TL;DR: This drone volumetric coal stockpile QLD case study covers an ROM and product survey of 14 piles at a Bowen Basin metallurgical coal operation near Moranbah, where a 9% gap between surveyed inventory and plant reconciliation was eroding monthly financial reporting confidence. ISS replaced a monthly GPS walkover with RTK drone photogrammetry, capturing all 14 piles in a single 38-minute flight and tightening the reconciliation gap to under 2% against an audited base surface. The result was defensible inventory figures, faster month-end reporting, and a recurring survey programme delivered within 24 hours each cycle.
Key takeaways
- A single 38-minute DJI Matrice 350 RTK flight at 95 m AGL captured all 14 ROM and product stockpiles — work that previously took a two-person GPS crew the better part of a day across an active coal handling pad.
- Surveyed inventory had been running roughly 9% adrift of plant throughput reconciliation; tightening edge definition, ground control, and a properly audited base surface brought the discrepancy to under 2%, well inside the operation's ±3% target.
- Five permanent ground control marks were established in GDA2020 / MGA2020 Zone 55 with AHD heights via a 20-minute RTK observation off the operation's existing control, giving check residuals of 14 mm horizontal and 21 mm vertical.
- Deliverables — a Propeller-hosted 3D model, per-pile volumes, applied bulk densities and a signed accuracy statement — were issued within 24 hours of the flight, against the prior 3-5 day turnaround.
- Locking in a fixed base surface and a standard monthly flight made each subsequent survey directly comparable, turning volumetrics from a monthly headache into a reconciliation input the finance team could actually rely on.
The operation and the challenge
The client operates an open-cut metallurgical coal mine in the Bowen Basin, with a coal handling and preparation plant (CHPP) serviced by a run-of-mine (ROM) pad and product stockpiles railed to the Port of Hay Point. Like most Queensland coal operations, the site reports stockpile inventory monthly, and that figure feeds production reconciliation, financial statements, and rail nomination planning.
The problem was trust in the number. Month after month, the surveyed stockpile inventory was diverging from the plant's mass-balance reconciliation — feeder-belt weightometer totals reconciled against product despatch and surveyed inventory movement. The gap was consistently in the order of 9%, and it sat well outside the operation's ±3% tolerance. On a combined ROM and product holding of roughly 480,000 tonnes, a 9% error is in the tens of thousands of tonnes — enough to materially misstate inventory in a quarterly report and enough to make the mine planner and the financial controller distrust every survey that landed on their desk.
The incumbent method was a monthly GPS walkover. A two-person crew traversed each accessible pile face on foot with an RTK rover, capturing surface points on a grid and breaklines along crests and toes. On a live coal pad, that approach has three structural weaknesses: the surveyors cannot safely walk steep or actively-built faces, so pile geometry is interpolated across the gaps; coverage is point-by-point, missing the surface undulation between observations; and the survey competes for the same airspace and ground as front-end loaders, dozers and haul trucks, which limits when and where the crew can work. The result was a surface model that systematically smoothed off real volume — and a reconciliation gap nobody could close.
Why this needed a different method
A 9% discrepancy is rarely one error. It is usually several smaller ones stacking in the same direction. Before mobilising, ISS hypothesised three contributors, all consistent with a GPS-walkover workflow on a coal pad:
- Edge definition. The toe-to-ground boundary is the single most error-prone part of any volumetric survey. Coal piles spread, feather and reflow at the base; a sparse walkover grid blurs that boundary and the calculated footprint drifts.
- Inaccessible faces. Actively-built and steep faces cannot be walked, so their geometry was being interpolated rather than measured — a one-directional error that almost always under-captures volume on a heaped pile.
- An assumed base surface. Inventory was being computed against a flat assumed pad level rather than a measured, as-stripped base. On a pad that flexes and accumulates fines over time, an assumed base is a moving target.
Drone photogrammetry directly addresses the first two: it measures the entire visible surface, including faces no surveyor can safely stand on, at a density orders of magnitude higher than a walkover. The third — the base surface — is a discipline problem, not a technology one, and we treated it as such (see below). The aim was not simply to fly the site; it was to remove the structural reasons the number had been wrong.
Equipment and methodology
Ground control and datum
Work was referenced to GDA2020, MGA2020 Zone 55, with heights on AHD to match the operation's existing mine grid and design surfaces. ISS established five permanent ground control points (GCPs) distributed around and through the stockpile pad — coded survey targets on stable, loader-clear ground — observed by RTK off the site's primary control network with a 20-minute occupation per mark. A sixth independent point was held back as a blind check. Network check residuals came in at 14 mm horizontal and 21 mm vertical, comfortably finer than the 2-3 cm the photogrammetric model needed, satisfying the rule that control should be two to three times more accurate than the target survey.
Aerial capture
| Parameter | Value |
|---|---|
| Aircraft | DJI Matrice 350 RTK |
| Sensor | Zenmuse P1, 45 MP full-frame |
| Flight altitude | 95 m AGL |
| Ground sample distance | ~1.2 cm/pixel |
| Overlap | 80% front / 75% side |
| Flight time | 38 minutes, single sortie |
| Piles captured | 14 (ROM + product) |
| Positioning | Network RTK, GCP-verified |
The flight was planned for early morning to minimise loader movement and capture flat, diffuse light — coal is a dark, low-texture surface, and harsh midday shadow degrades photogrammetric matching. All operations were conducted under CASA Part 101 within ISS's Remotely Piloted Aircraft Operator's Certificate (ReOC), by a licensed Remote Pilot, inside the mine's site-specific RPAS procedures and with a ground spotter coordinating loader exclusion zones over the radio. A handheld total station check on three pile crests provided an independent on-site verification before demobilisation.
