Pit lake water quality evolution after mine flooding
Decommissioned open pits that flood can produce stratified, acidic, metal-laden lakes whose surface chemistry shifts over years. Multispectral and hyperspectral sensors retrieve pH proxies, dissolved iron, turbidity and algal indicators without a boat.
Sensors
- Sentinel-2 MSI: 13 spectral bands from 443 nm to 2190 nm at 10–60 m spatial resolution; 5-day revisit at mid-latitudes. Red-edge bands (Band 5, 705 nm; Band 6, 740 nm) and the SWIR pair are particularly useful for chlorophyll proxies and suspended-sediment retrieval over small pit lakes. Many pit lakes are 0.1–2 km across, so 10 m visible bands are often marginal; 20 m SWIR bands can miss fine spatial structure.
- Landsat 8/9 OLI: 30 m multispectral resolution across seven reflective bands; 16-day single-satellite revisit (8-day combined for Landsat 8 and 9). The coastal-aerosol band (Band 1, 443 nm) is useful for distinguishing dissolved organic matter from suspended sediment. At 30 m, pits smaller than roughly 90 m in diameter are effectively unresolvable without mixed-pixel corrections.
- AVIRIS-NG (airborne): Approximately 5 m spatial resolution; 425 contiguous bands from 380 nm to 2510 nm at ~5 nm sampling. Enables full radiative-transfer inversion for dissolved iron speciation, acid mine drainage pH proxies and mineral identification in shoreline precipitates. Airborne campaigns are episodic, not continuous, but provide the calibration anchor for spaceborne time series.
- PlanetScope: 3 m spatial resolution in four to eight bands (depending on generation); near-daily revisit globally. Useful for tracking bloom extent and turbidity plume boundaries in small pits where Sentinel-2 pixels straddle the water edge. Radiometric calibration is less stable than Sentinel-2 or Landsat, which complicates quantitative retrieval; best used for change detection and spatial delineation rather than absolute concentration estimates.
What a pit lake's colour is actually telling you
The surface colour of a pit lake is a function of its optical constituents: dissolved iron (Fe²⁺ and Fe³⁺), suspended particulates, chromophoric dissolved organic matter and phytoplankton pigments. Each absorbs and scatters light differently. Ferric iron (Fe³⁺) produces the rust-orange tint characteristic of highly acidic pit lakes (pH below roughly 3.5), because iron hydroxide precipitates scatter strongly in the 600–700 nm range. At slightly higher pH, iron flocs settle and the water can shift to grey or pale blue. These transitions are detectable in Sentinel-2 Band 4 (665 nm) to Band 3 (560 nm) ratios, though the relationship is empirical and site-specific.
Algal blooms complicate the picture. In pit lakes that have neutralised sufficiently to support biological activity, cyanobacteria or acidophilic algae can produce chlorophyll-a signals in the red-edge region. Sentinel-2's Band 5 (705 nm) relative to Band 4 (665 nm) is a standard chlorophyll-a proxy used in inland water remote sensing. The honest caveat: a single ratio cannot distinguish a dense thin bloom from a sparse deeper one. Depth of the optically active layer matters, and that is not retrievable from passive reflectance alone without additional assumptions.
Radiative-transfer inversion versus empirical band ratios
Two broad approaches exist for retrieving water-quality parameters from reflectance. Empirical band ratios are fast and transferable once calibrated: you regress in-situ measurements of turbidity, chlorophyll-a or iron concentration against ratio values from coincident imagery. Published studies in journals such as Remote Sensing (MDPI) have demonstrated reasonable retrievals of turbidity (R² often 0.7–0.9 for well-calibrated sites) and chlorophyll-a in inland waters using Sentinel-2 and Landsat. The problem is that coefficients derived at one pit lake do not transfer reliably to another with different mineralogy or dissolved organic matter background.
Radiative-transfer inversion, using models such as the quasi-analytical algorithm (QAA) or Hydrolight-based approaches, estimates the inherent optical properties of the water body from the full reflectance spectrum without needing site-specific calibration data. This is where AVIRIS-NG earns its place: 425 contiguous bands allow proper spectral decomposition of absorption and backscattering coefficients. The trade-off is computational cost and the need for accurate atmospheric correction, which over small pit lakes surrounded by highly reflective waste rock is genuinely difficult. Adjacency effects from pit walls can contaminate water-leaving radiance by several percent reflectance.
The stratification problem: sensors see only the surface
Pit lakes frequently develop meromictic stratification, where a dense, saline, metal-rich lower layer (monimolimnion) sits permanently beneath a fresher upper layer (mixolimnion) with little seasonal mixing. The surface may appear relatively benign while the lower water column remains severely contaminated. Passive optical sensors cannot see below the first optical depth, which in turbid or iron-rich pit lakes may be less than one metre. This is not a limitation that better satellites will solve; it is a physical constraint of passive reflectance.
Thermal infrared can indicate surface temperature gradients that hint at upwelling or mixing events. Landsat 8/9 carries a thermal infrared sensor (TIRS) at 100 m resolution, sufficient to detect temperature anomalies of roughly 0.5°C or more across a large pit lake surface. Sentinel-3 SLSTR offers 500 m thermal resolution, which is too coarse for most pit lakes. Neither replaces in-situ profiling for understanding the full water column, but thermal time series can flag mixing events worth investigating on the ground.
