Ocean surface wind speed retrieval via GNSS reflectometry
Reflected GPS and Galileo signals carry a measurable imprint of ocean roughness. GNSS-R turns that accidental return into wind speed and wave height at roughly 25 km resolution, globally, with no transmitter of your own.
Sensors
- NASA CYGNSS constellation: Eight microsatellites in a 35-degree inclination orbit, each carrying a Delay Doppler Mapping Instrument (DDMI) with four simultaneous specular reflection tracks. Mean revisit over the tropics and subtropics is roughly 7 hours; individual footprints are approximately 25 km in the along-track direction. Wind speed retrieval range is published as 0–70 m/s, with best accuracy in the 5–20 m/s regime.
- ESA TechDemoSat-1 (TDS-1): UK-built technology demonstrator launched 2014, carrying the SGR-ReSI GNSS-R receiver. Published studies from TDS-1 validated delay-Doppler map processing against ECMWF wind fields and Jason-2 altimetry, establishing the empirical basis for later operational missions. No longer operational, but its open dataset remains a reference for algorithm development.
- Spire Global GNSS-R payload: Spire's LEMUR-2 cubesat fleet carries dual-use GNSS receivers capable of reflectometry mode alongside radio occultation. The commercial constellation offers higher revisit than CYGNSS at mid-latitudes, though published independent validation of wind retrievals from Spire's GNSS-R data is less extensive than for CYGNSS as of public literature.
- ESA SMOS (experimental GNSS-R mode): SMOS was designed for L-band passive radiometry, not reflectometry, but experimental processing of its data has demonstrated sensitivity to ocean surface roughness. This is a research-grade product; it is not an operational GNSS-R mission and the retrievals carry higher uncertainty than dedicated receivers.
Why a navigation signal becomes a wind gauge
A GPS L1 signal (1575.42 MHz) or Galileo E1 signal reaches the ocean surface at a grazing angle and partially reflects toward space. The geometry is bistatic: transmitter (GNSS satellite), reflecting surface, and receiver (the GNSS-R spacecraft) form a triangle, with the specular point sitting at the triangle's base. The receiver is not the intended destination of the signal. It is an eavesdropper on a reflection.
Ocean roughness scatters the reflection. A glassy sea produces a sharp, coherent return concentrated near the specular point. A wind-roughened sea smears the return across a broader range of delays and Doppler shifts, because waves at different heights and velocities each contribute a slightly different path length and frequency. The delay-Doppler map (DDM) captures this spreading as a two-dimensional power distribution. The shape of that distribution, specifically the slope of its trailing edge and the width of its Doppler ridge, encodes surface roughness, which is in turn a proxy for near-surface wind speed.
From raw DDM to a wind speed number
The standard retrieval chain starts with the normalised bistatic radar cross-section, sigma-naught, computed from the DDM peak power after correcting for GNSS transmit power, receiver gain, path losses, and the bistatic radar equation. Sigma-naught is then mapped to wind speed through an empirical geophysical model function (GMF), analogous to the GMFs used in conventional scatterometry. CYGNSS uses a GMF trained against ECMWF ERA5 reanalysis winds and validated against moored buoy observations; published root-mean-square errors for CYGNSS Level 2 wind speed retrievals are in the range of 1.4–2.0 m/s for winds between 5 and 20 m/s.
Significant wave height (SWH) adds a second retrieval. The DDM waveform's leading-edge slope is sensitive to the distribution of sea surface heights, not just mean roughness, so SWH can be estimated independently of wind speed, though with somewhat higher uncertainty. Published CYGNSS SWH retrievals show RMS errors of roughly 0.5–1.0 m against Jason-series altimeter data, depending on sea state regime.
Calibration is an ongoing concern. The GNSS transmit power varies by satellite and block type; the bistatic geometry changes continuously; and the empirical GMF is trained on open-ocean conditions that may not transfer well to fetch-limited coastal seas or extreme weather. CYGNSS science team publications document these issues openly, which is worth knowing before treating the product as equivalent to a calibrated scatterometer.
Resolution limits and what they mean operationally
The effective footprint of a single GNSS-R specular reflection is not a point. The first Fresnel zone at L-band, for a CYGNSS-like orbit at roughly 520 km altitude, is approximately 6–7 km in diameter under calm conditions. In practice, DDM integration over one second smears the footprint to around 25 km along-track, and the across-track extent depends on sea state. This is coarser than a C-band scatterometer such as ASCAT, which achieves 12.5 km gridded products, and far coarser than a synthetic aperture radar.
What GNSS-R trades in resolution it partially recovers in coverage and cost. CYGNSS's eight spacecraft together produce roughly 32 simultaneous specular tracks across the tropics. The constellation does not cover polar regions above about 38 degrees latitude, which is a hard geometric limit set by the GNSS satellite inclinations and CYGNSS's own 35-degree orbit. Mid-latitude and polar users need to look elsewhere. Spire's higher-inclination orbits extend coverage, but with the validation caveats already noted.
