Setting a Site Up Once So Every Later Capture Is Comparable

Published 2026-11-05 · UAV Imaging Inc.

Key takeaways

Accuracy and comparability are different problems. A series of individually accurate flights can still produce a volume series you cannot trust, because each flight was accurate to its own frame.
A monitoring series needs a frame that does not move. That means physical monuments, one named datum, a held base coordinate and a frozen processing recipe, decided once at the start.
Monuments go outside the work area, not in it. Control inside an active pit, fill or laydown area gets buried, pushed or paved, and the series loses its reference exactly when it has become valuable.
A base station on an autonomous position is the most common comparability killer. Each visit the base derives a slightly different coordinate, and the whole dataset shifts with it.
Datum, epoch and geoid model have to be written down, not assumed. A geoid model upgrade between captures produces a smooth vertical offset that reads as material movement across the whole site.
Freeze the base surface and the boundaries too. A volume series is only comparable if every term was measured above the same surface inside the same footprint.
The deliverable is a control sheet, not a conversation. Monument descriptions, photos, coordinates, datum, epoch, base point and processing settings in one document that outlives the people involved.
This is what makes a vendor change survivable. A new contractor handed a real control sheet continues the series. One handed nothing starts a new one and the step shows up as fake movement.

Most of what gets written about ground control answers a single-flight question: how many points does this flight need, and where do they go, to hit the accuracy the job requires. That question has a decent answer and we have written about it separately.

It is also the wrong question for a site that will be flown again. A pit measured quarterly, a landfill cell reported annually, a construction site captured monthly for progress payment, a containment facility with a regulator-facing record: on all of these the number that matters is not any single volume, it is the difference between captures. And a difference inherits the errors of both terms plus every change of method between them.

That produces the failure mode behind most of the strange volume series we get asked to explain. Every individual flight was competently done and defensibly accurate. Each one was accurate to its own frame. The series is still unusable, because the frame moved underneath it.

Accuracy to a specification, versus comparability across a series

Take a site flown four times a year by a careful contractor who hits a tight accuracy specification every time. If each visit establishes control independently, each capture sits in a slightly different position relative to the others. The flights are all accurate. The differences between them contain a mix of real material movement and the drift of the frame.

The uncomfortable part is the arithmetic. A frame offset affects a volume in proportion to the area it is applied over, so a vertical shift that is tiny against any accuracy specification turns into a large and entirely fictional volume once it is spread across a big footprint. That is why a series can look wrong in a way no single flight does, and why checking the accuracy of the latest flight never finds it.

The fix is not a better flight. It is a frame that is established once and then held.

What a site setup actually consists of

Five components. Each is ordinary work, and the value is in doing them together and once.

1. Physical monuments outside the work area

Permanent marks, placed deliberately, that will still be there and still be in the same place in five years. The practical requirements:

Outside the active area. On the perimeter, on undisturbed ground, outside the ultimate pit or fill limit and off the haul roads and laydown areas. Control placed where work will reach is control that will be destroyed, and it always gets destroyed in the middle of the series rather than at the end.
Set to survive frost. On Alberta sites, seasonal heave is the quiet killer of a vertical series. A mark set shallow in fine-grained soil will move between a December capture and a June one. Set below the local frost depth or in competent material, and treat any monument that has heaved as suspect rather than as evidence.
Visible from the air and identifiable on the ground. A mark that cannot be found in imagery does not serve a photogrammetric flight, and a mark that cannot be found by a crew on foot does not serve a check survey. Both matter.
Distributed, not clustered. Spread around and across the area of interest, including at the elevation extremes. Control all at one end leaves the far end unconstrained and a tilt in the surface can go undetected.
Enough of them to lose some. Redundancy is the whole point. If a site has four monuments and loses one, the remaining three cannot tell you which of them moved. Six or eight makes the series robust to attrition and lets a disturbed monument be identified and retired rather than quietly corrupting a capture.
Documented physically. A photo of each mark, a photo of its surroundings, a written description of how to find it, and a sketch. This is the part everyone skips and the part a crew in year four actually needs.

2. One coordinate set, from one control survey, with the frame named

The monuments get coordinates once, from a single properly designed control survey rather than from whichever flight happened to be first. What has to be written down alongside the numbers:

The horizontal coordinate system and, specifically, which realisation of the reference frame.
The epoch of the coordinates.
The vertical datum.
The geoid model, by name and version, if orthometric heights are in use.
The projection and any scale factor or grid-to-ground handling.
How the coordinates were established and the accuracy achieved.

Every one of those items has caused a real-world problem. The geoid model is the most reliably destructive, because a model change between captures introduces a smooth, slowly varying vertical offset across a site. Smooth offsets are exactly what material movement is not, so a difference map of the whole site is usually what reveals it, and two volume numbers never will.

Whether a site needs a professional land surveyor to establish this control is a separate question with its own answer, and it depends on what the data will be used for. For a measurement series feeding internal decisions, a well-executed control survey is what is needed. For anything that touches legal boundaries or gets filed in certain regulatory contexts, the sign-off question comes first.

3. A fixed base station point with a held coordinate

This is the single highest-value item on the list and the one most often missing. If the site uses an RTK or PPK workflow with its own base station, the base should set up on the same physical point every visit, and that point should have a coordinate that was determined once and is then entered, not re-derived.

The alternative, which is extremely common, is a base station that computes its own position each visit. Every one of those positions is a little different, and the entire capture is rigidly tied to it. The result is that each visit's dataset is internally excellent and externally shifted by a different amount. Nothing in the flight's own accuracy check detects this, because the check is internal.

