Mine water discharge turbidity plume mapping in river systems
Satellite-derived turbidity indices track suspended sediment plumes from mine dewatering and runoff events in river systems, using red and near-infrared band ratios at 3–10 m resolution. Timing and spatial origin distinguish mine-sourced pulses from background fluvial sediment loads, within the hard constraint of cloud-free overpasses.
Sensors
- Sentinel-2 MSI: 10 m resolution in the red (Band 4, 665 nm) and near-infrared (Band 8, 842 nm) bands used for turbidity ratios; 5-day revisit at the equator with both satellites, reduced to 2–3 days at mid-latitudes. Free archive from 2015.
- Planet SuperDove: 3 m resolution, 8-band multispectral including red-edge and NIR; near-daily revisit over most land surfaces. Commercial licence required. Best choice for resolving narrow discharge channels and early-stage plume geometry.
- Landsat-9 OLI: 30 m resolution, 16-day revisit (8 days combined with Landsat-8). Red (Band 4) and NIR (Band 5) support the same ratio-based turbidity approach; free archive extends the record back to 1972 across the Landsat family, useful for establishing pre-mining baseline.
- MODIS Terra/Aqua: 250 m resolution in Band 1 (red), 500 m in Band 3 (blue); 1–2 daily overpasses per platform. Spatial resolution is too coarse for most mine drainage channels but useful for tracking plume propagation over tens of kilometres in large river systems such as the Amazon or Congo.
What suspended sediment looks like from orbit
Water laden with fine sediment reflects strongly in the red portion of the spectrum and absorbs in the near-infrared. Clear water does the opposite. The ratio of red-band reflectance to NIR reflectance, or variants such as the Normalised Difference Turbidity Index (NDTI), exploits this contrast to produce a dimensionless turbidity proxy across an entire river reach in a single overpass.
The physics is well established. Nechad et al. (2010) and subsequent Sentinel-2 calibration work have shown that red-band surface reflectance correlates reliably with total suspended matter concentrations in the range of roughly 5 to 1,000 mg/L, which covers most mine-discharge plumes. Above that upper bound the signal saturates. Below about 5 mg/L, the contrast against background river water becomes too small to resolve confidently at 10 m. Those are real limits, and any honest turbidity product should report them.
Telling a mine plume from a rain event
Natural sediment pulses and mine-sourced plumes can look identical in a single image. The distinguishing evidence is spatial origin and timing. A plume that originates precisely at a known dewatering outfall, on a day with no significant upstream rainfall, is difficult to attribute to anything other than the mine. Cross-referencing discharge timing with rainfall radar records and upstream gauge data tightens that attribution considerably.
Colour matters too, though it is not diagnostic on its own. Iron-rich mine drainage often produces a characteristic orange-ochre signature in true-colour composites, distinct from the grey-brown of clay-dominated natural floods. Combining the turbidity index with a simple red-to-blue ratio can flag iron-elevated water, though confirming the chemistry still requires in-situ sampling. Satellite data narrows the search; it does not replace the laboratory.
Multi-temporal stacking is the most reliable approach. If a plume appears at the same spatial origin across several overpasses, including periods of low natural flow, the mine attribution strengthens significantly. Conversely, a single anomalous image during a regional storm event warrants caution before regulatory reporting.
Resolution, revisit and the cloud problem
The 10 m pixels of Sentinel-2 can resolve plume boundaries in rivers wider than roughly 30–40 m. Planet SuperDove at 3 m extends that to channels as narrow as 10–15 m, which matters enormously for the small tributaries that connect mine sites to main-stem rivers. Landsat-9 at 30 m is best reserved for historical baseline construction or large rivers where the plume already spans hundreds of metres.
Revisit is the harder constraint. A five-day Sentinel-2 cycle means that a discharge event lasting 12 hours may never be captured at all, or may be captured only in its tail. Planet's near-daily tasking reduces that risk substantially, though at commercial cost. MODIS provides daily coverage but at 250 m, which is functionally useless for most mine drainage channels.
Cloud cover is the binding limit in tropical mining regions. The Congo Basin, Indonesian Borneo, and the Peruvian Amazon, where many of the highest-impact mine-water discharge concerns are located, routinely experience cloud cover exceeding 70% of days. A practical monitoring programme must plan for multi-week gaps and design its compliance reporting around what satellite data can and cannot guarantee.
Calibrating the index to local conditions
A turbidity index is a relative measure unless it is calibrated against in-situ total suspended solids (TSS) or nephelometric turbidity unit (NTU) measurements taken close in time to a satellite overpass. Without that calibration, the product is useful for detecting anomalies and tracking spatial extent, but cannot reliably report concentrations in units that regulators recognise.
