Illegal artisanal and small-scale mining pit expansion
Unpermitted artisanal and small-scale mining leaves spectral, structural and hydrological signatures detectable from orbit. Open-source and commercial sensors can map pit expansion, turbidity plumes and mercury stress in fringing vegetation at operationally useful cadence.
Sensors
- Sentinel-2 MSI: 10 m resolution in visible and near-infrared bands, 5-day revisit at the equator with both satellites. Red-edge bands (B5, B6, B7 at 20 m) detect chlorophyll stress in vegetation fringing active sites. SWIR bands distinguish wet tailings from dry sediment. Cloud cover over tropical sites is the principal operational constraint.
- Landsat 8/9 OLI: 30 m multispectral resolution, 8-day combined revisit. Longer archive (Landsat 5 back to 1984) makes it the standard instrument for multi-decade pit-expansion chronologies. SWIR2 band (2.2 µm) separates sediment mineralogy. Coarser than Sentinel-2 but free and globally consistent.
- Sentinel-1 SAR C-band: Cloud-penetrating synthetic aperture radar at 10 m IW mode resolution. Detects surface roughness change when forest is cleared and soil is disturbed, regardless of persistent cloud. Backscatter decrease in VV polarisation marks newly exposed bare ground; coherence loss between repeat passes confirms fresh disturbance. 6-day revisit in most tropical regions.
- Planet SuperDove: 3 m resolution, daily revisit in most regions. Eight spectral bands including two red-edge channels. Useful for confirming individual pit boundaries identified in coarser imagery and for monitoring rapid day-to-day expansion. Commercial licence required; not open access.
- VIIRS DNB (Suomi NPP / NOAA-20): Nighttime light detection at roughly 375 m pixel scale. Generator-powered mining camps in remote forest produce persistent artificial light anomalies detectable in monthly composites. Useful as a low-cost screening layer to prioritise areas for higher-resolution follow-up.
What bare sediment and milky water give away
Alluvial gold mining strips topsoil and vegetation to reach placer deposits, exposing highly reflective pale sediment and clay. In Sentinel-2 true-colour composites, active pits appear as bright white or yellow patches against the dark background of intact forest canopy. The contrast is stark enough that experienced analysts can identify new clearings at 10 m resolution within days of disturbance. The spectral signature is not unique to mining: landslides and legal construction also produce bare soil. Discrimination requires contextual analysis, including remoteness from roads, proximity to river systems, and the characteristic geometry of artisanal pits, which tend to cluster in elongated chains along buried palaeochannels.
Tailings ponds are a more specific indicator. The combination of fine suspended sediment, residual mercury and processing chemicals produces a distinctive milky turquoise or grey-white reflectance in visible wavelengths, with elevated turbidity in the red and red-edge bands. Published work using Sentinel-2 imagery over the Madre de Dios region of Peru has shown that turbidity indices derived from Band 4 (red) and Band 5 (red-edge, 705 nm) can track sediment plumes tens of kilometres downstream from active sites, providing a hydrological footprint far larger than the pit itself.
SAR fills the gap that clouds create
The Amazon basin and the West African forest belt, two of the regions with the highest concentrations of illegal ASM activity, are cloud-covered for large fractions of the year. Optical sensors go blind for weeks at a time. Sentinel-1 C-band SAR does not. Forest canopy produces a characteristically high and stable backscatter return; when it is cleared and the surface becomes smooth bare soil or water, backscatter in VV polarisation drops sharply. Interferometric coherence between two passes six days apart collapses over disturbed ground, providing a second independent signal.
The practical limit is spatial resolution. At 10 m in Interferometric Wide Swath mode, Sentinel-1 resolves clearings of roughly 0.1 hectares and above reliably. Smaller individual pits may fall below detection, but the aggregate footprint of an active mining camp rarely does. SAR change detection is therefore most useful as a screening and alerting layer, with optical confirmation tasked when cloud permits.
Mercury stress in the canopy margin
Mercury is the dominant gold-separation reagent in artisanal mining globally. It volatilises during amalgam burning and settles on surrounding vegetation, causing measurable physiological stress. Chlorophyll content drops, and the red-edge reflectance slope, the steep rise in plant reflectance between roughly 700 and 740 nm, flattens or shifts. Sentinel-2 red-edge bands and the equivalent channels on Planet SuperDove can detect this stress signature in the canopy fringe within tens to hundreds of metres of active burning sites.
This is a secondary and probabilistic indicator, not a definitive one. Drought, disease and other pollution sources produce similar spectral responses. Mercury stress mapping is most credible when it spatially coincides with the geometric and hydrological signatures of confirmed pit activity. Used in combination, the converging evidence is substantially more persuasive to enforcement agencies than any single layer alone. Published studies from IPAM and UNODC have used exactly this multi-indicator approach to estimate active site counts across the Brazilian Amazon.
Building the expansion chronology
Enforcement and legal proceedings benefit from documented history, not just a current snapshot. The Landsat archive, continuous from 1984 and globally consistent, allows analysts to reconstruct pit expansion year by year and, in some cases, season by season. A site that has grown from two hectares to forty over a decade tells a different legal and environmental story than one that appeared last month.
