Marine fish cage structural change and expansion detection
Very-high-resolution optical and SAR imagery can track the footprint, count and layout of offshore fish cages over time, flagging unauthorised expansion, relicensing drift and storm damage before biosecurity breaks down.
Sensors
- Maxar WorldView constellation (WorldView-2, WorldView-3, WorldView-4): Panchromatic resolution of 0.31 m (WorldView-3) allows individual cage collars, mooring lines and walkways to be resolved and measured. Multispectral bands at 1.24 m support colour differentiation of net material and surface fouling. Revisit at a given point varies from 1 to 4.5 days depending on latitude and tasking priority.
- Planet SkySat: 0.5 m panchromatic, 1 m multispectral. Rapid tasking (typically same-day or next-day) makes it practical for post-storm damage surveys. The constellation of 21 satellites provides flexible scheduling, though cloud cover over coastal sites remains a hard limit for optical collection.
- ICEYE SAR constellation: X-band synthetic aperture radar at 0.5 m spotlight resolution. Metal cage frames and galvanised floating collars produce strong double-bounce and specular returns that stand out sharply against the surrounding water clutter. Operates through cloud, rain and darkness, which matters enormously for storm-damage assessment when optical windows may be closed for days.
- Sentinel-1 SAR (ESA): C-band SAR at 5 x 20 m resolution in Interferometric Wide Swath mode, or 5 x 5 m in Extra Wide Swath. Free and open, with a 6-day repeat at the equator (3-day with both satellites). Sufficient to detect large cage arrays and track gross footprint change or cage removal, though individual cage geometry requires the higher-resolution commercial options.
What a metal collar gives away to a radar
Offshore fish cages are, from a radar physics perspective, almost ideal targets. The circular or square floating collar, typically fabricated from high-density polyethylene or galvanised steel, sits at the air-water interface and presents a corner-reflector geometry to an obliquely illuminated SAR. The result is a bright, stable return that persists across passes and across seasons. A 120-metre diameter cage circle in ICEYE spotlight mode is not ambiguous. It is a distinct disc of elevated backscatter surrounded by relatively dark open water.
Sentinel-1 can detect cage arrays at the scale of a licensed farm block, which is enough for regulatory monitoring of whether a concession boundary has been respected. Distinguishing one cage from the next, or measuring whether a 25 m cage has been replaced by a 40 m cage, requires the sub-metre spotlight products from ICEYE or comparable X-band systems. The physics does not change with cloud cover or time of day, which is the primary operational advantage over optical sensors in high-latitude or monsoon-affected sites.
Where optical imagery earns its place
SAR tells you where cages are and roughly how large they are. Optical imagery at 0.3 to 0.5 m resolution adds structural detail: walkway configuration, mooring anchor points, the presence or absence of feed barges, and visible net deformation after a storm event. WorldView-3 at 0.31 m panchromatic can resolve the collar width and spacing with enough precision to cross-check declared cage dimensions against what is physically in the water.
The honest limit is cloud. Coastal aquaculture sites in Norway, Chile, Scotland and Southeast Asia all experience extended overcast periods. A post-cyclone optical survey may be delayed by several days while cloud persists, precisely when damage assessment is most urgent. This is why a combined SAR-plus-optical workflow is standard practice for serious monitoring programmes: SAR confirms the event and approximate geometry within hours; optical provides the structural detail when skies clear.
Neither sensor type can determine stocking density, fish biomass or disease status from imagery alone. Net colour and surface disturbance patterns have been studied as proxies, but no published method has demonstrated operationally reliable stocking-status classification without ground-truth data from the operator. That limit is worth stating plainly before a procurement decision is made.
Change detection: the method behind the map
The analytical backbone for cage monitoring is multi-temporal object-based image analysis. A baseline image establishes the licensed footprint: cage count, cage centroids, total water-surface area occupied, and the polygon of the concession boundary. Subsequent acquisitions are co-registered to the baseline and segmented using the same object parameters. Differences in centroid position, cage count or total occupied area beyond a defined threshold trigger a change flag.
For SAR time series, coherence change detection adds a second layer. Permanent structures such as fixed cage collars maintain high coherence between passes; temporary surface features such as wave patterns do not. A new cage that appears between two passes will show as a coherence anomaly, separable from environmental noise with appropriate filtering. Published work using Sentinel-1 coherence over salmon farm sites in Norway has demonstrated detection of cage installation and removal events at the farm-block level.
Positional accuracy matters for licensing compliance. A cage cluster that has drifted 200 metres from its licensed coordinates may cross into a marine protected area or a shipping lane. Sub-metre SAR and optical products, when orthorectified against accurate coastal DEMs, can locate cage centroids to within 2 to 5 metres, which is more than adequate for most concession boundary checks.
Storm damage and the biosecurity argument
When a cyclone or severe winter storm passes through a cage aquaculture area, the immediate regulatory concern is net integrity. A breached net releases fish into the surrounding environment, with consequences ranging from genetic contamination of wild stocks (Atlantic salmon escapes in Norway and Scotland are a documented problem) to introduction of disease into wild populations. Rapid damage assessment is therefore not just an insurance question.
