Mine tailings dam surface deformation monitoring
InSAR techniques applied to tailings storage facilities can detect millimetre-scale precursor deformation weeks before visible distress, giving operators and regulators a measurable early-warning signal that visual inspection alone cannot provide.
Sensors
- Sentinel-1 (ESA): C-band SAR at 5.6 cm wavelength; Interferometric Wide Swath mode gives 5 × 20 m resolution and 6-day repeat at the equator (12-day for a single satellite). Free and open archive from 2014. Adequate for slow deformation trends but susceptible to temporal decorrelation over wet tailings surfaces.
- COSMO-SkyMed (ASI): X-band SAR at 3.1 cm wavelength; Spotlight mode achieves better than 1 m resolution with repeat passes configurable to 1–4 days. Higher frequency improves sensitivity to small displacements but also increases phase noise over low-reflectivity wet tailings.
- TerraSAR-X / TanDEM-X (DLR): X-band SAR; Staring Spotlight mode reaches sub-metre resolution. The TanDEM-X pair enables rapid DEM generation for volumetric change detection of embankment geometry. Tasking is commercial and latency is typically 24–48 hours from acquisition to delivery.
- ICEYE (ICEYE Oy): X-band SAR microsatellite constellation; sub-1 m Spotlight resolution with daily or sub-daily revisit to any latitude achievable through constellation scheduling. Particularly useful for urgent monitoring after a trigger event or during heavy rainfall periods.
What the Brumadinho and Cadia failures told the industry
The January 2019 collapse of Córrego do Feijão Dam I at Brumadinho, Brazil, killed 270 people and released approximately 12 million cubic metres of tailings. Post-failure analysis, including work published by the Brazilian National Mining Agency and independent geotechnical reviews, identified that the upstream construction method produced a liquefaction-susceptible structure with little warning displacement visible at the surface before sudden collapse. Cadia, in New South Wales, suffered two upstream-constructed tailings dam failures in 2018. Neither event produced the kind of slow, progressive surface deformation that would have been obvious to a periodic visual inspection programme.
These events accelerated regulatory change. The Global Industry Standard on Tailings Management, published in 2020, explicitly calls for independent monitoring and real-time instrumentation. Several jurisdictions now require or are moving toward continuous deformation surveillance. That regulatory pressure is the direct commercial driver for satellite InSAR programmes on tailings facilities. The physics of InSAR had been available for decades; the political will to mandate it took two disasters.
Upstream versus downstream: why construction method changes the failure geometry you are looking for
Upstream-constructed tailings dams raise successive embankment lifts over the deposited tailings beach. The starter dam is progressively buried, and the phreatic surface (the internal water table) can rise close to the embankment face. This geometry is prone to liquefaction failure: rapid, with little precursor displacement at the crest. InSAR on upstream facilities must therefore be sensitive to subtle crest settlement and toe heave, and must be interpreted against piezometric data. A 5–10 mm displacement anomaly at the toe of an upstream embankment is a meaningful signal.
Downstream and centreline-constructed facilities build outward over competent ground. They are generally more stable and, crucially, tend to exhibit more progressive deformation before failure. The failure geometry is more amenable to InSAR detection because deformation develops over weeks to months rather than hours. Monitoring programmes on downstream facilities can tolerate longer revisit intervals and still catch precursor signals. The distinction matters when specifying a satellite programme: an upstream facility in a high-rainfall region warrants daily tasking from ICEYE or COSMO-SkyMed; a well-drained downstream facility may be adequately served by Sentinel-1's free 6-day archive.
Why wet tailings surfaces defeat standard persistent-scatterer processing
Persistent-scatterer InSAR (PS-InSAR) identifies pixels that maintain stable radar reflectivity across dozens of SAR acquisitions. Urban areas are ideal: metal structures, concrete edges and hard corners return strong, coherent signals. Tailings impoundment surfaces are the opposite. Wet, fine-grained slurry absorbs microwave energy and changes its dielectric properties with moisture content and surface roughness. PS density over active tailings ponds can drop to near zero, leaving the analyst with no measurement points precisely where the material is most hazardous.
Small-baseline subset (SBAS) processing helps by using distributed scatterers rather than point targets, averaging phase across statistically similar pixel clusters. This recovers some signal over the embankment crest and beach areas, but coherence still degrades rapidly in wet conditions. The practical solution for critical monitoring is corner-reflector augmentation: trihedral metal reflectors bolted to the embankment at known positions. A standard 60 cm trihedral reflector produces a radar cross-section of roughly 20–30 dBm² at X-band, creating a guaranteed PS regardless of surface conditions. Corner reflectors are inexpensive relative to the monitoring contract value and are now common on high-consequence facilities. Their positions must be surveyed to centimetre accuracy and logged in the processing chain.