Cloud contamination remains the main obstacle to dense time series in tropical regions. Gap-filling methods, combining Landsat and Sentinel-2 at compatible spectral bands, and using SAR-derived disturbance dates to anchor optical observations, can substantially improve temporal resolution. The honest caveat is that in persistently cloudy areas, annual rather than monthly precision is often the realistic expectation for optical-based chronologies. SAR provides higher temporal density but coarser spatial discrimination.
Detection floors and honest limits
The minimum reliably detectable clearing depends on sensor, cloud conditions and landscape context. In clear-sky conditions, Sentinel-2 at 10 m can flag individual pits of around 0.05 to 0.1 hectares. Under persistent cloud, SAR detection floors rise to roughly 0.1 to 0.5 hectares for isolated disturbances. Planet SuperDove at 3 m can resolve individual excavations of a few hundred square metres, but commercial tasking costs and data-volume constraints mean it is generally reserved for site-specific confirmation rather than basin-wide surveillance.
Forest canopy overhang conceals pits that have been partially revegetated or deliberately camouflaged with cut branches, a documented evasion tactic in the Amazon. Shallow alluvial workings in open savannah or semi-arid regions are generally easier to detect than deep jungle sites. Nighttime VIIRS screening helps identify camps that would otherwise be invisible under cloud and canopy, but its 375 m pixel scale means it can only confirm camp presence, not pit geometry. No single sensor or method is sufficient; the operational standard is multi-sensor fusion.
From pixels to enforcement evidence
Satellite evidence is increasingly accepted in environmental prosecutions and in supply-chain due diligence processes under frameworks such as the OECD Due Diligence Guidance for Responsible Mineral Supply Chains. The evidentiary standard requires documented methodology, sensor metadata, geometric accuracy assessments and clear chain of custody for the imagery. Analysts running this work for government clients typically deliver georeferenced shapefiles with confidence ratings, change-detection reports with before-and-after imagery strips, and pit-count estimates with stated uncertainty bounds.
Satellize runs multi-sensor fusion analytics of this type on open constellations, with commercial tasking added on client licence. The workflow is broadly consistent with methods published by IPAM for Brazilian Amazon ASM monitoring. For clients assessing supply-chain exposure rather than enforcement, the output is a site-risk register keyed to concession boundaries and artisanal mining zones, updated on a defined cadence. The next concrete step for a government or compliance team is a scoped pilot over a defined area of interest, which establishes realistic detection rates before any long-term commitment.
Typical figures
| Typical spatial resolution | 10 m (Sentinel-2 visible/NIR, Sentinel-1 IW SAR); 30 m (Landsat 8/9); 3 m (Planet SuperDove); 375 m (VIIRS DNB nighttime screening) |
| Revisit cadence | 5 days (Sentinel-2, equatorial); 6 days (Sentinel-1); 8 days combined (Landsat 8+9); daily (Planet SuperDove); daily composites (VIIRS) |
| Minimum detectable clearing (optical, clear sky) | ~0.05–0.1 ha at 10 m; ~0.1 ha at 30 m; ~0.005 ha at 3 m |
| Minimum detectable disturbance (SAR) | ~0.1–0.5 ha in IW mode; cloud-independent |
| Key spectral bands | Red-edge 705 nm and 740 nm (chlorophyll stress); SWIR 1610 nm and 2190 nm (sediment/tailings); VV/VH C-band SAR (surface roughness change) |
| Cloud impact | Severe for optical in Amazon and West Africa; SAR unaffected; VIIRS DNB affected by cloud and moonlight |
| Archive depth | 1984 to present (Landsat); 2015 to present (Sentinel-2); 2014 to present (Sentinel-1) |
| Typical change-detection latency | 2–5 days after clear-sky overpass for optical alert; near-real-time for SAR with automated processing |
| Delivery formats | GeoTIFF change maps, GeoJSON/Shapefile pit polygons, PDF site reports with imagery strips, GIS-ready risk registers |
Analytics Satellize can run
| Active pit mapping | Supervised classification of bare-soil and tailings-pond spectral signatures in Sentinel-2 and Landsat imagery; NDVI and turbidity indices | Georeferenced polygon layer of active pit extents with confidence scores; updated per clear-sky overpass |
| SAR disturbance alert | Backscatter change detection and coherence loss in Sentinel-1 IW time series; threshold-based flagging of new clearings | Near-real-time alert feed of candidate disturbance polygons for optical follow-up; cloud-independent |
| Pit expansion chronology | Multi-decadal Landsat time series analysis; gap-filling with Sentinel-2; annual and seasonal area estimates | Expansion timeline report with area-per-year table and before/after imagery strips; suitable for legal proceedings |
| Downstream turbidity plume mapping | Red and red-edge band turbidity index applied to Sentinel-2 river reaches; plume extent and intensity quantification | GIS layer of affected river reaches with turbidity intensity classes; linked to upstream pit polygons |
| Mercury vegetation stress index | Red-edge slope and chlorophyll index anomaly detection in canopy fringe zones around confirmed pit sites | Secondary indicator layer flagging stress anomalies within defined buffer zones; included as supporting evidence in site reports |
| Nighttime camp screening | VIIRS DNB monthly composite anomaly detection against stable baseline; persistent new light sources in forested areas | Screening layer of candidate camp locations for priority tasking of higher-resolution sensors |
| Supply-chain site-risk register | Overlay of detected active ASM sites against concession boundaries, protected-area polygons and declared mine permits | Tabular risk register with site coordinates, area, expansion rate and permit-status flag; updated on agreed cadence |
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.