Mining-induced subsidence and void migration monitoring
Underground longwall coal and hard-rock mining can drop the surface by metres in weeks. Satellite InSAR and offset-tracking methods, chosen to match deformation rate and geology, turn that motion into regulatory-grade displacement maps.
Sensors
- Sentinel-1 A/B (C-band SAR): 6-day repeat at mid-latitudes with both satellites active; 5 x 20 m IW mode resolution. Coherent for subsidence rates up to roughly 2-3 cm per day in the line-of-sight direction before phase wrapping becomes unresolvable. Free and open archive from 2014.
- ALOS-2 PALSAR-2 (L-band SAR): 23.6 cm wavelength tolerates roughly four times the deformation gradient of C-band before decorrelation, making it the preferred sensor above active longwall panels where weekly settlements of 5-15 cm are common. Nominal 14-day repeat; 3 x 6 m fine-beam mode.
- COSMO-SkyMed (X-band SAR): Constellation of four satellites enables revisits of 1-4 days over a tasked area; Spotlight mode reaches 1 m resolution. X-band phase is sensitive to very small displacements (sub-millimetre in stable rock) but decorrelates rapidly over disturbed or vegetated ground, limiting utility on the most active longwall faces.
- TerraSAR-X / TanDEM-X (X-band SAR): StripMap at 3 m, Spotlight at 1 m; 11-day nominal repeat, reducible by tasking. Useful for monitoring subsidence troughs that have passed peak settlement and are entering the residual phase, where deformation rates are low enough for X-band coherence to be maintained.
Why longwall mining breaks standard InSAR
Longwall coal extraction removes panels typically 200-400 m wide and up to several kilometres long. The overlying strata cave progressively, and the surface above follows: vertical settlements of 0.5-1.5 m are routine, sometimes achieved within four to eight weeks. That rate, concentrated in a narrow trough migrating at walking pace along the panel, can easily exceed 3 cm per day in the radar line-of-sight direction.
C-band InSAR (Sentinel-1 at 5.6 cm wavelength) can resolve phase differences up to half a wavelength per pixel per revisit interval. With a 6-day repeat, the practical deformation limit is roughly 14 mm per day before adjacent pixels become ambiguous and phase unwrapping fails. Active longwall panels routinely breach this. The result is not a gap in the data but a confidently wrong map, which is worse. Analysts who do not account for this produce subsidence contours that underestimate settlement by large factors.
Matching sensor to deformation regime
L-band sensors such as ALOS-2 PALSAR-2 operate at 23.6 cm wavelength. The decorrelation threshold scales with wavelength, so L-band can track deformation gradients roughly four times steeper than C-band before phase ambiguity sets in. Over an active longwall panel, this is often the difference between a usable interferogram and noise. PALSAR-2's 14-day revisit is a constraint, but the longer coherence window compensates on moderately vegetated British or German coalfield terrain.
Where even L-band fails, pixel-offset tracking abandons phase entirely and measures the shift of intensity features between two SAR images. Accuracy is typically one-tenth to one-twentieth of the pixel spacing, so 0.3-0.6 m for PALSAR-2 fine-beam mode. That is too coarse for regulatory reporting of residual subsidence but adequate for detecting the gross trough position and migration rate above a fast-moving face.
X-band (TerraSAR-X, COSMO-SkyMed) offers the opposite trade-off: high spatial resolution and short revisit, but the shortest coherence length. It is most useful in the post-active phase, when the panel has been extracted and settlement is decelerating toward the residual limit. In hard-rock mines with low deformation rates, X-band can detect millimetre-scale annual displacements above old workings, which is precisely the regime relevant to void migration and crown-hole precursor detection.
What the data actually measures, and what it does not
InSAR measures displacement in the radar line-of-sight direction, which is a projection of the true three-dimensional motion vector. Longwall subsidence is predominantly vertical, but horizontal strains of 5-10 mm/m are common at the trough edges and are what damage surface infrastructure. Separating vertical from horizontal components requires either ascending and descending pass combinations or the addition of GNSS ground truth. Without both geometries, the horizontal strain field is estimated, not measured.
Atmospheric phase delay is the other persistent limit. Tropospheric water vapour introduces apparent range changes of up to several centimetres per scene, which is comparable to the signal of interest in slow residual subsidence. Correction methods exist, including ERA5 reanalysis data and generic atmospheric correction algorithms, but they reduce rather than eliminate the error. For regulatory submissions, satellite data is best treated as a dense spatial interpolator between a sparser network of levelling benchmarks or GNSS stations, not as a standalone measurement system.
Archive depth matters for mining contexts. Sentinel-1 data runs from April 2014, covering the closure and post-closure phases of many UK and German coalfields. ALOS-1 PALSAR (2006-2011) extends L-band coverage further back. Combining epochs allows analysts to reconstruct the full settlement history of a panel and compare it against the National Coal Board empirical prediction model or the more recent SDPS software outputs.
