Ground subsidence mapping over underground mines
InSAR time-series analysis turns Sentinel-1 and ALOS-2 radar archives into precise maps of surface subsidence above longwall, block-cave and room-and-pillar workings, detecting displacement rates and spatial extents that ground-based survey networks routinely miss.
Sensors
- Sentinel-1 C-band SAR (ESA/Copernicus): 5 m × 20 m resolution in Interferometric Wide Swath mode, 250 km swath, 6-day repeat at the equator (12-day for a single satellite). C-band (5.4 GHz) gives good coherence over bare or sparsely vegetated ground. Phase-unwrapping breaks down above roughly 14 cm of line-of-sight displacement per 6-day interval, so fast-moving longwall fronts can alias. Archive runs continuously from 2014.
- ALOS-2 PALSAR-2 L-band SAR (JAXA): L-band (1.2 GHz) maintains interferometric coherence through light vegetation and over longer temporal baselines where C-band decorrelates. Stripmap mode delivers 3 m × 3 m resolution; ScanSAR offers 100 m resolution across a 350 km swath. Revisit is 14 days for a single pass geometry. Particularly useful for mines under secondary forest or scrub.
- COSMO-SkyMed X-band SAR (ASI): X-band (9.6 GHz) at 1 m resolution in Spotlight mode resolves individual infrastructure features such as rail lines, conveyor foundations and shaft collars within the subsidence bowl. Shorter wavelength reduces the displacement threshold before aliasing, so it is best suited to slow, post-mining consolidation rather than active longwall advance. Tasked commercially; not open data.
- ICEYE X-band SAR constellation: Growing commercial constellation offering sub-metre Spotlight imagery with same-day or next-day revisit through multi-satellite tasking. Useful for urgent post-event assessment when a subsidence event threatens surface infrastructure and a rapid second acquisition is needed to measure incremental change.
What a subsidence bowl reveals before the survey crew arrives
Underground extraction removes material. The overlying rock mass adjusts, and the surface settles in a characteristic bowl whose width, depth and edge angles are governed by geology, mining depth and the panel geometry. For longwall coal operations, subsidence typically reaches 60–90 % of seam thickness at the bowl centre, with the trough migrating laterally as the face advances. That progression is predictable in direction but not always in rate, and it rarely waits for the next scheduled levelling survey.
InSAR measures line-of-sight displacement between two SAR acquisitions by comparing the phase of the returned radar signal. A full phase cycle corresponds to half the radar wavelength of displacement: roughly 2.8 cm for C-band, 11.8 cm for L-band. Stack dozens of interferograms and apply a time-series algorithm such as SBAS or PS-InSAR, and you recover a displacement history at every coherent pixel, often at millimetre-per-year precision over stable ground. The spatial density is the critical advantage: a single Sentinel-1 frame covers tens of thousands of hectares and produces displacement measurements at far greater point density than a conventional survey network.
Where the physics sets hard limits
C-band coherence degrades quickly over vegetated ground, particularly after rainfall or crop growth between acquisitions. A field of maize between two June acquisitions can decorrelate completely, leaving a gap in the displacement map precisely where a subsidence trough may be developing. L-band from ALOS-2 is substantially more tolerant, but its 14-day revisit means each interferogram spans at least two weeks, which integrates rather than resolves rapid displacement events.
The phase-unwrapping limit is the most operationally significant constraint. Longwall faces in productive mines can advance 5–15 metres per day, and the subsidence trough directly above the active goaf can drop several centimetres within a single 6-day Sentinel-1 interval. If the displacement gradient across adjacent pixels exceeds half a fringe, phase unwrapping fails and the measurement either decorrelates or underestimates the true movement. Analysts should treat InSAR-derived rates in the active trough zone as lower bounds during peak extraction, and cross-check against prism or GNSS monuments where infrastructure is at risk.
Atmospheric path delay, particularly from water vapour, introduces apparent displacement signals of up to 5–10 cm in a single interferogram. Time-series averaging suppresses this, but individual event interferograms used for rapid assessment carry that uncertainty. ERA5 reanalysis data or GACOS corrections reduce the effect but do not eliminate it.
Reading the bowl: what the displacement map actually tells an engineer
The spatial gradient of the subsidence field, the tilt and curvature, is often more consequential for surface structures than the absolute vertical displacement. A pipeline crossing a subsidence bowl at an oblique angle experiences differential settlement that can induce bending stress well before any single point has moved enough to trigger a visual inspection. InSAR-derived curvature maps, calculated from the second spatial derivative of the displacement field, can be overlaid directly on infrastructure asset registers to prioritise inspection resources.
The horizontal extent of the subsidence bowl, defined conventionally by the angle of draw from the panel edge, is visible in the fringe pattern as the zone where displacement gradients taper to background noise. Tracking this boundary through successive time steps shows whether subsidence is still propagating laterally, which matters for permitting decisions about surface land use adjacent to the worked area. Residual subsidence after mining ceases can continue for months to years in weak sedimentary sequences; the Sentinel-1 archive is long enough to characterise that tail.
