Ship-to-ship cargo transfer detection at sea
At-sea cargo transfers between tankers are a primary sanctions-evasion route for oil and grain. Fusing SAR imagery, RF monitoring and AIS behavioural analysis exposes them even when both vessels go dark.
Sensors
- Sentinel-1 SAR (ESA/Copernicus): C-band synthetic aperture radar at 5 m resolution in IW mode, 250 km swath. Revisit of 6 days at the equator for a single satellite, roughly 3 days with both Sentinel-1A and 1C combined. Detects vessel hulls as bright radar returns regardless of cloud or darkness. Cannot distinguish vessel type without additional context.
- ICEYE SAR (commercial): X-band SAR with spotlight mode reaching sub-1 m resolution, tasked on demand. Revisit can be daily or better over a target area using the full ICEYE constellation. Useful for confirming hull geometry and relative positioning of two vessels once a candidate pair is flagged by wider-area sensors.
- Spire Global RF/AIS (commercial): A constellation of over 100 LEO cubesats receiving AIS transmissions from space and detecting RF emissions. Provides global AIS collection with latency typically under 30 minutes. Critically, Spire's RF layer can detect a vessel transmitting on AIS frequencies even when its reported position is spoofed, because signal-of-opportunity geolocation differs from the declared position.
- HawkEye 360 RF (commercial): Cluster-based RF geolocation constellation that passively locates emitters, including AIS, radar and communications signals, without relying on the content of those transmissions. Geolocation accuracy is typically 500 m to 2 km depending on geometry. Identifies vessels transmitting in a location inconsistent with their AIS-declared position.
What a floating roof gives away
A ship-to-ship (STS) transfer requires two vessels to lie alongside each other, usually for several hours, while cargo pumps or cranes do their work. That physical act is detectable from orbit even when both captains have switched off their AIS transponders. SAR sees the hull as a point or distributed bright target in the radar backscatter image. Two such targets in close proximity, separated by a distance consistent with mooring alongside (typically 30 to 80 metres between centrelines for VLCC-class tankers), with no port infrastructure nearby, is an immediately suspicious geometry.
Sentinel-1's Interferometric Wide Swath mode covers 250 km at 5 m ground resolution, which is sufficient to resolve individual large vessels and to measure the gap between them. The sensor operates in all weather and at night, which matters because STS transfers in sanctioned trades tend to happen in remote ocean areas chosen partly for poor visibility. The honest limit is revisit: 3 to 6 days depending on latitude and constellation status, which means a transfer lasting only a few hours may fall between passes. That gap is where commercial SAR and RF data earn their place.
The AIS gap is a signal, not an absence
AIS was designed for collision avoidance, not compliance. Switching it off is trivially easy, and the maritime industry has known for years that gaps in the AIS record correlate with illicit activity. The analytical challenge is that AIS gaps also occur legitimately: poor satellite geometry, coastal congestion and equipment failure all produce them. A gap alone proves nothing.
The combination that raises confidence is a gap coinciding with anomalous speed and heading. A vessel that slows to under 2 knots in open ocean, holds that speed for three or more hours, and then resumes passage speed is behaving like a vessel at anchor or alongside another ship. Spire's space-based AIS collection catches transmissions that terrestrial networks miss, so a vessel that appears dark to coastal receivers may still be visible to a satellite overhead. When Spire's RF layer detects an emission at a position that contradicts the vessel's last known AIS track, that discrepancy is itself evidence of spoofing.
HawkEye 360 adds a further layer. Its clusters geolocate emitters by time-difference-of-arrival without decoding the message content. A vessel broadcasting a false position on AIS will still be physically located by the RF geometry. The mismatch between declared and geolocated position, when it exceeds the system's accuracy floor of roughly 500 m to 2 km, flags the vessel as a spoofing candidate.
Confidence comes from convergence, not from any single sensor
No single observation is conclusive. SAR sees two hulls close together but cannot read the cargo manifest. RF detects an anomalous emission but cannot confirm physical contact. AIS behaviour analysis identifies suspicious loitering but cannot rule out a legitimate anchorage. The analytical value comes from requiring all three conditions to be satisfied within the same time window and the same geographic cell.
A practical detection pipeline works roughly as follows. AIS behavioural screening across the global fleet identifies candidate pairs: two vessels whose tracks converge, both showing speed anomalies, within a defined ocean area. RF monitoring confirms whether those vessels are still transmitting and whether their declared positions are credible. SAR imagery, either archived Sentinel-1 or tasked commercial SAR, then provides the physical confirmation of proximity. When all three align, the probability of a coincidental false positive drops sharply, though it does not reach zero. Legitimate STS operations do occur, including authorised lightering of very large crude carriers in designated anchorage zones. Distinguishing authorised from illicit transfers requires knowing the vessel identities, their flag states, their last port calls and the applicable sanctions regime.
