Illegal large-scale driftnet gear detection at sea

Driftnets longer than 2.5 km are banned by UN resolution, yet AIS tracks only the boat, not the gear. Synthetic aperture radar can image the net itself, and dark-vessel correlation closes the attribution gap.

Sensors

Sentinel-1 C-band SAR (ESA): 5 m x 20 m resolution in IW mode, 6-day exact repeat at mid-latitudes (12-day single-satellite). Detects elongated low-backscatter anomalies associated with net floatlines and suppressed Bragg scattering in their wake. Free and open archive from 2014.
ALOS-2 PALSAR-2 L-band SAR (JAXA): L-band (1.27 GHz) penetrates surface chop differently from C-band, providing complementary contrast for submerged floatline structures. Spotlight mode reaches 3 m resolution; standard stripmap is 10 m. Tasked commercially; revisit roughly 14 days without tasking.
ICEYE X-band SAR constellation: X-band (9.65 GHz) is more sensitive to small surface discontinuities than C-band. Sub-metre spotlight mode available. Commercial tasking enables same-day revisit over a priority area, which is critical for a drifting target that moves with current.
Copernicus Marine Service (CMEMS) ocean-current data: Provides gridded surface-current and wind-field analysis at roughly 1/12-degree resolution, updated daily. Used to project net drift trajectory between SAR acquisitions and to validate whether a detected anomaly moves consistently with passive drift rather than powered vessel motion.

Why AIS cannot see the problem

The 1992 UN moratorium on large-scale pelagic driftnets, reinforced by subsequent General Assembly resolutions, prohibits nets exceeding 2.5 km. Enforcement depends almost entirely on vessel monitoring systems and AIS, both of which track the ship. A driftnet, once deployed, is cut loose and left to fish autonomously for hours. The vessel may loiter nearby, steam away to avoid inspection, or switch off its transponder entirely. The gear and the boat become legally separate objects on the water.

This is not a data-quality problem that better AIS coverage solves. It is a structural gap: no transponder is attached to the net. Patrol aircraft can spot floatlines visually, but coverage is sparse and weather-dependent. What changes the calculus is a sensor that images the ocean surface regardless of cloud or darkness and that can detect the physical signature of the net itself.

What a floatline looks like to a radar

Synthetic aperture radar measures the microwave backscatter of the sea surface. Open water produces a characteristic Bragg-resonance return that varies with wind speed and incidence angle. A driftnet floatline disrupts this in two ways. The floats themselves create discrete point scatterers. More diagnostically, the monofilament mesh damps capillary waves immediately downwind of the line, producing an elongated low-backscatter shadow, a dark stripe against the surrounding sea clutter.

Published research using Sentinel-1 IW data has demonstrated detection of floatline signatures several kilometres long in sea states up to roughly Beaufort 4. Above that threshold, wave-breaking noise begins to mask the capillary-damping signal. JAXA researchers working with ALOS-2 have reported similar results in the North Pacific, where the L-band penetration of surface foam provides a marginal advantage in moderate swell. The honest limit: nets shorter than about 3 km are difficult to distinguish reliably from natural slick features such as biogenic surfactant films, which produce visually similar low-backscatter streaks. Attribution requires corroborating evidence.

Closing the attribution gap with dark-vessel correlation

Detecting a net-like anomaly is necessary but not sufficient for enforcement. The analytical step that converts a SAR observation into an actionable case is correlating the gear signature with a vessel position. SAR ship detection, applied to the same scene, identifies all vessels within a defined radius regardless of whether they are broadcasting AIS. A vessel that appears in SAR but not in AIS is classified as dark.

If a dark vessel sits within a few kilometres of a floatline anomaly, and if the anomaly's orientation and drift vector are consistent with local current data from CMEMS, the combined evidence substantially narrows the attribution range. It does not constitute legal proof on its own. Patrol interception or photographic confirmation remains the evidentiary standard for prosecution. What the satellite analysis provides is a prioritised cue: where to send the patrol, at what time, and with what expectation of finding gear still in the water.

Drift modelling matters here. A net deployed at 0200 local time may have moved 15 to 30 km by the time a morning SAR pass acquires the scene, depending on current speed. Projecting the net's probable position backward to the deployment point, and forward to where a patrol vessel should intercept, is a quantitative problem that ocean-current reanalysis data makes tractable.

Revisit, latency and the moving-target problem

A driftnet is not a fixed installation. It moves continuously, it is recovered or lost within a day or two, and enforcement windows are short. Sentinel-1's 6-day revisit at mid-latitudes is adequate for monitoring persistent fishing grounds and building pattern-of-life intelligence, but it is rarely fast enough to cue a same-day patrol interception. Commercial X-band SAR, such as ICEYE, can be tasked to a priority area with same-day or next-orbit acquisition, which compresses the detection-to-interception timeline to a matter of hours.

Latency from acquisition to processed alert is typically 2 to 4 hours for a Sentinel-1 scene through automated pipelines, and somewhat less for commercial operators with dedicated processing infrastructure. That window is workable if patrol assets are pre-positioned in the region. It is not workable if the nearest patrol vessel is 12 hours away. Satellite detection is most powerful when integrated into a broader maritime domain awareness system that already has assets at sea.

