Fishing Vessel Gear-Type Classification from SAR and Optical
SAR movement signatures and sub-metre optical imagery together distinguish trawlers, longliners, purse-seiners and pot vessels with enough specificity to support quota enforcement and IUU prosecution. Neither sensor alone is sufficient; the combination is.
Sensors
- Sentinel-1 SAR (C-band, ESA): 20 m resolution in Interferometric Wide Swath mode, 250 km swath, 6-day exact repeat at the equator (roughly 2-3 days with both satellites). Detects vessel movement patterns, slow-speed trawl tracks and circular purse-seiner manoeuvres through radar backscatter. Cloud-independent but cannot resolve deck hardware at this resolution.
- Planet SuperDove (PlanetScope): 3-4 m resolution, 8 spectral bands, near-daily revisit over most ocean areas. Resolves gross hull geometry and large deck structures such as net drums and outriggers. Useful for confirming gear-type hypotheses generated from SAR movement analysis. Sunglint and sea state can obscure small vessels.
- Maxar WorldView-3: 31 cm panchromatic, 1.24 m multispectral resolution on tasking. Resolves individual deck equipment: trawl doors, gallows frames, longline coilers, power blocks. Revisit at a given ocean point is irregular and tasking-dependent, so this sensor is used for confirmation of priority targets rather than systematic fleet monitoring.
- Airbus Pleiades Neo: 30 cm panchromatic resolution, tasked. Comparable to WorldView-3 for deck-hardware identification. Stereo collection pairs can provide height estimates of deck structures, adding a further discriminator between gear types. Same caveat on irregular revisit applies.
- Global Fishing Watch AIS-derived track data (public): Not a satellite sensor, but AIS-derived speed and heading data from GFW's published algorithms provide independent movement-pattern context, particularly the characteristic 2-4 knot sustained speed of active trawling versus the stationary periods of longlining. Useful as a cross-check where AIS is present; absent where vessels go dark.
What the gear actually looks like from orbit
Each fishing method imposes a distinct physical signature on both the vessel and its behaviour. A bottom trawler tows one or two large otter doors that spread a net across the seabed; those doors, the trawl warps and the stern gantry are visible at 30 cm resolution as angular protrusions from the hull's working deck. A purse-seiner carries a power block suspended from a boom at the stern quarter and a large net drum; the block is a dense, roughly cylindrical object that WorldView-3 or Pleiades Neo resolve clearly against the deck. Longliners are harder: their gear is largely below the rail, but the coiling machinery and bait-prep tables create a characteristic clutter pattern along the working side of the vessel.
Pot fishers are the most ambiguous in optical imagery alone. Their deck stacking of pots or traps can resemble cargo, and their hull forms overlap with small supply vessels. Movement pattern becomes the primary discriminator for this class.
Movement signatures in SAR: the physics of slow work
Active fishing imposes speed and heading constraints that passive transit does not. Bottom trawlers operate at 2 to 4 knots on a consistent heading for tow durations of 30 minutes to several hours, then execute a turn and repeat. In a sequence of Sentinel-1 passes separated by 12 or 24 hours, this produces a distinctive back-and-forth track pattern in the area of detections. Paired-vessel midwater trawling (two vessels towing a single net between them) shows two bright SAR returns maintaining a fixed separation of roughly 100 to 300 metres on parallel headings, a pattern that is essentially unmistakable.
Purse-seiners are the most visually striking. Net deployment involves a high-speed encirclement of a fish school, typically completed in 10 to 20 minutes. A single SAR pass is unlikely to capture the full circle, but the disturbed wake ring persists for 30 to 60 minutes in calm conditions and is detectable in the backscatter texture of the surrounding sea surface. Longliners set gear and then drift or hold station for hours; their SAR signature is a stationary or very slow-moving return, which is ambiguous without optical confirmation. Pot fishers make repeated short transits between fixed positions, producing a star-shaped track pattern over a sequence of passes.
Honest limits: where the method breaks down
Resolution is the first constraint. Sentinel-1's 20 m pixels are sufficient to detect a vessel and characterise its movement but cannot resolve deck hardware. The gear-type inference from SAR alone carries genuine ambiguity: a slow-moving vessel on a consistent heading could be a trawler or a vessel manoeuvring in congested water. Optical confirmation from Planet, WorldView or Pleiades reduces that ambiguity but introduces cloud dependence. Tropical fishing grounds in the western Pacific and Indian Ocean can have cloud cover exceeding 70% on any given day, making optical tasking unreliable for time-sensitive enforcement.
Vessel length matters too. WorldView-3's 31 cm resolution can resolve deck equipment on vessels above roughly 25 metres. Artisanal vessels below 15 metres, which account for a large share of small-island fishing fleets, are detectable in SAR but their gear type is often indeterminate from imagery alone. Speed-profile analysis from AIS helps where transponders are active, but many small vessels do not carry AIS or switch it off. The method is most reliable for industrial-scale vessels above 30 metres operating in open ocean.