What the numbers actually mean in practice
Sentinel-1 C-band InSAR can detect line-of-sight displacement of roughly 2–5 mm per epoch under good coherence conditions, with cumulative precision improving as the time series lengthens. TerraSAR-X Staring Spotlight has demonstrated sub-millimetre precision per epoch in controlled studies. These figures assume adequate PS or distributed-scatterer density. Over a tailings embankment with augmented corner reflectors, X-band systems routinely achieve 1–3 mm epoch precision.
Revisit rate determines how early a precursor is caught. A facility that accelerates from 2 mm/month background settlement to 15 mm/month over six weeks will be detectable within two or three Sentinel-1 passes if the analyst is running change-detection alerts rather than waiting for quarterly reports. Latency from SAR acquisition to processed displacement map is typically 12–36 hours for commercial tasking pipelines, depending on downlink scheduling. That is fast enough to trigger a ground inspection before a slow-moving failure becomes irreversible. It is not fast enough to respond to the kind of sudden liquefaction collapse seen at Brumadinho.
Integrating satellite data into a monitoring programme
Satellite InSAR is a surveillance layer, not a replacement for in-situ piezometers, inclinometers and survey prisms. Its value is spatial coverage. A single tailings facility may have an embankment perimeter of several kilometres; instrumenting every metre with physical sensors is impractical. InSAR provides a spatially continuous displacement map that flags anomalous zones for targeted ground investigation.
A well-designed programme combines Sentinel-1 time series (free, continuous, good for trend detection) with on-demand commercial tasking from COSMO-SkyMed, TerraSAR-X or ICEYE triggered by rainfall events, seismic activity or anomalies in the Sentinel-1 data. Outputs are delivered as GIS-compatible displacement maps, time-series plots at corner-reflector positions, and threshold-based alerts when velocity exceeds a defined limit. Satellize structures this kind of tiered monitoring architecture and has applied similar multi-source InSAR logic in its analytics work, including the Tonga crop-estimation programme, which required combining open and commercial imagery under constrained revisit conditions. The principle of blending free and tasked data to cover gaps is the same.
Regulators increasingly want auditable records. Processed displacement maps with full metadata, uncertainty estimates and version control satisfy that requirement in a way that a verbal inspection report does not. Archive depth is a genuine advantage: Sentinel-1 data runs back to 2014, allowing retrospective analysis of whether a facility was already deforming before a current monitoring contract began.
Typical figures
| Spatial resolution (Sentinel-1 IW) | 5 × 20 m (range × azimuth); resampled to 10–20 m for interferometric processing |
| Spatial resolution (TerraSAR-X Staring Spotlight) | ~0.25 m azimuth; typically processed at 0.5–1 m |
| Spatial resolution (ICEYE Spotlight) | Better than 1 m |
| Revisit interval | 6 days (Sentinel-1 two-satellite); 1–4 days (COSMO-SkyMed, ICEYE with tasking) |
| Minimum detectable displacement (per epoch, good coherence) | 2–5 mm (C-band); 1–3 mm (X-band with corner reflectors) |
| Radar frequency | C-band 5.405 GHz (Sentinel-1); X-band ~9.6 GHz (COSMO-SkyMed, TerraSAR-X, ICEYE) |
| Displacement measurement direction | Line-of-sight (LOS); ascending and descending geometries combined for vertical/horizontal decomposition |
| Processing latency (commercial tasking) | 12–36 hours from acquisition to displacement map, depending on downlink and processing pipeline |
| Archive depth (Sentinel-1) | 2014 to present; enables retrospective baseline analysis |
| Delivery formats | GeoTIFF displacement maps, CSV time-series at reflector positions, GeoJSON alert polygons, PDF monitoring reports |
Analytics Satellize can run
| PS-InSAR displacement time series | Persistent-scatterer InSAR (Stanford Method for PS or similar); requires minimum ~20 SAR acquisitions for stable solution | GeoTIFF velocity map and per-point time-series CSV; updated each new acquisition |
| SBAS distributed-scatterer deformation map | Small-baseline subset (SBAS) processing over coherent pixel clusters; tolerates lower PS density than classical PS-InSAR | GeoTIFF mean velocity and displacement stack; suitable for embankment crest and beach areas with variable coherence |
| Corner-reflector position monitoring | Sub-pixel phase extraction at known reflector coordinates; positions surveyed to centimetre accuracy as ground truth | Time-series plot per reflector with uncertainty bounds; threshold alert when velocity exceeds operator-defined limit |
| Acceleration anomaly alert | Velocity change detection on rolling time-series window; flagged when displacement rate increases beyond statistical threshold (e.g. 3-sigma from background) | Automated alert (email or API webhook) with map of anomalous zone and recommended ground inspection priority |
| Embankment geometry change detection | Multi-temporal DEM differencing using TanDEM-X or stereo optical pairs; detects crest settlement or toe bulge at decimetre scale | Volumetric change map (GeoTIFF) and summary table of embankment section elevations |
| Regulatory monitoring archive | Versioned processing chain with full metadata, uncertainty estimates and audit log; aligned with GISTM monitoring documentation requirements | Quarterly PDF report with displacement maps, time-series charts and signed data provenance 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.