Conflict mineral extraction site monitoring for supply-chain due diligence
Sentinel-2 and PlanetScope change detection can document pit expansion, road construction and processing-site activity at named conflict-mineral sites, giving supply-chain auditors satellite evidence to set alongside IPIS ground data.
Sensors
- Sentinel-2 MSI: 10 m resolution in visible and near-infrared bands, 5-day revisit at the equator with both satellites. Detects bare-soil expansion, vegetation clearance and the spectral signature of iron-oxide-rich tailings. Free archive back to 2015.
- PlanetScope SuperDove: 3 m resolution, near-daily revisit globally. Eight multispectral bands including red-edge. Best available open-commercial option for resolving individual pit clusters and tracking week-on-week changes in active workings. Commercial licence required.
- Sentinel-1 SAR: C-band SAR at 10 m (IW mode), 6-day revisit, cloud-independent. Coherence change detection identifies surface disturbance even under persistent cloud cover common in DRC's eastern provinces. Free archive back to 2014.
- Landsat 8/9 OLI: 30 m resolution, 16-day revisit per satellite (8-day combined). Useful for long-baseline trend analysis and cross-calibration with Sentinel-2. Free USGS archive back to 1972 for earlier sensors, OLI from 2013.
What the regulations actually ask for
The OECD Due Diligence Guidance for Responsible Supply Chains of Minerals from Conflict-Affected and High-Risk Areas (3TG edition) requires companies to identify and assess risks in their supply chain, including the geographic origin of minerals and conditions at extraction sites. The EU Conflict Minerals Regulation, which became fully applicable in January 2021, extends mandatory due diligence to all EU importers of tin, tantalum, tungsten and gold above defined volume thresholds.
Neither instrument specifies satellite imagery as a required input. What they require is credible, documented evidence of site conditions. Satellite change detection fills a specific gap: it can confirm or contradict claims about whether a named site was active during a given period, whether its footprint expanded, and whether access infrastructure consistent with organised extraction was present. That is a narrower claim than full traceability, but it is a verifiable one.
What a pit scar gives away
Artisanal and small-scale mining of coltan, cassiterite and wolframite in eastern DRC typically produces distinctive spectral and structural signatures. Bare laterite soil, often iron-oxide red, contrasts sharply with surrounding forest in Sentinel-2 band combinations using SWIR (band 11 or 12) alongside NIR and red. Tailings piles and processing water carry different reflectance from undisturbed soil. These signatures are detectable at 10 m resolution when the disturbed area exceeds roughly 0.01 hectares, though individual pit dimensions below about 5 m cannot be reliably measured from open-source sensors.
PlanetScope's 3 m resolution pushes the minimum detectable disturbed patch down considerably, allowing individual pit clusters to be distinguished from spoil heaps and shelter structures. SAR coherence from Sentinel-1 adds a cloud-independent layer: active digging or vehicle movement decorrelates the signal relative to a reference scene, flagging surface change even when optical imagery is obscured. Eastern DRC's cloud climatology makes this combination, rather than optical alone, the practical standard for continuous monitoring.
Cross-referencing against IPIS mine databases
The International Peace Information Service publishes georeferenced databases of artisanal mining sites in eastern DRC, updated periodically through field missions. Each site record includes mineral type, estimated number of workers, armed-group presence and GPS coordinates. These coordinates are the anchor for satellite analysis: change detection is run on named, located sites rather than as a blind landscape scan.
The workflow is straightforward in principle. A baseline image pair from the IPIS visit date is compared against subsequent imagery at regular intervals. Expansion of the bare-soil footprint, new track segments connecting to the main site polygon, or new roofed structures at processing areas all constitute observable change. The satellite record does not confirm who controls a site or whether minerals from it entered a specific supply chain. It confirms whether the site was active and whether its physical extent changed. That distinction matters for honest reporting to auditors.
Access roads are often the first signal
New extraction sites rarely appear fully formed. The leading indicator is usually a track cut through forest to an area of interest, followed by initial clearance, then progressive pit expansion over weeks to months. Sentinel-2 at 10 m can detect track widths of roughly 10 m or more when the canopy is sufficiently cleared; PlanetScope resolves tracks narrower than that.
