Roof Surveys by Drone: What the Client Actually Needs to See

An adjuster in Gresham once sent back a set of 180 drone frames I had delivered with a single line: “which four should I look at?” That was a fair question and a bad delivery on my part. I hold a Part 107 certificate and most of my week is roof and facade survey work for contractors and insurers, and the lesson from that job has shaped every set since. A roof survey is not an archive of everything the aircraft saw. It is an argument, made in the smallest number of frames that can carry it.

The four frames that do the work

Almost every residential roof survey resolves to the same four: a nadir of the whole roof showing plan geometry and the location of every penetration; an oblique of the slope in question at about thirty degrees, close enough that individual courses are distinguishable; a tight frame of the specific defect with a scale reference in it; and a frame of the same defect in context, wide enough that a reader can point to where on the nadir it sits. Everything else in the flight is supporting material. When I deliver, those four lead, numbered and captioned, and the remaining 150 frames go in a folder behind them labeled as the full capture. The adjuster reads four images and knows what happened. If he wants more, it is there.

Resolution is a distance problem, not a camera problem

People buy bigger sensors to see smaller defects and then fly at the same altitude. Ground sample distance is what matters: at 100 feet with a typical 1-inch sensor you are resolving somewhere around a third of an inch per pixel, which is enough to see a lifted shingle tab and not enough to judge whether a nail head has backed out. Cut the altitude in half and you double the resolution. On a Gresham cedar shake roof — mossy, textured, low contrast — I fly the survey grid at 100 feet for the overview and then drop to 30 to 40 feet for the detail obliques on the slopes that matter. Flying low over a roof requires more attention to obstacles, particularly the conifers that crowd lots east of Portland, so I walk the property and note tree heights before launch.

Light, moss and the Pacific Northwest problem

Overcast is normally a gift for roof survey work because it removes the hard shadows that hide damage in a valley. In western Oregon it is also the default, and the trap is that flat light makes moss, algae and genuine granule loss look identical — all three read as dark irregular patches. Low sun raking across the slope separates them: moss has height and casts a shadow, granule loss does not. So I try for the first clear hour after sunrise on any roof where surface condition is the question, and I accept overcast for geometry and penetration mapping. After a windstorm, get up quickly; the difference between a survey flown the morning after and one flown a week later is that the second one has tarps, foot traffic and rain damage layered on top of the original event.

Captions carry more weight than images

An uncaptioned drone frame is nearly worthless six months later. Each of my lead frames gets slope orientation, altitude, capture time, and a plain sentence describing what is visible — not what it means. “North slope, 35 feet AGL, 08:12, three courses of shake displaced approximately 14 inches above the valley flashing” is a caption. “Wind damage” is a conclusion, and it is not mine to draw. I keep the original metadata intact and I do not crop the lead frames, because cropping invites the question of what was outside the edge. The handoff works best when the person receiving it can act on the images without calling me, and a roofing company that reads drone imagery will usually come back with a scope and a question about one specific frame rather than a request for the whole set again.

Frequently asked questions

Does a drone survey replace walking the roof? For steep, wet, or shake roofs it often should — but it cannot lift a shingle or probe soft decking, so some jobs still need boots.

What file format should survey imagery be delivered in? Full-resolution JPEG for review plus the raw files archived. Keep GPS and timestamp metadata intact.

How many frames is a typical residential survey? Between 80 and 200 captured, of which 10 to 20 are delivered as the working set and four lead it.