Sensors
- CYGNSS constellation (NASA): Eight microsatellites in a 35-degree-inclination orbit, each carrying a Delay-Doppler Mapping Instrument (DDMI) that records reflected GPS L1 C/A signals. Designed for tropical ocean wind retrieval, but published studies show it captures specular reflections from flat polar ice where geometry permits. Inclination limits coverage to roughly 38 degrees latitude, excluding Greenland's interior and all of Antarctica.
- ESA HydroGNSS: ESA's dedicated GNSS-R mission, targeting hydrological variables including sea-ice and ice-sheet surface properties. Designed to receive reflected signals from GPS, Galileo and GLONASS across L-band. Published mission documentation cites expected soil-moisture and freeze-thaw products; ice-sheet altimetry is a secondary science objective with anticipated vertical sensitivity in the decimetre range over smooth surfaces.
- Spire GNSS-R payload: Spire operates a commercial LEO constellation carrying GNSS-R receivers alongside radio-occultation payloads. The constellation exceeds 100 satellites across a range of inclinations, including polar orbits, which extends coverage to high latitudes inaccessible to CYGNSS. Spire publishes GNSS-R data products commercially; ice-surface retrievals are an active research area rather than a fully validated operational product.
- TechDemoSat-1 and UK DMC (historical record): TechDemoSat-1 (2014-2017) carried the SGR-ReSI GNSS-R receiver and produced early delay-Doppler maps over sea ice, establishing that specular geometry over smooth ice yields coherent waveforms distinguishable from ocean returns. This archive underpins the theoretical framework used to interpret CYGNSS and HydroGNSS data over ice.
Why reflected GPS signals know something about ice elevation
Conventional satellite altimetry fires a radar pulse downward and times the echo. GNSS reflectometry is passive and bistatic: the transmitter is a navigation satellite, the receiver is a separate LEO spacecraft, and the signal of interest is the GPS or Galileo carrier that bounced off the surface. Over the ocean, the reflected waveform is smeared by wave roughness, and that smearing encodes wind speed. Over flat ice, the geometry changes fundamentally.
A smooth, flat ice surface produces a coherent specular reflection. The delay-Doppler map, the two-dimensional representation of signal power as a function of code delay and Doppler shift, becomes sharp and concentrated rather than diffuse. The position of that specular point in delay space corresponds directly to surface elevation relative to the GNSS satellite orbits, whose positions are known to centimetre accuracy from precise orbit determination. The technique therefore inherits the geometric precision of the GNSS constellation itself, limited in practice by receiver noise, surface roughness, and uncertainty in the specular-point location.
What the delay-Doppler map reveals beyond simple height
Elevation is the headline product, but the shape of the delay-Doppler map carries additional information. Over crevassed or sastrugi-covered ice, the waveform broadens because multiple sub-facets at different heights contribute incoherently. That broadening is a roughness proxy. Published studies using CYGNSS data over the Greenland periphery have used waveform trailing-edge slope to distinguish smooth refrozen ice from rougher ablation-zone surfaces.
Melt-pond fraction is a second retrievable. During Arctic summer, melt ponds form on sea ice and on the lower margins of ice sheets. Their flat, liquid surfaces produce anomalously strong coherent returns, raising the effective reflectivity well above dry snow. A sudden increase in reflected signal power at a known location and time is a reliable indicator of ponding, even when optical sensors are cloud-obscured. This is not a trivial point: melt-pond fraction is a leading predictor of summer sea-ice extent, and cloud cover is near-permanent over active melt zones.
Surface slope can also be inferred from the spatial gradient of retrieved elevation across successive specular points. Over the flat interior of the Greenland or Antarctic ice sheets, this works well. Near the margins, where slopes steepen and crevasses interrupt the surface, the specular assumption breaks down and retrievals become unreliable. Honest interpretation requires knowing which regime you are in.
Precision, coverage and the honest limits
Over flat interior ice, published validation studies comparing CYGNSS-derived surface heights to ICESat-2 laser altimetry report vertical RMS differences in the range of 0.3 to 0.8 metres, depending on surface slope and averaging window. That is not competitive with ICESat-2, which achieves centimetre-level precision along its ground track, or with CryoSat-2's synthetic-aperture radar altimeter. GNSS-R's advantage is not precision. It is coverage density and revisit.
CYGNSS produces millions of specular-point observations per day globally, with a median revisit of roughly six hours at mid-latitudes. Its 35-degree inclination means it never sees Greenland's interior above about 38 degrees north, and it cannot observe Antarctica at all. Spire's polar-orbiting GNSS-R receivers partially fill this gap, but their ice-altimetry products are not yet at the validation maturity of CYGNSS ocean products. HydroGNSS, once operational, is expected to extend validated coverage to higher latitudes.
Cloud cover is irrelevant to GNSS-R. L-band signals penetrate cloud, precipitation and polar darkness without attenuation. This is the method's clearest operational advantage over optical altimetry and over passive microwave systems that must resolve surface emission rather than a reflected navigation signal. The limit is terrain, not weather.
