GNSS radio-occultation payloads
GNSS radio-occultation receivers derive vertical atmospheric profiles from signal refraction, delivering temperature and humidity data accurate enough to improve numerical weather prediction without a single imaging pixel.
A signal that curves is a measurement
When a GNSS satellite sets behind Earth's limb as seen from a low-orbit receiver, its signal passes through progressively deeper layers of atmosphere before it disappears. Refraction bends the ray path. The receiver measures the bending angle as a function of time, and from that angle profile it recovers the refractivity of each atmospheric layer. Refractivity is a function of pressure, temperature and water-vapour partial pressure, so with a hydrostatic assumption and an independent moisture estimate, you extract a full vertical temperature profile from roughly 40 km down to the lower troposphere.
The measurement is called a radio occultation, and it has one property that distinguishes it from almost every other remote-sensing technique: it is self-calibrating. The GNSS signal frequency is known to atomic-clock precision. There is no instrument drift, no need for vicarious calibration against a reference scene, no degradation over the satellite lifetime. A profile taken on day one of the mission is directly comparable to one taken five years later. For climate monitoring, that is not a convenience; it is the whole point.
Vertical resolution no passive sounder can match in the upper troposphere
A conventional infrared or microwave sounder retrieves temperature by inverting radiance measurements across broad weighting functions that typically span several kilometres in the vertical. Radio occultation inverts a phase measurement whose vertical resolution is set by the Fresnel zone diameter and the signal processing method. With full-spectrum Abel inversion, vertical resolution in the stratosphere reaches 0.5 to 1 km. In the moist lower troposphere, multipath and super-refraction degrade this to 1 to 2 km and introduce retrieval ambiguity, but the upper troposphere and lower stratosphere, precisely the region that matters most for numerical weather prediction (NWP) initialisation, is where RO performs best.
COSMIC-2, operated by NOAA and Taiwan's NSPO and launched in 2019, demonstrated what a constellation of six satellites in low-inclination orbits can deliver: roughly 5,000 occultation profiles per day globally, with temperature accuracy in the upper troposphere of around 0.5 K and dry-atmosphere refractivity accuracy of 0.1 to 0.2 per cent. Spire Global's commercial constellation, which at various times has operated more than 100 3U and 6U cubesats, uses a similar GNSS receiver approach and sells occultation profiles commercially. Both programmes draw on GPS, GLONASS, Galileo and BeiDou signals, which roughly triples the profile count compared with GPS-only receivers.
Receiver design: small hardware, demanding signal processing
The flight hardware is deceptively simple. A GNSS radio-occultation receiver consists of an open-loop tracking firmware layer, a low-noise amplifier, and one or two patch or helix antennas pointed toward the limb. The antenna gain pattern needs to be broad enough to track an occulting satellite across tens of degrees of elevation change in under a minute, yet reject zenith signals that would alias into the measurement. Mass for a cubesat-class receiver sits between 0.3 and 1.5 kg depending on the number of GNSS constellations supported and whether a zenith antenna for precise orbit determination is included. Power draw is typically 3 to 8 W during an occultation event.
The difficulty is in the signal processing, not the hardware. Below about 10 km, GNSS signals in the moist troposphere suffer severe multipath: multiple ray paths arrive simultaneously, and a phase-locked loop loses lock. Open-loop tracking, where the receiver records raw intermediate-frequency samples and processes them on the ground, was the solution pioneered by the COSMIC programme and is now standard. The raw data volume per occultation event is modest, perhaps 10 to 50 MB, but a constellation generating thousands of events per day requires a well-dimensioned ground pipeline. That pipeline, not the receiver, is often the bottleneck a new operator underestimates.
Where the measurement fails, and by how much
Radio occultation is not a complete weather sensor. The technique has a hard lower-boundary problem: below 1 to 2 km in the boundary layer, super-refraction gradients can cause ducting that makes the Abel inversion non-unique. Profiles that terminate above 1 km altitude are common over warm oceans and in the tropics, exactly where boundary-layer moisture is most important for convective forecasting. Several research groups have developed statistical methods to extend retrievals into the boundary layer, but these introduce prior assumptions that compromise the self-calibration advantage.
The technique also cannot distinguish temperature from moisture without an independent constraint. The standard approach assimilates RO as a refractivity or bending-angle observation directly into NWP models, letting the model's background field provide the temperature-moisture separation. This works well in the stratosphere, where moisture is negligible, but in the lower troposphere the retrieval is genuinely ambiguous without collocated radiosonde or microwave sounder data. A national programme that wants standalone moisture profiling will need a complementary sensor; RO alone is not sufficient.
Finally, coverage is a function of constellation size. A single satellite in a 550 km circular orbit generates roughly 400 to 600 occultation profiles per day. That is useful for climate trend monitoring but too sparse for operational NWP assimilation over a specific region. A national programme seeking regional weather benefit typically needs either a multi-satellite constellation or a commercial data-purchase agreement to supplement its own profiles.
Why NWP centres pay for this data above almost anything else
ECMWF and NOAA have both published observing-system experiments showing that removing all radio-occultation data from their assimilation systems degrades 24-hour forecast skill more than removing any single conventional satellite instrument type. The reason is the combination of global coverage, high vertical resolution in the upper troposphere, and the absence of instrument bias. Radiosonde networks are dense but geographically clustered over land in the northern hemisphere. Infrared sounders are blinded by cloud. Microwave sounders have coarse vertical resolution. RO has none of these problems in the upper troposphere and lower stratosphere, and NWP centres have learned to trust it accordingly.
For a government operating a small constellation, the practical implication is that even two or three satellites in complementary orbital planes can generate profiles that ECMWF, NOAA or a regional NWP centre will assimilate in near-real-time. Several commercial operators, including Spire, have monetised exactly this: selling occultation profiles under data licences to national met services that cannot yet justify their own constellation. A sovereign programme changes that equation: the data stays domestic, the processing pipeline is under national control, and the profiles can be withheld from international dissemination if the government chooses. That option has strategic value that a data-purchase contract cannot replicate.
Engineering parameters
| Receiver mass (cubesat class) | 0.3 to 1.5 kg depending on constellation support and POD antenna |
| Power consumption (during occultation) | 3 to 8 W |
| GNSS constellations supported (modern receivers) | GPS, GLONASS, Galileo, BeiDou (quad-constellation increases profile count ~3×) |
| Profiles per satellite per day (550 km orbit) | Approximately 400 to 600 |
| Vertical resolution (upper troposphere/stratosphere) | 0.5 to 1 km (Abel inversion); degrades to 1 to 2 km below 10 km |
| Temperature accuracy (dry atmosphere, 5 to 30 km) | ~0.5 K (COSMIC-2 validated) |
| Refractivity accuracy (5 to 30 km) | 0.1 to 0.2 per cent |
| Raw data volume per occultation event | 10 to 50 MB (open-loop IF samples) |
| Lower boundary limit | Retrievals frequently terminate at 1 to 2 km in moist tropical boundary layer |
| Instrument calibration requirement | None (self-calibrating via atomic-clock GNSS reference) |
One contract, one accountable engineer
Commissioned as one programme, not a stack of contracts: spacecraft, launch, ground segment, mission control, training and handover are priced together. Source-access terms and audit rights are agreed in writing before signature. Request an RO constellation sizing study.