Regulatory reporting and the subsidence prediction loop
In Great Britain, the Coal Authority requires operators and affected landowners to demonstrate that surface movements remain within predicted limits and to report anomalous settlement. In Germany, the Bergämter (mining authorities) impose similar obligations under Landesbergrecht. Both frameworks were written around terrestrial survey methods, but satellite-derived displacement maps are increasingly accepted as supplementary evidence, particularly for demonstrating spatial extent and trough migration velocity.
The most productive workflow is not to replace prediction models with satellite data but to calibrate them. The NCB subsidence prediction method and its successors estimate maximum settlement as a function of seam thickness, depth, and panel width. Satellite time-series allow analysts to fit the predicted trough shape to observed data and identify panels where actual settlement exceeds prediction, which is the regulatory trigger for further investigation. Discrepancies of more than 10-15% between predicted and observed maximum settlement are a credible threshold for escalation, though the appropriate value is jurisdiction-specific.
Satellize can structure this calibration workflow as a repeating delivery, producing updated displacement maps on each new SAR acquisition cycle and flagging panels where the observed-to-predicted ratio crosses a defined threshold.
Void migration: what satellite data can and cannot see
Shallow abandoned workings, particularly room-and-pillar coal mines at depths of less than 50-100 m, are prone to pillar collapse and crown-hole formation at surface. The surface expression before collapse is often a subtle bowl of a few centimetres over months, which is within InSAR detection capability if coherence is maintained. Sentinel-1 time-series using persistent-scatterer or small-baseline subset methods can detect precursor motion at the 5-10 mm/year level over stable urban or semi-urban terrain.
The honest limit is that not every crown-hole is preceded by detectable surface motion. Collapse can be sudden and localised, with no precursor signal above the atmospheric noise floor. Satellite monitoring reduces risk but does not eliminate it. The appropriate role is systematic screening of large areas to prioritise ground investigation, not to certify that a specific location is safe.
Typical figures
| Spatial resolution (InSAR products) | 5-20 m (Sentinel-1 IW); 3-6 m (ALOS-2 fine-beam); 1-3 m (TerraSAR-X Spotlight / COSMO-SkyMed) |
| Minimum detectable displacement (stable terrain) | ~5 mm per year with PS-InSAR on Sentinel-1; ~1-2 mm per year with X-band on dense scatterer networks |
| Maximum trackable deformation rate (InSAR) | ~14 mm/day (C-band, 6-day repeat); ~55 mm/day (L-band, 14-day repeat); offset tracking extends to decimetre-scale weekly motion |
| Revisit interval | 6 days (Sentinel-1 dual-satellite); 14 days (ALOS-2); 1-4 days (COSMO-SkyMed tasked); 11 days (TerraSAR-X nominal) |
| Radar frequency / wavelength | C-band 5.6 cm (Sentinel-1); L-band 23.6 cm (ALOS-2); X-band 3.1 cm (TerraSAR-X, COSMO-SkyMed) |
| Displacement components resolved | Line-of-sight (single geometry); vertical + east-west (ascending + descending combination); north-south poorly constrained by InSAR alone |
| Archive depth | Sentinel-1 from April 2014; ALOS-1 PALSAR from 2006-2011; TerraSAR-X from 2007; COSMO-SkyMed from 2007 |
| Atmospheric correction | ERA5 reanalysis or GACOS service; residual error typically 1-3 cm per scene after correction |
| Delivery formats | GeoTIFF displacement maps, GIS-ready shapefiles, time-series CSV per point, PDF regulatory summary report |
Analytics Satellize can run
| Longwall trough migration map | Small-baseline subset (SBAS) InSAR time-series using L-band or C-band interferograms; trough geometry fitted to NCB/SDPS prediction model | GeoTIFF displacement map per acquisition epoch, updated on each satellite pass; PDF overlay against predicted trough for regulatory submission |
| Observed-to-predicted settlement ratio alert | Automated comparison of maximum InSAR-derived settlement against panel-specific NCB empirical prediction; threshold-based flagging | Alert report when ratio exceeds client-defined threshold (e.g. 1.15); includes spatial extent of exceedance zone |
| Horizontal strain field at trough edges | Gradient of vertical displacement field from ascending/descending InSAR combination; strain expressed in mm/m | GIS layer of compressive and tensile strain zones for infrastructure risk assessment |
| Precursor motion screening above shallow abandoned workings | Persistent-scatterer InSAR (PS-InSAR) on Sentinel-1 time-series; velocity map compared against Coal Authority abandonment plan extents | Annual velocity map (mm/year) with anomaly polygons ranked by magnitude for ground-investigation prioritisation |
| Pixel-offset tracking displacement field | Sub-pixel cross-correlation of SAR intensity images (ALOS-2 or COSMO-SkyMed); applicable where phase decorrelation prevents InSAR | East-west and vertical displacement grids at 0.3-0.6 m accuracy per epoch pair |
| Multi-decade settlement history reconstruction | Multi-temporal InSAR combining ALOS-1 PALSAR, Sentinel-1, and available commercial SAR archives; time-series stacking | Cumulative displacement profile from 2006 to present; CSV time-series per benchmark location for comparison with levelling records |
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.