Conflict-minerals supply chain monitoring via site activity tracking
Satellite SAR and optical imagery can track pit expansion, track formation and equipment presence at artisanal 3TG mining sites, giving compliance teams verifiable, dated evidence to meet EU and US due-diligence obligations.
Sensors
- Sentinel-1 SAR (C-band, ESA): 10 m ground range resolution in Interferometric Wide Swath mode, 6-day repeat at the equator with both satellites active. Cloud-penetrating and illumination-independent, making it the primary continuity sensor for equatorial forest regions. Backscatter change between passes reveals soil disturbance, new track formation and the appearance of metal roofing on camp structures.
- Sentinel-2 MSI (ESA): 10 m in visible and near-infrared bands, 5-day revisit at the equator. Bare-soil exposure from pit expansion produces strong spectral contrast against forest in the red-edge and SWIR bands. Usable only on cloud-free acquisitions; in high-humidity equatorial zones this can mean weeks between clean observations.
- Planet SuperDove: 3–4 m resolution, daily revisit globally in the 8-band PlanetScope constellation. Higher spatial detail resolves individual equipment and spoil heaps that Sentinel-2 merges into single pixels. Cloud cover remains a hard constraint; the daily cadence raises the probability of catching a clear window, but does not eliminate it.
- Capella Space SAR (X-band): Spotlight modes deliver 0.35–1 m resolution, sufficient to distinguish individual excavators, sluice boxes and vehicle tracks. Tasked commercially on demand; latency from order to delivery is typically under 24 hours. X-band is more sensitive to surface roughness and small metallic objects than C-band, but the narrow swath (5–10 km in spotlight) means targeted tasking rather than area surveillance.
What the regulations actually require you to prove
The EU Conflict Minerals Regulation (2017/821, in force from January 2021) and US Dodd-Frank Section 1502 both require smelters and downstream companies to conduct supply-chain due diligence against OECD guidance. In practice, that means documenting the origin of tin, tantalum, tungsten and gold and demonstrating that sourcing decisions are informed by credible risk assessments. Neither regulation mandates satellite monitoring, but both create a demand for dated, auditable evidence of site-level conditions that ground visits in active conflict zones cannot reliably supply.
Satellite activity tracking fills a specific gap: it can establish whether a named site was actively worked during a given period, whether its footprint expanded, and whether access infrastructure changed. That is not a complete chain-of-custody record, but it is a defensible, third-party-verifiable input to a risk assessment. The honest caveat is that imagery tells you about surface disturbance, not about the mineral being extracted or the identity of the operator.
Activity signatures that show up in imagery
Artisanal and small-scale mining (ASM) sites have a recognisable physical grammar. Pit expansion appears as progressive removal of vegetation and exposure of bare soil or laterite. Access tracks are narrow, often sinuous, and extend from existing roads toward the workings; they appear in SAR as linear low-backscatter features cutting through high-backscatter forest. Equipment, particularly metal-roofed shelters, generators and sluice structures, produces strong SAR returns. Water-filled pits have very low backscatter and are easily distinguished from surrounding terrain.
In optical imagery, the exposed soil of active workings has a distinctive spectral signature. The normalised difference vegetation index (NDVI) drops sharply at cleared areas, and bare laterite soils in central Africa and the eastern DRC show high reflectance in the red and SWIR bands. Tailings and spoil heaps cast shadows that help distinguish them from natural clearings. The combination of SAR-derived change and optical-derived spectral classification is more reliable than either alone, particularly for distinguishing active sites from abandoned ones.
Cloud cover is not an excuse, but it is a real constraint
The eastern DRC, Rwanda and parts of the Central African Republic, where many conflict-affected 3TG sites are concentrated, sit in equatorial and sub-equatorial climate zones. Cloud cover exceeds 70% of daylight hours through much of the year. A Sentinel-2 or Planet optical archive for a given site may have usable cloud-free imagery only every three to eight weeks, even with daily tasking. This is not a solvable problem; it is a physical one.
Sentinel-1 SAR operates at 5.4 GHz (C-band) and penetrates cloud and rain without signal degradation. It is the backbone of any monitoring programme in these regions. The limitation of C-band SAR is that dense, wet forest canopy attenuates the signal before it reaches the ground, so sub-canopy mining activity, common in the early stages of a new artisanal site, may not be detectable until clearing is sufficient to break the canopy. X-band SAR from Capella is more sensitive to surface detail but shares the canopy-penetration limit. No current commercial system reliably detects sub-canopy ground disturbance at ASM scale.
