Quarry volumetric change detection using stereo and SAR DEM differencing
DEM differencing from stereo optical and SAR acquisitions converts elevation change into cubic-metre volume estimates per reporting period, giving quarry operators, regulators and royalty auditors an independent production record.
Sensors
- Airbus Pléiades Neo: Tri-stereo mode collects three along-track views in a single pass, enabling photogrammetric DEM generation at 50 cm posting. Tasking revisit is nominally daily at mid-latitudes. Cloud is the binding constraint on optical stereo.
- Maxar WorldView-3: 0.31 m panchromatic resolution; stereo and tri-stereo modes yield DEMs at roughly 50 cm to 1 m posting depending on convergence angle and scene contrast. Strong archive depth from 2014 onward.
- TanDEM-X / TerraSAR-X: Bistatic X-band SAR pair provides the global TanDEM-X DEM at 12 m posting (0.4 m relative vertical accuracy over smooth terrain). New bistatic acquisitions can be commissioned as an independent elevation baseline, unaffected by cloud. Coherence degrades on wet or freshly blasted surfaces.
- ICEYE X-band SAR constellation: Sub-metre resolution stripmap and spotlight modes with revisit measured in hours rather than days. Useful for tracking rapid bench advance; coherence-based DEM generation requires careful scene-to-scene pairing.
What a shrinking pit reveals
Every tonne of rock extracted from a quarry leaves a measurable void. If you know the density of the material, the void volume converts directly to tonnage. That relationship is the physical basis for satellite-derived production auditing: no weighbridge, no truck manifest, no operator declaration required.
The method is DEM differencing. A digital elevation model captured at time T1 is subtracted from one captured at time T2. Negative elevation change in the working area represents material removed; positive change elsewhere can indicate stockpile growth or overburden placement. The result is a volumetric change map in cubic metres, attributed to each part of the pit.
Two routes to an elevation model, and why both matter
Stereo optical photogrammetry and SAR-based elevation retrieval are complementary rather than redundant. Stereo optical, particularly tri-stereo from Pléiades Neo or WorldView-3, produces very high resolution DEMs: 50 cm posting is achievable with Pléiades Neo tri-stereo under good contrast conditions. The geometry is well understood, the processing chain is mature, and the output integrates naturally with photogrammetric workflows that quarry surveyors already use. The limitation is cloud. A single overcast day voids the acquisition, and in humid or seasonally wet regions that can mean multi-month gaps.
SAR sees through cloud. TanDEM-X bistatic acquisitions, using the two-satellite formation that produced the global baseline DEM, can be retasked to generate new elevation snapshots over a specific quarry. The global TanDEM-X DEM carries a relative vertical accuracy specification of around 2 m at 90th percentile over non-forested terrain, though smooth, high-contrast quarry benches typically perform considerably better. The catch is coherence. Fresh blasting leaves a rubble surface whose scattering geometry changes between passes, decorrelating the SAR phase signal and degrading or destroying the interferometric elevation estimate in precisely the areas of greatest interest. Wet surfaces have the same effect. For active quarry faces, SAR-derived DEMs are best treated as a cross-check on optically derived results, not a replacement.
ICEYE and similar commercial X-band constellations add high-revisit SAR capability. Their value is in detecting rapid change, flagging that a new acquisition campaign is warranted, rather than in generating the primary volumetric estimate.
Accuracy: what the numbers actually mean
Vertical accuracy in a differenced DEM depends on the accuracy of each individual DEM and on how errors correlate between epochs. For stereo optical DEMs, the height-to-base ratio of the stereo pair is the dominant geometric factor. Pléiades Neo tri-stereo, with its three-view geometry, achieves root-mean-square vertical errors in the range of 20 to 50 cm over well-textured terrain with ground control or precise orbit data. Over a quarry bench of, say, 10,000 square metres, a 30 cm vertical uncertainty integrates to a volume uncertainty of roughly 3,000 cubic metres. Whether that matters depends on the size of the extraction event being measured.
Minimum detectable volume change is not a fixed number. It scales with pit area, surface roughness, the quality of ground control, and the time interval between acquisitions. Published studies using Pléiades stereo over active quarries have reported volume uncertainties in the low single-digit percentage range for large extraction events. Smaller quarries or short intervals between acquisitions, where the elevation change is small relative to the noise floor, produce wider relative uncertainties. Honest reporting requires stating the uncertainty alongside the estimate, not just the estimate.
Ground control points, whether from GNSS survey, corner reflectors, or stable reference surfaces within the scene, are the most effective way to reduce systematic bias between epochs. Without them, any tilt or offset in one DEM propagates directly into the volume estimate.
Regulatory and royalty audit applications
The most commercially significant application is independent production verification. Quarry royalties and extraction levies are typically assessed on declared tonnage. Satellite-derived volume estimates, converted to mass using the bulk density of the extracted material, provide regulators with a cross-check that requires no access to the site and no cooperation from the operator.
The method is not a substitute for a legal weighbridge record. Courts and regulators require calibrated instruments and chain-of-custody documentation that satellite data cannot provide. What satellite DEM differencing does provide is a statistically independent signal that can flag significant discrepancies between declared and observed extraction, prompting targeted inspection. Several national mining regulators have explored or adopted remote-sensing cross-checks for exactly this reason, though the specifics of each programme vary considerably by jurisdiction.
