Field note · Position · 14 October 2026
A drone record sits under a signed survey, not in place of one
A registered surveyor commented under one of my posts that flying the drone is the easy part, and that knowing, proving and taking responsibility for accuracy is what the surveying profession is for. He is right. He also caught a real mistake on this website. Here is what I sell, what I don't, and how the two fit on the same site.
90% of hard ground agreed within this between two visits 24 hours apart. Agreement with itself, not with a survey.
Surveyed marks held back from processing and checked afterwards. The study that would measure accuracy hasn't been run yet.
How far the drone's own height reference moved overnight. Why I don't sell levels.
A tape-measured box, read off the map. Its height under-read and its volume over-read 2.7×.
The objection, and why I agree with it
In August I posted a difference map: the same site flown twice, a day apart, with nothing moved in between, so everything left over was error. Most of the replies were from surveyors. The sharpest was from a registered surveyor with a mining endorsement.
His point was that ground sample distance, RTK and a dense point cloud don't prove accuracy. Accuracy is proven with independent control, check points, quality checks and statistics, and in his world those datasets feed volumes, reconciliations and statutory plans. Then:
"Flying the drone is the easy part. Knowing, proving and taking responsibility for the accuracy of the resulting survey is where the surveying profession comes in."
I could have argued with that. I didn't, because there was nothing to argue with. It describes the line I already draw around my own work: no RTK, no certified levels, no setting out, no audited stock figures. What I fly is a record of change, and a record of change belongs underneath a signed survey. It doesn't replace one.
His comment also caught a real error. This website was calling one of my numbers an "accuracy statement", and it wasn't one.
Accuracy and repeatability are different numbers
The figure I publish, 90% of bare ground within 0.24 m, comes from flying the same site twice and comparing the two results. That measures repeatability: how well the method agrees with itself from one visit to the next.
Accuracy is agreement with the real ground, in a coordinate system other people use. You can only measure it against marks that were surveyed independently and held back from processing, so the software never had a chance to fit to them. I haven't run that study yet.
Two flights can agree with each other to five centimetres and both sit forty centimetres off the true ground. Repeatability can't see that. Only independent checkpoints can.
So the number is now labelled repeatability everywhere it appears: on this site, in the sample pack and in the documents I hand over. It's still the right number for what I sell, which is whether a site changed between visits. "Accuracy" was the wrong word for it.
For scale, HELIX Innovations, who fly with surveyed ground control, replied in the same thread. On a construction site flown monthly for over a year, with the same control and the same method, their flights differed by up to 100 mm. They kindly agreed to be quoted. Their figure and mine are different statistics on different sites, so don't read them as a head-to-head. But it shows what ground control buys, and that even controlled repeat flights don't land in exactly the same place.
What each one is for
A signed survey
- Control tied to the national grid and to Ordnance Datum
- Levels a contractor can set out and build to
- Volumes of record: valuations, reconciliations, year-end stock, statutory plans
- A stated uncertainty, checked against independent marks
- A professional who puts their name, and their indemnity, behind it
My monthly record
- The same saved mission, flown from the same marked take-off spot every visit
- Each visit aligned to the first, so the months compare like for like
- An orthomosaic: where things are, and their size in plan
- What changed since the first visit, measured relative to it
- Labelled stills, on your desk by the third working day
The drone isn't what decides which column a job falls in. The question is whether someone will rely on an absolute number, a level, a coordinate or a certified volume, for money, for setting out, or in a dispute. If they will, that's a surveyor's job. My drone's own idea of height moved 27.78 m overnight with the weather. Alignment removes that between visits, but nothing I do turns it into a level.
How the two fit on one site
A survey is a precise snapshot taken at chosen moments, usually key stages or when a number has to be certified. Between those moments there are weeks when nobody records anything. A monthly record covers those weeks: the same view, flown the same way, showing what moved.
- The survey sets the reference. Levels, control and certified quantities come from the surveyor, and stay theirs.
- The record covers the months in between. What was built, cleared, stockpiled or moved, seen in plan and compared with the first visit.
- The record says where to look. If a change matters, the surveyor goes and measures it properly. Nobody prices a valuation off my map.
The record also has a size limit, and you should know it before relying on it. On a tape-measured 0.5 × 0.3 × 0.3 m box, the plan area off the orthomosaic came within 4%. The height off the elevation model read 0.20 m against a true 0.30, and the volume over-read 2.7×. Automated change detection, with nobody looking, picks things up at about the size of a car and not below. The full numbers, including the results that went badly, are in the repeatability study.
Four questions for any drone operator
- "Is that figure accuracy or repeatability?" If they can't say which, it's probably neither. It's likely the pixel size.
- "Checked against what?" An accuracy figure needs surveyed marks that were held back from processing. Ground control the software fitted to doesn't count as a check.
- "Who signs it?" If you'll rely on a level or a volume, someone has to put their name to it. That person is a surveyor.
- "What's the smallest change you can see?" A good answer comes with a measured object behind it, not a camera specification.
Limits, stated plainly
- The 0.24 m figure is repeatability, measured on one site over two dry summer mornings. It says nothing about agreement with Ordnance Survey coordinates or heights.
- No independent checkpoint study has been run. Until it has, I don't use the word accuracy about my own work.
- Doming is unquantified. Flights pointing straight down can bend a shallow bowl or dome into the surface, worst in the middle of a site. Cross-hatched lines help, but nobody has measured how much is left.
- The box test is one small object. The edge effects that inflate its volume should matter less on a large stockpile. That hasn't been measured, so I don't assume it.
Figures last updated 14 October 2026.