Mining output estimation from ore stockpile volume change
Repeat stereo imagery and SAR height models let analysts measure net material movement at open-pit mines independently of operator reporting, giving commodity desks and project lenders a physical cross-check on declared production.
Sensors
- Airbus Pléiades Neo: 30 cm native panchromatic resolution with tri-stereo collection capability in a single pass; produces dense point clouds suitable for DSM differencing at stockpile scale. Revisit at a given site is typically 1–2 days with programming priority.
- Maxar WorldView-3: 31 cm panchromatic, 1.24 m multispectral. Supports along-track stereo and tri-stereo modes. The multispectral bands (including SWIR) help distinguish ore lithology from waste material by surface mineralogy, adding interpretive context beyond pure geometry.
- Planet SkySat: 50 cm panchromatic, with video and stereo tasking available. Revisit is more frequent than Pléiades or WorldView-3 but point-cloud density and absolute height accuracy are lower, making SkySat better suited to change detection than to first-principles volume calculation.
- Capella Space SAR (X-band): Spotlight mode delivers roughly 50 cm resolution imagery. SAR is unaffected by cloud cover, which matters at tropical or high-latitude mine sites. Repeat-pass InSAR or radargrammetry can track surface height change, though layover and foreshortening on steep stockpile faces introduce geometric ambiguity that optical stereo does not.
What a stereo pair actually measures
When a satellite collects two images of the same scene from slightly different angles, photogrammetric software matches conjugate points and back-projects them into three-dimensional space, producing a digital surface model (DSM). Differencing two DSMs from different dates gives a height-change raster: positive values indicate material added, negative values indicate material removed. At a mine site, the analyst integrates those height changes over the footprint of each identifiable stockpile to obtain a net volume change in cubic metres.
The limiting factor is not the satellite's ground sampling distance but the accuracy of the DSM itself. For Pléiades Neo tri-stereo collections, published independent assessments place height accuracy (RMSE) in the range of 0.2–0.5 m on flat terrain. On the angular faces of a stockpile, accuracy degrades because matching fails on shadow-filled slopes and near the crest. Practically, a single-epoch volume estimate carries an uncertainty of a few percent for large stockpiles and can be meaningfully worse for small or narrow ones. Differencing two DSMs compounds those errors, so the minimum detectable volume change at commercial stereo resolution is roughly 5,000–15,000 cubic metres, depending on stockpile geometry and image quality. Below that threshold, the signal is inside the noise.
Ore pile or waste dump? The classification problem
Volume change alone does not tell you what moved. An open-pit mine generates at least three distinct material streams: run-of-mine ore, low-grade ore stockpiled for later processing, and waste rock dumped permanently. All three form surface accumulations that look geometrically similar in a DSM. Getting the attribution right matters enormously: a growing waste dump misread as ore inventory overstates production; a shrinking ore stockpile fed to the mill is revenue already realised, not future production.
Several signals help separate them. Location relative to the pit and the processing plant is the first filter: waste dumps sit away from the mill, ore stockpiles sit adjacent to it. Surface colour and spectral signature in multispectral or SWIR bands can distinguish oxidised waste from sulphide ore at many deposit types, though this is site-specific and requires calibration against known lithology. Temporal pattern matters too: waste dumps grow monotonically while ore stockpiles cycle as material is drawn down for processing. An analyst working across multiple observation epochs can read those cycles. None of these proxies is definitive on its own; the honest answer is that attribution requires combining geometric change with site-specific knowledge of the mine layout.
The compaction problem: from cubic metres to tonnes
Even a perfectly measured volume change does not convert directly to tonnes without knowing bulk density and compaction state. Freshly blasted and truck-dumped ore has a swell factor of roughly 1.25–1.40 relative to in-situ rock, meaning the same mass occupies 25–40 per cent more volume once broken. As material sits on a stockpile and fines settle, that swell factor decreases. Rain accelerates compaction. For a copper porphyry ore with an in-situ density of around 2.6–2.8 t/m³, the bulk density of a stockpile might range from 1.8 to 2.2 t/m³ depending on age and moisture content.
Commodity analysts must apply a conversion factor, and that factor is uncertain unless the operator publishes it or the analyst can infer it from publicly available geological data on the deposit. The satellite measurement gives you volume; the tonne figure is a derived estimate with its own error bar. Presenting a single-point tonnage figure without acknowledging this is misleading. The appropriate output is a range, with the conversion assumptions stated explicitly.
SAR as a cloud-resilient complement, not a replacement
Tropical mine sites, including much of the copper belt in the DRC and many Indonesian nickel operations, sit under persistent cloud cover for months at a time. Optical stereo fails entirely when cloud obscures the scene. X-band SAR from systems like Capella penetrates cloud and can be collected at night, giving continuous temporal coverage that optical cannot match.
