Vessel-to-vessel proximity event logging for insurance and liability
Spaceborne AIS and SAR together produce an auditable log of close-approach events between vessels, giving insurers and liability investigators a time-stamped, position-referenced record that shore-based systems routinely miss.
Sensors
- Spire Global spaceborne AIS: A constellation of over 100 LEO cubesats receiving AIS messages globally. Typical revisit for any ocean cell is 20–40 minutes in mid-latitudes, shorter near poles. Position accuracy depends on the originating vessel transponder, nominally ±10 m but degraded by multipath and message collision in dense traffic to ±100 m or more.
- exactEarth spaceborne AIS: Complementary spaceborne AIS network with a focus on reliable message deconfliction in high-density areas such as the English Channel and Singapore Strait. Combining exactEarth and Spire feeds reduces revisit gaps and cross-validates position fixes.
- Sentinel-1 SAR (C-band, ESA): Provides independent vessel detections at 5–20 m resolution in Interferometric Wide Swath mode, with a 250 km swath. Revisit is 6–12 days at the equator, shorter at higher latitudes. SAR detects physical presence regardless of AIS broadcast, so it can confirm or contradict the AIS-derived positions of both vessels at the moment of imaging.
- ICEYE SAR (X-band): Commercial SAR constellation offering sub-1-metre resolution in spotlight mode and tasking latency of a few hours. Useful for post-event tasking to image the exact location of a reported proximity event and recover vessel identifiers from hull geometry when AIS data is disputed.
What counts as a proximity event, and why the definition is contested
A proximity event is any occasion on which two vessels come within a defined distance threshold during a defined time window, as reconstructed from their recorded tracks. The thresholds are not universal. Marine insurers and P&I clubs typically treat 0.5 nautical miles as a collision-risk threshold in open water; port-approach zones often use 0.1 nm. The International Regulations for Preventing Collisions at Sea (COLREGs) do not specify a numerical distance, which means the threshold applied to a logged event will be chosen by the party commissioning the analysis. That ambiguity is worth stating plainly at the outset.
The practical complication is that AIS is a self-reported system. Vessels broadcast their own GPS position, speed and heading, and the accuracy of that broadcast depends on the quality of the vessel's own GNSS receiver, antenna placement and whether the crew have manually overridden the position. IMO performance standards require AIS position accuracy to match the vessel's own GNSS, which is typically better than 10 m under open-sky conditions, but real-world audits have documented errors of 50–200 m in port approaches and congested straits. For a proximity event defined at 100 m separation, that error budget matters enormously.
How satellite AIS reconstructs the encounter
Shore-based AIS receivers have a line-of-sight range of roughly 40–60 nautical miles. Beyond that, only spaceborne receivers capture the broadcast. A spaceborne AIS constellation such as Spire or exactEarth samples each vessel's position at intervals determined by orbital geometry, typically every 20–40 minutes in mid-ocean. The track between two consecutive fixes is interpolated, usually by linear dead-reckoning using the last reported speed and course. The interpolation error grows with time since the last fix and with any change in the vessel's actual heading or speed.
For a proximity event reconstruction, the workflow is: ingest raw AIS message streams for both vessels over the relevant period; flag all epochs at which the interpolated separation falls below the chosen threshold; compute the closest point of approach (CPA) and the time of CPA; and attach confidence bounds derived from the position error and interpolation uncertainty. The output is a time-stamped log entry with a separation estimate and an honest uncertainty range. An event at 0.08 nm separation with a positional uncertainty of ±0.05 nm is a very different evidential product from one at 0.08 nm with ±0.01 nm uncertainty. Good logging systems report both.
Message collision is a specific problem in busy waters. When many vessels transmit simultaneously on the same VHF channel, satellite receivers decode fewer messages per vessel per pass, increasing the gap between fixes. The English Channel, the Malacca Strait and approaches to major ports all exhibit this effect. Combining two independent AIS networks reduces but does not eliminate the problem.
SAR as an independent witness
SAR imagery does not care whether a vessel has its transponder on. A Sentinel-1 Interferometric Wide Swath scene covers 250 km at 5–20 m resolution and can detect vessels down to roughly 300–500 gross tonnes under typical sea-state conditions, with smaller targets detectable in calm water. When a SAR acquisition coincides with a reported proximity event, it provides an independent position fix for each vessel that is entirely separate from the AIS record.
The limit is revisit. Sentinel-1 returns to a given ocean cell every 6–12 days. The probability that a SAR pass coincides with a specific proximity event is low unless the event occurs in a frequently imaged area or tasked commercial SAR is used. ICEYE and similar commercial X-band systems can be tasked within hours of a reported incident, but they capture the scene after the fact. Their value in liability investigation is to image the exact location, confirm bathymetry and sea state, and sometimes recover hull markings at sub-metre resolution. They cannot recreate a past encounter directly, but they can corroborate or contradict the AIS-derived positions by cross-referencing with any near-coincident archive passes.
The revisit gap problem and what it means for liability records
The most important honest caveat in this entire use case is the revisit gap. A 30-minute AIS revisit interval means the track between fixes is inferred, not observed. Two vessels could meet, manoeuvre and separate in the interval between satellite passes, leaving no direct observation of the closest approach. The logged CPA is then a reconstruction from the positions before and after the gap, with uncertainty that can span hundreds of metres.
