Wind farm wake effect mapping from SAR-derived surface wind fields
Synthetic aperture radar backscatter from the sea surface, inverted through the CMOD geophysical model function, produces wind speed maps at roughly 500-metre resolution. Those maps reveal the wind speed deficits that offshore turbines impose on one another and on neighbouring planned developments.
Sensors
- Sentinel-1 A/B (C-band SAR, ESA): IW mode at 10-metre pixel spacing, swath 250 km. Revisit over European offshore zones is 6 days per satellite; with both satellites operational it was 3 days. Wind retrieval is performed at 500-metre posting after spatial averaging to reduce speckle. Free and open archive from 2014.
- RADARSAT-2 (C-band SAR, MDA): ScanSAR Wide mode covers 500 km swaths at 50-metre resolution, suitable for large-scale wind field mapping. Fine mode at 8-metre resolution can resolve individual turbine wakes close to the rotor plane. Tasking is commercial; revisit depends on orbit and latitude.
- ICEYE SAR constellation (X-band): Sub-metre spotlight and 3-metre stripmap modes. X-band backscatter responds to shorter Bragg-resonant waves than C-band, making CMOD calibration less mature, but the high resolution is useful for near-field wake structure within roughly 2 km of the rotor. Revisit can be daily over a fixed site with commercial tasking.
- RADARSAT Constellation Mission (C-band, CSA): Three-satellite constellation achieving daily revisit at high latitudes. Medium-resolution ScanSAR at 50-metre posting covers offshore areas in Canada and the North Atlantic. Supports the same CMOD-based wind retrieval workflow as RADARSAT-2.
What a dark streak on the ocean surface is actually measuring
C-band radar backscatter from the sea surface is governed by Bragg scattering off centimetre-scale capillary and gravity waves. Those small waves are generated by local wind stress. Where wind speed drops, wave amplitude drops, backscatter falls, and the image darkens. The CMOD geophysical model function, developed through decades of comparison with scatterometer and buoy records, formalises that relationship: given backscatter intensity, incidence angle, and wind direction from a numerical weather model, CMOD5.N or CMOD7 returns a wind speed estimate. The retrieval is typically valid between roughly 2 and 25 metres per second; below 2 m/s the surface is too smooth to distinguish speed differences reliably, and above 25 m/s wave breaking introduces ambiguity.
Offshore wind turbines extract momentum from the airflow and deposit a turbulent, slower-moving column of air downwind. At sea surface level, this appears in a SAR image as a darker elongated region, sometimes extending 20 km or more behind a large farm under stable atmospheric conditions. The shape, length, and recovery gradient of that region are the primary data products.
Wake length is not fixed. Atmospheric stability is the variable that matters most.
Published studies using Sentinel-1 imagery over the Horns Rev and Anholt farms in the Danish North Sea have documented wake lengths ranging from under 5 km in convective, unstable conditions to beyond 20 km under stable stratification. The physical reason is straightforward: a stable atmosphere suppresses vertical mixing, so the momentum deficit persists further downwind before ambient flow replenishes it. SAR captures this directly because it integrates the actual surface wind field rather than inferring it from turbine SCADA data, which only reports hub-height conditions.
Atmospheric stability cannot be read from the SAR image alone. The standard workflow combines the SAR-derived wind speed map with coincident sea surface temperature data (from MODIS, SLSTR, or VIIRS) and reanalysis profiles to estimate the Monin-Obukhov length. That stability parameter then contextualises each wake observation. A single SAR pass is a snapshot. A multi-year archive of passes, binned by wind direction and stability class, builds a statistical wake climatology with practical engineering value.
Turning a wind field map into a layout decision
The immediate application for a developer planning a farm adjacent to an existing one is quantifying the expected wind resource reduction at candidate turbine positions. A SAR-derived wake climatology, weighted by the wind-direction frequency distribution at the site, produces a spatial map of expected annual energy production loss attributable to the upstream farm. This is not a substitute for computational fluid dynamics modelling or met-mast measurement, but it provides an empirical constraint that CFD models often lack: observed wake behaviour under real atmospheric variability rather than idealised inflow.
Regulators in several European jurisdictions now request independent evidence of inter-farm wake effects during consent processes. A multi-year Sentinel-1 archive, freely available from 2014, can support that evidence base without additional measurement campaigns. The spatial resolution of 500 metres is sufficient to distinguish wake effects at the scale of turbine spacing in modern large farms, where inter-turbine distances are typically 700 metres to 1.5 km. It is not sufficient to resolve the near-rotor flow structure or to attribute energy loss to individual turbines within a dense array.