So: a base point marked as a monument like the others, a coordinate established in the control survey, the datum and epoch of that coordinate written down, and a standing instruction that the base is held on it. Where a network correction service is used instead of an on-site base, the same discipline applies to the service, the mount point and the frame it delivers, and a change of provider is a change of frame.

4. Frozen base surfaces, boundaries and a processing recipe

Control fixes where the data sits. It does not fix what gets measured. A volume series needs the measurement definition frozen too:

The base surface, as a file, with a note on how it was derived. A captured pre-disturbance surface where one exists, a surveyed pad, or a defined toe surface. Not a fresh judgement each visit.
The boundaries, as files, one per stockpile, cell, pay area or reporting unit. A perimeter redrawn by eye each quarter adds and removes a thin layer around a long edge, and the effect is larger than it sounds.
The processing settings, held as a template rather than reconstructed. Same software and version, same density and filtering choices, same surface generation method, same classification approach where LiDAR is in use.
The flight plan, as a saved mission. Same altitude, same overlap, same sensor and lens where possible. Comparable inputs produce comparable outputs.
A note of what changed, when it changes. Something will eventually: a sensor retires, a software version is forced. Record it on the capture it first applied to, so a step in the series has a documented cause instead of becoming an argument.

5. A control sheet that outlives everyone

All of the above exists in one document, held by the site owner, delivered to every contractor who flies the site. Monument table with coordinates, datum, epoch and geoid model. Monument photos and descriptions. The base point and its held coordinate. Base surface and boundary files. The processing template and flight plan. A revision history. A log of each capture with its date, who flew it, and anything that deviated.

The test of a control sheet is whether a contractor who has never seen the site can read it and produce a capture that drops cleanly into the existing series. If answering that requires a phone call with the person who flew it last year, the sheet is not finished, and it will fail in the year that person is unavailable.

What this costs and what it returns

Honestly: it is a real first-visit cost. Setting monuments, running a control survey over them, documenting it and building the sheet is more work than turning up and flying. On a site that will be captured once, it is not worth doing, and we will say so.

It pays back in three places. Later captures get cheaper, because established control means less field work per visit. Differences become trustworthy, which is the entire reason for the series. And the series survives disruption, specifically a change of contractor, a change of your own staff, or a gap of a year or two, any of which otherwise resets the record.

The decision rule we use: if a site will be flown three or more times, or if any single capture will be differenced against another, set the frame up at the start. The threshold is not about site size, it is about whether anyone will ever subtract one capture from another.

Doing it late

Most sites asking this question are already several captures in. The position is recoverable but partially:

Set the monuments and run the control survey now. The series from this point forward is comparable, which is better than continuing without it.
Earlier captures can sometimes be brought into the new frame, if their original control or base coordinates and datum were recorded. This is where the value of asking a previous contractor for their control data shows up, and it is much easier while you are still working with them.
Captures with no recorded frame are a historical record rather than a measurement series. Treat them as imagery, not as the earlier term in a volume calculation.
Mark the transition explicitly in the record, so the step between the old and new frames is documented as a method change rather than being read later as material that appeared or disappeared.

If you are running a repeating capture program on an Alberta site and have not set the frame, that is the thing to fix before the next flight rather than after it. We will quote the setup separately from the capture so you can see what it costs, and we will tell you if your site does not need it.

Frequently Asked Questions

How is this different from asking how many ground control points a flight needs?

That question is about the accuracy of one capture, and the answer is driven by the specification, the site and the method. This is about whether twenty captures can be compared to each other. A site can satisfy the per-flight question perfectly on every visit and still have an unusable volume series, because each flight was accurate to a frame that was established independently. Comparability needs the frame fixed once and held.

Can we not just use the same contractor every time and rely on them?

It helps and it is not sufficient. A single contractor still has to be running a fixed base point, a stated datum and epoch, a frozen base surface and a saved processing template, or the frame drifts between their own visits. And relying on a contractor's internal memory concentrates the risk in one company and often in one person. The control sheet exists so the site owner holds the frame rather than renting it.

How many monuments does a site need?

Enough to stay solvable after losing some, which in practice means more than the minimum. Four monuments on a site that loses one leaves you unable to tell which of the remaining three moved. Six to eight, spread around and across the area of interest and including the elevation extremes, is a reasonable starting point for a working site, and the right number depends on the size, the shape and how much of the ground is going to be disturbed.

What happens if a monument gets destroyed or heaves?

With adequate redundancy, you detect it: the disturbed monument stops fitting the others, the others still agree, and the bad one is retired and replaced. With minimal control you cannot tell which mark moved, and the error distributes silently into the surface. This is the main argument for both redundancy and for placing marks outside the work area and below frost depth, and it is why heave on a monument should be treated as suspicion rather than as a measurement.

Does the geoid model really matter?

On a vertical series, yes, and it is one of the more common causes of a volume step that looks like movement. A change of geoid model between captures introduces a smooth, gently varying vertical offset across the whole site. Because it is smooth it looks nothing like material movement, which is how it gets diagnosed, but it has to be looked for in a difference map and no pair of volume numbers will reveal it. Write the model name and version into the control sheet and keep it fixed.

We are changing drone contractors. What do we hand the new one?

The control sheet, in full: monument coordinates with the datum, epoch and geoid model named, monument photos and descriptions, the base point and its held coordinate, the base surface and boundary files, and the processing and flight templates. Handed that, a competent new contractor continues the existing series. Handed a folder of past deliverables and a verbal briefing, they will start a new one, and the step between the two will be read as material movement by whoever looks at the trend next year.

Book a commercial drone flightMapping, inspection, multispectral and thermal across Alberta.Request a quote →or call 587-532-9000