Sensors
- Sentinel-1 SAR (ESA): C-band (5.405 GHz), 10 m ground resolution in Interferometric Wide Swath mode, 250 km swath, 6-day repeat at the equator (1-3 days at higher latitudes with both satellites). Detects vessels as bright radar cross-section returns against sea clutter; length estimation reliable to roughly ±10 m on vessels above 50 m. Free and open data with 1-3 hour latency to archive.
- ICEYE SAR (commercial): X-band, spotlight mode achieves 25 cm azimuth resolution with approximately 1 m ground resolution, enabling beam-width estimation on mid-size vessels. Tasking latency can be under 24 hours; revisit on a single target achievable multiple times per day across the constellation. Useful for resolving ambiguity when Sentinel-1 cannot separate vessels of similar class.
- Capella Space SAR (commercial): X-band, spotlight mode to approximately 0.5 m resolution. Fine enough to resolve hull superstructure features, funnel placement and deck equipment layout, which serve as physical fingerprints independent of any transmitted identity. Constellation revisit is multiple passes per day over a fixed point.
- Planet SkySat (commercial optical): 0.5 m panchromatic, 0.8 m multispectral. On clear days, confirms hull colour, funnel markings, flag, and visible name lettering. Useless through cloud and unreliable at night, making it a complement to SAR rather than a substitute. Tasking latency typically under 24 hours.
What the radio says versus what the radar sees
The Automatic Identification System was designed for collision avoidance, not identity verification. Any vessel operator with a programming cable and ten minutes can overwrite the Maritime Mobile Service Identity number, vessel name, flag state, and reported dimensions stored in a Class A transponder. The IMO number, which is supposed to be hull-stamped and permanent, is equally easy to falsify in the broadcast. This is not a theoretical vulnerability: UN Panel of Experts reports on the Democratic People's Republic of Korea, published between 2017 and 2023, document dozens of cases in which vessels transmitted MMSI numbers belonging to other, legitimate ships while conducting ship-to-ship oil transfers.
Satellite SAR does not listen to radio. It illuminates the target with microwave energy and measures the return. The radar cross-section of a steel hull is a function of its physical dimensions, aspect angle, and surface geometry. A 180-metre tanker cannot impersonate a 90-metre product carrier in the SAR image, regardless of what its transponder says. That physical contradiction is the foundation of the detection method.
The physics of hull measurement from orbit
In a SAR image, vessel length is estimated from the bright return along the ship's axis; beam (width) is estimated from the cross-track extent, though it is systematically underestimated because the vessel's sides are less reflective than the superstructure. At Sentinel-1's 10 m resolution, length estimation carries an uncertainty of roughly ±10 m, which is adequate to separate a VLCC from a Panamax tanker but insufficient to distinguish two vessels of the same class separated by 15 m in length. Published studies in Remote Sensing (MDPI) and work cited in CSIS open-source investigations confirm this resolution floor as a genuine operational constraint.
X-band SAR from ICEYE or Capella narrows that uncertainty considerably. At sub-metre resolution in spotlight mode, the analyst can resolve individual deck features: the placement of cranes, the number and spacing of cargo hatches, the shape of the bridge superstructure. These features constitute a physical fingerprint that persists across name changes and repaintings. Comparing a high-resolution SAR image against Lloyd's Register or Equasis hull records for the claimed identity is the core verification step documented in open-source maritime intelligence practice.
One genuine limit: SAR measures the vessel as it sits in the water at the moment of the pass. Ballast condition changes apparent draft and can shift the radar cross-section. A fully laden tanker and the same vessel in ballast produce different returns. Analysts must account for this, and should treat length estimates as a range rather than a precise figure.
AIS spoofing patterns that precede identity switching
Before a vessel goes dark or switches identity, its AIS track often shows characteristic anomalies. Positional jumps inconsistent with the vessel's reported speed suggest GPS spoofing. Repeated transmission of an MMSI that is simultaneously active in a distant ocean indicates cloning. Gaps of several hours in otherwise continuous tracks, particularly in the vicinity of known transfer anchorages in the Gulf of Oman, the Bering Sea, or off the west coast of Africa, are documented precursors to illicit ship-to-ship transfers in UN expert-panel reporting.
Fusing these AIS anomalies with SAR-detected vessel positions is the standard cross-check. If SAR detects a vessel at coordinates where AIS shows no traffic, or shows a vessel of incompatible dimensions to the claimed identity, the discrepancy is flagged for further investigation. The method does not prove guilt; it generates a prioritised list of vessels warranting port-state inspection or additional imagery collection.
Flag-of-convenience patterns visible from orbit
Flag-of-convenience registries are legal. The problem arises when a vessel cycles through multiple flag states in a short period, a pattern associated with sanctions evasion documented in CSIS and UN reporting. Flag changes are recorded in public registries such as Equasis, but the change is often not reflected in the AIS broadcast for days or weeks. Optical imagery from SkySat, when cloud permits, can confirm the flag actually flying at the stern, which can then be compared against the AIS-reported flag state and the current registry record.
