Volumetric surveys using drones have improved tremendously! Typical stockpile measurements with modern RTK-equipped drones and photogrammetric processing reassure accuracy to within 1-2%, a massive improvement from the 5–10%+ margin of error associated with walking a pile with GPS rovers. And yet there are still plenty of quarry and mine managers that hit the wall come year-end reconciliation time, where the survey says one number and the production ledger says another, and nobody can quite explain why.
The problem, however, is usually not the accuracy of the survey. Something else is.
What "Stockpile Shrinkage" Actually Costs
These are not theoretical financial stakes, they are real. At year-end, a quarry that misjudges a stockpile by 10,000 tons could get hit with a $60,000-$70,000 write-down; in at least one recorded instance tightening up volumetric calculations showed $100,000 worth of material had been left unaccounted. To put this in perspective a construction or aggregate site running around $1 million worth of stock can suffer inaccuracies close to $150,000 when relying on traditional estimation techniques.
Volumetric surveys based on drones really solve much of this. In relation to the 5-10%+ error characteristic of land-based walk-and-measure methods, drone photogrammetry and RTK positioning routinely reduce that down between about 1-2%. That's an actual, quantifiable progress — but it is also where a lot of marketing copy on this topic ends, as though solving the measurement puzzle solves the reconciliation challenge by extension. It doesn't.
Well, an accurate survey doesn't jibe with the books; so what do you do?
This is due to the fact that a given amount of physical material occupies different volumes based on its state, and a drone survey captures one specific projection — not necessarily the projection underpinning your production ledger.
Un-disturbed in the ground is its bank state. When excavated, it turns into loose particulate with air between particles, taking up more space for a given mass — this is the loose state and what is actually being measured by drone survey method over a stockpile. If you then compact that same material — for example, place and roll in a fill — you squeeze it into an additional different third volume (a third, more massively stressed volume). Caterpillar's own reference material explains it this way: holding the same weight of material can take roughly 30% more volume once it's excavated and loose, if the swell for that material is 30%. Your drone survey that returned an accurately polished loose-volume measurement and a bank-volume based ledger can still be in sharp disagreement about the size of their nugget and the problem has nothing to do with the drone!
The Swell Factor, by Material
Keep in mind, this is not a set number—it differs largely by material type — which not only causes confusion with one-size-fits-all converters. For example, a 40% bulking factor means bank-state clay with volume of only 300 cubic meters swells to a blistering 420 cubic meters after excavation — nearly a third more volume for the same mass of material. Typical swell factor ranges used are 10-15% for sand, gravel and crushed stone; 20-30% for topsoil; 20-40% for clay; and as much as 40-65% for blasted rock. So a pile of blasted rock, in other words, can sit loose on the surface at well over one and a half times its original bank volume — a difference that is simply too large to ignore when we're reconciling the amount of material detected by drones with calculated tonnages which assume bank volume.
The real-world effect: the same properly surveyed stockpile may even translate to qualitatively different tonnage numbers depending entirely on which swell factor is applied when converting — and if nobody has verified that factor against the actual material on site, that number is a guess in fancy clothes riding along with a drone survey.
Where the Real Error Starts to Show
But this is where it starts to get a little less discussed: those errors of anything up to ±10% for bulk density, and up to ±33% for swell or shrinkage factor, are all possible if those assumptions haven't been verified against the physical samples of the actual material on site. A theoretical swell factor for "clay" is really only a first guess, not a replacement for measuring the clay on your job site — you can make that number extremely different than the range published in textbooks by altering moisture content and particle size distribution. Highway-engineering guidelines also build in such a margin, recognising that shrink and swell factors for earthwork tend to vary by actually ±33% — even using standard design tables.
The error propagates: a perfectly measured volume, times even 10–15%•density/or swell assumption wrong gives us an incorrect tonnage number that seems precise (even with the precision of a drone survey) but has an inherent error/property nobody accounted for.
So How Do You Actual Reconcile This With The Production Records?
Good reconciliation compares the tonnage surveyed to what an operation's own data systems show as having moved between survey dates (self-sourcing these additional data sources includes fleet management system [FMS[] loader and truck records; weighbridge tickets, conveyor belt scale totals, and shift loading tallies). Each of these sources has its own error potential − the calibration of payload records on FMS must be accurate; weighbridge readings drift without regular calibrations, and belt scales will both wear and shift with moisture through time.
The true fix isn't more precision drone data — the drone side of this is already the most accurate input in the chain. That means, for example, treating the assumptions around swell and bulk density with the same rigour as the survey itself: physically sampling that material every now and then instead of just applying a published range, and then actually stating which volume state — bank, loose or compacted — what number represents before you compare it to anything else.
An Example of How This Looks in Practice
This is exactly the type of work we did for Dutch drone mapping and data engineering company SkyWise Geodata generating heatmap-style volumetric stack analysis across an active facility — output that indicates where material has heaped or thinned across a site (down to individual stockpile zones) using the same underlying topographic survey discipline behind any accurate terrain model. That data is only as good as the assumptions pinned on top of it. This same checkpoint-driven verification discipline present for any drone deliverable — checking outputs against something independent rather than trusting the model on its own — applies just as directly here: the geometry can be exactly right, and tonnage figure built off it completely wrong if the material assumptions underlying it were never verified.
It is the same basic concept that a cut/fill report must be valid — an accurate model you still need, but by itself it does not satisfy since the conversion from volume to a real world number relies on an untested assumption.
Where This Leaves You
Drone volumetrics cracked the measurement issue — at least that piece of the business has really progressed. The reconciliation gap that often appears at year-end is not an indication that the survey is wrong, it tends to be an indication that a swell or bulk density assumption has never been validated against material on site. Next time you immerse into a discrepancy tracing back to the drone data itself, look for the conversion factors first: that's where most of these gaps actually lives.
1 comment
Well done team