Sentinel-1 SAR is particularly useful here because forest road construction produces a linear backscatter anomaly even before bare soil is visible in optical bands. A combination of SAR-based track detection and optical bare-soil mapping gives earlier warning than either sensor alone. For sites already in the IPIS database, this means monitoring can catch reactivation of dormant workings within a single revisit cycle rather than after a season of activity.
Honest limits of the method
Cloud cover is the dominant operational constraint. Eastern DRC experiences persistent cloud for months at a time, and a single Sentinel-2 clear observation may be all that is available in a given quarter. SAR mitigates this but does not eliminate ambiguity: coherence change from rainfall or vegetation growth can resemble surface disturbance, and careful baseline selection is necessary to reduce false positives.
Resolution sets a hard floor on what can be claimed. A site assessed as 'active' based on Sentinel-2 data could in practice range from a few dozen workers to several hundred; the imagery cannot distinguish those cases. PlanetScope narrows the uncertainty but does not resolve it entirely. No open-source sensor currently operational can confirm the mineral species being extracted, the identity of armed actors present, or the downstream chain of custody. Satellite evidence is one input to due diligence, not a substitute for ground verification or chain-of-custody documentation. Presenting it as more than that would undermine its credibility with auditors.
Turning site monitoring into a compliance deliverable
For a supply-chain compliance team, the useful output is not a raw image but a structured site-status report: a dated polygon showing the current disturbed footprint, a change log against a defined baseline, and a confidence rating tied to the number and quality of cloud-free observations in the period. That report can be attached to a due-diligence file alongside IPIS field data and smelter audit records.
Satellize structures this kind of output as a periodic GIS layer and accompanying PDF summary, cross-referenced against IPIS site identifiers, with honest confidence intervals where cloud cover or resolution limits the interpretation. The approach is similar in structure to the crop-area estimation work Satellize runs for the Kingdom of Tonga, where the discipline is the same: quantify what the imagery actually supports, state the uncertainty, and stop there. If your compliance programme needs site-specific baseline imagery from a defined date, the next step is to specify the IPIS site codes and the reporting period.
Typical figures
| Best optical resolution (commercial) | 3 m (PlanetScope SuperDove) |
| Best optical resolution (open) | 10 m (Sentinel-2 MSI) |
| SAR resolution | 10 m IW mode (Sentinel-1) |
| Revisit frequency | Near-daily at 3 m (PlanetScope); 5-day at 10 m (Sentinel-2 dual satellite); 6-day SAR (Sentinel-1) |
| Minimum detectable disturbed area | Approx. 0.01 ha at 3 m; approx. 0.1 ha reliably at 10 m |
| Individual pit dimension floor | Below approx. 5 m pit width, open-source sensors cannot reliably measure |
| Cloud-independent coverage | Sentinel-1 SAR operates through cloud; coherence change detects surface disturbance |
| Archive depth | Sentinel-2 from 2015; Sentinel-1 from 2014; Landsat from 1972 (30 m) |
| Spectral bands relevant to tailings | Sentinel-2 SWIR (B11 1610 nm, B12 2190 nm), NIR (B8 842 nm), Red (B4 665 nm) |
| Delivery formats | GeoTIFF change layers, GeoJSON site polygons, PDF site-status reports with confidence ratings |
Analytics Satellize can run
| Site footprint baseline and change polygon | Supervised bare-soil classification and binary change detection on multitemporal Sentinel-2 or PlanetScope composites | Dated GeoJSON polygon per IPIS site ID, with area in hectares and change delta from baseline |
| Access-road extension detection | Linear feature extraction on SAR backscatter difference images combined with optical NDVI loss | GIS line layer of new track segments with detection date and length |
| SAR coherence activity flag | Sentinel-1 InSAR coherence differencing between reference and monitoring epoch | Binary activity raster per site, cloud-independent, flagged for compliance log |
| Processing-area structure count | Object-based image analysis on PlanetScope imagery to identify roofed structures at processing zones | Structure count and footprint area per site, with date and image source |
| Periodic site-status compliance report | Synthesis of optical change, SAR activity flag and IPIS site metadata into structured summary | PDF report per site or site cluster, suitable for attachment to OECD due-diligence file, with explicit confidence rating |
| Long-baseline trend analysis | Landsat 8/9 OLI time series from 2013 to present, cross-calibrated with Sentinel-2 | Annual footprint area chart per site, showing expansion or contraction over the full available record |
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.