Where the method reaches its limits
Wind direction ambiguity is the most persistent problem in CMOD retrieval. The model function has a 180-degree directional ambiguity that must be resolved using external wind direction information, typically ERA5 reanalysis or ECMWF analysis fields. If the reanalysis direction is wrong by more than roughly 20 degrees, the retrieved speed can be biased by 1 to 2 m/s, which is material for wake detection where the deficit may itself be only 1 to 3 m/s at distances beyond 10 km.
Rain contamination is a second limit. Rainfall roughens the sea surface independently of wind, inflating backscatter and corrupting the retrieval. Standard practice is to flag passes with precipitation using coincident passive microwave data or reanalysis precipitation fields and exclude them from the climatology. In high-latitude sites with frequent frontal rainfall, this can reduce the usable pass count significantly. Finally, the method is strictly offshore. Land surface roughness is orders of magnitude larger than wind-induced sea surface roughness, and no equivalent geophysical model function exists for terrestrial surfaces.
Building a wake climatology in practice
A credible climatology requires a minimum of two to three years of SAR passes to sample the wind direction distribution adequately. Over a typical North Sea site, Sentinel-1 delivers roughly 50 to 100 usable passes per year after rain flagging and low-wind exclusion. Passes are binned by wind direction in 10- or 15-degree sectors, and within each sector by stability class. The median wind speed deficit at each grid point, relative to an upwind reference area, is then mapped. The result is a set of sector-specific wake footprints that can be convolved with a long-term wind rose to produce an annualised energy loss estimate.
Satellize applies this workflow using the open Sentinel-1 archive and adds commercial RADARSAT-2 or ICEYE tasking where higher spatial resolution or more frequent revisit is required for a specific site. The analytic output is a GIS-ready wake climatology layer with associated uncertainty bounds, delivered alongside a methodology report suitable for submission to planning authorities.
Typical figures
| Wind field spatial resolution (typical) | 500 m posting after speckle averaging (Sentinel-1 IW, RADARSAT-2 ScanSAR) |
| Near-field resolution (ICEYE / RADARSAT-2 Fine) | 8–25 m pixel; wind retrieval less mature at X-band |
| Revisit (Sentinel-1, single satellite) | 6 days over most European offshore zones; 3 days when two satellites operational |
| Revisit (commercial tasking, ICEYE) | Daily or sub-daily over a fixed site with tasking agreement |
| Radar frequency | C-band 5.405 GHz (Sentinel-1, RADARSAT-2); X-band ~9.6 GHz (ICEYE) |
| Wind speed retrieval range (CMOD5.N / CMOD7) | Approximately 2–25 m/s; accuracy degrades below 2 m/s and above 25 m/s |
| Minimum detectable wake deficit | Approximately 0.5–1 m/s at 500 m resolution under moderate wind conditions; higher uncertainty in rain or low-wind scenes |
| Archive depth (Sentinel-1) | From April 2014 (Sentinel-1A launch); continuous open access |
| Delivery formats | GeoTIFF wind speed grids, NetCDF climatology files, GeoPackage wake-footprint polygons, PDF methodology report |
| Applicable domain | Offshore only; onshore land surface roughness prevents equivalent wind retrieval |
Analytics Satellize can run
| Single-pass wind field map | CMOD5.N or CMOD7 inversion of calibrated Sentinel-1 or RADARSAT-2 sigma-naught, with wind direction from ERA5 | GeoTIFF wind speed grid at 500 m posting, delivered within 3 hours of scene availability |
| Wake footprint per wind sector | Multi-pass compositing binned by ERA5 wind direction in 15-degree sectors; median deficit relative to upwind reference box | GeoPackage layer set, one polygon per sector, with median deficit and 25th/75th percentile bounds |
| Stability-stratified wake climatology | Monin-Obukhov length estimated from coincident SST and reanalysis profiles; passes binned into stable, neutral, and unstable classes | NetCDF climatology with three stability-class dimensions; accompanying PDF summarising pass counts and exclusion rates |
| Annual energy production loss estimate | Convolution of sector wake footprints with long-term wind rose (ERA5 or site mast); percentage deficit mapped at candidate turbine positions | Spatial AEP-loss map (GeoTIFF) and tabular summary per candidate position, with uncertainty range |
| Wake length exceedance statistics | Per-pass automated detection of wake trailing edge using gradient threshold on wind speed profile; exceedance curves by stability class | CSV table of wake length percentiles by stability class and wind speed bin; suitable for regulatory submission |
| Change detection after farm commissioning | Comparison of pre- and post-commissioning SAR wind field distributions at downwind grid points using Mann-Whitney test | Before/after wind speed difference map (GeoTIFF) with statistical significance mask |
Who does the work
We can get this done for you. Satellize runs its own analyst desk and a strong science team. You do not buy a data feed and work out what it means; our people source the imagery, run the analysis described on this page, and hand you the answer with its confidence limits stated. Discuss this requirement.