Blog

Field notes: the acquisition where the cruise and the canopy model were twelve feet apart

Ground-level view through a dense stand of tall, straight pine trunks.

A crew finishes a variable-radius cruise on a 40-year-old loblolly block, writes up average total height for the stand, and it comes back twelve feet shorter than what the canopy surface model said for the same polygon. Nobody's wrong, exactly. They measured two different things and called them both "height."

The cruiser's number comes from a hypsometer shot to a handful of sample trees, usually the dominants and codominants a crew picks out because they're easy to sight on from the plot center. That's a point measurement to a specific tree's apex, averaged across maybe a dozen trees per plot. The canopy surface model is a different animal entirely. It's the top-of-crown elevation at every cell in a raster, built from how two overlapping stereo images resolve the same patch of canopy texture, then differenced against a ground surface. It doesn't pick dominant trees. It takes whatever the matching algorithm locks onto, which in a closed canopy is often the outer edge of a crown cluster, not the tallest stem underneath it.

On a tight, even-aged plantation row that gap usually closes to something small. The crowns are uniform, the texture is cooperative, and the top of the canopy surface roughly tracks the top of the stand. Where it opens up is on older, more irregular stands, or anywhere there's a gap in the overstory letting a tall emergent or a dead snag punch through a surrounding lower canopy.

The cruiser's eye filters that out instinctively, sending a hypsometer shot to a real merchantable tree. The stereo model has no such filter. It reports the highest matched surface in the cell, snag included, and on a stand with scattered relic trees above a younger cohort, that single artifact can drag a block average up by a meaningful margin.

There's a second, quieter cause: ground surface error under dense cover. Standing volume is calculated from height above ground, and any error in the ground surface model carries straight through into the canopy height layer sitting on top of it. On flat pine ground this rarely matters. On broken terrain with heavy understory, the stereo-derived ground surface can sit a few feet off from reality in spots, and that bias rides straight through into the final height number for every tree above it.

The cruise and the model are answering different questions, and treating one as the correct answer misses the point. The cruise is a sample, extrapolated across acres from a small number of plots and subject to whoever's sighting the hypsometer that day. The model is a wall-to-wall surface with its own blind spots around snags, gaps, and ground error on rough terrain. Knowing which failure mode is in play tells you whether the twelve feet is a real problem or just a difference in what got counted.

For a REIT holding or a carbon project area, that distinction matters more than it sounds like it should, because the volume regression downstream of the height layer inherits whatever bias is sitting in the surface model. A snag-inflated height estimate on a few stands is unlikely to swing a portfolio-level valuation on its own, though it's worth flagging before it goes into a reporting cycle, and flagging it is easier when you're refreshing the layer annually and can watch a stand's height trend year over year instead of reading a single flight's output in isolation. That's the shape of the standing-volume layer we build from VHR stereo at Canopy Height Mapping, meant to sit alongside the cruise data as a second read on the stand, refreshed every year.

If your stands have the kind of irregular overstory where this gap tends to show up, it's worth seeing what a stereo-derived height layer looks like against your own cruise numbers.

Start a pilot

← Back to the blog