Five years ago, drone surveying at most Australian mines meant a contractor flying a DJI Phantom over a stockpile once a quarter and producing a volumetric report a week later. Today, the better operations have permanent fleets in hangars on site, automated flights running daily, and orthomosaic outputs feeding into pit design software within hours. The gap between leaders and laggards has widened sharply, and it’s no longer just a tier-one versus mid-tier story.

I spent April visiting three operations across the Pilbara and Goldfields and came away with a clearer picture of where this technology has actually settled. The headline: the engineering case is well-proven, but the regulatory, IT, and skills picture is messier than vendors will admit.

What’s working

Stockpile volumetrics is the obvious win. A site I visited near Kalgoorlie has moved from monthly truck-based surveys to weekly drone runs, and the variance against load-in/load-out reconciliation has tightened from roughly 4% to under 1.5%. That difference matters when you’re talking about high-grade gold ore or lithium concentrate stockpiles where a percent or two of volume can move millions of dollars in inventory accuracy.

Highwall monitoring is the second area where adoption has accelerated. Photogrammetric change detection between flights, when properly post-processed, will pick up displacements down to a couple of centimetres. That’s nowhere near the precision of slope stability radar, but it covers areas that radar doesn’t — and at a fraction of the cost. Several operations now run drone-based highwall inspection as a complement, not a replacement, to ground-based monitoring.

Tailings dam surface inspection has also become routine, particularly post-Brumadinho when board-level scrutiny got serious. A weekly drone flight with multispectral imagery picks up seepage, vegetation stress, and surface deformation patterns that human inspectors miss. ABC reported in February that several Queensland coal operations had reduced their on-foot dam inspection hours by over 60% after standardising on drone-based programs.

Where the friction sits

CASA’s Part 101 framework is workable but the BVLOS (beyond visual line of sight) approvals process is still slow and operation-specific. Sites that want to run fully autonomous flights from a fixed dock without a visual observer present need an instrument approval that takes 6–9 months in my experience. A few of the larger operators have negotiated site-wide BVLOS approvals, but for mid-tier miners it remains a significant barrier.

The bigger problem, honestly, is data management. A modern survey-grade drone flight produces 80–200GB of imagery per hectare of detailed coverage. Multiply that across a typical open pit operation flying daily, and you’re generating tens of terabytes a week. Most sites I’ve visited haven’t thought through the storage architecture. They’re filling up local NAS units, struggling with bandwidth back to corporate networks, and producing photogrammetric outputs that take 12–18 hours to render on undersized workstations.

Operations that have done this well — Rio Tinto’s Western Range development, BHP’s Olympic Dam — have built proper edge-to-cloud pipelines with automated processing in Azure or AWS. That’s where firms doing Microsoft Fabric consulting work have started to find genuine demand in the resources sector. Getting drone telemetry, photogrammetric output, and operational data joined up in one analytics environment is harder than the vendor demos suggest.

The skills problem

Finding licensed RePL (Remote Pilot Licence) operators isn’t difficult. Finding people who can do RePL flights AND understand mine surveying AND interpret photogrammetric outputs AND troubleshoot the processing pipeline is genuinely hard. A senior surveyor I’ve known for fifteen years told me in March he’s been offered three jobs in the past year specifically because he sits in that overlap.

Most graduate surveyors coming out of Curtin or UNSW have some drone exposure now, but the depth varies wildly. The TAFE-level RePL training has improved but doesn’t cover mining context. Site training programs are inconsistent, and the better ones are essentially internal IP that companies don’t share.

Vendor landscape

The hardware market has consolidated. DJI’s Matrice series and Wingtra’s fixed-wing platforms cover probably 80% of Australian mine deployments. Sensefly went away. Quantum-Systems is making inroads on the German engineering side. Hovermap by Emesent (Australian, Brisbane-based) has carved out the underground LiDAR niche almost entirely.

Software is more fragmented. Pix4D, Agisoft Metashape, and Bentley’s ContextCapture each have meaningful share. Propeller Aero has done well in the volumetric-reporting workflow specifically — they’ve built a useable interface that non-surveyors can pull stockpile reports from. Their integration with truck weightometer data is genuinely good.

What I’d push for in 2026

The regulatory ask is straightforward: CASA needs a faster path for site-bounded BVLOS operations at gazetted mining tenements. The current approvals process treats every site as a unique case, which doesn’t reflect operational reality.

The industry ask is harder: standardised data formats and metadata for survey deliverables. Right now every site stores point clouds, orthomosaics, and DTMs slightly differently, and merging across operations or sharing with consultants is painful. AusIMM’s Surveying Society has been talking about this for two years without much progress.

Drone surveying isn’t novel anymore. The question is whether sites are going to actually run it as a discipline with proper data engineering, or keep treating it as a quarterly contractor visit dressed up in better hardware.