This is not a reason to dismiss satellite AIS as evidence. It is a reason to report it correctly. A well-constructed proximity event log states the time of the last fix before the estimated CPA, the time of the first fix after, and the interpolated CPA with its uncertainty. Courts and arbitration panels have accepted AIS-derived track reconstructions as evidence, but the credibility of that evidence depends on transparent methodology. Logs that report a single distance figure without uncertainty are, at best, incomplete.
Underwriting and port-approach applications
For marine insurers, a proximity event log serves two distinct purposes. In underwriting, historical proximity frequency for a vessel or a route quantifies exposure in a way that port-call records alone cannot. A tanker that repeatedly transits congested straits at night accumulates a different risk profile from one on open-ocean routes, and that profile is now extractable from multi-year AIS archives going back to roughly 2009 for some providers.
In post-incident liability investigation, the log becomes evidence. The question is typically who altered course, when, and whether COLREGs obligations were met. Satellite AIS can establish the broad geometry of an encounter; it cannot resolve the second-by-second manoeuvring that COLREGs compliance requires. That granularity still comes from the vessel's own voyage data recorder (VDR) if one is fitted and recovered. Satellite data narrows the hypothesis space and cross-checks the VDR record against an independent source.
Port-approach risk assessment is a third application. Proximity event frequency in the approaches to a specific port, aggregated over months, identifies the berths, anchorage lanes and traffic separation scheme segments where close approaches cluster. That analysis informs pilotage risk models and port authority safety reviews. Satellize runs this kind of aggregated spatial analysis on AIS archive data, drawing on the same open and commercial feeds used for its crop-estimation programme in Tonga and its broader analytics work.
Building an auditable log: format and chain of custody
An auditable proximity event record needs four things: a raw message archive with original timestamps and signal metadata; a documented interpolation method; explicit uncertainty quantification at each logged event; and an immutable delivery format. GeoJSON or CSV with cryptographic hashes on each event record satisfies the last requirement. The raw AIS messages should be retained separately from the derived track, so that any party can re-run the reconstruction and verify the output.
SAR detections, where available, should be stored as separate records linked to the AIS log by time and position, not merged into a single position estimate. Merging obscures the provenance of each data point, which is exactly the information a liability investigator needs. The goal is a log that a marine surveyor or arbitration expert can read, interrogate and, if necessary, reproduce independently.
Typical figures
| AIS position accuracy (open ocean) | Typically ±10 m (vessel GNSS-limited); degrades to ±50–200 m in congested areas due to multipath and message collision |
| Spaceborne AIS revisit (mid-latitude ocean) | 20–40 minutes typical per vessel, combining Spire and exactEarth feeds |
| Sentinel-1 SAR resolution (IW mode) | 5 × 20 m (range × azimuth); 250 km swath |
| Sentinel-1 revisit | 6 days at high latitudes, 12 days near equator (single satellite) |
| ICEYE SAR resolution (spotlight) | Sub-1 m; tasking latency a few hours for commercial orders |
| Minimum detectable vessel (Sentinel-1, calm sea) | Approximately 300–500 gross tonnes; smaller in very low sea states |
| AIS archive depth | From approximately 2009 for some commercial providers; Spire and exactEarth archives vary by contract |
| Proximity event threshold (typical insurance use) | 0.1–0.5 nautical miles separation, configurable; time window configurable per investigation |
| Delivery formats | GeoJSON event log, CSV with hash verification, KML track overlay, PDF investigation report |
Analytics Satellize can run
| Proximity event log for a named vessel or voyage | AIS track interpolation with closest-point-of-approach (CPA) computation and positional uncertainty propagation | Time-stamped GeoJSON or CSV log of all events below chosen threshold, with separation estimate and uncertainty range per event |
| SAR-corroborated position fix at time of incident | SAR vessel detection (CFAR algorithm on Sentinel-1 or ICEYE) cross-referenced to AIS interpolated position at nearest satellite pass | Georeferenced SAR chip with detected vessel positions and offset from AIS-reported position, delivered as GeoTIFF and PDF summary |
| Historical proximity frequency profile for a vessel | Multi-year AIS archive query with rolling CPA computation across all vessel pairs within configurable search radius | Risk profile report showing encounter frequency by route segment, time of day and sea area, suitable for underwriting review |
| Port-approach proximity heat map | Spatial aggregation of CPA events within a defined port-approach polygon over a chosen time window | GIS layer (GeoJSON or shapefile) of encounter density, with tabular breakdown by traffic lane and vessel class |
| Track reconstruction report for liability investigation | AIS dead-reckoning interpolation with documented methodology, uncertainty bounds and comparison to any available SAR or VDR data | Signed PDF report with methodology appendix, raw message archive, and track visualisation suitable for submission to arbitration or P&I club review |
| AIS message-gap alert for real-time monitoring | Streaming AIS ingestion with configurable gap-detection threshold; flag when a vessel's fix interval exceeds a set duration in a designated watch area | Near-real-time alert feed (JSON webhook or email) with vessel MMSI, last known position and gap duration |
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.