Processing and the base surface
Imagery was processed in Pix4Dmapper to a dense point cloud and digital surface model, then taken into Propeller Aero for volume computation and client delivery, with cross-checks in 12d Model. The decisive step was the base surface. Rather than carry the assumed pad level, ISS surveyed the as-stripped pad during a low-stock window and locked that measured surface as the fixed base datum for every pile and every future survey. Per-pile bulk densities were applied from the client's own sampling regime — separate values for ROM, washed product and weathered fines — and every density used was stated explicitly in the report rather than buried in an assumption.
The result
The reconciliation gap closed from approximately 9% to under 2% in the first cycle, inside the ±3% target with margin to spare. The drivers were measurable and unsurprising in hindsight: comprehensive surface capture recovered volume on steep and actively-built faces the walkover had been interpolating; high-density edge capture sharpened every toe boundary and corrected the footprint; and the audited base surface removed the systematic offset baked into the assumed pad level.
| Metric | Before (GPS walkover) | After (RTK drone) |
|---|---|---|
| Reconciliation gap vs plant | ~9% | < 2% |
| Field time on pad | ~6 hours, 2 crew | 38 min flight |
| Faces measured directly | Accessible only | All visible surfaces |
| Base surface | Assumed pad level | Audited as-stripped surface |
| Report turnaround | 3-5 days | 24 hours |
| Surface points per pile | Hundreds | Millions |
Just as important, the figures became comparable. Because the base surface and flight plan were now fixed, each month's model sat directly on top of the last, so month-on-month movement reflected real coal movement rather than survey noise. The finance team stopped treating the survey number as a debate and started treating it as an input.
Outcome and ongoing programme
The drone volumetric is now a standing monthly programme. ISS mobilises on a fixed cycle, flies the established plan against the locked base surface and existing GCPs, and delivers a Propeller-hosted 3D model, per-pile volume and tonnage table, applied densities, site photographs and a signed accuracy statement within 24 hours. Because the control network and workflow are established, each repeat survey carries a lower mobilisation overhead than the original — a standard pattern for recurring volumetric contracts.
Beyond the headline reconciliation improvement, the operation gained safety and schedule benefits that are easy to overlook. Surveyors no longer climb live coal faces or share the pad with loaders for hours at a time, removing a recurring working-at-height and mobile-plant interaction risk. Month-end reporting is no longer gated by a multi-day survey turnaround. And when an auditor or the mine planner questions a number, there is a defensible, traceable model behind it — datum, control residuals, base surface, densities and accuracy all documented.
Frequently asked questions
How accurate is a drone volumetric survey on coal stockpiles?
With well-distributed ground control and RTK positioning, drone photogrammetry on stockpiles typically achieves 2-3% volumetric accuracy. Coal is darker and lower in texture than most materials, so capture conditions matter — flat, diffuse light and adequate image overlap are what keep a coal survey at the better end of that range. In this case study, comprehensive surface capture and an audited base surface brought reconciliation against plant throughput to under 2%.
Why was the surveyed inventory 9% out before the drone survey?
The discrepancy was not a single fault. A GPS walkover cannot safely measure steep or actively-built faces, so that geometry was interpolated and under-captured; sparse edge points blurred the toe boundaries and distorted each pile footprint; and inventory was computed against an assumed flat pad level rather than a measured base surface. Each error pushed the same way. Fixing all three together closed the gap.
Why does the base surface matter so much for stockpile volume?
Volume is calculated as the space between the surveyed top surface and a defined base. If the base is wrong, every volume sitting on it is wrong by the same systematic amount. On a coal pad that flexes and accumulates fines, an assumed flat level drifts from reality. Surveying the as-stripped pad once and locking it as a fixed datum removes that offset and makes every subsequent monthly survey directly comparable.
What equipment and standards did ISS use?
A DJI Matrice 350 RTK with a Zenmuse P1 sensor, flown under CASA Part 101 within ISS's ReOC by a licensed Remote Pilot. Control was established in GDA2020 / MGA2020 Zone 55 on AHD heights, processed in Pix4Dmapper with volumes computed in Propeller Aero and cross-checked in 12d Model. A total station provided independent on-site crest checks.
Can the same approach work at other Bowen Basin or Queensland coal sites?
Yes. Multi-pile ROM and product pads at operations across the Bowen Basin — Moranbah, Dysart, Blackwater and the Mackay coal corridor — share the same reconciliation challenge and suit the same workflow. The method scales from a handful of piles to a full mine site of 50-plus, and recurring monthly programmes are the most common engagement once a control network and base surface are established.
Talk to us about your stockpile reconciliation
If your surveyed coal inventory is drifting from plant reconciliation, the cause is usually fixable — and usually a combination of edge definition, inaccessible faces and an assumed base surface, exactly as it was here. Industrial Spatial Solutions delivers CASA-compliant drone volumetric surveys across the Bowen Basin and Queensland, with results referenced to GDA2020 / MGA2020 and AHD, delivered within 24 hours and backed by a documented accuracy statement. Call us on 0407 057 015 to scope your stockpiles and arrange a fixed-price quotation for a one-off survey or a recurring monthly programme.
Industrial Spatial Solutions — volume measured, inventory defensible, reconciliation closed.
Related reading: UAV and aerial surveys, how stockpile volume is calculated, coal mining survey services, drone survey cost in Australia.