Vessel Loitering Pattern Detection for Maritime Smuggling
Low-speed loitering and repeated course reversals in open water are behavioural signatures of contraband transfer. Trajectory anomaly scoring on AIS tracks, confirmed by SAR when transponders go dark, gives maritime authorities a defensible, satellite-sourced evidence trail.
Sensors
- Spire Global spaceborne AIS: Constellation of more than 110 LEO satellites receiving Class-A and Class-B AIS messages globally. Typical message latency under 20 minutes for most ocean areas; polar and mid-ocean coverage is meaningfully better than coastal VHF networks, which is where smuggling activity concentrates.
- exactEarth satellite AIS: Complementary LEO AIS feed that improves detection probability in high-density shipping lanes through independent receiver geometry. Cross-referencing two independent feeds reduces the risk of a missed transmission being misread as deliberate AIS-off behaviour.
- Sentinel-1 SAR (C-band, ESA): Dual-polarisation C-band radar at 5.405 GHz. Interferometric Wide Swath mode delivers 250 km swaths at 5 x 20 m ground resolution, independent of cloud and darkness. Revisit at mid-latitudes is approximately 6 days for a single satellite; the two-satellite constellation (Sentinel-1A and 1C as of 2024) can reduce this to 3 days over priority areas, though on-demand tasking is not available through open access.
- ICEYE SAR (X-band, commercial): X-band SAR constellation offering on-demand tasking with collection windows typically within 24 hours for most ocean areas. Spotlight mode achieves approximately 1 x 1 m resolution; Strip mode approximately 3 x 3 m. X-band is more sensitive to small metallic superstructure details than C-band, useful for distinguishing vessel class during confirmation passes.
What loitering looks like in a trajectory record
A vessel engaged in a legitimate voyage moves with purpose: consistent heading, speed close to its declared service speed, deviations explainable by weather routing or traffic separation schemes. A vessel waiting to transfer cargo behaves differently. Speed drops to one or two knots, sometimes less. Course oscillates. The vessel may describe slow circles or back-and-forth transits across a patch of water a few kilometres wide. This pattern can persist for hours.
AIS records this precisely. Every Class-A transponder is required by IMO SOLAS regulations to transmit position, speed over ground, course over ground and navigational status at intervals as short as 2 seconds when under way. A loitering vessel therefore leaves a dense, anomalous trajectory segment in the record: low speed-over-ground variance, high course-change frequency, near-zero net displacement per hour. These are measurable quantities, not impressions.
Scoring anomaly against a baseline, not against a rule
Simple rule-based filters fail quickly. A vessel anchored in an uncharted roadstead looks identical to a loiterer under a speed threshold alone. A fishing vessel working a set looks similar under a course-reversal count. The correct approach is to score each trajectory segment against a statistical baseline for that vessel type, in that ocean region, at that time of year.
Published methods in the remote-sensing and maritime-surveillance literature use a combination of kinematic features: speed distribution, turning rate, stop duration, net displacement ratio and distance from the nearest declared anchorage or fishing ground. Isolation forest and autoencoder-based anomaly detectors have been applied to AIS data at scale in peer-reviewed work, and both can flag loitering clusters without requiring labelled smuggling events as training data, which are rarely available to civilian analysts.
The output is an anomaly score per trajectory segment, not a binary accusation. Scores above a defined threshold trigger a secondary review step rather than an automatic alert. That distinction matters legally and operationally.
When AIS goes dark, SAR takes over
Vessels involved in contraband transfer frequently disable or suspend AIS transmission during the rendezvous itself. This is the critical gap. A trajectory that shows a vessel loitering and then disappearing from AIS is more suspicious than one that simply shows loitering, but it is also the moment when the satellite record becomes hardest to read.
SAR fills that gap. Sentinel-1 detects vessels as bright radar cross-section returns against the ocean background. At 5 x 20 m IW resolution, a vessel of 50 metres or more is reliably detectable in moderate sea states; smaller vessels become ambiguous as sea clutter rises above Beaufort 4 or 5. ICEYE Spotlight at 1 m resolution can resolve hull geometry well enough to estimate vessel length and, in some cases, distinguish a cargo vessel from a fishing boat, which helps rule out false positives.
The confirmation workflow is: anomaly score breaches threshold, last known AIS position and drift estimate define a search box, SAR collection is tasked over that box, detection or non-detection is logged against the AIS gap. A confirmed SAR detection of two vessels in proximity during an AIS-off period, in a location with no declared anchorage or fishing purpose, is the strongest available satellite evidence short of an optical image with sufficient resolution to read hull markings.
Honest limits of the method
Revisit is the binding constraint. Sentinel-1 open-access revisit of 3 to 6 days means that a rendezvous lasting only a few hours will often fall between passes. Commercial SAR from ICEYE improves this significantly, but tasking must be triggered quickly after the AIS anomaly is scored, which requires near-real-time AIS processing and an automated tasking pipeline. Even then, cloud is irrelevant for SAR, but orbital geometry is not: a pass may not be available within the required window.
