Deep-sea and shallow seabed mining sediment plume detection
Satellite ocean colour sensors detect suspended sediment plumes from seabed mining and dredging by resolving blue-green reflectance anomalies tied to elevated particulate matter. Coverage is global and near-daily, but the method is physically bounded by the photic zone.
Sensors
- NASA PACE OCI: Hyperspectral ocean colour imager with continuous spectral coverage from 340 to 890 nm at roughly 5 nm resolution, 1 km spatial resolution, daily global revisit. Its fine spectral sampling allows separation of mineral-particle backscatter from phytoplankton absorption in ways that broadband sensors cannot, making it the most capable open sensor for this application since its February 2024 launch.
- Sentinel-3 OLCI: 21 spectral bands from 400 to 1020 nm at 300 m spatial resolution, with a two-satellite constellation providing roughly daily revisit at mid-latitudes. The 400–510 nm bands are well suited to quantifying suspended particulate matter (SPM) concentrations through established bio-optical inversion algorithms published by EUMETSAT and ESA.
- MODIS-Aqua: 36 spectral bands; ocean colour bands at 250–500 m resolution with near-daily global coverage. A 20-plus-year archive makes it the baseline for long-term plume climatology. Spatial resolution limits detection to plumes wider than roughly 500 m, and the sensor is past its design life, so continuity is not guaranteed.
- Landsat-9 OLI: 30 m multispectral resolution with coastal/aerosol band (Band 1, ~435–451 nm) useful for nearshore and shallow-water SPM mapping. Revisit is 16 days for a single satellite, which limits operational monitoring of fast-moving plumes. Best suited to shallow dredging in coastal or lagoonal settings where fine spatial detail matters more than frequency.
What ocean colour actually measures
When a seabed mining or dredging operation lifts sediment into the water column, suspended mineral particles scatter upwelling light differently from clean water or phytoplankton. The ratio of remote-sensing reflectance in the blue (~443 nm) and green (~560 nm) bands shifts measurably as suspended particulate matter (SPM) concentrations rise. At concentrations above roughly 1–2 g per cubic metre, this shift is detectable by sensors with adequate spectral resolution. PACE OCI and Sentinel-3 OLCI both meet that threshold in open-ocean conditions; MODIS-Aqua does too, though with coarser spatial discrimination.
The physics matters because it defines the honest scope of the method. Satellites measure water-leaving radiance from the surface and the photic zone, typically the upper 10–30 m depending on water clarity. A deep-sea mining operation working at 1,000 to 5,000 m depth generates a primary plume near the seafloor that is entirely invisible to any current satellite sensor. What satellites can detect is the secondary plume: sediment that rises buoyantly or is advected upward into the photic zone, plus any surface discharge from the mining vessel's dewatering operations. That is a real and significant environmental signal, but analysts and regulators should be clear about which part of the system they are observing.
Separating mineral plumes from everything else the ocean does
Open-ocean turbidity has many natural causes: phytoplankton blooms, resuspension by storms, river outflow, and carbonate sediment in shallow tropical waters. Distinguishing a mining-derived mineral plume from these backgrounds is the central analytical challenge. PACE OCI's hyperspectral capability is particularly valuable here because mineral particles (silicates, metal oxides, polymetallic nodule fines) have distinctive spectral backscatter shapes that differ from organic particles. Published bio-optical algorithms, including the quasi-analytical algorithm (QAA) family developed for MODIS and adapted for OLCI, decompose total backscatter into particulate and dissolved components. When combined with particle size proxies derived from spectral slope, it is possible to flag anomalies consistent with mineral-rich resuspension rather than biogenic turbidity.
Temporal analysis strengthens the attribution. A plume that appears at a known vessel position, persists over days, and advects consistently with local current fields is a different signal from a storm-resuspension event. Vessel AIS data, when cross-referenced with plume origin points, provides the corroborating evidence that turns a spectral anomaly into an operational finding. Cloud cover is the principal operational constraint: persistent cloud in tropical mining regions can block optical observation for days at a time, and no amount of spectral sophistication overcomes an opaque atmosphere.
Shallow dredging is a different problem from deep-sea mining
Aggregate dredging in coastal waters, sand extraction for land reclamation, and coral-reef-adjacent dredging for port construction all generate plumes in water shallow enough that Landsat-9 OLI's 30 m resolution becomes the right tool. At that scale, the spatial footprint of the plume relative to sensitive habitats (seagrass beds, coral reefs, mangrove fringes) can be mapped with enough precision to support compliance monitoring against permitted dredge corridors.
Deep-sea polymetallic nodule or seafloor massive sulphide mining, by contrast, operates in international or exclusive economic zone waters at depths where the satellite's view is limited to the secondary surface signal described above. The International Seabed Authority's environmental regulations for contractors in the Clarion-Clipperton Zone require monitoring of both near-bottom and surface plumes, and satellite data addresses only the latter. Any monitoring programme that relies solely on satellite observation for deep-sea operations is incomplete; the satellite layer is best understood as a wide-area screening tool that triggers targeted in-situ sampling when anomalies appear.
Archive depth and the baseline problem
Detecting a plume is straightforward compared to proving it is anomalous. That requires a credible baseline: a statistical description of normal SPM variability at the site before mining began. MODIS-Aqua's archive extends to 2002, Sentinel-3A's to 2016, and Landsat's ocean-colour-usable archive to the early 1980s for coastal work. For most proposed deep-sea mining areas in the central Pacific or Indian Ocean, this archive captures pre-disturbance conditions with reasonable fidelity, since industrial seabed mining has not yet begun at commercial scale.
