Wind-speed vertical profile estimation from SAR and spaceborne lidar fusion
Satellite SAR retrieves wind speed at 10 m above the surface with high spatial resolution, but wind turbines operate at 100–200 m. Bridging that gap requires atmospheric stability correction and spaceborne wind-profile data, and the method is less straightforward than it looks.
Sensors
- Sentinel-1 SAR (C-band, 5.4 GHz): Retrieves 10 m neutral-equivalent wind speed via CMOD geophysical model functions. IW mode delivers 5 × 20 m ground resolution across a 250 km swath; EW mode covers 400 km at 20 × 40 m. Revisit is 6 days per satellite (12-day single-satellite), reduced to roughly 3–4 days with both Sentinel-1A and 1C active.
- Aeolus ALADIN Doppler wind lidar: Provided direct line-of-sight wind component profiles from the surface to 30 km altitude, with vertical resolution of 0.25–2 km depending on height range. Horizontal sampling roughly every 87 km along track. Mission ended July 2023; archive spans 2018–2023.
- ASCAT on MetOp (C-band scatterometer): Measures 10 m neutral wind speed and direction at 12.5 or 25 km resolution across two 500 km swaths. Revisit approximately 1–2 days globally, better at high latitudes. Complements Sentinel-1 with higher temporal frequency at coarser spatial resolution.
- ERA5 reanalysis (ECMWF): Provides hourly atmospheric state including boundary-layer height, surface heat flux, sea-surface temperature and pressure at 31 km horizontal resolution. Used to derive Monin-Obukhov length and atmospheric stability class for wind-shear correction.
Why 10 metres is the wrong answer
Satellite SAR retrieves wind speed at the standard meteorological reference height of 10 m. Modern offshore turbines have hub heights between 100 m and 200 m, and rotor diameters that span another 150–220 m above that. The wind speed at hub height is not simply a scaled version of the 10 m value. The vertical profile depends on surface roughness, atmospheric stability and the height of the boundary layer, all of which vary by location, season and time of day.
Over open water the surface is aerodynamically smooth and roughness is relatively predictable, which is why SAR-derived offshore wind fields are genuinely useful as a starting point. Onshore, variable terrain and land cover make the extrapolation far noisier, and SAR wind retrieval over land is not physically well-founded in the same way. This page is therefore primarily about the offshore case, where the method has the strongest scientific basis.
Monin-Obukhov similarity theory: what it does and where it breaks
The standard framework for extrapolating wind speed from a reference height to hub height is Monin-Obukhov similarity theory (MOST). Under MOST, the wind profile in the surface layer follows a logarithmic form modified by a stability correction function. The key parameter is the Obukhov length, L, which characterises whether the boundary layer is stable (negative buoyancy suppresses turbulence), neutral (mechanical mixing dominates) or unstable (convective mixing enhances turbulence). Stable conditions, common at night and in winter over cold seas, produce wind profiles that increase more steeply with height than the neutral log-law predicts. Unstable conditions, common in summer when warm air sits over cooler water, produce flatter profiles.
The practical consequence is that assuming neutral stability, as early wind-resource assessments routinely did, can underestimate hub-height wind speed by several per cent in stable regimes and overestimate it in unstable ones. Several per cent of annual energy production is commercially significant. ERA5 provides the surface sensible heat flux and boundary-layer height needed to estimate L at hourly resolution, and this is now standard practice in satellite-derived wind-resource studies. The honest caveat is that ERA5 itself carries uncertainty in its boundary-layer parameterisation, particularly in coastal zones where sea-surface temperature gradients are sharp.
What Aeolus added, and what its absence means
Aeolus, launched by ESA in August 2018, carried the ALADIN instrument: a direct-detection Doppler wind lidar operating at 355 nm (ultraviolet). It measured the line-of-sight wind component in the atmospheric column, producing vertical profiles with vertical resolution as fine as 250 m in the lower troposphere. This was the first spaceborne instrument capable of directly observing wind at altitudes relevant to turbine rotors, not just at 10 m.
For wind-profile research, Aeolus provided a way to validate the shape of the extrapolated profile, not just the surface anchor. Studies comparing Aeolus retrievals with offshore met-mast data and numerical weather prediction found biases typically in the range of 1–2 m/s in the lower troposphere, with random errors larger than that. Aeolus was a research mission, not an operational wind-resource tool, and its 87 km along-track sampling meant it could not resolve spatial variability at the scale of a single wind-farm lease area.
The mission ended in July 2023 when the spacecraft exhausted its fuel and was de-orbited. Its five-year archive remains accessible. ESA is developing Aeolus-2 as an operational follow-on, with improved performance and a target launch in the early 2030s, but no confirmed date has been published at the time of writing. The gap matters: without a direct spaceborne wind-profile observation, the profile shape must come entirely from reanalysis and similarity theory, both of which carry larger uncertainties than a direct measurement.
Fusion workflow: from SAR pixels to hub-height wind maps
A practical fusion workflow proceeds in four steps. First, Sentinel-1 backscatter is converted to 10 m neutral-equivalent wind speed using a CMOD model function (CMOD5.N is the current standard for C-band VV polarisation). Wind direction must be supplied externally, typically from ERA5 or ASCAT, because SAR alone cannot determine direction without ambiguity. Second, the neutral wind speed is converted to actual wind speed by correcting for the atmospheric stability estimated from ERA5 surface fluxes. Third, the stability-corrected 10 m wind speed is extrapolated to hub height using the MOST profile with ERA5-derived Obukhov length. Fourth, where Aeolus archive data overlap the study period and area, profile shape validation is applied to check whether the reanalysis-based extrapolation is systematically biased.