For operators, the same data supports internal reconciliation: comparing pit survey volumes against mill feed records to identify losses in the comminution circuit or errors in density assumptions.
Practical constraints worth stating plainly
Cloud cover is the single largest operational risk for optical stereo programmes. In tropical regions, usable stereo acquisitions may be available only a few times per year. SAR partially mitigates this but introduces coherence constraints of its own. A combined optical-SAR strategy, using whichever sensor delivered a clean acquisition in a given period, is more reliable than either alone.
Steep pit walls cast shadows that degrade stereo matching. Benches below the shadow line may be missing from the DEM or carry larger errors than benches in direct illumination. Acquisition planning should account for solar elevation and azimuth relative to the pit orientation.
Stockpile areas adjacent to the pit require separate treatment. Material moved from the pit face to a stockpile appears as negative change in one zone and positive change in another. If the stockpile is outside the DEM footprint, the apparent extraction volume will be overestimated. Defining the analysis boundary carefully, to include both the working face and all associated stockpile areas, is essential to a coherent mass balance.
Satellize integrates stereo tasking and SAR acquisition planning for clients who need periodic volumetric reporting rather than one-off snapshots; the Tonga crop-estimation programme demonstrated the same multi-epoch cadence logic applied to a different physical quantity.
From elevation difference to a number you can act on
The analytic pipeline has four stages: DEM generation, co-registration, differencing, and volume integration. Co-registration is the step most often underestimated. Two DEMs that are not precisely aligned in X, Y and Z will produce spurious elevation differences even over stable terrain. The standard check is to compute the mean and standard deviation of elevation difference over areas that should not have changed between epochs. A non-zero mean indicates a systematic offset; a large standard deviation indicates either poor co-registration or genuine change across the reference area.
Volume integration uses the prismatoid formula or simple column summation over the raster grid. At 50 cm posting, each cell represents 0.25 square metres of area; multiplied by the elevation difference and summed across the change polygon, the result is cubic metres. Conversion to tonnes requires a bulk density figure, which must come from laboratory measurement or published values for the specific lithology. Limestone, for example, has a bulk density typically in the range of 1.5 to 2.0 tonnes per cubic metre depending on void fraction, and the choice of figure materially affects the mass estimate.
Deliverables can be structured as a GIS layer showing spatially distributed elevation change, a tabular report of net volume and estimated mass per reporting period, or an alert when cumulative change exceeds a threshold between scheduled surveys. The right cadence depends on the extraction rate: a large aggregate quarry moving tens of thousands of tonnes per week warrants monthly or quarterly reporting; a smaller decorative-stone operation might need only annual reconciliation.
Typical figures
| Best achievable DEM posting (optical stereo) | 50 cm (Pléiades Neo tri-stereo, WorldView-3 stereo) |
| Vertical accuracy, optical stereo | 20–50 cm RMSE over well-textured terrain with ground control; degrades on shadowed or featureless surfaces |
| TanDEM-X global DEM vertical accuracy | ~2 m at 90th percentile (relative); smooth quarry benches typically better |
| SAR coherence constraint | Coherence lost on wet or freshly blasted surfaces; limits interferometric DEM quality at active faces |
| Optical revisit (tasked) | Pléiades Neo: nominally daily at mid-latitudes; cloud-dependent |
| SAR revisit (ICEYE constellation) | Sub-daily to daily, cloud-independent |
| Minimum detectable volume change | Scene- and area-dependent; published studies report low single-digit percentage uncertainty for large extraction events over quarries of several hectares |
| Archive depth | WorldView-3 from 2014; Pléiades from 2012; TanDEM-X global baseline from 2010–2015 acquisition campaign |
| Delivery formats | GeoTIFF DEM, change-raster GeoTIFF, polygon shapefile or GeoJSON, tabular CSV volume report |
| Coverage | Global tasking; cloud risk highest in tropical and monsoon-affected regions |
Analytics Satellize can run
| Epoch-pair volume change estimate | Stereo photogrammetric DEM generation followed by co-registered DEM differencing and prismatoid volume integration | Tabular report: net volume (m³) and estimated mass (tonnes) per reporting period, with stated uncertainty |
| Spatially distributed elevation change map | Raster subtraction of co-registered DEMs; change classified by magnitude and sign | GeoTIFF change raster and polygon GIS layer, coloured by elevation delta, suitable for overlay in mine planning software |
| Stockpile and pit mass balance | Separate volume integration over defined pit and stockpile polygons; net balance computed from signed volumes | CSV table and summary report distinguishing extraction, stockpile growth and overburden placement |
| Co-registration quality assessment | Stable-surface residual analysis: mean and standard deviation of elevation difference over non-change reference zones | QA appendix included with each delivery, flagging systematic offsets or high noise in the reference zone |
| SAR coherence change alert | Interferometric coherence magnitude computed between successive X-band SAR acquisitions; low-coherence zones flagged as active-face indicators | Alert layer (GeoJSON) identifying zones of recent surface disturbance, triggering priority optical tasking |
| Multi-epoch production time series | Repeated DEM differencing across an archive of tasked acquisitions; cumulative and per-period volumes plotted against time | Interactive chart and underlying CSV for integration into operator or regulator reporting dashboards |
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.