What resolution imagery resolves individual tree crowns?

If you've ever tried to count stems off a Sentinel-2 scene, you already know the answer: you can't. A 10-meter pixel covers roughly the footprint of three or four mature crowns fused into one blur of green. To pull individual tree crowns apart, as opposed to just classifying "forest" versus "not forest," you need ground sample distance well under a meter, and for anything resembling a stem count or crown-diameter measurement, you're really working at 0.3 to 0.5 m or finer.

That's not an arbitrary line. A dominant loblolly or Douglas-fir crown in a well-stocked stand runs somewhere in the 4-8 m diameter range at maturity. At 10 m GSD, that crown is a fraction of a pixel. At 1 m, it's maybe 20-50 pixels, enough to see that something is there but not enough to trace its edge against the neighbor crowding it. Drop to 0.5 m or below and you start getting a crown outline with real geometry: the rounded or irregular boundary, the gaps where light reaches the forest floor, the shadow line that tells you where one crown ends and the adjacent one begins.

Why crown closure changes the math

The resolution you need isn't fixed. It depends on how closed the canopy is.

In an open, recently thinned stand, or a plantation on wide spacing, crowns sit apart enough that even 1 m imagery will separate most dominants. Shadow and gap space do a lot of the work for you. In a fully stocked, unthinned natural stand, crowns interlock, overtop each other, and cast shade across their neighbors. That's where sub-meter GSD stops being a nice-to-have and becomes the only way to tell two stems apart instead of counting them as one blob.

Species mix matters too. A mixed hardwood stand with irregular, broad crowns delineates differently than a planted conifer row with tight, conical crowns and visible row spacing. Suppressed and understory trees, sitting below the dominant canopy, are mostly invisible from any optical pass regardless of resolution, since the overstory occludes them from above. This is a structural limit of passive optical sensing, not a resolution problem you solve by flying closer.

Pan-sharpening and why spectral bands still matter

Most VHR platforms deliver a high-resolution panchromatic band alongside coarser multispectral bands, then pan-sharpen the multispectral data to the panchromatic GSD. The pan band is what actually resolves crown edges; the multispectral bands add the spectral contrast that helps separate live crown from shadow, from bare soil, from understory vegetation showing through gaps. For crown delineation work, you want both: fine enough GSD to see the shape, and enough spectral separation to tell a crown edge from a shadow edge, which otherwise look similar in a single panchromatic band.

What this means for a stand inventory or MRV layer

None of this is really about counting crowns for their own sake. For a timberland holding or a carbon project area, crown delineation is a step toward something else: a canopy surface model, and from there a height and standing-volume estimate you can defend in a valuation cycle or an MRV report. The resolution threshold that resolves individual crowns is also roughly the threshold that gives a photogrammetric stereo pass enough detail to build a usable canopy height model without a LiDAR campaign.

If you're scoping how imagery resolution maps to a volume or biomass layer for a specific holding or project area, Canopy Height Mapping builds that canopy surface and standing-volume estimate from overlapping VHR stereo, refreshed on an annual cadence rather than a one-off flight.

Get in touch to talk through what a pilot on your parcel would look like.