Sensors
- Sentinel-1 IW (Interferometric Wide Swath) mode: C-band SAR at 5.405 GHz, 250 km swath, 5 × 20 m ground resolution in IW mode. Revisit 6 days at the equator with one satellite, 3 days with two. Provides spatially detailed wind fields resolving individual squall lines and frontal boundaries.
- MetOp ASCAT: Ku-band fan-beam scatterometer operating at 5.255 GHz. Delivers wind speed and direction at 12.5 km and 25 km grid spacing with near-global daily coverage. Three look angles per swath enable direction retrieval but leave a 180-degree ambiguity requiring NWP disambiguation.
- CYGNSS (Cyclone Global Navigation Satellite System): Eight micro-satellites receiving GPS signals reflected from the ocean surface (GNSS-R). Median revisit near 7 hours in the tropics, 0.5–25 km effective resolution depending on processing. Particularly useful inside tropical cyclones where conventional radar saturates.
- RapidScat (heritage, 2014–2016): Ku-band scatterometer mounted on the International Space Station. Non-sun-synchronous orbit gave diurnal sampling diversity unavailable from polar platforms. Decommissioned but its data archive remains a useful reference for algorithm validation.
Why a radar sees wind it cannot directly measure
Wind does not return a radar pulse. What does return it is the ocean surface roughened by wind stress. At wind speeds above roughly 2–3 m/s, surface stress generates centimetre-scale capillary and short gravity waves whose crests align roughly perpendicular to the wind direction. These crests act as Bragg scatterers for C-band microwave energy, reflecting energy back to the sensor in proportion to their amplitude. The stronger the wind, the rougher the surface, the higher the normalised radar cross-section (sigma-naught).
The relationship is not linear and it depends on geometry. Incidence angle changes the fraction of the surface visible to the radar. Polarisation matters: VV polarisation is more sensitive to wind speed than HH across most incidence angles, which is why Sentinel-1 IW mode in VV is the standard configuration for wind retrieval. Rain contaminates the signal by roughening the surface independently of wind and by attenuating the beam, so rain-flagging is a necessary pre-processing step, not an optional refinement.
CMOD: the equation that does the work
The geophysical model function CMOD converts sigma-naught to wind speed given incidence angle, azimuth angle relative to wind direction, and polarisation. CMOD5.N is the current operational standard for VV-polarised C-band retrievals; it was developed and validated against buoy and dropsonde data over many years by ECMWF and EUMETSAT. The function is empirical, not derived from first principles, which means it carries inherited biases at the extremes of its training distribution, particularly above 25 m/s where sea state becomes complex and buoy measurements themselves become unreliable.
For HH polarisation a polarisation ratio correction is applied before CMOD, introducing additional uncertainty. At very low wind speeds, below about 2 m/s, the surface approaches specular reflection and the Bragg mechanism breaks down; CMOD retrievals in this regime are unreliable. Honest operational use means flagging retrievals below that floor rather than reporting them as valid data.
The 180-degree problem and how NWP solves it
A scatterometer measures backscatter from multiple look angles across its swath. Solving CMOD for wind direction yields two or four candidate solutions that fit the observations equally well. In practice the dominant ambiguity is 180 degrees: the radar cannot distinguish a 270-degree westerly from a 90-degree easterly on backscatter alone. This is a physical limitation of the measurement geometry, not a processing deficiency.
The operational solution is a first-guess field from a numerical weather prediction model, typically ECMWF's analysis or short-range forecast. The NWP wind vector is used to select the ambiguous solution closest to the model, a technique called two-dimensional variational ambiguity removal (2DVAR). The result is a scatterometer wind field that is meteorologically consistent but is no longer fully independent of the NWP model. For research applications where model independence matters, this dependency must be declared. For operational maritime routing and storm tracking, the combined product is well-validated and routinely used by national meteorological services.
Sentinel-1 as a storm-cell microscope
At 5 × 20 m resolution in IW mode, Sentinel-1 can resolve wind speed gradients across frontal boundaries, coastal jets, and the eyewall of a tropical cyclone at a spatial detail no scatterometer can approach. A 250 km swath acquired in roughly 25 minutes captures a large storm system in a single pass. The trade-off is revisit: 3 to 6 days is too slow to track a fast-moving cyclone continuously.
