Fish Aggregating Device Detection and Drift Tracking
Drifting fish aggregating devices concentrate tuna but evade routine monitoring. Optical satellites and spaceborne AIS can detect, identify and track dFADs across EEZs, exposing fishing effort that never appears in catch records.
Sensors
- Sentinel-2 MSI: 10 m resolution in visible and near-infrared bands; 5-day revisit at the equator with both satellites. Capable of resolving FAD surface structure and detecting the anomalous spectral signature of netting or tarpaulin against open water. Free and open archive from 2015.
- Planet Dove: 3–4 m resolution, near-daily revisit in most tropical latitudes. Sufficient to resolve individual FAD components and confirm object identity. Tasked commercially; archive depth varies by licence.
- NOAA VIIRS (SNPP and NOAA-20): 375 m resolution in the I-bands; daily global coverage. Too coarse to resolve individual FADs but useful for detecting anomalous surface reflectance clusters and for night-light signatures where vessels fish around FADs under artificial illumination.
- Spire spaceborne AIS: Global AIS message collection from low-Earth orbit, with latency as low as a few minutes per pass. Links vessels to FAD positions when skippers transmit; exposes gaps where AIS is switched off near known FAD locations.
What a FAD actually looks like from orbit
A drifting FAD is typically a raft of bamboo, rope, netting and sometimes tarpaulin, ranging from a few square metres to around 20 m across. Attached beneath it hangs a subsurface appendage, sometimes 50–100 m of netting, designed to attract fish by mimicking a floating log. At the surface, the object presents a spectrally distinct target: tarpaulin and synthetic netting reflect differently in the near-infrared from open water, and the shadow cast by the raft is detectable at 10 m resolution on a calm, low-glint day.
Sentinel-2's 10 m bands can resolve the surface expression of a large FAD under favourable conditions, though smaller or partially submerged devices sit at or below the detection floor. Planet Dove at 3–4 m improves confidence substantially. The practical limit is not resolution alone but sun-glint geometry: equatorial oceans in the FAD deployment zones of the western and central Pacific produce strong specular reflection that can mask small objects. Analysts schedule imagery acquisitions around low-glint windows, typically early morning passes, and apply glint-correction algorithms before thresholding.
Radio buoys: the RF signature that optical misses
Most industrial-scale dFADs now carry GPS radio buoys, which transmit position data to fishing companies via satellite or cellular relay. These buoys are the monitoring gap that matters. The fishing company knows exactly where its FADs are; the coastal state whose EEZ the FAD has drifted into does not. Some buoy types also carry echosounder payloads that estimate fish biomass beneath the device, giving operators real-time catch forecasts.
Spaceborne AIS receivers pick up vessel transmissions near FAD locations, which allows analysts to correlate vessel presence with known or suspected FAD positions. Where a vessel loiters for several hours at an isolated open-ocean coordinate and then departs, and where optical imagery subsequently shows a surface object at that coordinate, the inference is strong. The RF emissions from the buoys themselves are not routinely collected by open civil satellites, but the vessel-FAD association is recoverable from AIS and optical in combination.
Drift modelling: reconstructing paths through an EEZ
A FAD set in international waters does not stay there. Surface currents in the tropical Pacific and Indian Ocean carry drifting objects at typical speeds of 0.2–0.5 m/s, meaning a FAD can travel 200–400 km in a fortnight. NOAA's Global Drifter Programme and the Copernicus Marine Service publish near-real-time ocean current products that allow analysts to propagate a FAD's last known position forward in time and estimate where it will enter or exit an EEZ.
The method is probabilistic. Current fields have uncertainty, wind drag on the surface structure adds a component that varies with FAD geometry, and the subsurface appendage creates a drag differential that is difficult to parameterise without knowing the exact construction. Ensemble drift forecasts bracket the likely position with a confidence ellipse that grows over time, typically reaching tens of kilometres of uncertainty after five to seven days without a fresh satellite fix. Repeat optical passes at the forecast location then confirm or correct the trajectory.
Counting effort: what FAD density reveals about a fishery
The Inter-American Tropical Tuna Commission and the Western and Central Pacific Fisheries Commission have both documented the scale of dFAD deployment: estimates from observer programmes suggest tens of thousands of dFADs are active in the Pacific at any one time, though the figure is genuinely uncertain because reporting is incomplete. Satellite detection offers an independent count.