Where the method reaches its limits
Revisit is the most serious operational constraint. Sentinel-1's 3-to-6-day repeat means a short transfer in a poorly covered ocean area may leave no SAR evidence. Tasking commercial SAR on demand addresses this, but only if the candidate pair is identified quickly enough from AIS and RF signals to allow a tasking request before the vessels separate. Latency in the RF and AIS data feeds, typically 20 to 60 minutes for space-based AIS, sets the practical floor on how fast that loop can close.
SAR resolution at 5 m is sufficient to detect vessels of VLCC size (around 300 m length) and most Aframax tankers (around 245 m). Smaller vessels, such as the coastal tankers sometimes used as intermediaries, produce weaker and less distinctive radar returns. Sub-1 m commercial SAR improves this, but at significantly higher cost per image. In very calm sea states, the contrast between vessel and sea clutter improves; in rough seas above Beaufort 5 or 6, sea clutter can obscure smaller targets entirely.
RF geolocation accuracy of 500 m to 2 km means that for vessels separated by only a few hundred metres, the RF layer alone cannot confirm physical contact. It can confirm that a vessel is not where it claims to be. That is useful but not sufficient on its own.
Building a monitoring programme around this method
The most effective deployments treat STS detection as a persistent screening programme rather than a reactive investigation. A standing watch over defined ocean areas of interest, such as the Gulf of Oman, the waters north of the Lacadive Sea, or the Atlantic approaches to West African ports, uses continuous AIS and RF feeds to generate daily candidate lists. SAR tasking is then reserved for the highest-confidence candidates, keeping imagery costs proportional to intelligence value.
Satellize runs analytics on open constellations including Sentinel-1 and integrates commercial RF and SAR tasking on client licence. The workflow is similar in structure to the crop-estimation programme run for the Kingdom of Tonga, where combining open satellite data with targeted commercial tasking allowed a small island government to build a credible monitoring capability without a dedicated space programme of its own. For a compliance team or a government with sanctions-enforcement responsibilities, the equivalent is a curated alert feed: vessel pairs, timestamps, geographic coordinates, and the evidence chain behind each flag, delivered as a structured report or a GIS layer that feeds directly into an existing case-management system.
The public record on STS detection methodology is well established. Published work in journals such as Remote Sensing (MDPI) has demonstrated hull detection in Sentinel-1 imagery at operational scale. Global Fishing Watch has published open methods for AIS gap analysis. The building blocks are not proprietary. What a monitoring programme requires is the integration of those blocks into a coherent, low-latency pipeline with appropriate thresholds for the specific sanctions regime and vessel classes of interest.
Typical figures
| SAR spatial resolution (Sentinel-1 IW mode) | 5 m range, 20 m azimuth (detected vessel footprint typically 1–3 pixels for large tankers) |
| SAR spatial resolution (ICEYE spotlight) | Sub-1 m, tasked on demand |
| Sentinel-1 revisit (combined 1A + 1C) | 3–6 days depending on latitude; shorter at higher latitudes |
| Commercial SAR revisit (ICEYE constellation) | Daily or better over a tasked area of interest |
| Space-based AIS latency (Spire) | Typically under 30 minutes from transmission to delivery |
| RF geolocation accuracy (HawkEye 360) | 500 m to 2 km, geometry-dependent |
| Minimum detectable vessel size (Sentinel-1) | Approximately 50–100 m length in moderate sea states; smaller in calm conditions |
| Sentinel-1 archive depth | From 2014 (Sentinel-1A launch); accessible via Copernicus Data Space |
| Coverage | Global ocean; Sentinel-1 prioritises Europe, Arctic and conflict-adjacent areas in its acquisition plan |
| Delivery formats | GeoTIFF (SAR chips), GeoJSON or Shapefile (vessel detections), structured JSON alert feed, PDF investigation report |
Analytics Satellize can run
| STS candidate pair alerts | AIS behavioural screening for co-located speed anomalies and simultaneous gap events, cross-referenced with RF geolocation | Daily alert feed with vessel IMO numbers, coordinates, timestamps and confidence tier (Low / Medium / High) |
| SAR vessel proximity confirmation | CFAR (Constant False Alarm Rate) ship detection on Sentinel-1 GRD imagery; inter-vessel distance measurement | Annotated SAR chip with detected centroids, separation distance and bearing, delivered as GeoTIFF and PDF |
| Spoofing discrepancy report | Comparison of AIS-declared position against RF-geolocated position from HawkEye 360 or Spire; offset distance calculated per observation | Per-vessel spoofing log with declared vs. actual position, offset in kilometres and time of observation |
| Historical transfer pattern analysis | Retrospective AIS gap clustering over a defined ocean area and time window, matched against Sentinel-1 archive detections | GIS layer of historical candidate STS events with frequency heatmap; suitable for sanctions-case background reports |
| Vessel identity dossier | Cross-referencing detected vessel characteristics (length, beam estimated from SAR) against IMO registry, flag state records and last-port-call AIS history | Structured PDF dossier per flagged vessel pair, including ownership chain where publicly documented |
| Persistent area watch | Standing automated monitoring of defined ocean polygons using continuous AIS and RF ingestion with SAR tasking triggered on threshold breach | Weekly summary report plus real-time alerts when a new candidate pair enters the watch area |
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.