Building a change-detection baseline that holds up to scrutiny
A credible monitoring product starts with a baseline: a dated, georeferenced characterisation of each site's extent and activity state at a known point in time. From that baseline, change is measured at each subsequent acquisition. The method class is bitemporal SAR coherence change detection combined with optical NDVI differencing on cloud-free pairs. Coherence between two Sentinel-1 passes drops sharply where the surface has been disturbed between acquisitions; a coherence map can flag new activity even when the absolute backscatter values are ambiguous.
The output is not a binary active/inactive label. It is a graded activity indicator, updated on each usable acquisition, with confidence intervals that reflect data quality and cloud contamination. Sites are assigned to a watch tier based on their regulatory risk profile. High-risk sites, those in areas flagged by the OECD or UN Group of Experts reports as conflict-affected, receive priority commercial SAR tasking to fill gaps in the Sentinel-1 archive.
Satellize structures this kind of tiered monitoring programme as a standing analytics service, drawing on open Sentinel data and adding Capella tasking for priority sites. The Tonga crop-estimation programme demonstrated the same underlying architecture of open-data baseline plus commercial gap-fill, applied to a very different problem.
What the imagery cannot tell you, and what that means for compliance
Satellite monitoring establishes activity, extent and change. It does not identify the mineral being extracted, the operator, the trading chain, or whether violence or coercion is present. A site that appears active in imagery may be legally licensed artisanal mining, unlicensed but non-conflict mining, or mining under armed group control. The imagery is silent on this distinction.
For compliance purposes, imagery evidence is most useful when combined with other data: OECD and UN Group of Experts reports, government mine-registration databases, NGO field assessments and smelter audit records. The satellite layer provides the temporal and spatial anchor. It can confirm that a site was active during the period when a smelter claims to have sourced from it, or flag that a site previously assessed as inactive has reopened. That is a meaningful contribution to a due-diligence programme, but it is an input, not a conclusion.
Typical figures
| SAR spatial resolution (Sentinel-1 IW mode) | 10 m range × 10 m azimuth (multi-looked) |
| SAR spatial resolution (Capella spotlight) | 0.35–1 m depending on mode |
| Optical spatial resolution (Sentinel-2) | 10 m (VIS/NIR), 20 m (SWIR/red-edge) |
| Optical spatial resolution (Planet SuperDove) | 3–4 m |
| Sentinel-1 revisit (equatorial, both satellites) | 6 days |
| Planet SuperDove revisit | Daily (cloud-limited in equatorial zones) |
| Minimum detectable cleared area (SAR coherence) | Approximately 0.1 ha at Sentinel-1 resolution; smaller features detectable with Capella |
| Archive depth (Sentinel-1 and Sentinel-2) | From 2014 (Sentinel-1A launch) and 2015 (Sentinel-2A launch) |
| Latency (standing monitoring product) | 24–72 hours after satellite acquisition, depending on processing queue |
| Delivery formats | GeoTIFF change layers, GeoJSON site polygons, PDF activity reports, API feed |
Analytics Satellize can run
| Site activity status update | Bitemporal SAR backscatter change detection and coherence differencing (Sentinel-1) | Per-site activity indicator (active / reduced / inactive / new) updated on each usable acquisition; GeoJSON feed |
| Pit footprint expansion measurement | Optical NDVI differencing and bare-soil spectral classification on cloud-free Sentinel-2 and Planet pairs | Polygon shapefile of current cleared extent with area in hectares and date-stamped change since baseline |
| Access track formation alert | SAR coherence change detection for linear feature emergence, confirmed on optical where cloud permits | Alert report with track geometry, estimated length and first-detection date |
| Equipment and structure presence flag | X-band SAR bright-target detection (Capella spotlight) combined with optical visual confirmation | Point-feature GIS layer with object class, confidence score and acquisition date |
| Historical activity reconstruction | Time-series analysis of Sentinel-1 archive from 2014 and Sentinel-2 archive from 2015 | Site-level activity timeline chart and accompanying GeoTIFF stack for the period of interest |
| Quarterly compliance summary report | Aggregation of all analytic layers against a client-supplied site list, cross-referenced with OECD risk-area designations | PDF report suitable for submission to auditors or regulatory filings, with methodology annex |
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.