Practical limits and honest caveats

Several sources of ambiguity deserve direct acknowledgement. Natural surfactant slicks, ship wakes, and internal wave surface expressions all produce elongated low-backscatter features that can resemble floatline signatures. Discrimination relies on shape analysis (nets tend to be straighter and longer than biogenic slicks), drift consistency with current vectors, and co-location with vessel detections. A single SAR scene is rarely conclusive. Two acquisitions 12 to 24 hours apart, showing a feature that has drifted coherently with the current, substantially increase confidence.

Cloud cover is not a limitation for SAR, which operates at microwave frequencies that pass through cloud and rain (heavy rain introduces some noise in X-band but is not disabling). Darkness is equally irrelevant. The binding constraints are sea state, revisit cadence, and the availability of patrol assets to act on detections. High sea states that mask the capillary-damping signal are also the conditions that make patrol operations difficult, so the operational window for both detection and interception is similar.

Satellize builds detection pipelines on Sentinel-1 open data and can add commercial tasking from ICEYE or similar operators on client licence. The same analytical architecture used in the Kingdom of Tonga crop-estimation programme, correlating multi-temporal rasters with auxiliary environmental data, applies directly to the drift-modelling component of this workflow.

What a monitoring programme actually looks like

A functional driftnet surveillance programme for a coastal state or regional fisheries management organisation typically combines three layers. First, automated SAR processing over a defined exclusive economic zone or high-seas convention area, running on every available Sentinel-1 acquisition and supplemented by tasked commercial SAR during high-priority fishing seasons. Second, a dark-vessel detection layer applied to the same scenes, cross-referenced against AIS feeds and vessel monitoring system data held by the client authority. Third, a drift-projection service that ingests CMEMS current data and generates intercept coordinates for patrol coordination.

Outputs are delivered as GIS layers, timestamped alert reports, and, where requested, a daily or weekly summary digest. The alert report for a candidate detection includes the SAR scene ID, the detected anomaly's coordinates and estimated length, the nearest dark vessel position, the projected drift track, and a confidence classification based on the number of corroborating indicators present. Enforcement decisions remain with the client authority. The satellite layer provides the intelligence; the legal and operational response is theirs.

Typical figures

Primary sensor frequency	C-band 5.405 GHz (Sentinel-1); L-band 1.27 GHz (ALOS-2 PALSAR-2); X-band 9.65 GHz (ICEYE)
Spatial resolution (operational)	5 m x 20 m (Sentinel-1 IW); 10 m stripmap / 3 m spotlight (ALOS-2); 0.5–3 m spotlight (ICEYE)
Revisit cadence	6-day (Sentinel-1 dual satellite, mid-latitudes); ~14-day without tasking (ALOS-2); same-day tasking available (ICEYE commercial)
Minimum detectable net length (published estimates)	Approximately 3 km in Beaufort ≤4 conditions; shorter nets ambiguous against natural slicks
Sea-state operating limit	Reliable detection to approximately Beaufort 4; performance degrades in higher sea states due to wave-breaking noise
Processing latency (detection to alert)	2–4 hours from SAR acquisition for automated pipelines
Ocean-current auxiliary data	CMEMS global ocean physics analysis, 1/12-degree grid, daily update
SAR archive depth	Sentinel-1: from October 2014. ALOS-2: from 2014. ICEYE: from 2018.
Coverage per scene	Sentinel-1 IW swath: 250 km wide. ICEYE spotlight: ~5 km x 5 km tasked.
Delivery formats	GeoTIFF anomaly layers, GeoJSON vessel and net detections, PDF alert reports, drift-track KML

Analytics Satellize can run

Floatline anomaly detection	SAR low-backscatter streak segmentation using adaptive thresholding and morphological filtering, consistent with published ESA and JAXA driftnet detection literature	GeoJSON layer of candidate net signatures with length estimate, orientation, and confidence score, per SAR acquisition
Dark-vessel detection and AIS correlation	SAR ship detection via constant false alarm rate (CFAR) processing; cross-matched against AIS broadcast records to identify non-reporting vessels	Timestamped dark-vessel position list with distance to nearest floatline anomaly, delivered as GeoJSON or CSV
Net drift projection and patrol intercept coordinates	Lagrangian particle advection using CMEMS surface-current analysis; net position projected forward and backward from detection time	Intercept coordinate report with estimated net position at user-specified future time, plus uncertainty ellipse
Pattern-of-life fishing-ground mapping	Multi-temporal stacking of SAR scenes over a season; frequency mapping of dark-vessel and anomaly co-occurrences to identify persistent high-risk zones	Seasonal heatmap GeoTIFF and summary PDF for fisheries authority strategic planning
Combined evidence confidence classification	Rule-based scoring across four indicators: anomaly morphology, drift consistency, dark-vessel proximity, and vessel behaviour history from AIS archive	Per-event confidence tier (low / medium / high) appended to alert report, with indicator breakdown
EEZ-wide surveillance digest	Automated ingestion of all available Sentinel-1 acquisitions over a defined polygon; aggregated weekly into a sortable event log	Weekly PDF digest and accompanying GIS package for patrol scheduling

Sources

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.