SAR is the first tool deployed after a storm because optical collection is usually blocked. A comparison of pre-storm and post-storm SAR backscatter can identify cages that have been displaced, submerged or destroyed. Displaced cages show as new bright returns outside the licensed boundary polygon. Destroyed cages show as the disappearance of a previously stable return. The interpretation requires care: storm-driven surface roughness elevates background clutter and can mask weak returns, so analysis should use the calmest available post-storm pass rather than the first available one.
Honest limits and what they mean for programme design
Three limits deserve explicit attention before specifying a monitoring programme. First, very-high-resolution commercial tasking is expensive per square kilometre, and a large aquaculture jurisdiction may have hundreds of licensed sites spread across a wide coastline. A practical programme tiers the response: Sentinel-1 provides the baseline surveillance across all sites on a 6-day repeat; commercial SAR or optical tasking is triggered only for sites that show a change flag or that are subject to a specific compliance investigation.
Second, archive depth varies by sensor. Sentinel-1 data is publicly available from 2014, which allows a decade of change history to be reconstructed. Commercial VHR archives are patchier over coastal aquaculture sites, which are rarely priority tasking targets for general archive collection. Baseline establishment for a new monitoring programme may require a dedicated tasking campaign.
Third, the method detects structural change. It does not detect what is happening inside the cage. Feed conversion ratios, mortality events, parasite loads and water quality at the cage boundary all require in-situ instrumentation or supporting data from the operator. Satellize's analytics work, as in the Tonga crop-estimation programme, pairs satellite-derived structural and environmental layers with ground-truth data to improve inference. The same principle applies here: the satellite layer is most valuable when it is integrated with the operator's own operational records, not treated as a standalone compliance oracle.
Typical figures
| Best available spatial resolution (optical) | 0.31 m panchromatic (Maxar WorldView-3); 0.5 m (Planet SkySat) |
| Best available spatial resolution (SAR) | 0.5 m spotlight (ICEYE X-band); 5 m stripmap (Sentinel-1 IW) |
| Revisit (Sentinel-1) | 6 days single satellite, 3 days combined A+B at equator; shorter at higher latitudes |
| Revisit (commercial VHR optical/SAR) | 1 to 4.5 days (WorldView constellation); same-day to next-day (SkySat tasking); sub-daily possible with ICEYE constellation |
| Radar frequency | C-band 5.405 GHz (Sentinel-1); X-band ~9.6 GHz (ICEYE) |
| Minimum detectable cage structure | Individual cage collars from ~10 m diameter in VHR SAR spotlight; farm-block level (~100 m aggregate) in Sentinel-1 IW |
| Positional accuracy (orthorectified VHR) | 2 to 5 m CE90 against accurate coastal reference, sufficient for concession boundary checks |
| Archive depth | Sentinel-1: from 2014 (open access); commercial VHR: variable, typically 2016 onwards with gaps over aquaculture sites |
| Cloud sensitivity | SAR: none. Optical: complete blockage under cloud; coastal sites may experience multi-day gaps |
| Delivery formats | GeoTIFF change maps, GeoJSON cage-centroid layers, PDF compliance reports, GIS-ready polygon shapefiles |
Analytics Satellize can run
| Baseline cage inventory | Object-based image analysis (OBIA) on VHR optical or SAR imagery to segment and classify cage structures, extract centroids, diameters and total occupied area | GeoJSON polygon layer with per-cage attributes (centroid coordinates, estimated diameter, cage count) and PDF summary report |
| Periodic change-detection alert | Multi-temporal OBIA comparing current acquisition to baseline; flags additions, removals and positional shifts beyond user-defined threshold | Automated alert with change-flag polygons overlaid on licensed concession boundaries, delivered as GIS layer and email notification |
| Concession boundary compliance check | Spatial intersection of detected cage centroids and footprints against official licensed area polygons supplied by the regulatory authority | Compliance report identifying cages outside licensed boundaries, with distance-to-boundary measurements and timestamped imagery evidence |
| Post-storm damage assessment | Pre/post SAR backscatter comparison using intensity differencing; coherence change detection to identify displaced or destroyed structures | Damage classification map (displaced, destroyed, intact) per cage unit, delivered within 24 to 48 hours of suitable post-storm SAR pass |
| Long-term expansion trend analysis | Annual or quarterly time-series analysis of total cage footprint and cage count using Sentinel-1 archive from 2014 and available VHR tasking | Time-series chart and GIS layer showing footprint growth or contraction per licensed site over the analysis period |
| Unauthorised installation screening | Sentinel-1 wide-area surveillance combined with VHR confirmation tasking; new bright SAR returns outside registered concession polygons flagged for follow-up | Flagged-site report with coordinates, imagery thumbnails and recommended inspection priority ranking |
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.