Block caving and room-and-pillar: different geometries, same physics
Longwall coal is the most studied case, but block caving at large copper and gold mines produces subsidence bowls of comparable or greater areal extent, often with a more circular planform and steeper edge gradients. The Chuquicamata and El Teniente operations in Chile have been studied using InSAR in the published literature, demonstrating that displacement rates in block-cave crown pillars can reach several centimetres per month and are detectable at Sentinel-1 resolution. The challenge is that many block-cave operations are in high-relief terrain where layover and shadow in SAR geometry create data voids on steep slopes.
Room-and-pillar workings present a different problem. Pillar failure, when it occurs, can be sudden rather than progressive, and the precursory displacement signal may be small and spatially confined. PS-InSAR over urban or industrial surfaces near old room-and-pillar workings, where coherent point scatterers are abundant, has detected precursory subsidence of a few millimetres per year above weakening pillars in several documented European cases. That signal is detectable; the difficulty is distinguishing it from other sources of slow ground movement such as groundwater withdrawal.
Turning displacement maps into regulatory and operational products
Regulators in most mining jurisdictions require operators to demonstrate that subsidence remains within predicted limits and that surface structures are not being damaged beyond agreed thresholds. A time-series displacement product delivered as a GIS layer, updated on each satellite pass, gives a mine operator a defensible, spatially continuous record that supplements point-based survey data. It also provides an early warning when the bowl edge approaches a protected structure or property boundary.
Satellize processes Sentinel-1 and ALOS-2 stacks for clients who need this kind of continuous monitoring without building an in-house InSAR processing chain. The workflow sits on open-constellation data, which keeps the archive cost low, with commercial tasking added when higher resolution or faster revisit is needed for a specific event. The Overhead column has covered the methodological choices between PS and SBAS approaches for readers who want the technical background before commissioning an analysis.
A practical monitoring programme typically combines three products: a basin-scale displacement velocity map updated quarterly to track the evolving bowl, a near-real-time alert triggered when displacement rate at a flagged asset exceeds a defined threshold, and an annual compliance report with uncertainty quantification suitable for submission to a mining regulator. Each rests on published, peer-reviewed methods and sensor specifications that any competent reviewer can audit.
Typical figures
| Spatial resolution (Sentinel-1 IW mode) | 5 m × 20 m (range × azimuth); multi-looked to ~20 m for interferometry |
| Spatial resolution (ALOS-2 Stripmap) | 3 m × 3 m; ScanSAR 100 m |
| Revisit interval | 6 days (Sentinel-1 two-satellite constellation); 14 days (ALOS-2); 1–3 days (ICEYE commercial tasking) |
| Minimum detectable displacement rate (PS-InSAR, stable ground) | 1–2 mm/year over multi-year stacks; single-interferogram precision ~5–10 mm |
| Phase-unwrapping limit (C-band, 6-day interval) | ~14 cm line-of-sight per interval; faster displacement causes aliasing and underestimation |
| Swath width | 250 km (Sentinel-1 IW); 350 km (ALOS-2 ScanSAR) |
| Archive depth | Sentinel-1: from April 2014; ALOS-2: from 2014; ALOS-1 PALSAR extends L-band back to 2006 |
| Atmospheric correction uncertainty | 5–10 cm per interferogram before correction; reduced to ~1–2 cm with ERA5 or GACOS |
| Delivery formats | GeoTIFF displacement and velocity rasters, GeoPackage / Shapefile subsidence boundary polygons, CSV time-series at flagged asset locations, PDF compliance summary |
| Latency (operational monitoring) | 12–48 hours after SAR acquisition, depending on processing queue and data availability |
Analytics Satellize can run
| Displacement velocity map | SBAS or PS-InSAR time-series inversion over multi-year Sentinel-1 stack | GeoTIFF raster of mean line-of-sight velocity (mm/year) with uncertainty layer, updated quarterly |
| Subsidence bowl boundary delineation | Spatial gradient thresholding on cumulative displacement surface to identify zone of influence | Polygon GIS layer showing bowl extent and edge-angle contours, versioned by reporting period |
| Infrastructure tilt and curvature index | Second spatial derivative of displacement field computed along pipeline, road and rail alignments | Per-asset curvature time-series CSV with exceedance flags against user-defined thresholds |
| Near-real-time displacement alert | Single-pair interferogram processed within 24 hours of acquisition; displacement delta compared to rolling baseline | Automated alert (email or API) when displacement at a flagged point exceeds defined increment between consecutive passes |
| Residual subsidence characterisation | Post-mining time-series decomposition to separate primary and residual settlement phases | Report with projected time to stabilisation and confidence interval, suitable for mine closure documentation |
| L-band coherence assessment for vegetated sites | Temporal coherence mapping from ALOS-2 PALSAR-2 stack to identify reliable measurement zones | Coherence mask GeoTIFF indicating which areas support reliable InSAR measurement versus decorrelated zones |
| Annual regulatory compliance report | Aggregation of time-series products with uncertainty quantification and comparison against predicted subsidence model | Structured PDF report with methodology annex, suitable for submission to national mining regulator |
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.