GNSS radio occultation atmospheric profiling
GNSS radio occultation turns GPS and GLONASS signals into vertical slices of the atmosphere, yielding temperature, pressure and humidity profiles with roughly 100 m vertical resolution and no need for calibration targets.
Sensors
- COSMIC-2 / FORMOSAT-7: Six-satellite constellation in low-inclination orbits (~24°), operated jointly by NOAA and Taiwan's NSPO. Delivers roughly 5,000 globally distributed occultation profiles per day, with concentration in the tropics. Vertical resolution ~100 m in the lower troposphere, ~1 km in the stratosphere. Data latency for near-real-time assimilation feeds is typically under three hours.
- Spire GNSS-RO: Commercial constellation of over 100 small satellites in multiple orbital planes, providing more than 10,000 occultation profiles per day with improved polar and mid-latitude coverage relative to COSMIC-2. Spire sells near-real-time data feeds directly to national meteorological services and reanalysis programmes.
- Metop GRAS (GNSS Receiver for Atmospheric Sounding): Flown on Metop-A, -B and -C in sun-synchronous polar orbit. GRAS provides rising and setting occultations processed by EUMETSAT and ECMWF, with a long consistent archive stretching back to 2006 that underpins climate reanalysis work. Vertical resolution comparable to COSMIC-2.
- PlanetiQ: Commercial constellation designed specifically for high-quality GNSS-RO, targeting signal-to-noise performance that supports penetration deeper into the moist lower troposphere where water vapour retrievals are most uncertain. Operational data delivery began in the early 2020s.
What actually happens when a signal grazes the limb
A GNSS satellite transmits a precisely timed signal at L-band frequencies (roughly 1.2 and 1.6 GHz). As a LEO receiver satellite sets behind Earth's limb from the perspective of that transmitter, the signal path descends through progressively denser atmospheric layers. Each layer has a slightly different refractive index, which bends the ray and delays its arrival. The receiver measures the Doppler shift of the signal with extraordinary precision, from which the bending angle at each tangent height is calculated geometrically.
Bending angle profiles are then converted to refractivity profiles using an Abel transform, a mathematical inversion that assumes spherical symmetry of the atmosphere within the occultation geometry. Refractivity in the troposphere depends on temperature, pressure and water vapour partial pressure in a well-known physical relationship. With an independent pressure constraint, temperature and humidity can be separated. In the stratosphere, where water vapour is negligible, temperature is retrieved directly from refractivity alone with high confidence.
Self-calibrating: why meteorologists trust it
Most satellite sensors drift. Their calibration depends on onboard blackbodies, vicarious targets or cross-calibration with other instruments, all of which introduce uncertainty that accumulates over years. GNSS-RO does not work that way. The measurement is a time difference, derived from atomic clocks on the GNSS satellites whose performance is continuously monitored by ground networks. There is no instrument degradation term to account for, no radiometric drift. The bending angle is a geometric observable tied to fundamental constants.
This makes GNSS-RO one of the few satellite data types considered suitable for climate benchmarking without homogenisation corrections. The consistency between CHAMP (launched 2000), GRACE, COSMIC-1, Metop GRAS and COSMIC-2 records has been demonstrated in multiple reanalysis comparisons. For a buyer interested in long-term atmospheric trend monitoring, that stability is not a minor technical footnote. It is the point.
Vertical resolution and the honest limits of the retrieval
In the lower troposphere, GNSS-RO achieves vertical resolution of roughly 100 m, which is finer than most radiosonde ascent rates can resolve in practice. That figure degrades to around 1 km in the stratosphere, where the atmosphere is thinner and the bending signal weaker relative to noise.
Horizontal resolution is a different matter. Each occultation profile integrates signal along a ray path that is roughly 200 to 300 km long in the horizontal. The measurement is not a point observation; it is a weighted average along that path. In regions of sharp horizontal gradients, such as near fronts or coastlines, this smearing introduces errors that the Abel inversion cannot remove. Moist tropical boundary layers are particularly problematic: strong super-refraction can cause the Abel inversion to fail or produce spurious negative refractivity anomalies. Research groups at UCAR and ECMWF have published correction schemes, but the lower 1 to 2 km of the troposphere in humid conditions remains the most uncertain part of any GNSS-RO profile.
Coverage is also uneven. COSMIC-2's low-inclination orbits concentrate profiles in the tropics and subtropics, leaving polar regions relatively sparse. Spire's multi-plane architecture partially addresses this, but even at 10,000 profiles per day globally, the average spacing between profiles is several hundred kilometres. GNSS-RO supplements conventional observations; it does not replace them.