Hull repainting is a related indicator. A vessel that appears in one colour in archived imagery and a different colour in a recent pass, without any corresponding dry-dock record, has likely undergone a name-change operation. Planet's SkySat archive, combined with Sentinel-2 multispectral data at 10 m, can establish a colour-change timeline even when individual images lack the resolution to read hull lettering.
Where the method fails and what honest analysts say about it
Cloud cover renders optical confirmation impossible for days or weeks in tropical and high-latitude regions. SAR is cloud-independent but carries its own ambiguities: two vessels of similar class and similar loading condition may produce indistinguishable radar cross-sections at 10 m resolution. The 3 m resolution floor cited in CSIS and UN analyses is a real constraint, not a caveat inserted for legal protection.
AIS data quality varies significantly by provider. Gaps in satellite AIS coverage over open ocean mean that dark periods are not always sinister; they may simply reflect coverage holes. Analysts who treat every AIS gap as evidence of wrongdoing will generate a high false-positive rate that undermines the credibility of the whole programme. The method works best when multiple independent signals converge: a SAR-detected vessel, an AIS anomaly, a known transfer anchorage, and a prior history of flag cycling for the claimed identity.
Satellize applies this multi-layer fusion approach in its analytics work, drawing on open Sentinel-1 archives and commercial SAR tasking on client licence. Analysts familiar with the Tonga crop-estimation programme will recognise the same principle: a single sensor is a clue, not a verdict.
From image to actionable intelligence product
The output of a ghost-vessel investigation is not a map. It is a structured case file: the SAR-derived hull dimensions, the AIS-reported dimensions, the delta between them, the timestamp and position of each observation, the flag and MMSI history from public registries, and the optical imagery where available. That file goes to a port-state authority, a sanctions compliance team, or a flag-state administration. It is designed to meet the evidentiary standard required for a port inspection, not merely to satisfy analytical curiosity.
Revisit planning matters as much as the imagery itself. A vessel conducting a ship-to-ship transfer typically completes the operation in four to eight hours. Sentinel-1's 6-day repeat is too slow to catch the act; commercial SAR tasking with same-day or next-day collection is required. Knowing when to task, based on AIS anomaly alerts, is the operational skill that turns a capable sensor into a functioning enforcement tool.
Typical figures
| SAR spatial resolution (Sentinel-1 IW mode) | 10 m ground range; vessel length uncertainty ±10 m |
| SAR spatial resolution (ICEYE / Capella spotlight) | 0.5–1 m; resolves deck features and superstructure layout |
| Optical resolution (Planet SkySat) | 0.5 m panchromatic; sufficient for hull colour, flag, and visible lettering |
| Sentinel-1 revisit (equatorial) | 6 days single satellite; 1–3 days combined A+B at mid-to-high latitudes |
| Commercial SAR tasking latency | Under 24 hours for ICEYE and Capella; multiple passes per day possible |
| Minimum detectable vessel length (Sentinel-1) | Approximately 50 m in moderate sea state; smaller vessels lost in sea clutter |
| AIS satellite coverage latency | Varies by provider; open-ocean gaps of 30–90 minutes are common |
| Sentinel-1 archive depth | From 2014 (Sentinel-1A launch); accessible via Copernicus Data Space |
| Delivery format | GeoTIFF imagery, GeoJSON vessel-position layers, structured PDF case file |
Analytics Satellize can run
| SAR-AIS dimension discrepancy alert | Automated SAR vessel detection (CFAR algorithm) cross-referenced against AIS-reported length and beam from satellite AIS feed | Timestamped alert with SAR-derived dimension range, AIS-reported dimensions, and position fix; delivered as GeoJSON point feature or API push |
| Hull geometry fingerprint comparison | High-resolution SAR feature extraction (superstructure profile, hatch spacing, crane placement) compared against Lloyd's or Equasis reference imagery for claimed identity | Structured case file with side-by-side imagery panels and written assessment of match confidence |
| AIS anomaly pattern scoring | Time-series analysis of MMSI transmission gaps, positional jumps, simultaneous MMSI conflicts, and flag-state change frequency against public registry records | Vessel-level risk score with supporting evidence table; exportable as CSV or integrated into client maritime intelligence platform |
| Hull colour-change timeline | Multi-date optical and Sentinel-2 multispectral comparison to detect repainting events not correlated with dry-dock records | Annotated image time series with change-detection dates and confidence assessment; delivered as PDF report or GIS layer |
| Transfer-event confirmation package | SAR detection of two or more vessels in close proximity at a known transfer anchorage, cross-referenced with AIS tracks showing rendezvous; method consistent with UN Panel of Experts documented approach | Tasked SAR collect report with vessel positions, separation distance, duration estimate, and AIS concordance summary |
| Flag-state discrepancy report | Optical flag confirmation from SkySat imagery compared against AIS-broadcast flag and current Equasis registry record | Single-page evidence sheet suitable for port-state control submission, with imagery, registry extract, and AIS screenshot |
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.