Small vessels below roughly 20 metres are unreliable SAR detections in anything above calm sea conditions. Semi-submersibles and low-freeboard craft, which are used in some narcotics operations in the Eastern Pacific, present radar cross-sections close to the noise floor of current civil SAR systems. AIS carriage is not mandatory for vessels under 300 gross tonnes on international voyages, so the AIS baseline itself has gaps for smaller craft.
Finally, anomaly scoring produces candidates, not convictions. Legitimate reasons for loitering include mechanical breakdown, crew medical emergencies, waiting for a pilot, weather shelter and rendezvous with a supply vessel. Ground-truth from patrol assets or port intelligence is needed to close the loop.
How the analytics layer is structured
The practical workflow runs in three tiers. First, continuous AIS ingestion from Spire and exactEarth feeds, with trajectory segmentation and anomaly scoring running against a rolling baseline updated weekly by vessel type and region. Second, an alert queue ranked by anomaly score, filtered by proximity to known trafficking corridors (the Caribbean, West Africa, the Strait of Malacca and similar areas documented in UNODC reporting). Third, on-demand SAR tasking requests generated automatically for the highest-priority alerts, with results returned as georeferenced detections overlaid on the original AIS gap.
Satellize applies this framework on open constellations and adds commercial SAR tasking on client licence. The same anomaly-scoring architecture that underpins the Tonga crop-estimation programme for agricultural monitoring is adapted here to maritime kinematics: different features, same principle of scoring observations against a statistically grounded baseline rather than a fixed rule.
Deliverables to a client maritime operations centre would typically include a daily alert feed in GeoJSON or shapefile format, a weekly pattern report identifying persistent loitering zones, and a case package for each confirmed SAR match containing the AIS trajectory, the anomaly score breakdown, the SAR image chip and metadata, and a plain-language summary suitable for briefing a duty officer who is not a remote-sensing specialist.
Typical figures
| AIS positional update interval (Class-A, under way) | 2 to 10 seconds at speed; up to 3 minutes at anchor (IMO SOLAS requirement) |
| Spire / exactEarth message latency | Typically under 20 minutes for open-ocean positions |
| Sentinel-1 IW SAR ground resolution | 5 m range x 20 m azimuth; 250 km swath |
| Sentinel-1 revisit (two-satellite constellation) | Approximately 3 days at mid-latitudes; 6 days single satellite |
| ICEYE Spotlight SAR resolution | Approximately 1 x 1 m; on-demand tasking, typical collection window under 24 hours |
| Minimum detectable vessel (SAR, calm sea state) | Approximately 20 m length for reliable detection; smaller vessels ambiguous above Beaufort 4 |
| AIS archive depth (Spire / exactEarth) | Multi-year historical records available for retrospective trajectory analysis |
| Sentinel-1 archive depth | From 2014 (Sentinel-1A launch); freely accessible via Copernicus Data Space |
| SAR operating frequency | Sentinel-1: C-band 5.405 GHz; ICEYE: X-band approximately 9.65 GHz |
| Alert-to-SAR-tasking latency (commercial pipeline) | Dependent on operator workflow; technically achievable under 2 hours with automated tasking |
Analytics Satellize can run
| Loitering anomaly score per AIS trajectory segment | Kinematic feature extraction (speed distribution, turning rate, net displacement ratio, distance from declared anchorages) fed into isolation forest or autoencoder anomaly detector trained on vessel-type and region baselines | Daily GeoJSON alert feed ranked by anomaly score, with segment geometry and feature breakdown |
| AIS gap detection and gap characterisation | Cross-referencing Spire and exactEarth feeds to distinguish genuine transmission absence from single-feed dropout; gap duration, last-known position and estimated drift cone calculated for each gap event | Gap event log appended to alert feed, with search-box polygon for SAR tasking |
| SAR vessel detection over AIS-gap search boxes | Constant False Alarm Rate (CFAR) detection on Sentinel-1 IW or ICEYE Spotlight imagery; detected objects cross-referenced against AIS positions of known vessels to identify dark contacts | Georeferenced detection shapefile with vessel length estimate, image chip, and match/no-match flag against AIS |
| Vessel proximity event log during AIS-off periods | Spatial clustering of SAR detections within the gap search box; proximity threshold set at 500 m between centroids to flag potential rendezvous pairs | Case package per proximity event: AIS trajectory, anomaly score, SAR image chip, plain-language summary |
| Persistent loitering zone mapping | Kernel density estimation on historical anomaly-flagged positions aggregated over rolling 90-day windows; compared against UNODC-documented trafficking corridor polygons | Weekly PDF pattern report with hotspot map and trend commentary |
| Vessel re-identification across AIS gaps | Matching SAR-detected vessel length and heading against last-known AIS vessel profile; probabilistic re-identification score where multiple candidates exist in the search box | Re-identification confidence table appended to case package |
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.