RF emissions fingerprinting of vessels at sea
Signals-intelligence satellites in LEO detect and geolocate non-AIS radio-frequency emissions from vessels, exposing ships that deliberately hide. This page explains the physics, the geolocation maths, and the honest limits of the method.
Sensors
- HawkEye 360 Cluster Satellites (HE360): Clusters of three formation-flying satellites measure time-difference-of-arrival (TDOA) and frequency-difference-of-arrival (FDOA) across VHF, UHF and S-band. Published geolocation accuracy is on the order of hundreds of metres to low single-digit kilometres depending on signal duration and orbital geometry. Revisit at any given ocean point is typically several times per day across the growing constellation.
- Kleos Space RF Reconnaissance Constellation: Four-satellite clusters in formation flying, designed for TDOA/FDOA geolocation of VHF and UHF maritime emissions. Kleos targets the same frequency ranges used by ship-to-ship VHF radio and EPIRB-class beacons. Geolocation precision degrades with short signal bursts and with vessels in close proximity.
- Spire Global GNSS Radio Occultation (GNSS-RO) Satellites: Spire's LEO constellation carries software-defined radios capable of collecting AIS and some RF signals of opportunity. The GNSS-RO payloads are primarily atmospheric, but the platform architecture illustrates the broader trend of multi-mission RF collection from small satellites. Geolocation from a single Spire pass is less precise than a dedicated TDOA cluster.
- ELINT/SIGINT heritage systems (publicly documented): Orbital SIGINT has been documented in open literature since the Cold War. Modern commercial analogues inherit the same TDOA/FDOA geometry. The physics is well-established; what varies between operators is constellation size, baseline separation between satellites, and onboard processing latency.
What ships transmit when they think no one is listening
A vessel underway generates radio-frequency emissions continuously, whether its crew intends to or not. The X-band navigation radar sweeps every two to four seconds. The master may use VHF channel 16 for routine calls. A satellite phone uplink fires a burst when a crew member checks email. A fishing vessel's net-sonar pings on specific UHF frequencies. None of these emissions are AIS, and none of them stop simply because the AIS transponder has been switched off.
This is the premise of orbital RF fingerprinting: the electromagnetic environment around a vessel is a persistent, involuntary broadcast. Signals-intelligence satellites in low Earth orbit intercept these emissions and, by comparing arrival times and Doppler shifts across multiple receivers, compute a position for the emitting source. The technique is not new. What is new is that it is now available commercially, at maritime-surveillance prices rather than national-intelligence-programme prices.
TDOA and FDOA: the geometry behind the fix
Time-difference-of-arrival (TDOA) exploits the fact that a radio signal reaches two spatially separated receivers at slightly different moments. The difference in arrival time, multiplied by the speed of light, constrains the emitter to a hyperbola in space. A second pair of receivers produces a second hyperbola. Where the hyperbolas intersect is the emitter's position. In practice, a three-satellite cluster provides two independent TDOA measurements, which together yield a two-dimensional position fix on the Earth's surface.
Frequency-difference-of-arrival (FDOA) adds a second observable. Because the satellites are moving at roughly 7.5 kilometres per second relative to the vessel, each receiver sees a slightly different Doppler shift of the transmitted frequency. The difference between those shifts is sensitive to the emitter's position and velocity. Combining TDOA and FDOA tightens the position estimate and, in principle, resolves some of the geometric ambiguity that TDOA alone leaves open.
Published accuracy figures from HawkEye 360 suggest geolocation errors on the order of hundreds of metres for strong, sustained signals under favourable orbital geometry, degrading to several kilometres for brief bursts or when the satellite baseline is short. These are not GPS-class positions. A fix placing a vessel within two kilometres of its true location is still enormously useful for cueing optical or SAR collection, but it should not be confused with a precise track.
The proximity problem: when two vessels share a patch of ocean
The most practically difficult scenario is not the lone dark ship in open water. It is two or three vessels operating within a few kilometres of each other, all transmitting on similar frequencies. TDOA geometry cannot separate emitters whose positions differ by less than the system's resolution cell. The result is a blurred fix that could correspond to any of the vessels in the cluster.
This ambiguity is not a flaw in the mathematics; it is a physical limit imposed by the wavelengths involved and the baseline separations achievable from a small satellite cluster. Analysts address it by cross-referencing with AIS records (vessels that are transmitting AIS can be subtracted from the RF picture), by examining emission timing patterns, and by waiting for the vessels to separate before re-attempting a fix. Where vessels are deliberately co-located to obscure a transfer, the proximity itself becomes intelligence, even if the individual emitters cannot be resolved.
Frequency ranges and what they reveal
Different emission types carry different intelligence value. VHF maritime communications (156 to 174 MHz) are the most commonly intercepted class. A vessel using VHF channel 16 or a working channel places itself in a detectable state even if it transmits for only a few seconds. X-band radar (9.2 to 9.5 GHz) is harder to intercept from orbit because the beam is narrow and the antenna rotates, but the aggregate energy is detectable by sensitive receivers at close orbital ranges. Satellite phone uplinks, particularly Iridium and Inmarsat L-band bursts, are strong enough to be detected readily, and the burst pattern can sometimes distinguish vessel type by usage rhythm.