The trade-off is geometric. SAR images in slant range, and steep stockpile faces facing away from the sensor create layover or shadow that corrupts the height signal in exactly the areas that matter most for volume calculation. Radargrammetry, which derives height from two SAR images taken from different angles, partially addresses this but typically delivers lower height accuracy than optical stereo on structured surfaces. Repeat-pass InSAR is better suited to detecting subtle surface deformation on flat or gently sloping ground than to measuring the large discrete height changes of an active stockpile. In practice, a combined approach works best: optical stereo for periodic high-accuracy volume snapshots, SAR to fill cloud gaps and flag significant change events between optical collections.
What the archive adds that a single image cannot
Commercial satellite archives for Pléiades and WorldView go back roughly a decade at many mine sites, and the Maxar archive in particular is dense over economically significant operations. That history allows an analyst to reconstruct stockpile volume trajectories over multiple years, revealing seasonal patterns, production ramp-ups and shutdowns that a single observation cannot distinguish. A stockpile that appears large today might represent months of accumulation during a mill outage rather than current production, a distinction that matters to a lender assessing repayment capacity.
Archive depth also provides a baseline for fraud detection. If a company reports consistent production over a period during which satellite-derived stockpile volumes show no net change at the mine and no corresponding draw-down at the processing facility, the discrepancy is a flag worth investigating. The satellite does not prove fraud; it identifies an inconsistency that warrants a harder question. Satellize structures these multi-epoch analyses as periodic monitoring reports, with change layers delivered as GIS-ready rasters alongside a written interpretation. The Tonga crop-estimation programme uses a similar repeat-observation architecture, though the commodity and physics are different.
Honest limits and what they mean for buyers
Several conditions defeat or degrade this method. Heavy cloud at optical wavelengths is the most common. Stockpiles with very shallow slopes or smooth surfaces produce fewer stereo matching features, reducing DSM quality. Sites with active dust or spray suppression systems can obscure surface texture. Night-time operations that move material between satellite passes create a temporal aliasing problem: the DSM captures a snapshot, not a continuous ledger.
The method also cannot see underground. Run-of-mine ore drawn from the pit but not yet placed on a surface stockpile is invisible. Similarly, ore fed directly from truck to crusher without intermediate stockpiling leaves no surface signature. These are structural blind spots, not instrument failures. A sophisticated buyer uses satellite-derived stockpile volumes as one input alongside mill throughput data, power consumption records and shipping manifests, not as a standalone production audit. The value is in cross-checking, not in replacing the full data room.
Typical figures
| Spatial resolution (optical stereo) | 30–50 cm panchromatic (Pléiades Neo, WorldView-3, SkySat) |
| DSM height accuracy (flat terrain) | 0.2–0.5 m RMSE for Pléiades Neo tri-stereo; degrades on steep stockpile faces |
| Minimum detectable volume change | Approximately 5,000–15,000 m³ depending on stockpile geometry and image quality |
| SAR resolution (Capella spotlight) | ~50 cm; radargrammetric height accuracy typically 0.5–2 m |
| Revisit (tasked optical) | 1–2 days with programming priority (Pléiades Neo); 1–3 days (WorldView-3) |
| Cloud penetration | SAR (X-band) only; optical stereo fails under cloud cover |
| Archive depth | Up to ~10 years at major mine sites for Maxar and Airbus commercial archives |
| Delivery formats | GeoTIFF DSM differencing rasters, GeoJSON stockpile polygons, PDF monitoring report |
| Volume-to-mass conversion uncertainty | Bulk density range typically 1.8–2.2 t/m³ for broken hard-rock ore; conversion factor must be stated explicitly |
Analytics Satellize can run
| Stockpile volume change estimate | DSM differencing from repeat stereo photogrammetry (Pléiades Neo or WorldView-3 tri-stereo) | GeoTIFF height-change raster with per-stockpile volume summary table, delivered as periodic report |
| Material classification (ore vs waste) | Spatial relationship analysis combined with SWIR spectral signature comparison where multispectral data available | GeoJSON polygon layer with material-type attribution and confidence flag |
| Production trajectory reconstruction | Multi-epoch archive DSM series with time-series volume curve | Excel-compatible time series and PDF narrative covering selected historical period |
| Cloud-gap SAR change flag | Capella X-band radargrammetry or intensity change detection between cloud-affected optical windows | Alert report flagging significant stockpile change events during cloud-obscured optical periods |
| Reported-vs-observed discrepancy score | Comparison of operator-reported production volumes against satellite-derived volume change, converted using stated bulk density assumptions | Structured discrepancy table with sensitivity analysis on conversion factors, suitable for credit committee review |
| Site layout baseline map | Single high-resolution optical collect with manual and semi-automated feature delineation (pit boundary, stockpile zones, haul roads, processing facilities) | GIS layer set establishing monitoring zones for subsequent change-detection runs |
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.