GNSS reflectometry for soil moisture and sea surface roughness
GNSS-R exploits reflected navigation signals to infer surface dielectric properties and roughness. CYGNSS gives sub-daily tropical revisit that passive microwave cannot match at comparable scale, but a ~25 km resolution floor and vegetation ambiguity set real limits.
Sensors
- CYGNSS (NASA, 8-satellite constellation): Receives reflected GPS L1 C/A signals (1575.42 MHz) from a downward-facing antenna. Median revisit in the tropics (38°S–38°N) is roughly 2.8 hours; spatial footprint per specular point is approximately 25 km for soil moisture products. Launched December 2016.
- Spire GNSS-R (commercial, 100+ satellite constellation): Operates GPS and GLONASS reflectometry alongside radio occultation on the same smallsats. Offers global coverage beyond CYGNSS's tropical band, with revisit improving as the constellation grows. Spatial resolution comparable to CYGNSS at L-band.
- TechDemoSat-1 (UK, heritage): Carried the SGR-ReSI GNSS-R receiver; demonstrated ocean wind speed retrieval and soil moisture sensitivity from 2014 until end of mission. Established the delay-Doppler map processing chain that CYGNSS later scaled.
- SMAP (NASA passive microwave, complementary): L-band radiometer at 1.4 GHz; 36 km native soil moisture resolution, 2–3 day global revisit. Not a GNSS-R instrument, but its soil moisture retrievals are the standard validation reference and can be fused with CYGNSS to improve temporal density.
What a reflected navigation signal actually encodes
Every GPS or Galileo satellite broadcasts a ranging signal. Most receivers look upward and treat the reflected multipath as noise. GNSS-R turns that logic around: a downward-facing antenna captures the specular reflection from the surface below, and the receiver records a delay-Doppler map (DDM), a two-dimensional histogram of signal power against time delay and Doppler frequency shift.
The DDM shape carries two distinct pieces of information. Its peak power and the rate at which power falls away from the specular point encode surface roughness. The absolute magnitude of the reflected signal, expressed as the bistatic radar cross-section, depends on the dielectric constant of the surface. Liquid water has a dielectric constant near 80; dry soil sits around 3 to 5. Wet soil falls somewhere between, and the relationship between volumetric soil moisture and dielectric constant at L-band is well described by the Dobson mixing model, published in the 1980s and still the basis of most operational retrievals. That physics is what makes the technique work.
Why the tropics and why sub-daily
CYGNSS was designed primarily to measure ocean surface wind speed inside tropical cyclones, where conventional scatterometers saturate and passive microwave retrievals degrade under rain. The eight-satellite constellation achieves a median revisit of about 2.8 hours between 38°S and 38°N, a cadence no passive microwave sensor at comparable spatial scale can approach. SMAP revisits any given tropical point every two to three days; AMSR2 is similar.
That temporal advantage matters most for applications where soil moisture changes fast: flash-flood forecasting, irrigation scheduling in monsoon regions, and post-storm agricultural damage assessment. A soil moisture map that is 48 hours old after a convective event is often useless for operational decisions. CYGNSS, for all its resolution limitations, can capture the drying curve after a rainfall event in near-real time.
Outside the tropics, CYGNSS coverage degrades sharply. Spire's polar-inclusive constellation partially fills that gap, though its GNSS-R data products are less mature in the peer-reviewed literature than CYGNSS's.
The vegetation ambiguity: an honest account
Soil moisture retrieval from GNSS-R is not clean. Vegetation intercepts and scatters the reflected signal before it reaches the satellite antenna, attenuating the bistatic cross-section in a way that mimics drier soil. The vegetation optical depth at L-band is lower than at shorter wavelengths, which is why L-band was chosen for SMAP, but it is not zero. Dense tropical forest, tall maize at peak biomass, and flooded rice paddies all introduce retrieval errors that are difficult to separate from true soil moisture variation without auxiliary data.
The standard correction uses a vegetation water content (VWC) estimate derived from ancillary optical indices, typically NDVI from MODIS or Sentinel-2, fed into a two-parameter tau-omega model. The correction works reasonably well for sparse to moderate canopies. For VWC above roughly 5 kg/m², the uncertainty in the CYGNSS soil moisture product is large enough that the retrieval is better treated as qualitative. Published validation studies against in-situ networks such as SCAN and COSMOS-UK report unbiased RMSE values in the range of 0.04 to 0.07 m³/m³ for vegetated agricultural land, rising significantly under forest.
Sea surface roughness: the original application
Over open ocean, the DDM peak power and its spread encode mean square slope, which correlates with the 10-metre wind speed. CYGNSS wind speed retrievals are validated against buoys and numerical weather prediction reanalysis, with reported RMSE of roughly 1.4 to 1.8 m/s for wind speeds below about 20 m/s. Above that threshold, the relationship between roughness and wind speed becomes less well constrained, partly because wave age and swell complicate the sea surface geometry.