Building that baseline requires care. Seasonal cycles in productivity, episodic storm resuspension, and inter-annual variability driven by ENSO all need to be characterised before a monitoring threshold can be set. A plume detection system that triggers on any elevated SPM reading will produce false positives constantly in a productive upwelling region. One calibrated against the site's own historical distribution will be far more useful to a regulator or an operator managing their permit conditions.
What an operational monitoring programme looks like
A practical programme combines near-daily Sentinel-3 OLCI processing for wide-area screening with targeted Landsat-9 acquisitions at higher resolution when an anomaly is flagged. PACE OCI adds spectral discrimination for ambiguous cases. Outputs are typically SPM concentration maps in milligrams per litre, plume area estimates in square kilometres, and advection trajectories derived from ocean current models (CMEMS or HYCOM) that project where the plume will travel over the following 24–72 hours.
Satellize runs this class of ocean colour analytics on open constellations and can configure alert thresholds against a client-specific baseline, delivering findings as georeferenced GIS layers or structured reports timed to regulatory reporting cycles. The Tonga crop-estimation programme demonstrates that the same analytical pipeline approach, combining multi-sensor time series with local ground-truth calibration, transfers across very different domains. For seabed mining clients, the equivalent ground truth is water-column turbidity sensor data from the vessel or moored instruments, used to validate and periodically recalibrate the satellite-derived SPM retrievals.
Honest limits, stated plainly
Cloud cover can block observation for three to seven consecutive days in equatorial regions during wet seasons. No satellite sensor currently detects deep plumes below the photic zone. SPM retrieval accuracy degrades in highly absorbing waters (such as those influenced by coloured dissolved organic matter from river outflow) and at very low solar elevation angles. Sentinel-3 OLCI's 300 m pixel size means that narrow plume filaments or small dredge footprints in coastal waters may be underestimated in area or missed entirely. Landsat-9 closes that gap spatially but introduces a 16-day revisit gap that makes it unsuitable as the sole sensor for fast-moving operations.
The method also cannot distinguish between a mining-related plume and a natural resuspension event from seismic activity or a passing storm without the corroborating vessel and current data. Satellite ocean colour is a necessary component of seabed mining environmental monitoring. It is not sufficient on its own.
Typical figures
| Best spatial resolution (open ocean) | 300 m (Sentinel-3 OLCI); 1 km (PACE OCI, MODIS-Aqua) |
| Best spatial resolution (nearshore/coastal) | 30 m (Landsat-9 OLI) |
| Typical revisit (open ocean) | Near-daily with Sentinel-3A/B pair or PACE; 16 days for Landsat-9 alone |
| Key spectral bands for SPM retrieval | Blue (~443 nm), green (~560 nm), red (~665 nm), NIR (~865 nm); PACE adds continuous 340–890 nm hyperspectral coverage |
| Approximate SPM detection floor | ~1–2 g m⁻³ in clear open-ocean water under cloud-free conditions (published QAA-based retrievals) |
| Depth penetration (photic zone) | Typically 10–30 m depending on water clarity; deep-sea plumes below this are not detectable |
| Archive depth | MODIS-Aqua from 2002; Sentinel-3A from 2016; Landsat coastal band from 2013 (OLI); PACE from February 2024 |
| Cloud cover constraint | Full optical blockage; persistent cloud in tropical regions can cause 3–7 day observation gaps |
| Typical latency (operational products) | 6–24 hours after satellite overpass for near-real-time Sentinel-3 and MODIS products |
| Delivery formats | Georeferenced SPM concentration rasters (GeoTIFF), plume polygon shapefiles, time-series CSV, PDF compliance reports |
Analytics Satellize can run
| SPM concentration maps | Bio-optical inversion using quasi-analytical algorithm (QAA) applied to Sentinel-3 OLCI or PACE OCI water-leaving reflectance | GeoTIFF raster of SPM in mg/L, updated near-daily, with uncertainty layer |
| Plume area and extent time series | Thresholded SPM anomaly detection against site-specific historical baseline; plume polygon extraction | Shapefile or GeoJSON of plume boundary per acquisition, with area in km² and peak concentration tabulated |
| Plume advection forecast | Lagrangian particle tracking using CMEMS or HYCOM surface current fields initialised at detected plume origin | 72-hour advection trajectory map showing projected plume footprint, delivered as GIS layer with confidence envelope |
| Anomaly alert against permit baseline | Statistical exceedance detection: SPM retrieval compared to percentile thresholds derived from multi-year MODIS/OLCI archive at the licensed site | Automated alert report (PDF and email) triggered when SPM exceeds agreed threshold, with supporting imagery and vessel AIS cross-reference |
| Spectral particle-type classification | Hyperspectral backscatter slope analysis using PACE OCI to discriminate mineral-rich from biogenic particulates | Classified raster distinguishing mineral-dominated from phytoplankton-dominated turbidity, with confidence score per pixel |
| Long-term baseline characterisation | Multi-year MODIS-Aqua and Sentinel-3 time-series decomposition to separate seasonal, inter-annual and episodic SPM variability | Site baseline report with seasonal SPM climatology, ENSO sensitivity assessment, and recommended monitoring thresholds |
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.