Every time a cruise comes back with a stem count that doesn't match what the drone imagery suggests, someone asks whether the flight needed a NIR band. The honest answer depends on what's splitting the crowns apart on the ground.
What plain RGB gets you
A sub-10cm RGB pass gives you color, shape, and shadow. For a segmentation pass counting individual crowns, that's often plenty. In a well-stocked plantation with clean spacing and a closed but distinguishable canopy, the crown boundaries show up fine in visible light. You can see where one crown ends and the gap or the next crown starts, because the gaps are real gaps and the trees are roughly even-aged.
RGB struggles in a few specific situations. Dense, overlapping hardwood crowns in full leaf-out can blur together into one green mass with no visible seam between stems. Shaded crowns under a taller overstory lose contrast against the background. Species with similar canopy color, say a mixed stand of two hardwoods with near-identical leaf color in visible light, can confuse an automated segmenter even when a cruiser on the ground would tell them apart in a second.
Where the NIR band earns its keep
Near-infrared reflectance behaves differently than visible light does on live foliage. Healthy chlorophyll-rich leaves reflect NIR strongly, while bark, dead branches, shadow, and non-vegetated gaps reflect it weakly. That contrast doesn't depend on the leaf color a human eye would see. It depends on whether the material is live, photosynthesizing canopy or not.
For crown segmentation, that contrast sharpens the edge between a crown and the gap behind it, even when the gap sits in shadow and would read as "just dark green" in RGB alone. It also separates crowns that look similar in visible color but differ in vigor or canopy structure, since the NIR return shifts with leaf density and health even within a single species.
None of this fixes every hard stand. A true closed-canopy situation, where crowns are physically touching and overlapping at the top with no gap at all between individual trees, is still a hard count no matter what band you fly. Multispectral crown segmentation narrows the ambiguous cases. It doesn't invent a boundary where the canopy genuinely has none.
Deciding which pass to fly
For most even-aged plantation stands with normal stocking, an RGB-only pass at sub-10cm resolution gets you a clean stem-level crown count without the extra sensor weight and cost. Save the NIR band for stands where you already know the segmentation is going to be ambiguous: dense natural stands, understory encroachment into gaps, or mixed-species blocks where visible color alone won't separate the crowns cleanly.
When you're not sure which category a stand falls into, the safer default on a first flight is RGB+NIR. It costs more in payload and processing, but it won't miss a boundary that the color band alone would have smudged, and you only fly the stand once a year regardless of which sensor combination you choose.
This is the same tradeoff that sits behind a full-stand crown inventory flown once a year instead of a sampled cruise: the sensor choice decides how clean the per-tree boundaries come out, but either way you get one entry per crown across the whole stand, not an extrapolated per-acre number from a handful of plots.
If crown boundaries have been the weak point in past counts on a given stand, decide on the NIR band before you book the flight. Adding it later means a second pass over ground you've already flown once.
