GNSS radio occultation for atmospheric profiling
When a GNSS signal grazes Earth's limb before reaching a low-orbit receiver, its path curves in proportion to atmospheric density. Inverting that curve yields precise vertical profiles of temperature, pressure, and water vapour from the surface to roughly 60 km altitude, with no calibration drift and near-global coverage every six hours.
Sensors
- COSMIC-2 / FORMOSAT-7: Six satellites in low-inclination orbits (24°), jointly operated by NOAA and Taiwan's NSPO. Delivers roughly 5,000 occultation profiles per day concentrated in the tropics and mid-latitudes. Vertical resolution approximately 100–200 m in the lower troposphere, degrading to 1–2 km in the stratosphere. Data latency to NWP centres typically under three hours via direct broadcast.
- Spire Global Lemur-2: Commercial constellation of over 100 cubesats carrying GNSS-RO receivers alongside AIS and ADS-B payloads. Produces upwards of 10,000 occultation profiles per day across all latitudes, including polar regions underserved by COSMIC-2. Vertical resolution comparable to COSMIC-2; Spire licenses data commercially and contributes profiles to NOAA and EUMETSAT under data-sharing agreements.
- Metop GRAS (GNSS Receiver for Atmospheric Sounding): Flown on Metop-A, -B, and -C in sun-synchronous orbit. GRAS provides roughly 700 occultation profiles per satellite per day, with particularly good high-latitude coverage. Its long operational record (Metop-A launched 2006) anchors climate-trend studies because the measurement is self-calibrating against the stable GNSS signal frequency.
- GRACE-FO GPS receiver: Primarily a gravimetry mission, but the onboard GPS occultation receiver contributes atmospheric profiles as a secondary product. Useful for cross-validation and for demonstrating that commodity GNSS receivers on science missions can generate operationally useful soundings at negligible marginal cost.
- ESA Aeolus (context): Not a radio-occultation instrument. Aeolus used a UV Doppler wind lidar to profile tropospheric and lower-stratospheric wind fields, complementing RO temperature and humidity data in NWP assimilation. Aeolus re-entered in 2023 after demonstrating measurable forecast-skill improvement, particularly in the Southern Hemisphere where conventional upper-air observations are sparse.
What a bent signal reveals
A GNSS satellite transmits at precisely known frequencies. When a receiver in low Earth orbit tracks that signal as the transmitter sets below the Earth's limb, the signal passes through progressively denser air. Refraction bends the path; the receiver measures the resulting Doppler shift as a function of time. From that shift, the bending angle at each tangent altitude is computed geometrically. From bending angles, refractivity is retrieved by Abel inversion, a mathematical transform that assumes spherical symmetry of the atmosphere along the occultation path.
Refractivity in the troposphere depends on temperature, pressure, and the partial pressure of water vapour. In the stratosphere, where water vapour is negligible, refractivity depends almost entirely on density, which means temperature and pressure can be separated cleanly using the hydrostatic equation. The result is a vertical profile with roughly 100–200 m resolution near the surface, degrading to 1–2 km above 20 km. The measurement is traceable to the SI second via atomic clocks aboard GNSS satellites, which is why RO profiles do not drift over decades and are trusted as a climate benchmark.
The Abel inversion and its honest limits
The Abel inversion is elegant but carries assumptions. Spherical symmetry is violated whenever sharp horizontal gradients exist, such as at the edges of a frontal zone or inside a tropical convective system. In those cases, retrieved profiles can show artefacts, typically a spurious warm or cold layer a few kilometres thick. The error is usually small relative to NWP background uncertainty, but it matters for research applications that require sub-Kelvin accuracy.
The lower troposphere presents a separate problem. Multipath propagation occurs when the signal travels through a region where refractivity decreases sharply with height, a condition called a super-refraction layer. The receiver then receives signal energy along multiple paths simultaneously, and the single-path Abel inversion breaks down. Techniques such as Full Spectrum Inversion (FSI) and Phase Matching partially recover profiles through these layers, but below about 1 km in the tropics, data quality degrades and some profiles are simply discarded. Buyers who need boundary-layer humidity in tropical coastal environments should treat RO data with caution and cross-check against microwave sounders.
How profiles enter weather forecasts
Numerical weather prediction centres, including ECMWF, NCEP, and the UK Met Office, assimilate GNSS-RO bending angles or refractivity directly into their data assimilation systems. The measurement's self-calibrating character means it carries high observation weight relative to radiosonde data, which can have sensor biases and sparse geographic coverage. ECMWF has published that RO data from COSMIC and its successors produces the largest single-observation impact per profile of any satellite data type in their system, particularly in the Southern Hemisphere and over the oceans.
Latency is critical for operational NWP. COSMIC-2 achieves under three hours from occultation to delivery at NOAA's data centre via direct broadcast and near-real-time processing. Spire's commercial pipeline can match this. For climate reanalysis rather than operational forecasting, latency is irrelevant, and the focus shifts to long-term consistency. The CHAMP, GRACE, COSMIC-1, and Metop GRAS records, stretching back to 2001, provide a continuous, intercalibration-free dataset that reanalysis projects such as ERA5 use to anchor upper-tropospheric temperature trends.