Rain, ice, and other contamination the physics cannot hide
Rain is the most operationally significant interference. Heavy precipitation attenuates the L-band signal and adds surface roughness that is not wind-driven, biasing the sigma-naught retrieval high. CYGNSS published studies show that rain rates above roughly 10 mm/hour introduce wind speed overestimates of 2–5 m/s. Flagging rain-contaminated retrievals requires collocated precipitation data, typically from GPM IMERG, which adds latency.
Sea ice is a different problem. Ice surfaces reflect very differently from open water, and the retrieval algorithms are not designed for ice. The CYGNSS team publishes ice flags, but marginal ice zones in spring and autumn are ambiguous. Inland water bodies and coastal areas within a few kilometres of land also contaminate DDMs through mixed reflections, so coastal wind retrievals should be treated with caution inside roughly 25 km of the shoreline.
Ionospheric scintillation, which is covered in a separate reference page in this library, can corrupt the phase coherence of the reflected signal, particularly at low latitudes during solar maximum. The effect on DDM-based wind retrievals is modest compared to its impact on radio occultation, but it is not zero.
Where the product earns its place in an operational picture
The honest use case for GNSS-R wind retrievals is gap-filling, not replacement. In the tropics, where conventional scatterometer revisit is sparse and tropical cyclone intensity estimation is critical, CYGNSS was specifically designed to observe the inner wind field of hurricanes. Published results show CYGNSS successfully retrieved winds inside Hurricane Maria (2017) where conventional scatterometers saturate or lose coherence. That is a genuine capability advantage.
For routine maritime weather support, GNSS-R provides a useful independent check on NWP model winds, particularly in data-sparse ocean basins. Shipping route optimisation, offshore energy planning, and search-and-rescue drift modelling all benefit from denser wind observations, even at 25 km resolution, when the alternative is a six-hour model forecast with no observational anchor.
Satellize can ingest CYGNSS Level 2 and Level 3 products alongside scatterometer and altimeter data to produce blended wind analyses for specific ocean regions, as part of a broader maritime or environmental monitoring programme. The analytics methodology draws on the same collocation and bias-correction approaches published in the CYGNSS science literature. Buyers wanting to understand what a blended product would look like for their region of interest can request a demonstration run against a defined bounding box and time window.
Typical figures
| Effective footprint (single DDM) | Approximately 25 km along-track at 1-second integration; first Fresnel zone 6–7 km diameter under calm conditions |
| Gridded product resolution (CYGNSS L3) | 0.25-degree grid (~28 km at equator), daily composite |
| Mean revisit (CYGNSS, tropics) | ~7 hours median; degrades to ~12 hours at 30–35 degrees latitude |
| Latitude coverage (CYGNSS) | 38°S to 38°N; Spire GNSS-R extends to higher latitudes with less published validation |
| Signal frequencies used | GPS L1 (1575.42 MHz), GPS L2 (1227.60 MHz), Galileo E1 (1575.42 MHz) |
| Wind speed retrieval range | 0–70 m/s; published RMS error 1.4–2.0 m/s for 5–20 m/s regime (CYGNSS L2) |
| Significant wave height RMS error | Approximately 0.5–1.0 m against Jason-series altimetry |
| Latency (CYGNSS near-real-time product) | ~4–6 hours from observation to NRT Level 2 delivery via NASA PO.DAAC |
| Archive depth (CYGNSS) | Operational since March 2017; full mission archive publicly available |
| Rain contamination threshold | Retrievals biased high by 2–5 m/s at rain rates above ~10 mm/hour; GPM flagging recommended |
Analytics Satellize can run
| Tropical cyclone inner-core wind analysis | CYGNSS L2 wind retrievals collocated with best-track position data; sigma-naught inversion using published GMF | Gridded wind field GIS layer per storm passage, with rain-flag annotation and comparison to NWP background |
| Blended ocean wind analysis (GNSS-R + scatterometer) | Optimal interpolation or variational merging of CYGNSS L3, ASCAT, and ERA5 background; bias correction per published collocation studies | Daily 0.25-degree NetCDF wind speed and direction grid for a defined ocean region, with per-cell uncertainty estimate |
| Significant wave height time series for offshore asset planning | DDM leading-edge slope inversion against Jason/Sentinel-6 altimeter calibration, following published CYGNSS SWH retrieval methodology | Monthly SWH climatology and percentile tables for a defined bounding box, delivered as CSV and PDF summary |
| Data-gap assessment for NWP assimilation | Coverage density analysis of CYGNSS and Spire observation counts against model grid cells over a target region and period | Coverage statistics report identifying hours and grid cells with fewer than one observation per day, supporting assimilation feasibility decisions |
| Extreme wind event detection and alert | Threshold exceedance monitoring on CYGNSS NRT L2 wind retrievals within a defined ocean polygon; rain-flagged retrievals excluded | Automated alert (JSON feed or email) when retrieved winds exceed a client-defined threshold, with DDM quality metrics attached |
| Scatterometer cross-calibration validation | Collocation of CYGNSS L2 with ASCAT or HSCAT observations within 50 km and 30 minutes; statistical bias and scatter analysis per published validation protocols | Bias characterisation report for a specific ocean basin and season, suitable for input to a data assimilation quality-control scheme |
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.