Building a multi-year time series: archive depth and change detection
Landsat's archive extends to 1972, with consistent calibrated data from Landsat 5 TM onwards (1984). For pit lakes flooded in the 1990s or 2000s, this means a usable decadal record of surface-colour evolution is already available at no cost via USGS EarthExplorer. Sentinel-2 adds higher spatial and spectral resolution from 2015 onwards. Together they allow analysts to track the progression from initial flooding (often turbid, high suspended sediment) through iron oxidation phases to potential biological colonisation over years to decades.
Change detection in this context is not a simple threshold. The spectral signal of a pit lake evolves slowly and non-monotonically. A dry season can concentrate solutes and shift colour; a wet season inflow can dilute and reset the surface layer. Seasonal normalisation, using climatological baselines built from the full archive, is necessary before year-on-year trends are meaningful. At Satellize, this kind of long-archive time-series construction is standard practice; the Tonga crop-estimation programme required similar multi-year seasonal baseline work, and the same pipeline applies here.
Practical limits buyers should know before commissioning work
Pit lake size is the first filter. For Sentinel-2 at 10 m, a reliable water-quality retrieval generally requires a water body at least 50–100 m across after excluding mixed edge pixels. At Landsat's 30 m, the practical minimum is 150–200 m. Smaller pits require airborne or very-high-resolution commercial imagery, with corresponding cost implications. PlanetScope's 3 m resolution helps with spatial delineation but its radiometric consistency across time is not yet at the level required for quantitative geochemical retrieval without careful cross-calibration against Sentinel-2 or Landsat.
Cloud cover is a persistent constraint in high-rainfall mining regions. In parts of the tropics, Sentinel-2 may yield fewer than six cloud-free observations per year over a given pit. SAR (Synthetic Aperture Radar) can detect the water surface extent through cloud but carries no water-chemistry information. For regulatory compliance monitoring that requires monthly or quarterly water-quality snapshots, cloud gaps may require gap-filling with modelled estimates or planned airborne campaigns. Any honest scope of work should state the expected cloud-free observation frequency for the specific site latitude and season before promising a monitoring cadence.
Typical figures
| Spatial resolution (multispectral) | 10 m (Sentinel-2 visible/NIR), 20 m (Sentinel-2 red-edge/SWIR), 30 m (Landsat 8/9 OLI) |
| Spatial resolution (hyperspectral, airborne) | ~5 m (AVIRIS-NG); campaign-based, not continuous |
| Revisit frequency | 5 days (Sentinel-2, mid-latitudes); 8 days combined (Landsat 8+9); near-daily (PlanetScope, radiometric caveats apply) |
| Spectral coverage | 443–2190 nm (Sentinel-2 MSI); 435–2294 nm (Landsat 8/9 OLI/TIRS); 380–2510 nm at ~5 nm sampling (AVIRIS-NG) |
| Minimum resolvable pit lake | ~50–100 m diameter for Sentinel-2; ~150–200 m for Landsat; smaller pits require airborne or commercial VHR |
| Turbidity retrieval accuracy (published range) | R² 0.7–0.9 for well-calibrated empirical models on inland waters; degrades significantly without site-specific in-situ calibration |
| Thermal sensitivity (surface temperature) | ~0.5°C (Landsat 8/9 TIRS at 100 m); useful for detecting mixing events |
| Archive depth | Landsat: 1984 (TM) to present; Sentinel-2: 2015 to present; AVIRIS-NG: campaign-specific |
| Atmospheric correction requirement | Essential; adjacency effects from pit walls can introduce errors of several percent reflectance; site-specific validation recommended |
| Delivery formats | GeoTIFF parameter maps, time-series CSV, GIS-ready shapefiles for bloom/turbidity extents, PDF monitoring reports |
Analytics Satellize can run
| Surface iron concentration proxy map | Empirical band-ratio regression (Sentinel-2 B4/B3 and B4/B2) calibrated against published inland water iron retrieval studies; site-specific calibration improves accuracy | Quarterly GeoTIFF maps with concentration class zones and time-series trend chart |
| Turbidity time series | Nechad or Dogliotti single-band turbidity algorithms applied to atmospherically corrected Sentinel-2 and Landsat reflectance; well-documented in inland water remote sensing literature | Monthly turbidity index raster and CSV trend export per pit lake polygon |
| Chlorophyll-a and algal bloom extent | Sentinel-2 red-edge chlorophyll index (B5/B4 ratio) and maximum chlorophyll index; bloom boundary delineated by threshold on normalised difference chlorophyll index | Bloom area shapefile with alert flag when bloom exceeds defined threshold area |
| Surface colour change anomaly detection | Seasonal baseline normalisation across full Landsat/Sentinel-2 archive; z-score anomaly flagging on RGB composite and individual band reflectance | Automated alert report when surface colour departs significantly from seasonal climatology |
| Multi-decadal flooding and chemistry evolution narrative | Landsat archive time series (1984 to present) with spectral trajectory analysis; change-point detection to identify phase transitions (turbid, iron-rich, biologically active) | PDF baseline report with annotated spectral trajectory plots and key transition dates |
| Surface temperature anomaly map (mixing event indicator) | Landsat 8/9 TIRS thermal band calibrated to surface temperature; anomaly detection relative to seasonal mean | Thermal anomaly GeoTIFF flagged when temperature gradient exceeds defined threshold, indicating potential upwelling or mixing |
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.