Blog

What GSD do you need to spot a tree encroaching your Fresnel zone?

A birch branch with fresh green leaves reaching against a clear blue sky.

GSD gets treated as the whole answer. It's really just the first filter.

Ground sample distance tells you the size of one pixel on the ground, nothing more. Detecting an object reliably in an image takes more than one pixel touching it. Remote sensing practice has used a rough rule for decades: you want an object spanning at least two to three pixels across its narrowest dimension before you can call it "detected" rather than "maybe there's something there." A single pixel hitting a target is a hint, not a finding.

Apply that to a tree crown. A mature hardwood or conifer crown typically runs 4 to 8 meters across, often wider. At 0.5 m GSD, that crown spans eight to sixteen pixels in each direction, comfortably enough for a multispectral pass to resolve it as a distinct canopy feature rather than a smudge in the understory. Drop to 2 m or 5 m GSD, which is what a lot of free or low-cost satellite layers deliver, and that same crown shrinks to two or three pixels, maybe less once you account for shadow and mixed-pixel edge effects from neighboring foliage. At that resolution, a crown reads as a blurry patch of color, something a classifier struggles to call a tree with any confidence.

A young sapling or a single fast-growing leader branch, the kind that goes from fine to clipping your path in two growing seasons, is smaller still. Sub-meter GSD is the difference between catching that growth early and catching it after it's already a problem.

Resolution tells you where canopy is, not whether it's in your way

Here's where the question gets more interesting than "finer pixels, better detection." A 0.5 m ortho image, no matter how sharp, is a flat picture taken from directly overhead. It tells you where there's canopy. It does not tell you how tall that canopy is relative to the straight line between your two towers.

The Fresnel zone is a 3D volume, an ellipsoid wrapped around the line-of-sight path, widest at midspan and narrowing to almost nothing near each antenna. A tree standing 40 meters off the path near one tower might never touch the clearance envelope no matter how tall it gets. The same species growing near the midpoint of a long hop can intrude at a fraction of that height, because the Fresnel radius there is at its widest and your vertical margin is thinnest. Horizontal resolution alone, even very good horizontal resolution, can't tell you which of those two trees is the one to worry about.

That's the part a flat image skips. A canopy height estimate matters here, checked against the clearance envelope at that specific point along that specific path, because the Fresnel radius changes continuously from tower to midspan and a fixed buffer distance can't track that curve.

What this means for picking a data source

If you're evaluating imagery for vegetation encroachment work, the GSD question is worth asking alongside two others: does the pass build a 3D canopy height model along the path, and does the vendor check that height against your actual Fresnel geometry for each hop, including frequency and antenna height.

Sub-meter GSD gets you crown-level detection. Canopy height modeling gets you clearance math. Without both, you're either missing small growth or you're flagging trees that were never going to be a problem in the first place, which burns crew time just as fast as missing a real one does.

Link Vegetation Monitor runs an annual multispectral pass at 0.5 m GSD or finer along your link paths, builds a canopy height estimate from it, and checks that height against each link's own Fresnel clearance envelope to flag where growth is closing in. You can see how the per-link encroachment report is put together on the main page.

If a mid-span hop is due for its annual check, that's the place to start.

← Back to the blog