Aquaculture effluent and nutrient plume mapping in coastal receiving waters
Dense aquaculture installations elevate chlorophyll-a and suspended solids in receiving waters, creating optical signatures detectable by Sentinel-2 and Landsat-9. Band-ratio methods map plume extent and intensity, but atmospheric correction uncertainty in coastal water remains a real constraint regulators must understand before acting on the data.
Sensors
- Sentinel-2 MSI: 10 m resolution in visible and near-infrared bands (B2–B8A); 5-day revisit at the equator with both satellites. Red-edge bands B5 and B6 (705 nm, 740 nm) are particularly useful for chlorophyll-a estimation in optically complex coastal water. Free archive from 2015.
- Landsat-9 OLI-2: 30 m resolution across six reflective bands including a dedicated coastal-aerosol band (Band 1, 443 nm) that aids atmospheric correction over water. 16-day revisit solo, 8-day combined with Landsat-8. Free archive complements Sentinel-2 for multi-decadal trend analysis.
- PlanetScope SuperDove: 3 m resolution, near-daily revisit, eight spectral bands including red-edge. Useful for resolving sub-plume structure that Sentinel-2 misses, but the atmospheric correction pipeline over coastal water is less mature than for the ESA or USGS flagship sensors. Commercial licence required.
- MODIS Aqua (MODOCGA): 250–500 m resolution, twice-daily revisit. Too coarse for individual farm plumes but valuable for basin-scale chlorophyll anomaly context and for cloud-gap-filling in time series. Ocean colour products from MODIS are among the best-validated in the public record.
What a nutrient plume looks like to a satellite
Aquaculture effluent, whether from open net-pen salmon farms, coastal shrimp ponds or mussel longlines, raises dissolved inorganic nitrogen and phosphorus in the water column. Phytoplankton respond within days. The elevated chlorophyll-a shifts water-leaving radiance: absorption deepens in the blue (around 443 nm) and a secondary peak appears near 700 nm as pigment fluorescence and cell scattering combine. Suspended faecal pellets and uneaten feed add a separate turbidity signal, increasing backscattering across visible wavelengths and raising reflectance in the red and near-infrared.
Sentinel-2 captures both signals. The ratio of Band 5 (705 nm) to Band 4 (665 nm) is a well-established proxy for chlorophyll-a in Case-2 coastal water, where dissolved organic matter and sediment complicate the simpler blue-to-green ratios that work in the open ocean. Turbidity is commonly estimated from Band 4 or Band 6 (740 nm) reflectance. Neither ratio is a direct physical measurement: they are empirical relationships calibrated against in-situ data, and the published calibration coefficients vary by water type. Applying an open-ocean calibration to a turbid estuary downstream of a salmon farm will produce wrong numbers.
Atmospheric correction: the error that does not announce itself
Over land, atmospheric correction errors of a few percent in surface reflectance are tolerable. Over water, the water-leaving signal is so small, typically 5–10% of top-of-atmosphere radiance in the blue, that the same absolute error becomes dominant. The standard processors, ESA's Sen2Cor and USGS's LaSRC, are designed primarily for land. Dedicated water processors such as C2RCC (Case-2 Regional CoastColour) and ACOLITE have been shown in peer-reviewed literature to outperform land-targeted processors in coastal and inland water, but they introduce their own assumptions about aerosol type and adjacency effects from bright land pixels near the shore.
The practical consequence is that chlorophyll-a retrievals in coastal water carry uncertainties that can exceed 30–50% in absolute terms, depending on aerosol loading and proximity to land. Plume detection is more reliable as a relative anomaly, comparing downstream pixels to an upstream reference taken from the same image, than as an absolute concentration estimate. Regulators should treat satellite-derived chlorophyll maps as evidence of spatial pattern and temporal change, not as substitutes for in-situ sampling at licence compliance thresholds.
Building a time series that means something
A single cloud-free image tells you a plume exists. A two-year monthly composite tells you whether it is growing, shrinking or seasonal. The combined Sentinel-2A and 2B archive, free from the Copernicus Data Space, reaches back to 2017 for most coastal areas. Landsat-8 and -9 extend that to 2013 and, with Landsat-5 and -7, to the 1980s at 30 m resolution, though the spectral configuration differs and cross-sensor harmonisation requires care.
Seasonal stratification matters. Many coastal receiving waters are well-mixed in winter and stratified in summer; the same nutrient load produces a larger surface chlorophyll response when the water column is stratified. A time series that does not account for this will confuse seasonal physics with farm-level change. Cloud cover is a separate problem: in high-latitude fjord systems where salmon farming is concentrated, winter cloud cover can reduce usable Sentinel-2 acquisitions to fewer than four per month. MODIS Aqua's twice-daily revisit helps fill gaps at the cost of spatial detail.
Change detection is most defensible when the analysis compares the same season across years and uses a consistent atmospheric correction chain throughout. Mixing processors mid-series introduces offsets that can mimic real trends.
Spatial resolution limits and what they mean for compliance
Sentinel-2's 10 m pixels are fine enough to distinguish a plume extending 200 m from a cage cluster from background water, provided the plume has a detectable optical contrast. Sub-pixel mixing is a genuine issue near the cage boundary itself: a pixel that is half cage structure and half water will return an intermediate reflectance that belongs to neither. PlanetScope at 3 m reduces this problem but introduces its own atmospheric correction uncertainty.
For regulatory purposes, the relevant question is usually not 'what is the chlorophyll concentration at this pixel?' but 'does the spatial footprint of elevated chlorophyll extend beyond the licensed mixing zone?' That is a boundary-detection problem, and it is better suited to satellite data than absolute concentration retrieval. A plume that consistently reaches 800 m downcurrent when the licence specifies a 300 m mixing zone is a finding that satellite imagery can support, especially when corroborated across multiple acquisitions and seasons.
