Block cave subsidence crater and draw-zone surface expression mapping

Block-cave mining creates a measurable surface signature months before breakthrough. Time-series InSAR on Sentinel-1 and COSMO-SkyMed maps subsidence velocity, crater extent, and tension-crack propagation to give early warning where underground sensors cannot.

Sensors

Sentinel-1 A/B (C-band SAR): 5 m × 20 m IW mode resolution, 6-day repeat at mid-latitudes with two satellites. Free and open archive back to 2014. C-band penetrates cloud but is sensitive to atmospheric delay; coherence degrades over vegetated or disturbed surfaces. Suitable for detecting subsidence rates above roughly 10–20 mm per year in LOS.
COSMO-SkyMed (X-band SAR): Spotlight mode delivers 1 m resolution; StripMap delivers 3–15 m. Constellation of four satellites gives revisit of 12–24 hours on request. X-band is more sensitive to small displacements and maintains coherence better over bare ground, but loses it faster over disturbed cratered surfaces. Minimum detectable displacement per interferogram is sub-centimetre under good coherence conditions.
ICEYE X-band SAR: Sub-metre Spotlight mode (25 cm × 50 cm in the highest-resolution product), with a growing constellation that supports daily revisit over a fixed point. Particularly useful for resolving the sharp displacement gradient at the crater rim, where conventional InSAR phase unwrapping breaks down due to steep gradients.
Capella Space X-band SAR: Spotlight resolution down to approximately 0.35 m. On-demand tasking with same-day turnaround available. High-resolution coherent change detection between acquisitions separated by hours can identify fresh tension-crack opening events that slower revisit schedules would miss.

What block caving actually does to the surface

Block caving removes ore from below by letting gravity fracture and draw down a large ore column. As the cave propagates upward through the rock mass, the surface above subsides. The resulting deformation field is not uniform. Directly above the cave, a subsidence crater forms, characterised by inelastic, often discontinuous displacement. Outside it, an elastic zone develops where the ground deforms coherently but has not yet fractured. Beyond that, a tension-crack zone marks the boundary where extensional stress is opening the rock mass.

Each zone has a different radar signature. The elastic zone produces coherent interferometric fringes that can be unwrapped to give millimetre-scale displacement rates. The crater interior often decorrelates entirely as the surface fractures, which is itself a diagnostic signal. The transition between coherent and incoherent zones, tracked over time, maps the advancing crater rim with a precision that ground survey crews cannot match safely or frequently enough.

Why InSAR is the right tool, and where it struggles

Differential InSAR measures the phase difference between two SAR acquisitions of the same scene. Each 2π phase cycle corresponds to half the radar wavelength of displacement in the line-of-sight direction. For C-band (5.6 cm wavelength) that is 2.8 cm per fringe; for X-band (3.1 cm) it is 1.55 cm. Small-baseline subset (SBAS) and persistent-scatterer (PS-InSAR) methods stack dozens of interferograms to separate deformation signal from atmospheric noise, achieving millimetre-per-year precision over stable point targets.

The honest limits matter here. InSAR measures only the line-of-sight component of displacement, so vertical and horizontal motion must be separated using ascending and descending passes, or by combining with GNSS. Where subsidence rates exceed roughly 10 cm per day, fringes become too dense to unwrap and the signal is lost. This is precisely the regime that precedes unplanned breakthrough, so InSAR provides warning in the run-up, not a real-time record of the final collapse. Atmospheric water vapour introduces artefacts at the 1–3 cm level per acquisition; correcting for this requires auxiliary weather data or empirical atmospheric models. Vegetation cover degrades coherence significantly, though bare-ground mine sites are usually favourable.

Separating the crater from the elastic bowl: the analytic workflow

The standard processing chain begins with co-registration of a SAR image stack, interferogram generation, and filtering. Phase unwrapping converts the wrapped interferogram into a continuous displacement field. Atmospheric correction, using either ERA5 reanalysis data or the Generic Atmospheric Correction Online Service (GACOS), removes the dominant tropospheric signal. The corrected time series is then decomposed into vertical and east-west displacement components using ascending and descending geometries.

Spatial segmentation of the deformation field separates three zones operationally. The incoherent crater core is identified by coherence below a threshold, typically 0.3 in the literature. The inner elastic zone is defined by displacement rates above a site-specific background and by the presence of concentric fringe patterns. The outer tension-crack zone is identified by anomalous displacement gradients, where the rate changes sharply over short distances. Velocity maps and cumulative displacement time series are produced for each zone. The rate of outward migration of the coherence-loss boundary is the most direct proxy for cave propagation rate.

Early warning of unplanned breakthrough: what the data can and cannot tell you

Unplanned breakthrough occurs when the cave reaches surface ahead of schedule, often through a weaker geological structure. The precursor signal in InSAR is acceleration: the subsidence rate increases, and the incoherent zone expands faster than the historical trend. Automated threshold alerting on velocity anomalies, combined with coherence-loss rate, can flag this condition days to weeks in advance, depending on the rate of acceleration and the revisit cadence of the constellation.