Spatial resolution of the final hub-height wind map is set by the coarsest input. ERA5 at 31 km is the binding constraint on stability correction, so the apparent 5–20 m SAR resolution is not preserved through to the final product. In practice, hub-height wind maps derived this way are meaningful at scales of 5–10 km, which is sufficient for lease-area resource characterisation but not for turbine-by-turbine micrositing.
Stability matters more offshore, and here is why
Onshore, the diurnal cycle of surface heating creates a strong and relatively predictable stability pattern: unstable by day, stable by night. The atmosphere resets daily. Offshore, stability is driven by the difference between sea-surface temperature and air temperature, which can persist for weeks. In the North Sea, for example, cold upwelling events in spring can create persistently stable conditions that suppress turbulence and steepen wind profiles for extended periods. A wind-resource estimate that ignores this will overestimate the hub-height wind speed during those periods.
Conversely, when warm continental air advects over cooler coastal waters in summer, strongly unstable conditions develop rapidly. The wind profile flattens and hub-height speeds are lower than the neutral log-law would suggest. These transitions happen at spatial scales of tens of kilometres and time scales of hours, which is precisely why ERA5's hourly output at 31 km is useful here despite its resolution limitations. The stability correction is not a small refinement; in some offshore regimes it changes the estimated annual energy production by 5–10%, according to published studies using similar methods.
Honest limits and what comes next
Several sources of uncertainty compound in this workflow. SAR wind retrieval has a reported root-mean-square error of roughly 1.5–2 m/s against buoy observations under moderate wind conditions. ERA5 boundary-layer fluxes carry their own errors, particularly near coasts. MOST itself is only strictly valid in the surface layer, typically the lowest 10–20% of the boundary layer, and hub heights of 150–200 m frequently extend above that layer in stable conditions, where the theory's assumptions break down. When the boundary layer is shallow and stable, the profile above the surface layer can deviate substantially from the MOST prediction.
Aeolus-2 will address the direct-measurement gap when it arrives, but the early 2030s target means current projects must work with the 2018–2023 Aeolus archive for validation and ERA5 for operational stability data. Commercial radiosonde networks and offshore met-mast data, where available, remain the ground truth of choice for calibrating satellite-derived profiles at a specific site.
Satellize runs this fusion workflow on Sentinel-1 and ERA5 open data, with optional ASCAT integration, as part of its offshore wind analytics offering. The Aeolus archive is incorporated for historical validation where the study period overlaps. As with the Tonga crop-estimation programme, the aim is to make the method auditable rather than opaque: clients receive the stability-correction fields alongside the wind maps, not just the final number.
Typical figures
| SAR wind-field spatial resolution (native) | 5 × 20 m (Sentinel-1 IW mode); hub-height product meaningful at 5–10 km after ERA5 stability correction |
| SAR swath width | 250 km (IW mode), 400 km (EW mode) |
| Sentinel-1 revisit | 3–6 days (two-satellite constellation); single satellite 12 days |
| ASCAT wind resolution / revisit | 12.5 or 25 km; ~1–2 day global revisit |
| ERA5 spatial and temporal resolution | 31 km horizontal, hourly; archive from 1940 to present |
| Aeolus vertical wind profile resolution | 250 m–2 km (height-dependent); along-track sampling ~87 km; archive 2018–2023 |
| SAR wind-speed retrieval RMS error | ~1.5–2 m/s against buoy observations under moderate winds (5–20 m/s) |
| Hub-height extrapolation range covered | 100–200 m; MOST validity degrades above ~20% of boundary-layer height in stable conditions |
| Archive depth (Sentinel-1) | 2014 to present (Sentinel-1A); combined archive enables multi-year climatologies |
| Delivery formats | GeoTIFF wind-speed grids, NetCDF with stability-correction fields, CSV site-point time series |
Analytics Satellize can run
| Hub-height wind-speed climatology map | CMOD5.N SAR retrieval + MOST extrapolation with ERA5 Obukhov length; multi-year Sentinel-1 stack | GeoTIFF grid of mean annual and seasonal hub-height wind speed at 5–10 km resolution, with uncertainty layer |
| Atmospheric stability class frequency distribution | ERA5 surface heat flux and boundary-layer height binned into Pasquill-Gifford stability classes per grid cell | NetCDF file and summary report showing fraction of hours in stable, neutral and unstable conditions by month |
| Wind-shear exponent spatial map | Empirical power-law exponent fitted between ERA5 50 m and 100 m wind speeds, cross-checked against SAR surface anchor | GeoTIFF shear-exponent grid for the lease area; used as input to energy-yield models |
| Aeolus profile validation report | Co-location of Aeolus ALADIN L2B wind profiles with SAR-derived surface wind and ERA5 extrapolated profiles over the study area (archive period 2018–2023) | PDF report with bias and scatter statistics between Aeolus, ERA5 and SAR-derived profiles; flags systematic stability-correction errors |
| ASCAT / Sentinel-1 wind-speed cross-calibration | Temporal co-location of ASCAT and Sentinel-1 passes over the site; bias correction of SAR retrievals using higher-revisit ASCAT baseline | Corrected Sentinel-1 wind-speed time series in NetCDF; calibration coefficient table |
| Coastal stability transition zone mapping | ERA5 sea-surface temperature minus air-temperature gradient analysis; identification of persistent stable or unstable fetch zones | GIS layer showing stability-regime boundaries relevant to wind-shear enhancement or suppression, with seasonal breakdowns |
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.