How numerical weather prediction actually uses this data
Operational centres including ECMWF, NOAA's NCEP, the UK Met Office and others assimilate GNSS-RO bending angle or refractivity profiles directly into their four-dimensional variational (4D-Var) or ensemble Kalman filter systems. The data enter as Level 2 bending angle profiles rather than retrieved temperature or humidity, because assimilating the raw observable avoids compounding retrieval assumptions with model assumptions.
ECMWF has published impact studies showing that GNSS-RO is among the most positively impactful observation types in their system on a per-observation basis, particularly for upper troposphere and lower stratosphere temperature analysis over the southern hemisphere and oceans where conventional data are sparse. The improvement in 24- to 72-hour forecast skill from GNSS-RO is measurable and well documented in the operational record.
For governments building sovereign weather capability, this matters practically. A national meteorological service that ingests GNSS-RO feeds into even a regional NWP model gains bias correction anchor points that are independent of any foreign radiosonde network. That independence has real value for island nations, polar territories and any state operating in a data-sparse region.
From raw profiles to operational products
The processing chain from raw excess phase measurements to assimilation-ready profiles is well standardised. UCAR's COSMIC Data Analysis Center (CDAAC) and EUMETSAT's ROM SAF both produce open-access Level 1b through Level 2 products. Commercial providers such as Spire deliver equivalent products with contractual latency guarantees that open-access archives do not offer.
Beyond weather forecasting, GNSS-RO profiles are used for ionospheric electron density monitoring (the same geometry works at higher altitudes), detection of atmospheric gravity waves, tropopause height mapping and, with sufficient archive depth, stratospheric temperature trend analysis relevant to ozone recovery monitoring. Satellize draws on open GNSS-RO archives as an atmospheric context layer in analytics work where surface conditions alone are insufficient, including crop-stress assessments where upper-air moisture and temperature structure affects interpretation of surface indices.
The practical question for any buyer is latency versus cost. Open CDAAC and ROM SAF data are free but arrive with delays of several hours to days. Near-real-time commercial feeds from Spire or PlanetiQ carry licensing costs but support operational forecast cycles. The right answer depends entirely on whether the application is research, reanalysis or operational.
Typical figures
| Vertical resolution (lower troposphere) | ~100 m |
| Vertical resolution (stratosphere) | ~1 km |
| Horizontal smearing per profile | 200–300 km along ray path |
| Profiles per day (COSMIC-2) | ~5,000 globally, tropics-concentrated |
| Profiles per day (Spire constellation) | >10,000 globally, multi-plane coverage |
| Signal frequencies | L1 (~1.575 GHz) and L2 (~1.227 GHz) GNSS bands |
| Near-real-time latency (commercial feeds) | Typically <3 hours from occultation |
| Archive depth (climate record) | CHAMP from 2001; Metop GRAS from 2006; COSMIC-1 from 2006 |
| Calibration dependency | None (SI-traceable atomic clock timing; self-calibrating) |
| Primary retrieved parameters | Refractivity, temperature, pressure, water vapour partial pressure, electron density (ionosphere) |
Analytics Satellize can run
| Atmospheric profile time series for a defined region | Aggregation and spatial binning of open CDAAC or ROM SAF Level 2 refractivity and temperature profiles over a client-specified domain | Monthly NetCDF or CSV archive with profile density statistics and data-gap flags |
| Tropopause height anomaly map | Cold-point and lapse-rate tropopause detection algorithms applied to GNSS-RO temperature profiles, following published UCAR CDAAC methods | Gridded GIS layer (GeoTIFF or GeoJSON) at monthly or seasonal cadence |
| NWP assimilation readiness assessment | Latency, coverage and quality-flag analysis of available GNSS-RO feeds against a client's forecast cycle requirements | Technical report comparing open-access versus commercial feed options with honest gap analysis |
| Upper-air moisture anomaly indicator for crop-stress context | Tropospheric refractivity anomaly derived from GNSS-RO profiles, used as an independent moisture proxy alongside surface vegetation indices | Seasonal summary report; compatible with surface analytics outputs |
| Stratospheric temperature trend analysis | Linear trend fitting on multi-mission merged GNSS-RO temperature records (CHAMP, COSMIC-1, Metop, COSMIC-2) using published homogenisation approaches from ROM SAF | Trend coefficient table with uncertainty ranges and mission-overlap consistency checks |
| Sovereign NWP data feed specification | Requirements analysis mapping client forecast domain, model resolution and assimilation scheme to appropriate GNSS-RO product level and latency class | Procurement specification document and vendor comparison matrix |
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.