The technique also responds to surface films. Oil slicks, biogenic surfactants, and rain-roughened patches all alter the DDM in ways that are distinguishable from clean wind-driven roughness, at least statistically. This sensitivity is a research-stage capability rather than an operational product, but it points toward future applications in marine pollution monitoring that complement SAR-based slick detection.
Resolution floor and what it means for buyers
The 25 km figure quoted for CYGNSS soil moisture is not a pixel size in the optical sense. It is the effective spatial scale over which the bistatic reflection integrates, set by the geometry of the specular zone and the coherent integration time. You cannot sharpen it by processing more carefully: the physics of the glistening zone imposes the limit.
That floor rules out field-scale precision agriculture at current constellation sizes. A 10-hectare field is invisible. What GNSS-R does well is regional and catchment-scale soil moisture monitoring, drought early warning across large agricultural zones, and inputs to hydrological models that operate at 25 km or coarser. Buyers who need field-scale soil moisture should look at SAR backscatter retrieval at C- or L-band, which can reach 10 to 100 m resolution at the cost of lower revisit and greater sensitivity to surface roughness confounders. GNSS-R and SAR are genuinely complementary: one gives temporal density, the other spatial detail.
Satellize incorporates CYGNSS-derived soil moisture layers into regional agricultural analytics, including work adjacent to the Tonga crop-estimation programme, where soil water availability between optical observation windows is a key model input.
Processing chain from DDM to delivered product
Raw CYGNSS data are distributed as Level 1 DDMs through NASA's EOSDIS. The standard NASA Level 3 soil moisture product applies a land surface model-informed inversion, the UCAR/CU retrieval, which regresses the normalised bistatic radar cross-section against VWC-corrected surface reflectivity on a 0.1-degree grid with daily compositing. Independent groups have published alternative retrievals using machine learning trained against SMAP, which tend to reduce bias in densely vegetated regions at the cost of interpretability.
Latency from observation to Level 3 product is typically 24 to 48 hours for the standard NASA stream. Near-real-time applications can work from Level 1 data with custom processing, reducing latency to a few hours, but require in-house expertise in bistatic radar geometry and DDM inversion. Spire offers processed wind speed and soil moisture outputs via API with latency closer to one to two hours, though independent validation of those products is less extensive in the published literature than for CYGNSS.
Typical figures
| Effective soil moisture resolution (CYGNSS L3) | ~25 km (0.1° grid) |
| Tropical revisit (CYGNSS, 8 satellites) | Median ~2.8 hours between 38°S and 38°N |
| Coverage (CYGNSS) | 38°S to 38°N; Spire extends to polar regions |
| Signal frequency | GPS L1: 1575.42 MHz; GLONASS L1: ~1602 MHz (Spire) |
| Soil moisture retrieval accuracy (vegetated agricultural land) | Unbiased RMSE ~0.04–0.07 m³/m³ (published validation, SCAN/COSMOS networks) |
| Ocean wind speed RMSE (CYGNSS) | ~1.4–1.8 m/s for winds below 20 m/s |
| Standard product latency (NASA Level 3) | 24–48 hours; near-real-time custom processing ~2–4 hours |
| Archive depth (CYGNSS) | December 2016 to present |
| Data access | NASA EOSDIS (open); Spire via commercial API |
| Vegetation water content limit for reliable retrieval | Below ~5 kg/m²; high uncertainty above this threshold |
Analytics Satellize can run
| Tropical soil moisture anomaly maps | CYGNSS Level 3 normalised bistatic cross-section inversion with MODIS VWC correction (tau-omega model) | Weekly GIS raster layers (0.1° resolution) showing departure from climatological mean, flagged by vegetation density class |
| Post-rainfall drying curve time series | Daily CYGNSS composites fused with SMAP climatology to fill polar gaps and reduce noise | CSV or GeoTIFF time series per region of interest, suitable for hydrological model input |
| Drought early warning index | Percentile ranking of GNSS-R soil moisture against 2016-to-present CYGNSS archive, cross-validated against SMAP | Monthly bulletin with sub-regional drought severity classification and trend chart |
| Flood inundation extent proxy | Detection of anomalously high bistatic reflectivity indicating standing water, using DDM coherence metrics | Alert layer (GeoJSON) triggered when reflectivity exceeds threshold, with confidence flag and optical cross-check note |
| Ocean wind field analysis for tropical cyclone monitoring | CYGNSS Level 2 wind speed retrieval using minimum variance estimator, validated against ECMWF ERA5 | Gridded wind speed GeoTIFF per overpass, with storm-track annotation and gap analysis relative to conventional scatterometer coverage |
| Agricultural soil water availability index | GNSS-R soil moisture fused with Sentinel-2 NDVI and a simple crop water demand model, following methods used in Pacific island crop analytics | Bi-weekly report per agricultural zone with irrigation advisory flag and data-quality confidence score |
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.