Cumulative impact across a licensed bay: the harder problem
Individual farm plumes are tractable. Cumulative impact across a bay holding a dozen licensed sites is harder, because plumes overlap and the background chlorophyll field reflects regional oceanography as well as farm discharge. Separating the farm signal from natural variability requires a baseline: either a pre-farming archive (Landsat provides this for sites licensed in the last decade) or a spatial control, comparing the bay to an adjacent embayment with similar physical characteristics but no farms.
Neither approach is clean. Pre-farming baselines assume the regional chlorophyll regime has not changed for other reasons, such as climate-driven shifts in upwelling or river discharge. Spatial controls assume the two bays are oceanographically comparable, which is rarely exactly true. Satellize applies these methods to open-constellation data for regulatory clients, drawing on the same analytical approach used in its Tonga crop-estimation programme: pixel-level anomaly scoring relative to a climatological baseline, with explicit uncertainty bounds attached to each output.
The honest position is that satellite data can narrow the range of hypotheses about cumulative impact. It cannot replace a well-designed field monitoring programme. What it can do is make that field programme cheaper by directing sampling effort to the times and locations where the satellite signal suggests the largest anomalies.
What the data cannot see
Optical sensors see the surface. Benthic nutrient loading, the accumulation of organic matter beneath cages, is invisible to any satellite sensor currently in operation. Hyperspectral systems can detect changes in benthic reflectance in clear, shallow water, but the turbid conditions common beneath active farm sites defeat that approach. Sub-surface chlorophyll maxima, common in stratified water, are also invisible: a satellite chlorophyll map of a stratified fjord in summer may underestimate total phytoplankton biomass by a factor of two or more if the peak is below the optical depth of the sensor.
Night-time thermal infrared can detect warm discharge plumes from land-based recirculating aquaculture systems, but that is a different use case. Cloud is the most immediate operational constraint: in many high-value aquaculture regions, cloud-free revisit at Sentinel-2 resolution may average once every two to three weeks across a full year, which limits the temporal resolution of any compliance monitoring programme built on optical data alone. SAR can see through cloud but does not detect chlorophyll or turbidity directly.
Typical figures
| Spatial resolution (primary) | 10 m (Sentinel-2 MSI visible and NIR bands); 30 m (Landsat-9 OLI-2) |
| Spatial resolution (high-res supplement) | 3 m (PlanetScope SuperDove, commercial licence) |
| Revisit frequency | 5 days (Sentinel-2A+B combined, equatorial); 8 days (Landsat-8+9 combined); near-daily (PlanetScope); twice daily (MODIS Aqua) |
| Key spectral bands for plume detection | Sentinel-2 B4 (665 nm), B5 (705 nm), B6 (740 nm), B8A (865 nm); Landsat-9 Band 1 (443 nm), Band 3 (561 nm), Band 4 (655 nm), Band 5 (865 nm) |
| Minimum detectable plume extent (optical contrast dependent) | Approximately 1–2 Sentinel-2 pixels (~10–20 m) for high-contrast plumes in clear water; larger in turbid or mixed-pixel conditions |
| Chlorophyll-a retrieval uncertainty (coastal water) | Typically 30–50% absolute error depending on atmospheric correction processor and water type; relative anomaly detection is more reliable than absolute retrieval |
| Archive depth | Sentinel-2: 2015–present (global coverage from ~2017); Landsat: 1972–present; MODIS Aqua: 2002–present |
| Cloud limitation | Optical only; cloud-free revisit can fall below 4 usable scenes per month in high-latitude or monsoon-affected regions |
| Latency (open data pipelines) | Sentinel-2 Level-2A typically available within 3–5 hours of acquisition via Copernicus Data Space; Landsat within 24 hours via USGS EarthExplorer |
| Delivery formats | GeoTIFF chlorophyll/turbidity anomaly rasters, GeoJSON plume boundary polygons, time-series CSV per licensed site, PDF compliance summary report |
Analytics Satellize can run
| Chlorophyll-a anomaly map per acquisition | Sentinel-2 B5/B4 red-edge band ratio with C2RCC or ACOLITE atmospheric correction; anomaly scored against site-specific climatological baseline | GeoTIFF raster and GeoJSON plume boundary polygon per scene, delivered within 24 hours of cloud-free acquisition |
| Turbidity plume extent map | Sentinel-2 Band 4 or Band 6 reflectance threshold after atmospheric correction; spatial extent computed as contiguous pixels exceeding baseline by defined sigma threshold | GeoJSON polygon with area in hectares and maximum downcurrent distance from farm centroid |
| Monthly composite anomaly layer | Median compositing of all cloud-free acquisitions within calendar month; reduces single-scene atmospheric correction noise | GeoTIFF monthly composite with per-pixel observation count; suitable for regulatory reporting |
| Multi-year trend report per licensed site | Mann-Kendall trend test on monthly median chlorophyll anomaly time series; Landsat and Sentinel-2 harmonised using published USGS HLS (Harmonised Landsat Sentinel-2) methodology | PDF report with trend direction, significance level, and annotated time-series chart; uncertainty bounds explicit throughout |
| Mixing-zone exceedance alert | Automated comparison of plume boundary polygon against licensed mixing-zone perimeter loaded from client GIS; exceedance flagged when plume extends beyond boundary in two or more acquisitions within a rolling 30-day window | Email or API alert with scene date, exceedance distance, and link to supporting imagery |
| Cumulative impact baseline assessment | Pre-licensing versus post-licensing chlorophyll anomaly comparison using Landsat archive; spatial control bay included where available; seasonal stratification index applied to normalise inter-annual variability | GIS layer set and written assessment report with explicit caveats on confounding factors |
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.