Calibration campaigns need not be elaborate. A series of grab samples across a turbidity gradient, collected within two hours of a cloud-free overpass, is usually sufficient to fit a simple power-law or linear regression between band reflectance and measured TSS. Published calibration coefficients from comparable river systems can serve as a starting point, but sediment grain size and mineralogy vary enough between basins that local validation is worth the effort. The Tonga crop-estimation programme Satellize runs illustrates a similar principle: local ground-truth transforms a generic spectral index into a defensible operational product.
What the output actually looks like, and what to do with it
The core deliverable is a georeferenced turbidity anomaly map: pixels classified as elevated turbidity, with a continuous index value and, where calibrated, an estimated TSS concentration range. Overlaid on a river network, this shows plume extent, the downstream distance affected, and the approximate boundary where mine-sourced water mixes with background flow.
Time-series products are more useful than single-event maps for compliance purposes. A monthly or quarterly stack of all cloud-free overpasses, annotated with known discharge events and rainfall records, gives regulators and operators a defensible audit trail. Exceedance frequency, meaning the fraction of cloud-free observations on which turbidity exceeded a defined threshold at a downstream monitoring point, is a metric that translates directly into permit language.
Alert products are possible but require realistic expectations. An automated flag triggered when a new cloud-free image shows elevated turbidity at a designated outfall can be delivered within hours of the satellite downlink, but the false-positive rate from natural turbidity events means that every alert needs a rapid human review before it becomes a formal notification.
Typical figures
| Primary spatial resolution | 3 m (Planet SuperDove), 10 m (Sentinel-2 MSI red/NIR bands), 30 m (Landsat-9 OLI) |
| Minimum resolvable river width | ~10–15 m at 3 m pixel; ~30–40 m at 10 m pixel; ~90 m at 30 m pixel |
| Revisit cadence | Near-daily (Planet); 2–5 days (Sentinel-2, latitude-dependent); 8–16 days (Landsat-8/9 combined/single); 1–2 per day (MODIS, 250 m) |
| Spectral bands used | Red (~665 nm) and NIR (~842 nm) for NDTI; blue (~490 nm) added for iron-ratio screening |
| Detectable TSS range (calibrated) | Approximately 5–1,000 mg/L; saturation above ~1,000 mg/L; noise floor ~5 mg/L in clear water |
| Cloud constraint | Optical only; no usable data under cloud. Tropical regions may have >70% cloudy days |
| Temporal archive depth | Sentinel-2 from 2015; Landsat family from 1972; Planet SuperDove from ~2021 |
| Latency (alert mode) | 2–6 hours after satellite downlink for automated flag; add 1–24 hours for human QA review |
| Delivery formats | GeoTIFF turbidity index raster, vector plume extent polygon (GeoJSON/Shapefile), time-series CSV at defined monitoring points, PDF compliance summary |
| Sub-daily discharge resolution | Not achievable; discharge pulses shorter than the revisit interval will not be captured |
Analytics Satellize can run
| Turbidity anomaly map (single overpass) | NDTI or red/NIR band ratio applied to atmospherically corrected surface reflectance (Sen2Cor or 6S-based correction); threshold classification against local baseline | Georeferenced GeoTIFF and vector polygon of elevated-turbidity extent, delivered per cloud-free scene |
| Plume origin attribution score | Spatial back-tracking of plume centroid to known outfall coordinates; cross-referenced against upstream rainfall accumulation from ERA5 or IMERG reanalysis to separate mine-sourced from storm-driven events | Per-event attribution report with confidence rating (high/medium/low) and supporting rainfall overlay |
| Calibrated TSS concentration map | Power-law or linear regression between red-band reflectance and in-situ TSS grab samples; applied to all cloud-free scenes after calibration campaign | Raster with pixel-level TSS estimate in mg/L, uncertainty band reported per image |
| Downstream exceedance frequency product | Time-series extraction at regulatory monitoring point coordinates; fraction of cloud-free observations exceeding defined turbidity threshold, reported per quarter | Quarterly compliance summary table and chart, formatted for permit reporting |
| Multi-temporal plume extent stack | Cloud-free scene compositing over user-defined period; pixel-wise maximum and median turbidity index to distinguish persistent from episodic contamination | GeoTIFF composite layers (maximum, median, frequency-of-exceedance) with accompanying GIS symbology |
| Near-real-time discharge alert | Automated NDTI threshold trigger on incoming Sentinel-2 or Planet scenes at registered outfall polygons; human QA review before notification dispatch | Email or API alert with scene thumbnail, turbidity index value, and plume area estimate in hectares |
| Historical baseline and trend report | Landsat archive analysis (1972 to present) to establish pre-mining turbidity distribution at downstream monitoring points; Mann-Kendall trend test on annual median values | PDF report with time-series charts, trend significance statistics, and pre/post-mining comparison maps |
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.