What InSAR cannot do is tell you why the cave is accelerating. That requires integration with underground cave tracker data, seismic monitoring, and geological models. The satellite view is a boundary condition, not a process model. It is also blind to sub-surface geometry. A surface that looks stable may be bridging over a void. Conversely, rapid surface subsidence does not always mean imminent breakthrough; it may reflect shallow secondary collapse. The value of the method is in providing an independent, spatially continuous check on underground monitoring, not in replacing it.

Sensor selection by site condition

Sentinel-1 is the baseline choice for long time series and budget-constrained programmes. Its free archive since 2014 means that for any mine that has been operating for several years, a historical deformation baseline already exists. The 6-day repeat is adequate for detecting slow to moderate subsidence rates, and the IW mode covers the full mine area in a single pass.

COSMO-SkyMed or ICEYE become necessary when the subsidence gradient is steep, the crater rim is narrow, or the operation needs same-day tasking for incident response. X-band coherence is also better preserved over the short baselines used in high-rate monitoring. For the sharpest crater-rim delineation, sub-metre Spotlight acquisitions from ICEYE or Capella resolve features that C-band simply cannot. A practical programme often uses Sentinel-1 for routine monitoring and commercial X-band for targeted acquisitions when the velocity field changes character.

Satellize structures monitoring programmes around this tiered logic, with open-constellation baselines and commercial tasking added on client licence. The approach is the same one used in the Tonga crop-estimation programme: open data carries the routine load, and commercial sensors are triggered by analytic thresholds rather than on a fixed schedule.

What you receive, and when

A typical delivery cycle for an active block-cave site runs on a 6-day cadence aligned to Sentinel-1 passes, with X-band acquisitions triggered when velocity thresholds are exceeded. Each cycle produces a displacement velocity map in GeoTIFF format, a coherence map, a crater-boundary polygon updated from the coherence-loss contour, and a time-series chart for a set of user-defined monitoring points. Alert notifications fire when the outward migration rate of the crater rim exceeds a configurable threshold, or when any monitoring point records an acceleration above a set value.

Latency from SAR acquisition to delivered product is typically 24–48 hours for Sentinel-1 after the data appears on the Copernicus Data Space. Commercial X-band products with same-day tasking can be processed and delivered within 12 hours of acquisition. Archive analysis, covering the full Sentinel-1 record back to 2014, is delivered as a single historical deformation dataset within two to three weeks of project start.

Typical figures

Spatial resolution (Sentinel-1 IW)	5 m × 20 m (range × azimuth); resampled to 10–20 m grid for InSAR products
Spatial resolution (COSMO-SkyMed Spotlight)	~1 m; ICEYE and Capella sub-metre modes available down to ~0.35 m
Revisit cadence	6 days (Sentinel-1, two satellites); 12–24 hours on request (COSMO-SkyMed); daily or same-day (ICEYE, Capella)
Minimum detectable LOS displacement rate	~10–20 mm/year (SBAS/PS-InSAR on Sentinel-1 over bare ground); sub-mm/year with dense PS networks
Maximum measurable gradient before phase unwrapping fails	~10 cm per 6-day interval for C-band; ~5 cm per interval for X-band (half-wavelength ambiguity limit)
Radar frequency	C-band 5.405 GHz (Sentinel-1); X-band ~9.6 GHz (COSMO-SkyMed, ICEYE, Capella)
Sentinel-1 archive depth	2014 to present (Sentinel-1A launch April 2014)
Atmospheric correction	ERA5 reanalysis or GACOS service; residual error typically 1–3 cm per acquisition
Delivery formats	GeoTIFF displacement and coherence maps, Shapefile/GeoJSON crater-boundary polygons, CSV time-series, PDF summary report
Typical product latency	24–48 hours post-acquisition (Sentinel-1); 12 hours post-acquisition (commercial X-band, on-demand tasking)

Analytics Satellize can run

Subsidence velocity map	SBAS or PS-InSAR time-series inversion on stacked interferograms, atmospheric correction applied	GeoTIFF velocity raster (mm/year), updated each 6-day Sentinel-1 cycle
Crater boundary polygon	Coherence-loss contouring: pixels with mean coherence below 0.3 over a rolling 60-day window classified as crater interior	GeoJSON polygon updated per cycle, with area and centroid attributes
Elastic vs. inelastic zone classification	Spatial segmentation of displacement gradient field; zones defined by displacement rate thresholds and fringe continuity	Classified GeoTIFF and Shapefile with three-zone legend; updated monthly or on threshold trigger
Crater rim migration rate	Temporal differencing of crater boundary polygons; outward migration distance per unit time calculated for eight cardinal sectors	CSV time series and polar plot showing directional asymmetry; alert fired if rate exceeds configurable threshold
Displacement acceleration alert	Kalman-filter velocity estimation on monitoring-point time series; alert triggered when acceleration exceeds two standard deviations above site baseline	Automated notification (email or API webhook) with displacement chart and map attachment
Historical deformation baseline	Full Sentinel-1 archive reprocessing from 2014 or mine commissioning date, whichever is later; SBAS inversion over entire stack	Single GeoTIFF stack and NetCDF time series covering full archive period, delivered within 2–3 weeks of project start
Ascending/descending decomposition	Two-geometry combination to separate vertical and east-west displacement components; north-south component unresolvable from SAR alone	Separate GeoTIFF layers for vertical and horizontal displacement, with uncertainty map

Sources

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.