Sensors
- Spire Global AIS: LEO constellation of over 100 small satellites carrying VHF receivers on AIS channels 87B and 88B (161.975 MHz and 162.025 MHz). Global revisit typically under 20 minutes for high-traffic zones; polar regions covered more frequently. Message latency to ground can be under 30 minutes on near-real-time tasking. Provides the broadcast layer: what a vessel claims to be and where it claims to be.
- exactEarth spaceborne AIS: Commercial S-AIS operator with a dedicated LEO constellation and data-fusion agreements. Offers historical AIS archive depth of over a decade, useful for establishing behavioural baselines and detecting pattern breaks. Complements Spire data in gap analysis by increasing message-capture probability in congested areas where slot collisions degrade detection.
- Sentinel-1 SAR (C-band, ESA): C-band (5.405 GHz) synthetic aperture radar with a 250 km swath in Interferometric Wide mode at 5 x 20 m ground resolution, or 80 km swath in Extra Wide mode at 20 x 40 m. Six-day exact repeat at the equator, shorter at higher latitudes. Detects hull-sized metallic returns regardless of transponder state. Free and open data. The primary SAR layer for broad-area dark vessel search.
- ICEYE SAR (X-band): Commercial X-band (9.65 GHz) constellation offering sub-metre spotlight resolution and same-day tasking in some geometries. Revisit to a specific point can be under 12 hours with multi-satellite scheduling. Higher frequency improves small-vessel detectability compared to C-band, though at the cost of narrower swath. Used when Sentinel-1 flags a region of interest and finer discrimination is needed.
Why AIS was never a surveillance system
AIS was designed for collision avoidance, not accountability. The VHF broadcasts are self-reported: a vessel transmits its MMSI, position, speed and heading, and other ships and shore stations listen. There is no cryptographic verification. A captain who wants to disappear can simply switch the transponder off, or transmit false position data while the hull goes elsewhere. Both acts are illegal under IMO regulations in most jurisdictions, but the regulation is only enforceable if someone notices.
Spaceborne AIS changed the geometry of that problem considerably. Before LEO constellations, AIS coverage beyond 40 to 50 nautical miles from shore was patchy at best. Satellites receive the same VHF broadcasts from orbit, giving near-global coverage. But the fundamental weakness remains: the satellite can only hear what the vessel chooses to say.
What SAR sees that AIS cannot
Synthetic aperture radar illuminates the ocean surface with microwave pulses and measures the energy scattered back. A steel hull, regardless of its length or its transponder state, produces a distinctive bright return against the relatively dark sea clutter. At Sentinel-1's Interferometric Wide resolution of roughly 5 x 20 m, vessels above approximately 50 to 100 metres in length are reliably detectable in moderate sea states. Smaller vessels, including fishing boats under 30 metres, sit near the detection floor and can be masked by wave clutter in sea states above Beaufort 4 or 5. That is an honest limit worth stating plainly.
X-band systems like ICEYE push that floor lower. Sub-metre spotlight imagery can resolve vessel superstructure detail and improve discrimination of smaller targets, though the swath is narrow and tasking must be deliberate. The practical workflow is two-stage: Sentinel-1 provides broad-area screening across hundreds of thousands of square kilometres per pass, and commercial SAR provides confirmation or characterisation where the economics justify it.
The correlation step: where dark vessels become visible
Detection of a dark vessel requires not just a SAR return but the absence of a corresponding AIS broadcast. That sounds straightforward. In practice it involves several non-trivial matching problems. AIS messages arrive at different times than the SAR acquisition. Vessels move. Position uncertainty in S-AIS, after accounting for Doppler shift and slot collisions in congested channels, can reach several hundred metres. SAR ship detection algorithms produce their own false alarms, particularly near offshore platforms, buoys or breaking waves.
The standard published approach applies a time-gated spatial join: AIS positions are propagated forward or backward to the SAR acquisition timestamp using reported speed and heading, then a search radius is applied, typically 0.5 to 2 nautical miles depending on the time gap and position uncertainty. SAR detections that fall outside any AIS association window are flagged as candidate dark vessels. The candidate list is then filtered by target size, shape and radar cross-section to remove non-vessel clutter. What remains is a prioritised list of hulls that chose not to be seen.
Spoofing is a different problem, and harder
Switching off a transponder is detectable by its absence. Spoofing, transmitting false positions while the vessel is somewhere else entirely, is detectable only by contradiction. A vessel broadcasting a position in the middle of a port while SAR shows open water at those coordinates, or broadcasting movement at 12 knots while SAR captures it stationary, creates a logical inconsistency that the correlation step can flag. The AIS message manipulation patterns that support this analysis are covered separately in the sibling page on AIS spoofing and manipulation; this page focuses on the physical detection of hull presence.
One important ambiguity: a SAR detection with no AIS match is not automatically sinister. Vessels under 300 gross tonnes are not required by SOLAS to carry Class A AIS. Small fishing boats, recreational craft and some naval vessels are legitimately absent from the AIS record. Context, including vessel size estimated from SAR, proximity to known fishing grounds or shipping lanes, and historical behaviour at that location, is required before a dark detection carries operational weight.