Mapping FAD detections across an EEZ over a fishing season produces an effort-distribution layer that is independent of vessel logbooks. High FAD density in a particular zone correlates with high set frequency and, consequently, high bycatch exposure for non-target species including juvenile bigeye tuna, sharks and sea turtles. This layer is directly useful for fisheries management authorities setting spatial or temporal FAD restrictions. It is also useful for tuna buyers conducting supply-chain due-diligence audits, since FAD-caught tuna carries a different sustainability certification status from free-school-caught fish.
Honest limits of the method
Cloud cover is the dominant constraint. Tropical convergence zones produce persistent cloud over exactly the ocean areas where FAD deployment is heaviest. A Sentinel-2 pass over a cloud-covered scene returns nothing useful. Planet's higher revisit rate improves the probability of a clear acquisition, but even daily revisit does not guarantee a cloud-free window in the intertropical convergence zone during wet seasons.
Small FADs, particularly those made from biodegradable materials now encouraged by some RFMOs, may fall below the detection floor of any currently operational optical sensor. A bamboo raft without tarpaulin, partially waterlogged, presents very little spectral contrast. VIIRS is too coarse to resolve individual devices. The method works best for the large, tarpaulin-covered FADs deployed by industrial purse-seine fleets, which are also the devices of greatest management concern. Detection confidence drops sharply for devices smaller than roughly 5 m on a side, even with Planet imagery.
Putting the analysis to work
A fisheries ministry or an RFMO observer programme can commission a FAD-density survey for a defined EEZ by combining Sentinel-2 archive screening with targeted Planet tasking for cloud-gap filling, layered against Spire AIS vessel presence data and Copernicus Marine current products. The output is a georeferenced FAD detection catalogue with drift-trajectory estimates and vessel-association flags, updated on a schedule matched to the fishery's seasonal calendar.
Satellize runs exactly this kind of multi-source optical and AIS fusion, applying the same analytic pipeline it uses for its Tonga crop-estimation programme to maritime object detection problems. The tractable next step for a fisheries authority is a bounded pilot: select one EEZ, one fishing season, and commission a retrospective FAD census against the existing Sentinel-2 archive. That produces a baseline effort map before any new tasking budget is committed.
Typical figures
| Best optical resolution (FAD detection) | 3–4 m (Planet Dove); 10 m (Sentinel-2 MSI) |
| Revisit frequency | Near-daily (Planet Dove); 5 days at equator (Sentinel-2 two-satellite); daily (VIIRS, 375 m I-band) |
| Minimum detectable FAD size | Approximately 5 m surface extent under low-glint conditions at Planet resolution; larger at Sentinel-2 |
| Spectral bands used | Visible (RGB), near-infrared (NIR) for object contrast; VIIRS I1/I2 for surface anomaly screening |
| Cloud impact | High in tropical convergence zones; persistent cloud can block optical acquisition for days to weeks |
| AIS latency (Spire) | Minutes to low tens of minutes per orbital pass; not continuous |
| Drift forecast uncertainty | Tens of kilometres after 5–7 days without a fresh satellite fix; grows with time |
| Sentinel-2 archive depth | 2015 to present, free and open via Copernicus Data Space |
| Delivery formats | GeoJSON FAD detection catalogue, GeoTIFF density rasters, drift-trajectory shapefiles, AIS-association CSV |
Analytics Satellize can run
| FAD detection catalogue | Spectral anomaly detection and object segmentation on Sentinel-2 and Planet imagery, with glint correction pre-processing | Georeferenced point layer of confirmed and probable FAD detections with date, confidence score and estimated size |
| Drift trajectory reconstruction | Ensemble Lagrangian particle tracking using Copernicus Marine Service surface current fields, propagated from last optical fix | Polyline shapefile of reconstructed and forecast FAD paths with positional uncertainty ellipses at 24, 72 and 168 hours |
| EEZ intrusion alerts | Intersection of drift forecast envelopes with EEZ boundary polygons (VLIZ Maritime Boundaries) | Automated alert report when a tracked FAD is forecast to enter a designated EEZ within a configurable time window |
| Vessel-FAD association log | Spatial and temporal co-location of Spire AIS vessel positions with FAD detection coordinates, applying a configurable proximity radius and dwell-time threshold | CSV table linking MMSI, vessel name, flag state and timestamp to FAD detection events |
| Seasonal FAD effort density map | Kernel density estimation over a season's FAD detection catalogue, gridded at 0.1-degree resolution | GeoTIFF raster and summary report showing spatial concentration of FAD deployment effort within a defined EEZ or high-seas pocket |
| Bycatch-risk zone classification | Overlay of FAD density layer with published species habitat models (e.g. juvenile bigeye tuna distribution from WCPFC stock assessments) to flag high-overlap zones | Classified GIS layer and written assessment for fisheries management or supply-chain audit use |
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.