Three surveyors, three measurement methods, three different volumes on the same pile — usually within a few percent of each other, occasionally much further. The variance lives in the base-plane assumption, and the only honest fix is to stop assuming.
If you've ever sat through a reconciliation meeting where the production tonnes, the truck-count tonnes, the survey tonnes and the finance tonnes don't agree — and three of those four sit on a stockpile measurement — you've met the base-plane problem.
The fix isn't a better algorithm or a more accurate sensor. The fix is to stop treating the base as an assumption and start treating it as a measurement. This article is the engineering version of how that works, why drone LiDAR is the only practical way to do it at scale, and what to ask before the first capture is booked.
A stockpile is a 3D shape. Its volume is the integral of the surface over its footprint, measured against a reference base. Change the reference base and the volume answer changes — sometimes materially.
The surface itself is straightforward. A drone LiDAR pass captures the pile to ±20 mm vertical at point densities exceeding 300 pt/m² on a typical capture. That part is essentially solved.
The reference base is the loaded question. Three options exist:
Each base gives a different volume from the same captured pile surface. The honest answer to "how much material is in the pile?" requires picking the right base, and that requires having captured it.
Volume V is the trapezoidal integration of the height difference between pile surface S(x,y) and base plane B(x,y) across the footprint:
V = ∫∫ [ S(x,y) − B(x,y) ] dx dy for all (x,y) in footprint
In practice, both surfaces are rasterised at the same cell resolution (typically 50–100 mm), the per-cell difference is multiplied by cell area, and the sum produces the volume.
The math is identical for all three base assumptions. The volume isn't.
Consider a hypothetical 10,000 m³ stockpile sitting on slightly undulating ground:
The spread is 570 m³ — 5.8% of the headline number. On a single pile that's not enormous; on a 200,000-tonne ROM stockpile worth $15M at recovered metal value, 5.8% is real money, and it's the spread between three numbers all calculated honestly from the same captured surface.
Where does the spread come from?
The flat plane is an estimate of the ground beneath. If the ground was undulating (as ground tends to be), the flat plane either passes above the actual ground at the centre (over- estimating volume by including phantom material between the plane and the dipping ground) or below it (under-estimating by deducting phantom voids).
The prior LiDAR ground captures the actual undulations — it's the real surface that was there. Volume against this base is the defensible as-placed number.
The design pad captures the engineered surface the material was placed on. Volume against this base is the as-built number. If the pad has settled, eroded or had remedial work since construction, the design-pad volume is no longer the as-placed volume.
Three legitimate numbers, three legitimate base choices, one material pile. The right answer depends on the question you're asking.
Of the three base options, only one of them — the prior LiDAR ground capture — produces a base that's both defensible and universally applicable. The catch: it has to exist before the pile does.
Once a stockpile is placed, the ground beneath it is buried. No amount of subsequent capture can recover it. You're stuck with either the flat plane (assumed) or the design pad (only available if engineered surveys exist), neither of which is as defensible as the actual surveyed ground.
The cheap, simple thing to do on the first capture of any new stockpile site is to also capture the cleared, prepared ground in any locations where future piles will sit. A 30-minute extra flight, before piles are placed, locks in defensible reconciliation for every subsequent capture for as long as the site is operating.
For sites that already have stockpiles in place, the next-best option is to capture the ground in any cleared corners as a proxy and extrapolate the local terrain shape under the existing piles. Not as good as a true prior capture, but better than flat-plane assumption.
Once the first capture establishes baseline (control + flight plan
The result is a deterministic reconciliation — the same input data produces the same number from the same workflow every time. The argument between production, survey and finance moves from whose number is right to what changed materially since last period, which is the conversation that's actually useful.
A useful side-product of multi-period stockpile capture is back-calibration of in-situ bulk density.
The standard assumption for a given material is published by the geological model: e.g., 1.7 t/m³ for iron ore at typical moisture content. The actual in-situ bulk density depends on placement method, compaction, weathering, moisture and grain-size distribution — and can differ from the regional average by 5–10%.
Over several capture cycles, you have:
Dividing gives an inferred in-situ density. With enough captures, the inferred density converges on a site-specific number more accurate than the regional default. The correction is meaningful for inventory carrying value and for production planning.
This calibration is impossible without consistent geo-referenced volume measurements across periods — which is what the repeatable LiDAR workflow provides and ground-walked GPS surveys do not.
Five questions worth raising at the scoping call:
Do you capture a ground reference before the pile? If the site is new or under construction, the answer should be yes. If it's an existing site, the answer should be a discussion of what proxy can be used.
What base plane will you use, and can I see volumes against all three? The auditable workflow reports the same pile under all three bases simultaneously so the assumption stays visible rather than hidden in a single headline number.
Are you flying to the same flight plan every cycle? Consistent flight plan + same control + same processing = same number from the same data. Anything else introduces unaccounted variance.
Are AHD ground control checkpoints captured as independent marks? The QA story for stockpile reconciliation is the same as any other engineering capture — checkpoint residuals are the only honest accuracy measure.
Will the same processing pipeline run every capture? Algorithm changes between captures can move the volume number without any material having moved. A locked pipeline is what keeps the comparison apples-to-apples.
Any reputable mining-services provider will answer all five without difficulty.
Three patterns we see regularly when reviewing other vendors' work:
Mixed base assumptions across periods. Period 1 uses a flat plane, period 2 uses a prior LiDAR base captured in the meantime, period 3 reverts to flat plane because the prior base was forgotten. The reported "material moved" includes a phantom component from the base-assumption change, not just actual movement.
Captures at different point densities. Period 1 is 200 pt/m², period 2 is 80 pt/m² because the operator was rushing. The period-on-period difference includes resolution-induced variance that has nothing to do with material change.
Sparse breaklines on irregular piles. Stockpiles with steep batters and irregular crests need denser breakline definition in the surface than uniformly-conical piles. Sparse breaklines can chop several percent off the volume by missing local high points. Captures that report ±1.5% volume accuracy are using full surface generation, not breaklines.
The stockpile volume number is only as defensible as the base plane it's measured against. Three legitimate base options exist; only one is universally applicable; only one of those requires capture before the pile is placed.
The minimum-cost change that converts stockpile reporting from a recurring argument into a deterministic number is capturing the ground before the pile is placed and re-using that base for every subsequent reconciliation. The marginal cost is minutes of additional airtime on the first visit. The return is years of defensible volume numbers and tens-of-thousands of dollars of saved dispute time.
If you're scoping a stockpile programme right now and the answer to "what base will you use?" is anything other than a surveyed ground reference — flag it, ask, and budget for the extra flight on capture one.
Tell us the site, the reporting cycle and your base-survey history. We'll scope a capture programme that delivers audit-grade volumes from the first cycle — with the base reference locked in.
Interactive cross-section showing the same pile against flat, prior and design base planes — with live volume readouts and the spread between them.
Whole-of-site capture covering pit, dumps, stockpiles, haul roads and rehabilitation — with the monthly reconciliation workflow detailed.