Cosmetic vs. structural change on a national basemap

A golden stubble field stretching toward the horizon under a clear blue sky.

A change tile shows up red. Does that mean a crew needs to drive out there, or did someone just repave the parking lot? That's the question every update lead ends up answering by hand, tile by tile, and it's the one that decides whether your resurvey list has forty stops on it or four hundred.

The short answer: cosmetic change is anything that alters the pixels without moving the geometry your basemap carries. Structural change is anything that moves, adds, or removes a feature the basemap is supposed to represent. A change score tells you something is different. It doesn't tell you which kind of different you're looking at, and that distinction is the whole job.

What cosmetic change looks like on the ground

Cosmetic change is surface noise that a human eye reads correctly in about two seconds but a change detection pass flags anyway, because the pixels genuinely shifted.

A few that generate a lot of false pulls:

In every one of these cases, the road network, building footprint, and water boundary sit exactly where the basemap already has them. A crew sent out for any of these comes back having confirmed what the imagery could already tell you, if you'd read it right the first time.

What counts as structural change

Structural change moves or erases a feature's geometry: it shows up when a feature gets added, removed, or repositioned relative to where the vector layer says it should be. That's the test: did something change location or stop existing, measured against the current footprint on file?

That includes a new road segment or a realigned intersection, a demolished building where the basemap still shows a footprint, a new subdivision with streets and parcels that didn't exist at the last cycle, a reshaped shoreline or reservoir edge after a flood event, and a new rail spur or bridge. Any of these means the basemap is now wrong in a way that matters to whoever uses it next, whether that's an emergency routing system or a land registry.

A triage rule that holds up

The change score by itself doesn't sort these two groups for you, because a repainted runway and a demolished hangar can produce similar pixel-level deltas. What separates them is whether the footprint or alignment shifted. If the outline of the feature in the current imagery matches the outline in the vector basemap and only the surface texture changed, that's cosmetic, log it and move on. If the outline has moved, a feature has appeared where none existed, or a feature is gone that the basemap still shows, that's structural, and it goes on the worklist.

Reading the footprint against the vector layer takes the analyst maybe thirty seconds per tile, and it's the step that separates the two categories rather than the raw change score. Teams that skip that step end up resurveying reroofed buildings and repaved lots at the same rate as genuinely new construction, which is exactly the blanket-cycle problem a tile-level approach was supposed to fix.

This is the filtering step National Change Alerts is built around: a continuously updated change tile index that scores where the ground no longer matches the basemap, so the triage call starts from a worklist instead of a blank map.

If sorting cosmetic from structural by hand is eating your analysts' week, it's worth seeing what a prefiltered tile list does to that pile.

Get started

← Back to the blog