Frequency analysis also offers a crude form of equipment fingerprinting. Radar transmitters have characteristic pulse repetition intervals and pulse widths that differ between manufacturers and vessel classes. A fishing vessel's sonar operates on different frequencies from a tanker's depth sounder. These distinctions are probabilistic rather than definitive, and they require a library of reference signatures to be useful. Commercial operators have begun building such libraries, though the contents are not publicly documented in detail.
Orbital geometry and the revisit ceiling
A single LEO cluster passes over any given ocean point for a window of roughly four to eight minutes per orbit. During that window it can collect emissions and compute a fix. Outside that window, the vessel is unobserved by that cluster. With a constellation of multiple clusters, the aggregate revisit improves, but gaps remain. HawkEye 360's published revisit figures for mid-latitude ocean areas are several passes per day, which means a vessel can transit significant distances between observations.
Orbital inclination also matters. Polar or near-polar orbits provide good coverage of high-latitude fishing grounds and Arctic routes. Equatorial orbits are less useful for polar regions. The geometry of the satellite baseline at the moment of collection determines whether TDOA and FDOA produce a tight fix or a smeared one. An analyst receiving a position estimate should always ask what the orbital geometry was at collection time, not just what the nominal system accuracy is.
Latency from collection to delivered fix varies by operator and data pipeline. Near-real-time delivery (under an hour) is achievable when satellites have inter-satellite links or frequent ground-station contacts. Archive data, useful for retrospective investigations, can go back several years for established operators.
Putting the method to work: analytical products and honest caveats
RF emissions data is most powerful when fused with other sources. A TDOA fix placing an unknown emitter at a given position, combined with a SAR image showing a vessel at approximately that location, produces a much stronger identification than either source alone. Optical imagery can confirm vessel type if the timing aligns. AIS history can establish whether the vessel was transmitting normally before and after the RF detection, which is itself a behavioural indicator.
Satellize integrates RF detections from licensed commercial operators into multi-source maritime intelligence workflows, applying the same fusion logic it uses across open and commercial constellations. The honest caveat is this: RF geolocation at commercial accuracy levels is an indicator, not a verdict. It tells you a vessel was probably in a given area and was actively transmitting. It does not tell you the vessel's name, flag, or cargo. Those conclusions require corroborating evidence from AIS records, imagery, port call data, or ownership registry analysis.
For a government client building a maritime domain awareness picture, RF emissions data fills the gap between AIS coverage (voluntary and manipulable) and SAR coverage (high confidence but infrequent). It is particularly valuable in regions where shore-based VHF monitoring is absent and SAR tasking budgets are limited. The method is not infallible, but it is the closest thing available to a persistent, physics-based witness in open ocean.
Typical figures
| Frequency ranges covered | VHF 156–174 MHz; UHF 300–3000 MHz; L-band 1–2 GHz (satellite phone); X-band radar 9.2–9.5 GHz (operator-dependent) |
| Geolocation method | TDOA and FDOA using formation-flying satellite clusters (typically 3 satellites per cluster) |
| Typical geolocation accuracy | Hundreds of metres to low single-digit kilometres, depending on signal duration, strength, and orbital geometry at time of collection |
| Revisit rate (mid-latitude ocean) | Several passes per day per constellation; gaps of hours between observations are normal |
| Minimum detectable signal | Sustained VHF transmissions of a few seconds or more; brief bursts degrade fix quality significantly |
| Proximity resolution limit | Emitters within approximately 1–5 km of each other may not be individually resolved; figure is geometry-dependent |
| Latency (collection to delivered fix) | Under 1 hour for near-real-time pipelines; archive data available for retrospective analysis |
| Archive depth | Several years for established commercial operators (HawkEye 360 operational since 2018) |
| Coverage | Global ocean; polar coverage dependent on orbital inclination of specific constellation |
| Delivery formats | Geolocated emission events (lat/lon, timestamp, frequency, confidence); GeoJSON, CSV, API feed |
Analytics Satellize can run
| Dark vessel RF activity alert | TDOA/FDOA geolocation cross-referenced against AIS absence in the same area and time window | Near-real-time alert with position estimate, frequency class, and AIS gap duration; delivered as API push or email brief |
| Emission pattern timeline | Temporal aggregation of RF detection events for a defined ocean area, annotated by frequency class and estimated vessel count | Time-series chart and GIS layer showing emission density by hour and day; useful for identifying operational rhythms |
| Multi-source position reconciliation report | Fusion of RF fix with SAR detection and AIS record for the same vessel candidate, scored by positional overlap and timing consistency | PDF or structured JSON report per vessel candidate, with confidence tier (corroborated / partial / RF-only) |
| Proximity cluster analysis | Clustering of simultaneous RF detections within the geolocation resolution cell; flagged when no AIS traffic accounts for the observed emission count | GIS polygon layer marking unresolved multi-emitter zones, with timestamp and frequency metadata |
| Frequency-class equipment inference | Comparison of detected pulse characteristics against published radar and sonar frequency ranges for vessel categories (fishing, tanker, naval) | Probabilistic vessel-class annotation appended to each RF detection event; clearly labelled as indicative, not definitive |
| Retrospective track reconstruction | Sequential TDOA fixes from archive data connected into a movement track, gap-filled using dead reckoning where fixes are absent | KML or GeoJSON track file covering a defined historical period, with fix-quality flags on each node |
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.