Key takeaways
- Snow is not a thin layer over the truth. It changes both surfaces a volume depends on — the top of the pile and the ground it sits on — and it does not sit evenly on either.
- Fresh, smooth snow gives photogrammetry almost nothing to match on. The result is not a slightly noisier surface; it is holes, warped patches and a model that looks plausible and is not.
- The strongest winter move is to stop asking the winter flight to find the ground. Capture a snow-free base surface in autumn, reuse it, and let the winter flight measure only the top of the pile.
- Decide in writing, before the flight, how snow will be handled in the reported volume. An inventory number that quietly includes a season's snow is the kind of error that is only found during an audit.
Year-end inventory in Alberta lands in the middle of winter, which means a lot of stockpile counts and site surveys get flown over snow. The flying part is routine — cold-weather drone operations are a solved problem with a defined envelope. The measuring part is not, and it is where winter surveys quietly go wrong.
The problem is easy to state. A volume is the difference between two surfaces: the top of the material and the base it sits on. Snow interferes with both of them, and it does not interfere evenly.
Why snow breaks photogrammetry
Drone photogrammetry works by finding the same physical point in several overlapping photos and triangulating where it must be. That requires the point to be visually distinguishable from its neighbours. Gravel, soil, rock and vegetation are covered in distinguishable points. Fresh, wind-smoothed snow is not — it is a large area of near-identical white, and the matching step has nothing to lock onto.
The failure mode matters more than the fact of it. You do not get an obviously blank result that screams for attention. You get a surface with holes in it where matching failed outright, gently warped patches where it matched the wrong points, and noise that the software smooths into something that looks like a perfectly reasonable pile. It is a confident-looking model built on very little evidence, and nothing in the deliverable announces that.
Two secondary effects compound it. Snow is bright enough to push a camera into clipping, and a blown-out highlight carries no detail at all no matter how the model is processed. And flat overcast light over snow removes the shadows that give the remaining micro-relief what little contrast it had, which is the opposite of the intuition that soft light is always kinder.
What helps: capture with deliberate underexposure so the snow holds detail rather than clipping; fly when a low sun angle is raking across the surface and casting relief; and prefer a pile that has been worked recently, because a freshly cut face of exposed material is full of the texture the process needs.
Where LiDAR is different, and where it is not
LiDAR does not care about visual texture. It emits its own pulses and measures their return, so a uniform white surface is no harder for it than a gravel one. That removes the matching problem entirely, and it is the main reason winter capture over snow is a legitimate use case for LiDAR rather than photogrammetry.
What LiDAR does not do is see through the snow. The return comes off the snow surface, so you still measure a snow-covered pile rather than the material inside it. Wet, dense snow also absorbs more of the pulse than dry snow does, which affects return quality. LiDAR solves the measurement problem in winter; it does not solve the snow volume problem, and those are separate.
The pile measures bigger than the material
A snow-covered pile has a coat on. The measured top surface is the top of the snow, so the computed volume includes the snow as though it were product.
The instinct is to subtract an average snow depth. That fails because snow does not distribute evenly on a stockpile — it is the one place it distributes least evenly. Wind scours the crest and the windward face and drifts deep against the lee face and the toe. On an active pile, the worked face may be nearly bare while the untouched end carries a season's accumulation. A single average depth applied across that geometry moves material from one part of the calculation to another rather than removing it.
It also compounds between flights. Compare a January capture to a November one and the change you report includes the change in snow cover, which is a weather signal, not an inventory signal.
The base surface is the bigger problem
Volumes are computed against a defined base surface — the ground the pile sits on. In winter that ground is under snow too, and drifted, so a base surface derived from a winter flight is not the pad. It is the snow around the pile, which in a yard that gets ploughed is not even a natural snow surface — it is windrowed banks where equipment pushed it.
Now the error works in both directions at once: snow on top adds material, snow on the pad raises the floor and removes it. The two do not cancel in any controlled way, and there is no defensible way to reason about the residual.
This is why the single strongest winter practice is not a correction factor. It is to stop asking the winter flight to find the ground.
What to do instead
- Capture and keep a snow-free base surface. Fly the yard in autumn, before persistent snow, and store that base surface as an asset. Winter flights then measure only the top of the pile and are differenced against the known pad. This is the change that does the most work, and it costs one autumn flight.
- Schedule the reconciliation flight before persistent cover. If a number has to be defensible — financial reporting, contract payment, an audited count — put the flight in late autumn and treat winter captures as operational tracking rather than the number of record.
- Fly the working window. A pile that has been pushed, cut or loaded in the last few days has exposed material and a fresh face. That is both better texture and less accumulated snow.
- Use LiDAR where the schedule is fixed. If the count has to happen in February, an active sensor removes the texture problem and leaves you with only the snow-thickness question to manage.
- State the snow condition in the deliverable. Date, cover, whether the pile had been recently worked, whether the base surface came from this flight or a stored snow-free one. A volume with its conditions attached can be defended a year later. A bare number cannot.
Sites, not just stockpiles
The same logic applies to topographic capture, and the consequences are often worse because the deliverable feeds a design.
A winter surface model is a model of the snow. On open, level ground with thin, even cover the difference may be tolerable for a rough planning surface. On anything with relief, ditches, cut slopes or vegetation it is not — drifting fills the low ground preferentially, so ditches and swales measure shallower than they are, and that is precisely the geometry a drainage or earthworks design depends on.
For design-grade terrain, the honest answers are to wait for bare ground, to take critical points by conventional ground survey and use the aerial capture for context, or to accept the surface explicitly as a snow surface and label it that way in the deliverable. What should not happen is a winter surface quietly entering a design model as though it were ground.
Progress documentation, asset condition, inspection and imagery work all carry on through winter unchanged. It is measurement against a ground surface that needs the extra thought.
Put it in the scope
Almost every winter dispute we see is the result of something that was never written down. Before a winter measurement flight, settle in writing: whether the base surface comes from this flight or a stored snow-free capture, whether any snow allowance is being applied and on what basis, what the comparison capture is, and what the accuracy statement is being made against. Our drone survey RFP guide covers the rest of the clause list, and ground control points matter more in winter than in summer, because targets need to stay visible and findable on snow.
Winter does not stop the work. It changes what the flight can honestly claim to have measured, and the job is to make that claim precise rather than optimistic.