Key takeaways
- Smoke is two separate problems. There is an airspace restriction around an active fire that is not negotiable, and there is a data-quality problem from haze hundreds of kilometres away from any fire at all.
- The dangerous case is not the day too smoky to fly. It is the moderately smoky day that produces a full, ordinary-looking deliverable built on weak imagery, because nothing in the file announces the degradation.
- Smoke hits capture types unevenly. Solar thermal and multispectral index work are the first to become invalid, photogrammetry degrades gradually, and close-range visual inspection is often barely affected.
- The fix is contractual, not technical. Book summer capture against a window with stated flexibility and write down in advance who calls a re-fly and who pays for it.
Every Alberta summer now includes stretches where the sky goes flat orange and stays there for days. For anyone buying aerial data between May and September, that raises a practical question that has nothing to do with whether the aircraft can take off: on a smoky day, is the survey you paid for still the survey you are going to get?
The honest answer is that it depends entirely on what is being measured, and the difference between capture types is much larger than most buyers expect.
First, the part that is not a judgement call
There is a hard airspace restriction around an active forest fire in Canada. Aircraft, including drones, are kept out of the area over and immediately around a fire below a set altitude unless they have specific authorisation, because that airspace belongs to the water bombers, birddog aircraft and helicopters working the fire. This is enforced, and it exists for an obvious reason: an uncrewed aircraft in that block can ground an entire air attack.
This restriction is geographic and narrow. It applies near the fire itself. It has nothing to do with the smoke layer, which routinely drifts hundreds of kilometres and sits over ground with no fire activity anywhere near it. Most smoke-affected survey work in Alberta happens far outside any restricted area, which is exactly why the second problem catches people out. The airspace question has a clean yes or no answer. The data question does not.
What smoke actually does to an aerial image
Smoke is a suspension of fine particles between your camera and the ground, and it does three things at once.
It scatters light on the way down and on the way back up, which lifts the darkest parts of the scene and pulls down the brightest. The whole image compresses toward the middle. That reads as haze to a person, but to a photogrammetry engine it reads as lost local contrast, and local contrast is the raw material the whole process runs on. Matching the same physical point across overlapping frames gets harder as the point gets less distinguishable from its neighbours.
It shifts colour. Smoke scatters short wavelengths much more strongly than long ones, which is why a smoky sky is orange, and it means the light reaching your subject is no longer the light your camera profile assumes. Visually this is a nuisance. For anything that treats pixel values as measurements rather than pictures, it is a fault.
And it changes the light itself from directional to diffuse. Direct sun casts the shadows and highlights that give terrain and surfaces their visible relief. Heavy smoke replaces that with soft, shadowless light from everywhere, which flattens exactly the micro-relief a matching algorithm was using.
The failure mode is a plausible deliverable, not an obvious one
If it were bad enough to fail outright, smoke would be an easy problem. Photogrammetry rarely fails that way. As the imagery weakens, the software finds fewer reliable matches, fills the gaps by interpolating between the matches it trusts, and hands back a surface that is smooth, complete and confident looking. The holes are filled with assumption rather than evidence, and the deliverable does not say which parts are which.
Which is why the moderately smoky day is worse than the badly smoky one. Nobody accepts a survey they could not fly. Plenty of people accept a survey that looks exactly like every other survey they have received and was built on a fraction of the usual evidence. If a volume, a contour set or a design surface is coming off that model, the error is now inside the number and nothing on the page flags it.
Smoke does not hit every capture type equally
Solar thermal inspection: stop first
A solar array thermal survey depends on the array being driven hard by sunlight, because the faults being looked for only show up as a temperature difference when current is flowing and heat is being produced. Smoke cuts the irradiance reaching the panels, and below a working threshold the faults simply stop being visible. As we covered in the conditions that make or break a thermal inspection, a survey flown in the wrong conditions is not a weaker result, it is an uninterpretable one. A smoke-flattened array that reads clean is the worst possible outcome, because it gets filed as proof the array is healthy.
Multispectral and vegetation index work: stop or recalibrate
An NDVI or other vegetation index is a ratio computed from calibrated reflectance, which means the sensor has to know what the incoming light was. That is what the calibration panel and the incident light sensor are for. Smoke changes both the intensity and the spectral shape of that incoming light, and a drifting smoke layer changes it during the flight. Values captured under smoke are not comparable to values captured the week before under clear sky, which quietly destroys the thing index work is most useful for: watching a change over time. If the point of the flight is a season-long comparison, a smoke-day capture is worse than a missing week.
Mapping and volumes: degraded, sometimes usable
Photogrammetric mapping sits in the middle and depends on what is being flown. A working gravel pit with freshly cut faces, sharp equipment tracks and high-contrast material is full of matchable texture and survives light haze reasonably well. Bare uniform ground, dense uniform crop, fresh asphalt or a large water-adjacent site was marginal for texture on a clear day and goes first. Where a site has the texture to spare, flying lower to recover detail is a real option, at the cost of more flight lines and more time.
Close-range visual and structural inspection: usually fine
Roof, structure, tower, flare stack and facility inspections put the camera within metres of the subject. Whatever smoke is between the aircraft and the asset at that range is negligible compared to what sits between a mapping aircraft and the ground several tens of metres below. Confined space and internal inspections are unaffected entirely. If a smoke week is going to cost you a mapping window, this is the work to pull forward into it.
Building envelope and equipment thermal: usually fine
Unlike solar work, envelope thermography and electrical or mechanical inspection do not need sunlight. They need a temperature difference across a wall, or an asset under normal operating load. Smoke does not remove either. Long-wave infrared also penetrates haze considerably better than visible light does. These jobs generally keep running.
Scheduling around a season you cannot forecast
Smoke cannot be planned against the way a snow date or a growth stage can. It arrives with a wind shift, it can clear in an afternoon, and the multi-day forecast for it is weak. Three things actually help.
Book a window, not a date. Any summer capture that depends on image quality should be scoped with a stated flexibility window rather than a fixed day, and the window should be agreed when the job is quoted rather than negotiated on a bad morning.
Front-load the season. Baseline captures, the ones later flights will be compared against, are worth getting in the ground early rather than letting them drift into August. A missing baseline costs more than a missing update.
Judge the morning, not the forecast. Smoke often thins overnight and rebuilds through the afternoon as the day heats up, so an early window is frequently the usable one. The call is made on the ground against what the site looks like, not on a regional index from the day before.
What to put in the scope before summer
The contractual side matters more than the technical side here, because the technical answers are known and the disputes are always about who absorbs a wasted day.
- State the conditions under which the operator will stand down rather than fly, and state them before the season, not on the morning.
- Say who makes that call. An operator who will not fly a survey they cannot stand behind is protecting your data, and that only works if the authority is written down.
- Decide in advance who pays for a re-fly when a flight is completed and the data turns out to be unusable, which is a different case from a stand-down.
- Require that capture conditions are recorded in the deliverable. A flight log that notes visibility and sky condition is what lets a future reader tell a smoke-degraded model from a clean one.
- For any index or change-detection program, state that captures flown under materially different atmospheric conditions are flagged rather than silently compared.
The bottom line
Smoke season does not stop commercial drone work in Alberta, and an operator who tells you it does is either flying one narrow kind of job or has not thought about it. What it does is make the operator's judgement the most valuable part of the service. Anyone can put an aircraft up on a smoky day. Knowing which of your jobs is still valid that day, which one is quietly about to produce a confident, wrong number, and which one to reschedule into next Tuesday is the part worth paying for.