Fishing vessel activity inference from RF fingerprinting
Passive RF detection from low Earth orbit can locate fishing vessels that have disabled or never carried AIS, by capturing navigation radar pulses, VHF traffic, and fish-finder emissions. Vessel-type inference from RF signatures alone is probabilistic, not definitive, and this page explains exactly where the confidence holds and where it does not.
Sensors
- HawkEye 360 RF constellation: Clusters of three small satellites flying in formation detect and geolocate RF emitters by time-difference-of-arrival and frequency-difference-of-arrival. Geolocation accuracy is published at roughly 500 m to 2 km depending on geometry and signal duration. Revisit over any given ocean point is several times per day across the full constellation.
- Kleos Space RF constellation: Similar TDOA/FDOA passive-intercept architecture to HawkEye 360. Kleos targets maritime VHF and radar bands. Published geolocation accuracy is in the low-kilometre range. The constellation is still expanding, so revisit and coverage density vary by latitude.
- Spire Global AIS and GNSS: Spire's 100-plus satellite constellation collects AIS messages globally, providing the baseline of vessels that are broadcasting. AIS gaps identified against Spire's coverage are the starting point for dark-vessel suspicion before RF fingerprinting is applied.
- Sentinel-1 SAR (ESA): C-band synthetic aperture radar at 5.405 GHz, 10 m resolution in Interferometric Wide Swath mode, 250 km swath. Detects vessel-sized radar cross-sections regardless of cloud or darkness. Revisit is 6 days at the equator for a single satellite, shorter at higher latitudes. Used to corroborate RF-detected positions with a physical ship echo.
What a fishing vessel broadcasts whether it wants to or not
A vessel that switches off its AIS transponder removes itself from the cooperative tracking picture. It does not remove itself from the electromagnetic one. Navigation radars, typically X-band (9.2 to 9.5 GHz) or S-band (2.9 to 3.1 GHz), pulse continuously at characteristic repetition frequencies and rotation rates. Fish-finders operate in the 50 kHz to 200 kHz acoustic range but their topside electronics and transducer drivers produce RF byproducts. VHF radios on channels 16 and working channels produce short, geographically anchored bursts whenever crew communicate. Each of these emissions has a timing signature, a frequency, and a power envelope that can be captured by a passive RF receiver in orbit.
The key physical principle is that RF geolocation does not require the vessel's cooperation. A formation of satellites measuring the nanosecond-scale arrival-time differences of the same pulse can triangulate the emitter's position on the surface. This is TDOA, time-difference-of-arrival. When the satellites are also in relative motion, FDOA, frequency-difference-of-arrival from Doppler shift, adds a second independent position constraint. The combination is what HawkEye 360 and Kleos Space exploit commercially.
From emission to vessel type: where confidence is earned and where it is not
Navigation radar pulse repetition frequency and antenna rotation rate are partly diagnostic. Commercial fishing vessels in the 20 to 60 metre range typically carry X-band radars with rotation rates of 20 to 40 RPM and pulse repetition frequencies in the 1 to 4 kHz range, but these parameters overlap substantially with coastal cargo and passenger vessels of similar size. A radar emission alone does not prove a vessel is fishing. It proves a vessel is present and operating a radar.
Inference improves when multiple emission types are detected simultaneously from the same geographic point. A vessel emitting X-band radar pulses, VHF traffic on a working channel, and acoustic fish-finder RF byproducts at the same location, in an area with no AIS registration and no declared fishing licence, is a materially stronger candidate than any single emission would suggest. The inference is still probabilistic. A recreational vessel or a small cargo ship could produce a similar multi-emission signature. Analysts should treat RF fingerprinting as generating leads for further investigation, not as proof of illegal fishing.
One published limitation is geolocation accuracy. HawkEye 360 reports position uncertainties of roughly 500 m to 2 km. At that uncertainty radius, distinguishing a vessel inside a closed zone from one just outside it requires careful handling, particularly near zone boundaries. Decisions with legal consequence should corroborate RF position with Sentinel-1 SAR, which provides a physical radar cross-section at 10 m resolution.
Cross-referencing registries and zone geometry
The analytic value of an RF detection rises sharply when it is placed in regulatory context. Exclusive Economic Zones, seasonal closure areas, and marine protected areas have published boundary geometries. An RF-located emitter inside a closed zone, with no corresponding AIS record in Spire's global message archive, and no vessel registration matching the expected radar characteristics for that area, is a meaningful anomaly worth reporting to a flag state or fisheries authority.
Vessel registry cross-referencing is imperfect. Many small fishing vessels in developing maritime states are not registered in any database accessible to a commercial analyst. The absence of a registry match is therefore not itself evidence of wrongdoing. It is evidence of an information gap. Documenting that gap systematically, over time, across a defined zone, is itself analytically useful: it tells a fisheries authority where their registration data is weakest.
Combining RF detections with SAR and optical to build a case
RF fingerprinting works best as the first layer of a multi-source workflow, not as a standalone verdict. The sequence that produces the most defensible output is: RF detection flags a dark emitter at a position and time; Sentinel-1 SAR, tasked or archive-searched for the same area and date, confirms or denies a vessel-sized radar cross-section at that position; optical imagery, if cloud-free, adds a visual confirmation and potentially a vessel silhouette from which length and hull type can be estimated.
