How many stereo passes does a canopy height model actually need?

Ask a photogrammetrist how much overlap a stereo job needs and you'll usually get a bare-earth answer: 60% forward, 30% side, done. Ask the same question about a timber stand or a carbon project polygon and the answer changes, because crowns aren't flat ground.

Forward overlap, side overlap, and why canopy breaks the default numbers

Forward overlap is how much each frame along the flight line shares with the next one. Side overlap is the same idea across adjacent flight lines. The classic 60/30 split works fine over open ground, where every pixel has one clean, unobstructed return. A forest canopy doesn't offer that. Crowns overlap each other, branches cast shadow into neighboring gaps, and a convergent look angle that's perfectly fine over a parking lot can end up staring straight into the side of a crown instead of its top.

That's the matching problem stereo photogrammetry has to solve on every pixel: find the same point on two (or more) images taken from different positions, and triangulate height from the parallax between them. Over bare ground, two looks almost always find a match. Over a closed canopy, a single stereo pair can lose the match entirely in deep shadow or where one look angle sees a gap the other doesn't. The usual fix is to add a third look, straight down the middle, so the processing has a near-nadir reference it can cross-check against the two oblique passes. That's tri-stereo, and it's become the practical default for forested terrain rather than a specialty add-on.

So what's the actual pass count

For a single stand or a defined carbon project boundary, three convergent looks, forward overlap pushed up toward 80-90%, is the range most canopy work settles into. Side overlap matters more as the area of interest gets wider than one flight line's swath, since that's what stitches adjacent strips into one continuous surface model without a seam running through the middle of a compartment.

Pair count isn't really a fixed number you can quote for "a canopy height model" in the abstract, because it scales with the size and shape of the parcel, not with the fact that it's forested. A 40-hectare block might sit entirely inside one tri-stereo footprint. A holding that runs long and narrow along a drainage, or a project area stitched from several non-contiguous compartments, needs enough strips and side overlap to cover the boundary without gaps, and that number is a tiling exercise more than a photogrammetry rule.

There's a tradeoff underneath all of this that's worth knowing even if you're not the one flying the mission: base-to-height ratio. Wider angular separation between looks gives you better height precision, because small errors in matching translate into smaller height errors when the triangulation angle is steep. But push the angle too wide and you start losing matches again, because the two looks are seeing different sides of the crown rather than the same point. Canopy work tends to sit at a narrower B/H ratio than bare-earth stereo for exactly that reason. More overlap, tighter angles, and a third look to hold the match together where the first two lose it.

None of this changes year to year for a given holding, which is part of why an annual VHR stereo pass is enough cadence for standing-volume tracking rather than something that needs re-flying every quarter. Canopy Height Mapping builds the canopy surface model from that kind of overlapping pass set and regresses it to a standing-volume estimate for the parcel, without a ground LiDAR campaign behind it.

If you're scoping a parcel or a project boundary and want to see what that looks like for your acreage, that's the conversation worth starting.