Coverage, revisit, and what the numbers actually mean
COSMIC-2 alone produces roughly 5,000 profiles per day. Spire's constellation adds upwards of 10,000 more. Globally, the combined commercial and government RO constellation now generates on the order of 20,000 profiles per day. That sounds impressive until you map it: Earth's atmosphere contains roughly 500 million 100 km grid cells at 1° resolution. RO coverage is statistically uniform but individually sparse. Any given 500 km radius area receives perhaps two to five profiles per six-hour NWP window.
For synoptic-scale weather and climate monitoring, that density is sufficient. For mesoscale convective systems, tropical cyclone inner cores, or orographic precipitation events, it is not. RO is a complement to geostationary imagery and microwave sounders, not a replacement. Its unique value is vertical precision and calibration stability, not horizontal density.
Sovereign programme design considerations
A government building a national space programme faces a choice: buy RO data commercially, join a data-sharing agreement, or fly its own receiver. The receiver hardware is small. GNSS-RO payloads have flown on 3U and 6U cubesats. Taiwan's FORMOSAT-3 (the original COSMIC) demonstrated that a six-satellite constellation built for tens of millions of dollars could deliver operationally significant data. The marginal cost of adding an RO receiver to a satellite with another primary mission is low enough that several national agencies now treat it as standard.
The harder problem is processing. The Abel inversion pipeline, quality control, and NWP assimilation interface require sustained scientific infrastructure. Governments that lack that infrastructure typically access processed profiles through NOAA's COSMIC Data Analysis and Archive Center (CDAAC) or through commercial providers. Satellize works with clients to specify which processed data products, from raw excess phase to calibrated bending angles to final temperature profiles, are appropriate for a given operational requirement, and to integrate those products into existing forecast or monitoring workflows. For clients interested in building domestic profiling capacity, the Spire data-licensing model offers a practical entry point before committing to sovereign hardware.
Climate record integrity and what makes RO unusual
Most satellite instruments drift. Detectors age, optical coatings degrade, and successive generations of instruments require careful intercalibration to stitch together a climate record. GNSS-RO sidesteps most of this. The measurement depends on signal timing, which is referenced to atomic clocks, and on geometry, which is known precisely. There is no emitted energy to calibrate, no detector dark current to track. Successive constellations, CHAMP to GRACE to COSMIC-1 to Metop to COSMIC-2, show excellent agreement in the overlap periods without any post-hoc bias correction.
This makes RO data particularly valuable for detecting long-term trends in upper-tropospheric and lower-stratospheric temperature, a region where different climate models and different satellite records have historically disagreed. Published analyses using the merged RO record show warming trends in the tropical upper troposphere consistent with model predictions, though the magnitude and altitude structure remain active research questions. For treaty-verification or climate-reporting purposes, the self-calibrating nature of RO is a genuine advantage over instruments that require external reference targets.
Typical figures
| Vertical resolution (lower troposphere) | 100–200 m (Abel inversion; degrades below ~1 km in super-refraction conditions) |
| Vertical resolution (stratosphere) | 1–2 km above 20 km altitude |
| Altitude range | Surface to ~60 km (usable science product typically 2–50 km) |
| Profiles per day (combined RO constellation) | ~15,000–20,000 globally (COSMIC-2 ~5,000; Spire ~10,000+; Metop GRAS ~2,100) |
| Latency to NWP assimilation | Under 3 hours for near-real-time products (COSMIC-2, Spire) |
| Temperature accuracy (stratosphere) | ~0.5 K (self-calibrating; no instrument drift) |
| Horizontal footprint per profile | ~200–300 km along the occultation plane (not a point measurement) |
| Signal frequencies used | GPS L1/L2 (1575.42 / 1227.60 MHz); GLONASS, Galileo, BeiDou on newer receivers |
| Archive depth | Continuous global record from CHAMP (2001) onwards; COSMIC-1 2006–2020; Metop GRAS 2006–present |
| Standard data products | Excess phase, bending angle profiles, refractivity profiles, temperature/pressure/humidity profiles (netCDF via CDAAC or commercial API) |
Analytics Satellize can run
| NWP-ready bending angle profiles | Abel inversion of excess phase with FSI/Phase Matching in lower troposphere; quality-controlled against ECMWF background | NetCDF profile feed, latency under 3 hours, formatted for BUFR assimilation at national meteorological services |
| Tropical cyclone intensity assessment | RO profiles within 500 km of storm centre composited to characterise warm-core structure and tropopause height anomaly | Per-storm PDF report with annotated vertical temperature anomaly cross-sections, updated each NWP cycle |
| Upper-tropospheric humidity climatology | Refractivity-to-water-vapour retrieval in 0–15 km layer, monthly gridded composites at 2° resolution using merged COSMIC-2 and Spire profiles | GeoTIFF monthly grids and anomaly time series, suitable for climate-model validation |
| Tropopause height monitoring | Lapse-rate tropopause detection from RO temperature profiles following WMO definition; trend analysis over multi-year archive | Annual trend report with regional breakdowns; GIS layer of monthly tropopause height anomaly |
| Ionospheric total electron content profiles | Dual-frequency GNSS carrier phase differencing above 60 km; Abel inversion of ionospheric refractivity | Electron density profile dataset for space-weather monitoring or HF communications planning, delivered as CSV or netCDF |
| Climate-record consistency audit | Cross-comparison of RO temperature profiles against radiosonde archive and ERA5 reanalysis at client-specified stations and pressure levels | Statistical bias and trend report for use in national climate assessments or treaty reporting |
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.