Operational uses and honest limits
Maritime domain awareness agencies use dark vessel detection for several distinct purposes: monitoring sanctions compliance (vessels transporting prohibited cargo often go dark near transfer points), fisheries enforcement (industrial vessels fishing illegally in exclusive economic zones), and port state control (identifying arrivals with no prior AIS history). The method is well-established; the UN Office for Outer Space Affairs and the International Maritime Organization have both published guidance acknowledging spaceborne AIS and SAR fusion as a legitimate monitoring approach.
The limits are worth being direct about. Cloud does not affect SAR, which is an advantage over optical sensors. But temporal gaps do matter: Sentinel-1's six-day revisit means a vessel can transit an area, conduct an illicit operation and leave before the next pass. Commercial SAR reduces that gap but at cost. S-AIS message capture probability in very congested areas drops due to VHF slot collisions, meaning some legitimate vessels may appear dark simply because their messages were not received. And the correlation algorithm requires careful calibration: too tight a matching radius produces false positives; too loose a radius misses genuine dark targets.
Satellize runs this correlation workflow on Sentinel-1 open data combined with licensed S-AIS feeds, delivering prioritised vessel-of-interest lists. The same analytical infrastructure that supports the Tonga crop-estimation programme handles the geospatial fusion at scale. Clients receive outputs as GIS layers or structured feeds, not raw imagery.
What a detection actually tells you, and what it does not
A confirmed dark vessel detection tells you that a hull was at a given position at a given time and was not broadcasting AIS. It does not tell you why. Intent requires additional intelligence: flag state, ownership chain, cargo manifest, port call history, and the pattern of previous dark events at similar locations. The analytic product is a cue, not a verdict. Its value is that it compresses an enormous search space, directing human attention to the small fraction of ocean traffic that warrants a second look.
Revisit cadence is the binding constraint for time-sensitive operations. A six-day Sentinel-1 cycle is adequate for trend analysis and sanctions monitoring over weeks. It is not adequate for intercepting a vessel mid-transfer. Clients with real-time enforcement needs should plan for commercial SAR tasking on top of the open-data baseline, accepting that the cost per square kilometre rises sharply as revisit shortens.
Typical figures
| SAR spatial resolution (broad area) | 5 x 20 m (Sentinel-1 IW mode); 20 x 40 m (Sentinel-1 EW mode) |
| SAR spatial resolution (spotlight) | Sub-1 m (ICEYE spotlight); 3 x 3 m (ICEYE stripmap) |
| Sentinel-1 revisit at equator | 6 days (single satellite); shorter at higher latitudes |
| Commercial SAR revisit (ICEYE) | Under 12 hours to a specific point with multi-satellite scheduling |
| S-AIS global revisit (Spire) | Typically under 20 minutes for high-traffic zones |
| AIS frequency bands | VHF channels 87B and 88B: 161.975 MHz and 162.025 MHz |
| Minimum detectable vessel (SAR) | Approximately 50 to 100 m in moderate sea states (Sentinel-1 C-band); smaller targets detectable with X-band spotlight |
| Position matching uncertainty | 0.5 to 2 nautical miles depending on AIS-to-SAR time gap and slot collision rate |
| S-AIS archive depth | Over 10 years (exactEarth historical archive) |
| Delivery formats | GeoJSON, Shapefile, structured CSV feed, or API-delivered vessel-of-interest list |
Analytics Satellize can run
| Dark vessel candidate list | Time-gated spatial join of SAR ship detections against propagated S-AIS positions; unmatched detections filtered by radar cross-section and shape | Prioritised GeoJSON layer of unmatched SAR returns with estimated vessel length, position, timestamp and confidence score |
| AIS gap event log | Temporal analysis of S-AIS track continuity; gaps exceeding configurable thresholds flagged against last-known position and area of interest polygons | Structured CSV feed of gap events with vessel MMSI, gap duration, last-known position and re-appearance position |
| Sanctions-zone dark activity report | Spatial filter of dark vessel detections within user-defined exclusion or monitoring zones; cross-referenced against flag state and vessel type from AIS history | Weekly PDF and GIS layer report of detected activity within defined zones, with historical comparison |
| Vessel rendezvous detection | SAR detection of two or more co-located hulls with no corresponding AIS proximity event; ship-to-ship transfer signature based on relative position and size | Alert feed with imagery chip, coordinates, estimated vessel sizes and time of acquisition |
| Traffic density anomaly map | Comparison of SAR-derived vessel counts against S-AIS-derived expected density across a defined area; statistical deviation highlights under-reporting zones | Gridded density-difference raster (GeoTIFF) updated per Sentinel-1 pass cycle |
| Historical dark behaviour baseline | Multi-year S-AIS archive analysis to establish normal transponder-off patterns by vessel type, flag and route; anomalies scored against baseline | Vessel-level risk score table updated monthly, delivered as structured data file |
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.