aside from 2010) were within the range of pre-construction observations (Figure 6). Additionally, the variation in suspended matter concentration at Thanet showed coupled changes through time with the northwest control site (Figure 5A, 6). Such coupled intra- and inter-annual variability further suggests changes in sea surface turbidity within the Thanet site are likely not an anthropogenic response to the construction and operation of the wind farm, rather natural fluctuations.

In situ measurements support the hypothesis that colour contrasts in the surface waters of monopile wakes are caused by elevated suspended sediment concentrations in comparison to the surrounding waters. For example, water samples collected from the surface waters in the wakes of monopiles C15 and E01 showed a 50% increase in suspended matter compared to samples obtained upstream (Figures 9A, C). This observed change is less than the 500% increase measured at the Belwind 1 wind farm, located 46 km offshore from Zeebrugge, Belgium, however, much higher quantities of suspended particulate matter were observed in the surface waters of monopile wakes at Thanet (Baeye and Fettweis, 2015). Such disparity between monopile wake suspended sediment likely reflects wind farm site characteristics. For example, greater sediment in the monopile wakes at Thanet could be due to shallower water depth range (14–23 m compared to 15–37 m; Figure 3) or the wind farm site in much more turbid waters than Belwind 1 (14–22 mg L⁻¹ compared to 3 mg L⁻¹; Figure 2A; Baeye and Fettweis, 2015). In either instance, it is therefore likely that changes in suspended sediment in monopile wakes are site specific.

No increase in organic matter abundance was observed in monopile wakes at Thanet (Figures 9B, D), therefore providing evidence against colour contrasts arising from the release of organic material from epifauna-colonizing monopiles (e.g., Baeye and Fettweis, 2015). Similarly, optically significant water constituents such as chlorophyll (plankton) and cDOM show no enrichment within visible monopile wakes that would be outside the range observed in waters surrounding the Thanet site (Table 1). Elevated surface water turbidity and acoustic backscatter recorded by the ADCP in the wake of monopile E01 could provide further evidence to suggest colour changes were related to higher suspended sediment concentration (or an increase in grain size; Figure 10D). A lack of difference between spectral bands (from AC9 and BB9 measurements) indicated that increases in attenuation and scattering was spectrally-neutral, therefore unlikely to be caused by chlorophyll or cDOM. While our results demonstrate the colour contrast in monopile wakes are related to increased sediment concentration it is not possible to rule out bubble formation as a contributing factor to the observed enhanced surface acoustic backscatter (Figure 10D).

Averaged water column measurements (to 10 m water depth) within wakes of monopiles D03 (STN2A and STN2B) and E01 (STN3A and STN3B) showed no significant difference with the adjacent waters or to water column measurements collected to the north of the windfarm (STN1; Table 1). However, bottled water samplings, acoustic backscatter and turbidity measurements provide evidence for elevated suspended sediment concentration near surface (Figure 9, 10C). Additionally, a decrease in acoustic backscatter near-bed suggests that suspended sediment concentration was depleted in the wake at depth, in comparison to the surrounding water column (Figure 10E). This near surface enrichment and near bed depletion is consistent with suspended particulate matter measurements at monopile D03 (Figure 9A). No such depletion was observed at C15; however, three of the five samples taken in the waters adjacent to the monopile wake were collected at high tide (Figure 9C) when tidal velocity and therefore the potential to transport sediment was reduced. Assuming changes in the vertical gradient of sediment concentration in monopile wakes is the norm, the absence of an overall change in suspended particulate matter in the water column (Table 1) indicates elevated wake surface sediment concentration is due to sediment redistribution, rather than additional sediment being sourced.

Increased turbulence in monopile wakes is the driving factor in redistribution of suspended particulate matter from the lower water column to the surface. Although precise ADCP measurements of water column velocity were inconclusive due to variable signals from vessel motion, even when tethered to the monopile, acoustic measurements demonstrated the underwater shape of the wake was complex (Figure 10F). Despite this, time-averaged data indicates a clear shift to upwards motion throughout the water column on entering the wake of E01 during both passive drifts and a change to downwards motion when leaving the visible wake during Drift A and prior to entering during Drift B (Figures 10F, G). These observations are consistent with modeling and laboratory experiments of the hydrodynamics in offshore wind farms (e.g., Petersen et al., 2015; Grashorn and Stanev, 2016; Rivier et al., 2016; Miles et al., 2017; Zhang et al., 2021). For example, previous work has demonstrated that horseshoe-vortices are generated around the monopile base by downward flow upstream (Figure 12); subsequent vortex shredding results a pair of counter-rotating vortices forming an upward flow along the center of the monopile wake (Figure 12). At Thanet this influence of these counter-rotating vortices would equate to a minimum of ∼15 m above bed (i.e., at least three times diameter of monopiles in the wind farm), almost the entire water column at E01. Therefore, we suggest that counter-rotating vortices in monopile wakes drives sediment redistribution towards surface waters (Figure 12). Passive drifting away from the monopile during Drift A shows upwards motion is detectable downstream for a minimum of 100 m (Figures 4B, 10D). Once the surface waters in monopile wakes are enriched in suspended sediment, the development of Kármán vortex street (Figures 1, 12) provide an efficient mechanism to transport these waters downstream for several kilometers, producing the elongate sediment plumes observed.

Scouring at the base of monopiles has significant economic impacts, potentially requiring extra scour protection and wind farm maintenance to prevent erosion and extend operational life (e.g., Rezaei et al., 2018). While it is not possible to entirely rule out scour at monopile bases at Thanet (e.g., Vanhellemont and Ruddick, 2014), our results show that averaged sediment concentration in the water column upstream of the monopile is similar to that in the wake (Table 1). The consistent suspended sediment concentration suggests that additional sediment is not sourced from the seafloor, therefore unlikely that erosion around monopile bases is the control on producing optically-distinct sediment plumes in monopile wakes.