Sentinel-1's 6-day single-satellite revisit is a real constraint. Over a 250 km swath, the chance of a coincident SAR pass within a few hours of an RF detection is not guaranteed. At higher latitudes, where swath overlap increases revisit frequency, the geometry is more favourable. In tropical fishing grounds near the equator, analysts may wait several days for a confirming SAR pass, by which time the vessel has moved. This is an honest limitation of the current sensor architecture, and any reporting system should communicate it clearly to end users.
Practical outputs for a fisheries monitoring authority
A fisheries ministry or coast guard using this approach typically wants three things: a near-real-time alert when a dark emitter enters a regulated zone, a weekly pattern-of-life summary showing which areas have persistent unregistered RF activity, and an evidence package for specific incidents that can withstand scrutiny in an administrative or legal process.
The near-real-time alert is constrained by satellite revisit. RF constellation passes over a given ocean point may be separated by several hours. Alerts are therefore best framed as 'detected within this time window at this location' rather than as live tracking. The pattern-of-life summary, built from weeks or months of RF detections aggregated into heat maps and vessel-count estimates, is the more statistically reliable product. Satellize's analytics pipeline can run this kind of aggregation on RF detection feeds from commercial constellations, cross-referenced against open AIS archives, to produce zone-level activity indices. The approach is similar in structure to the crop-estimation methodology Satellize runs for the Kingdom of Tonga, where repeated satellite observations are aggregated into a statistical model rather than treated as single-point facts.
An evidence package for a specific incident should include: the raw RF detection record with timestamp, frequency band, and geolocation uncertainty ellipse; any coincident SAR or optical imagery; the AIS archive query result showing no registered vessel at that position; and a clear statement of what the data can and cannot prove. Presenting RF fingerprinting evidence without that last element is how enforcement actions get challenged and overturned.
Honest limits a buyer should know before committing
RF geolocation accuracy from current commercial constellations is sufficient for open-ocean zone attribution but marginal for disputes near zone boundaries. Coastal waters with many legitimate vessels produce cluttered RF environments where TDOA solutions can be ambiguous. Vessel-type inference from emission signatures alone carries irreducible uncertainty; the method identifies candidates, not culprits. SAR corroboration is not always available within a useful time window. And the entire workflow depends on RF constellation coverage, which is denser over high-traffic ocean corridors and thinner over remote areas, precisely where some illegal fishing is most concentrated. These are not reasons to avoid the method. They are the parameters within which it should be scoped.
Typical figures
| RF geolocation accuracy (TDOA/FDOA) | Approximately 500 m to 2 km, depending on satellite geometry, signal duration, and emitter frequency (HawkEye 360 published figures) |
| RF constellation revisit (open ocean) | Several passes per day per point for HawkEye 360 and Kleos; exact revisit varies with latitude and constellation size at time of tasking |
| Frequency bands monitored | X-band radar (9.2 to 9.5 GHz), S-band radar (2.9 to 3.1 GHz), maritime VHF (156 to 174 MHz); fish-finder RF byproducts in UHF range |
| SAR corroboration resolution | 10 m (Sentinel-1 IW mode); 250 km swath |
| SAR revisit (Sentinel-1, single satellite) | 6 days at equator; shorter at higher latitudes due to swath overlap |
| Minimum detectable vessel (SAR) | Radar cross-section detection reliable for vessels above approximately 20 m length in moderate sea states; smaller vessels may fall below noise floor in high sea states |
| AIS archive depth (Spire) | Multi-year global archive; message completeness varies by ocean region and historical constellation size |
| RF detection latency | Typically hours from satellite pass to processed detection; near-real-time alerting depends on ground station contact and processing pipeline |
| Delivery formats | GeoJSON emitter-position files, CSV detection logs, GIS-compatible zone-activity heat maps, PDF incident evidence packages |
Analytics Satellize can run
| Dark emitter alert | TDOA/FDOA geolocation of RF detections cross-referenced against AIS archive gaps and EEZ or closure-zone boundary geometry | Timestamped alert with position, uncertainty ellipse, frequency band, and zone-attribution flag; delivered as API push or email report |
| Zone-level unregistered activity index | Temporal aggregation of RF detections over a defined area, normalised against known registered vessel AIS density, to produce a weekly or monthly anomaly score | GIS heat-map layer and tabular index by zone, updated on agreed cadence |
| Vessel-type probability classification | Emission signature matching against published radar parameter ranges for vessel classes; multi-emission coincidence scoring (radar plus VHF plus fish-finder byproduct) | Per-detection probability breakdown (fishing vessel, cargo, recreational, unknown) with confidence band |
| SAR-corroborated incident package | RF detection matched to nearest Sentinel-1 SAR pass by time and position; radar cross-section extracted and compared against detection uncertainty ellipse | PDF evidence package including RF record, SAR chip, AIS query result, and written limitations statement suitable for administrative review |
| Pattern-of-life report for a defined maritime zone | Multi-week RF detection aggregation with temporal clustering to identify vessel behavioural patterns (repeated presence at same coordinates, dawn/dusk activity spikes consistent with net-setting) | Monthly analytical report with annotated maps and vessel-count estimates with stated uncertainty ranges |
| Registry gap assessment | Systematic comparison of RF-detected emitter counts against national vessel registry data and AIS message counts for a defined area, to identify zones of structural under-registration | Tabular gap analysis by sub-zone with recommendations for registry data collection priorities |
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.