Combining the sensors: a practical workflow
The operationally sound approach runs SAR as the wide-area screening layer. Sentinel-1's 250 km swath and 2-3 day revisit produce a manageable vessel-detection list for a given exclusive economic zone. Movement pattern analysis, applied to a sequence of passes, filters that list into gear-type hypotheses: probable trawlers, probable purse-seiners, ambiguous slow-movers. The ambiguous and high-priority cases then receive optical tasking. Planet SuperDove's near-daily 3-4 m imagery provides a first optical confirmation layer at low cost; WorldView-3 or Pleiades Neo are reserved for vessels where a legally defensible gear-type determination is needed.
The output of this workflow is a classified vessel list with a confidence tier attached to each entry. High confidence requires both a consistent SAR movement signature and an optical image showing characteristic deck hardware. Medium confidence rests on movement pattern alone. Low confidence flags vessels that are stationary or whose behaviour is ambiguous. Enforcement agencies can triage their response accordingly. Satellize applies this multi-sensor workflow in its analytics programmes, including work supporting Pacific island states where the gap between what AIS reports and what imagery shows is frequently the most important finding.
Archive depth adds a retrospective dimension that real-time monitoring cannot. Sentinel-1 data extends back to 2014. A vessel suspected of operating illegally in a closed season can be checked against the historical record to establish whether its movement patterns on earlier dates match active fishing. This is the kind of evidence that supports administrative proceedings, not just operational interception.
What classification actually enables
Gear-type classification is not an end in itself. Its value is what it enables downstream. A state that knows a vessel is trawling inside a zone closed to trawling has an actionable case; a state that knows only that a vessel was present does not. Classification also supports quota management: different gear types have different bycatch profiles and different target species, so a licence issued for longlining does not authorise trawling, and imagery that shows trawl-door geometry on a vessel licensed as a longliner is a direct compliance breach.
For insurers and flag states, gear-type records from imagery provide an independent check on what vessel operators self-report. The combination of SAR movement history and optical deck inspection is increasingly accepted as admissible evidence in fisheries tribunals, particularly where it corroborates other data sources. The method does not replace at-sea inspection, but it narrows the target list to a manageable number of vessels worth inspecting.
Typical figures
| SAR spatial resolution (Sentinel-1 IW mode) | 20 m (range) × 22 m (azimuth) |
| Optical resolution (WorldView-3 / Pleiades Neo) | 30-31 cm panchromatic; 1.2-1.24 m multispectral |
| Optical resolution (Planet SuperDove) | 3-4 m, 8 bands (coastal blue to NIR) |
| SAR revisit (Sentinel-1, both satellites) | 2-3 days at mid-latitudes; 6-day exact repeat per satellite |
| Optical revisit (Planet SuperDove) | Near-daily for most ocean areas |
| Minimum vessel length detectable in SAR | Approximately 15-20 m in calm sea state; larger threshold in high sea state |
| Minimum vessel length for deck-hardware resolution (WorldView-3) | Approximately 25 m for reliable gear-type identification |
| SAR archive depth (Sentinel-1) | From April 2014 (Sentinel-1A launch) |
| Cloud penetration | SAR: fully cloud-independent. Optical: cloud-dependent; tropical coverage often below 30% usable days |
| Typical analytic latency | SAR detection list: 1-3 hours post-pass. Optical-confirmed classification: 12-48 hours depending on tasking queue |
Analytics Satellize can run
| Gear-type classification layer | SAR movement-pattern analysis (speed, heading consistency, track geometry) fused with optical deck-hardware detection using object-based image analysis | GIS polygon layer with vessel detections attributed by gear-type class and confidence tier; updated per Sentinel-1 pass cycle |
| Trawl-track reconstruction | Sequential SAR detection linking across passes to reconstruct tow lines and estimate swept area; validated against published trawling-speed thresholds (2-4 knots) | Georeferenced tow-line vectors with timestamps, suitable for overlay against closed-area boundaries |
| Purse-seiner encirclement event detection | Wake-ring detection in SAR backscatter texture combined with speed anomaly in AIS or SAR-derived Doppler where available | Event log with position, estimated time and vessel identity (where AIS correlates); alert feed via API |
| Optical deck-hardware confirmation report | Tasked WorldView-3 or Pleiades Neo image interpreted against published gear-type morphology; annotated with identified structures (trawl doors, power block, net drum, pot stacks) | Annotated image report with structured findings, suitable for inclusion in enforcement dossier |
| Historical compliance audit | Retrospective SAR archive query (Sentinel-1 from 2014) to establish movement-pattern history for named vessels or licence areas during specified seasons | Tabular vessel-activity record with gear-type inferences per pass, covering requested date range |
| Fleet composition summary for an EEZ | Aggregated classification results across all detections in a defined zone and time window, broken down by gear-type class and flag where AIS correlates | Monthly or quarterly summary report with charts; delivered as PDF and machine-readable CSV |
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.