Sensors
- Sentinel-1 C-band SAR (ESA): 5 x 20 m resolution in Interferometric Wide Swath mode, 250 km swath, roughly 6-day repeat at mid-latitudes (shorter near poles). Free and open archive from 2014. Adequate for vessels above approximately 50 m length; smaller targets approach the noise floor in high sea states.
- ICEYE X-band SAR: Spotlight mode delivers sub-1 m resolution; Strip mode around 3 m. Constellation of more than 30 satellites enables same-day revisit over most maritime chokepoints. X-band is more sensitive to vessel superstructure detail than C-band, aiding vessel-type discrimination.
- Capella Space X-band SAR: Spotlight imagery at approximately 0.5 m resolution, tasked on demand. Particularly useful for confirming vessel identity and estimating waterline draught from freeboard geometry once a dark target has been flagged by a wider-area pass.
- Spire AIS receiver network: Spire operates more than 110 satellites carrying AIS receivers, providing global AIS coverage with latency under 30 minutes. Gaps in the Spire AIS record, when overlaid on SAR detections, are the primary indicator of deliberate transponder suppression.
Why radar does not care about the off switch
AIS is a cooperative system. A vessel must choose to broadcast, and choosing not to is trivially easy. SAR is not cooperative. A C-band or X-band radar pulse illuminates the sea surface and records the energy reflected back; a steel hull hundreds of metres long returns a strong, distinctive signal regardless of what the crew has done with the transponder switch. That physical independence is the entire basis of dark-vessel detection.
The relevant physics is radar cross-section (RCS). A large tanker in calm water can return an RCS of 30 to 50 dBsm, orders of magnitude above the surrounding sea clutter. Even in moderate sea states, where wave returns raise the noise floor, a vessel above roughly 50 to 80 m length remains detectable in Sentinel-1 Interferometric Wide Swath imagery. Smaller vessels, including many fishing boats used in ship-to-ship transfer operations, can fall below the detection threshold at high incidence angles or in rough conditions. That is an honest limit of the method and should be factored into any investigation workflow.
The gap-matching workflow: where the analysis actually lives
Raw SAR detection is a point in space and time: a bright pixel cluster at a given latitude, longitude, and acquisition timestamp. On its own it tells you a vessel is there. It does not tell you which vessel, or whether that vessel should have been broadcasting. The analytical value comes from fusion with the AIS record.
The workflow runs in three steps. First, a SAR detection algorithm, typically constant false alarm rate (CFAR) processing, identifies candidate targets in the scene. Second, those detections are matched spatially and temporally against the AIS position history from space-based receivers such as Spire. Detections with no corresponding AIS message within a configurable time window, typically 30 to 60 minutes, are flagged as dark. Third, vessel-identity inference uses the SAR-derived length and beam estimate, combined with port-call history from prior AIS records, to narrow the candidate list. A tanker that was broadcasting in Rotterdam three weeks ago, went dark in the eastern Mediterranean, and reappears in a Syrian anchorage is a very different risk profile from a vessel with a continuous clean record.
Port-call inference from SAR time series adds another layer. If a facility has no AIS traffic but repeated SAR passes show vessels mooring, loading, and departing over weeks, the time series itself becomes evidence of activity. This approach has been documented in open-source investigations into Iranian and Venezuelan crude exports.
False positives: sidelobes, ambiguities, and sea clutter
CFAR detection is not perfect. SAR imagery contains artefacts that mimic vessel returns. Range sidelobes from very bright targets, such as large vessels or offshore platforms, produce ghost echoes at fixed range offsets. Azimuth ambiguities create mirror images of bright targets displaced along the flight direction. Both can generate false detections that look, to an automated classifier, like a second vessel nearby.
Experienced analysts filter these by checking the geometric relationship between candidate detections and known bright targets, by examining the azimuth-to-range ratio of the bright cluster, and by comparing detections across multiple passes. A sidelobe artefact does not persist or move between acquisitions; a real vessel does. Published false-positive rates for CFAR on Sentinel-1 IW imagery vary widely with sea state and scene complexity, but figures of 5 to 20 percent for unfiltered detections are commonly reported in the peer-reviewed literature. Post-filtering with contextual checks typically brings operational false-positive rates below 5 percent, though that figure depends heavily on the analyst's threshold settings and the sea-state conditions at acquisition time.
Minimum detectable length and incidence angle
Sentinel-1 IW mode acquires at incidence angles between approximately 29 and 46 degrees across the swath. At shallower incidence angles, near-range, the sea clutter is higher and small targets are harder to separate from background. At steeper incidence angles, far-range, clutter is lower but the geometric foreshortening of vessel returns changes the apparent length. In practice, reliable detection of vessels shorter than 50 m is not guaranteed in Sentinel-1 IW imagery under anything other than calm sea conditions.