Sea ice versus ice sheet: different surfaces, different retrieval strategies
Sea ice and continental ice sheets both produce coherent GNSS-R returns, but the interpretation differs. Sea ice is thin (typically under three metres for first-year ice), fragmented, and drifting. The metric of interest is usually ice extent, concentration, and the presence of leads or melt ponds, rather than absolute elevation. GNSS-R reflectivity and waveform coherence are well-suited to distinguishing open water, new ice, and multi-year ice, because each surface type has a characteristic dielectric constant and roughness that shapes the delay-Doppler map.
Continental ice sheets are thick, slow-moving, and topographically complex. Here, elevation change over time is the primary geophysical signal: Greenland is losing mass at a rate that is well-documented by GRACE gravity measurements and ICESat-2 laser altimetry. GNSS-R cannot replace those systems for precise mass-balance accounting, but it can provide high-density surface-height snapshots that help interpolate between ICESat-2 ground tracks, flag rapid surface changes such as supraglacial lake drainage events, and monitor outlet-glacier surfaces where ICESat-2 repeat coverage is sparse.
Where this fits in a polar monitoring programme
No single sensor covers polar ice adequately. CryoSat-2 provides precise radar altimetry but with a 369-day exact repeat and a 30-day sub-cycle. ICESat-2 achieves centimetre precision but only along six ground-track pairs. Synthetic-aperture radar from Sentinel-1 maps surface velocity and detects calving events but does not directly measure elevation. GNSS-R fills a specific niche: frequent, weather-independent surface observations at decimetre-scale vertical precision over flat terrain, available from existing navigation infrastructure at near-zero marginal transmission cost.
Satellize builds analytics pipelines on open and commercial GNSS-R data streams for clients who need to integrate ice-surface monitoring into broader sovereign earth-observation programmes. The methodology is the same class used in the Tonga crop-estimation work: ingest, quality-filter, and fuse multi-source geophysical retrievals into decision-ready outputs rather than raw data files. For polar applications, that typically means a change-detection layer updated on each constellation pass, flagging anomalous reflectivity or elevation departures against a rolling baseline.
If your programme requires ice-sheet monitoring and you are deciding how GNSS-R fits alongside radar altimetry and optical assets, the right conversation starts with your revisit requirement and your tolerance for vertical uncertainty. Those two numbers determine whether GNSS-R is your primary sensor, a gap-filler, or a validation cross-check.
Typical figures
| Specular footprint diameter (flat ice) | Approximately 0.5 to 1 km for coherent returns over smooth ice; broadens to several km over rough or crevassed surfaces |
| Vertical precision (flat interior ice) | 0.3 to 0.8 m RMS against ICESat-2 reference, from published CYGNSS validation studies; degrades to >1 m over sloped or rough terrain |
| Revisit rate | CYGNSS: median ~6 hours at mid-latitudes (35°N/S limit); Spire polar orbits: variable, typically 2-6 passes per day at 80°N/S |
| Signal frequency | GPS L1 (1575.42 MHz), Galileo E1/E5, GLONASS L1; all L-band, cloud- and darkness-transparent |
| Latitude coverage | CYGNSS: ±38°; Spire and HydroGNSS: polar-inclusive; no GNSS-R system provides nadir coverage at the geographic poles |
| Minimum detectable surface change | Decimetre-scale height change detectable over flat ice with multi-pass averaging; sub-metre melt-pond reflectivity anomalies detectable in single passes |
| Archive depth | CYGNSS: from March 2017; TechDemoSat-1 GNSS-R: 2014-2017; Spire commercial: from approximately 2019 |
| Data latency (raw to product) | CYGNSS science data: typically 6-24 hours from observation to NASA archive; near-real-time experimental products available for some parameters |
| Delivery format | NetCDF (NASA CYGNSS archive); GeoTIFF or GIS-ready raster for processed elevation and reflectivity grids |
Analytics Satellize can run
| Surface elevation anomaly map | Delay retrieval from peak of delay-Doppler map, converted to height via precise GNSS orbit and receiver position; differenced against multi-pass baseline | GeoTIFF anomaly grid updated per constellation pass, flagging departures >0.5 m from 30-day rolling mean |
| Melt-pond fraction estimate | Reflectivity excess above dry-snow baseline; coherent-to-incoherent power ratio from delay-Doppler waveform shape, following published CYGNSS sea-ice classification methods | Weekly gridded melt-pond fraction layer at 1 km posting, with confidence flag based on specular-point density |
| Surface roughness index | Trailing-edge slope and waveform width of delay-Doppler map; calibrated against CryoSat-2 surface roughness retrievals over overlapping tracks | Monthly roughness change report distinguishing smooth ablation ice, sastrugi fields and crevasse zones |
| Sea-ice type classification | Reflectivity and coherence thresholds discriminating open water, new ice, first-year ice and multi-year ice, based on published GNSS-R dielectric contrast studies | Daily ice-type raster for defined area of interest, formatted for ingestion into national ice-service GIS |
| Supraglacial lake drainage alert | Rapid reflectivity drop at a previously high-return location, indicating loss of ponded water; corroborated against Sentinel-1 SAR backscatter where available | Event alert with location, estimated timing and magnitude, delivered within 24 hours of detection |
| Multi-sensor elevation fusion layer | GNSS-R height retrievals co-registered and gap-filled against ICESat-2 ATL06 land-ice product and CryoSat-2 LRM/SARIn data using kriging interpolation | Quarterly fused elevation grid at 500 m posting, with per-pixel uncertainty estimate, in GeoTIFF and NetCDF |
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.