Sentinel-1 wind retrievals are operationally produced by EUMETSAT and several national agencies. Typical reported accuracy against buoy data is around 1.5–2 m/s RMS in speed and 20–25 degrees in direction under moderate wind conditions. Accuracy degrades in heavy rain and at the very high speeds found in major tropical cyclones, where wave breaking and sea spray alter the surface in ways CMOD was not trained to handle. The archive extends to 2014, giving a decade of high-resolution ocean wind observations available at no data cost through the Copernicus Data Space.
ASCAT's daily global sweep and its operational value
Where Sentinel-1 is a scalpel, ASCAT is a wide brush. MetOp-A, B and C together provide near-daily global coverage. ASCAT data feeds directly into ECMWF and NOAA operational analyses and has done so since 2007. The 12.5 km coastal wind product resolves mesoscale features well enough to be useful for offshore wind energy resource assessment, port operations planning, and synoptic weather analysis. The 25 km product is the standard for open-ocean applications.
Latency from observation to operational product is typically under three hours through the Global Telecommunication System. For applications that need same-day wind fields over large ocean basins, ASCAT remains the workhorse. Its limitation is that it cannot see inside heavy precipitation, and its swath gaps at low latitudes mean some ocean regions are revisited less frequently than the headline daily figure implies.
Where this data fits in a real programme
Ocean surface wind data from SAR and scatterometry is used operationally for tropical cyclone intensity estimation, offshore infrastructure planning, maritime route optimisation, and as a constraint in global climate reanalyses. The data is open and well-documented; the value added by analytics lies in combining sensor streams intelligently, applying appropriate quality flags, and delivering outputs in formats decision-makers can act on.
Satellize runs wind-field analytics on Sentinel-1 and ASCAT data as part of broader maritime and environmental monitoring programmes. The Copernicus Data Space provides the underlying archive; the processing chain applies CMOD5.N with ECMWF first-guess disambiguation and outputs gridded wind speed and direction layers at the native resolution of each sensor. Buyers wanting to commission a scoped wind-monitoring programme should start by defining their area of interest, acceptable latency, and whether storm-cell resolution or daily global sweep is the primary requirement. Those two needs point to different sensors and different cost structures.
Typical figures
| Sentinel-1 IW spatial resolution (wind product) | 500 m to 1 km (after spatial averaging for noise reduction; native pixel 5 × 20 m) |
| ASCAT spatial resolution | 12.5 km (coastal product) / 25 km (standard product) |
| Sentinel-1 revisit (two-satellite constellation) | 3 days at mid-latitudes; 6 days with single satellite |
| ASCAT revisit | Near-daily global coverage (MetOp-A/B/C combined) |
| Radar frequency | C-band, 5.405 GHz (Sentinel-1); 5.255 GHz (ASCAT) |
| Wind speed retrieval range | Approximately 2–25 m/s reliable; degraded accuracy above 25 m/s and below 2 m/s |
| Typical wind speed accuracy | ~1.5–2 m/s RMS against buoy data under moderate conditions |
| Wind direction accuracy | ~20–25 degrees RMS after NWP-based ambiguity removal |
| ASCAT operational latency | Under 3 hours via Global Telecommunication System |
| Sentinel-1 archive depth | 2014 to present, freely accessible via Copernicus Data Space |
Analytics Satellize can run
| Gridded wind speed map | CMOD5.N applied to Sentinel-1 VV sigma-naught with incidence-angle correction | GeoTIFF layer at 500 m resolution, delivered per acquisition pass |
| Wind direction field with ambiguity removed | 2DVAR ambiguity removal using ECMWF short-range forecast as first guess | Vector field GIS layer (GeoJSON or NetCDF) with direction confidence flag |
| Daily basin-scale wind summary | ASCAT L2 product ingestion and spatial aggregation | Daily PDF report with annotated wind speed contour maps for defined area of interest |
| Storm-cell intensity time series | Sentinel-1 multi-pass wind retrieval over a defined storm track, keyed to best-track data | CSV time series of peak wind speed and spatial extent per overpass |
| Offshore wind resource screening layer | Multi-year ASCAT climatology aggregated to mean and percentile wind speed statistics | GIS polygon layer with wind speed statistics per grid cell, exportable to GIS or energy modelling tools |
| Rain-flagged wind quality mask | Co-located IMERG or ERA5 precipitation field applied as a quality filter to wind retrievals | Masked wind raster with rain-contaminated pixels excluded and flagged |
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.