Commercial X-band systems change this calculus. ICEYE and Capella in spotlight mode can resolve vessels down to roughly 20 to 30 m, which covers the small tankers and barges frequently used in ship-to-ship transfers near sanctioned ports. The trade-off is cost and coverage: a Sentinel-1 IW scene covers 250 km of swath for free; a commercial spotlight task covers a few kilometres and carries a per-image cost. The practical approach is to use Sentinel-1 for wide-area screening and task commercial SAR only on flagged targets.
Oil and grain: the two commodity contexts
Sanctions on crude oil exports, notably those applied to Russia, Iran, and Venezuela, have driven a documented increase in AIS manipulation. The techniques include transponder switch-off during ship-to-ship transfers at sea, spoofing of position to show a vessel in a permitted location while it is physically elsewhere, and identity switching using MMSI numbers belonging to other vessels. SAR cannot directly detect spoofing, but it can contradict a spoofed position by placing a vessel in a different location at the same timestamp.
Grain trade compliance is a newer but growing use case. Vessels carrying grain from sanctioned origins have been documented using similar dark-voyage techniques. The challenge here is that grain ports are often in shallow coastal waters where SAR sea clutter is higher and small vessels are harder to detect. Port-call inference from SAR time series, rather than open-water detection, is often the more reliable method for grain-loading facilities.
Satellize runs dark-vessel screening as part of its broader commodity-monitoring analytics stack, drawing on the same open-constellation infrastructure used in its Tonga crop-estimation programme. Compliance teams and insurers can commission periodic or continuous monitoring of specific chokepoints, anchorages, or vessel watch-lists.
What this analysis can and cannot prove
SAR-based dark-vessel detection is an investigative lead, not a legal proof. It establishes that a vessel was physically present at a location while not broadcasting AIS. It does not establish what cargo was aboard, whether a transfer occurred, or who instructed the crew to go dark. Corroboration from financial records, port documentation, and vessel inspection remains necessary for enforcement action.
The method is also retrospective in its most common form. Sentinel-1 has a 12-day exact repeat (6-day with both satellites), so a vessel that loads and departs within a 6-day window may be missed entirely unless a commercial task was placed in advance. Persistent monitoring of high-risk anchorages, using a combination of Sentinel-1 and tasked commercial SAR, reduces but does not eliminate this gap. Buyers of this analysis should understand that coverage is probabilistic, not continuous, and that the absence of a SAR detection is not evidence of absence.
Typical figures
| Sentinel-1 spatial resolution (IW mode) | 5 x 20 m (range x azimuth) |
| Commercial SAR spotlight resolution | 0.5 to 1 m (ICEYE, Capella) |
| Sentinel-1 revisit (mid-latitudes, both satellites) | Approximately 6 days |
| Commercial SAR revisit (ICEYE constellation) | Same-day tasking possible over most chokepoints |
| AIS latency (Spire space-based) | Under 30 minutes globally |
| Minimum detectable vessel length (Sentinel-1 IW, calm sea) | Approximately 50 m; smaller targets unreliable |
| Minimum detectable vessel length (commercial X-band spotlight) | Approximately 20 to 30 m |
| CFAR false-positive rate (post-filtering, typical) | Below 5 percent; higher in rough sea states |
| SAR archive depth (Sentinel-1) | From 2014 to present |
| Delivery formats | GeoJSON detection layers, timestamped vessel-gap reports, PDF investigation briefs |
Analytics Satellize can run
| Dark-vessel detection layer | CFAR SAR target detection cross-matched against space-based AIS gaps | GeoJSON point layer of unmatched SAR detections with timestamp, estimated vessel length, and confidence score |
| Port-call inference report | SAR time-series analysis of vessel presence at specific anchorages or terminals | PDF or spreadsheet log of inferred arrivals and departures at a named facility over a specified period |
| Vessel-identity candidate list | SAR-derived length and beam estimate matched against AIS historical fleet registry | Ranked list of candidate vessel identities with prior AIS track history |
| AIS spoofing contradiction report | Comparison of AIS-reported position against SAR-confirmed physical location at matching timestamps | Timestamped discrepancy table with geographic evidence, suitable for compliance or legal review |
| Chokepoint or anchorage watch | Scheduled Sentinel-1 pass monitoring supplemented by triggered commercial SAR tasking on flagged targets | Recurring alert feed with new detections flagged against a client-supplied vessel watch-list |
| Ship-to-ship transfer event log | Detection of vessel pairs within proximity threshold in open water, cross-checked against AIS absence for both vessels | Event log with SAR image chips, estimated vessel sizes, and gap duration for each candidate transfer |
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.