How to estimate greenhouse floor area from satellite imagery

If you're scoping a protected-cropping territory for input planning, you already know the problem with driving the region: you see the structures on the road, not the ones two farms back from the highway. Satellite imagery solves the visibility problem. Getting a usable floor-area number out of it is a different skill, and it's worth walking through how it actually works.

Start with resolution, not software

Floor area estimation lives or dies on ground sample distance. At 0.5-2 m GSD, a typical gutter-connect glasshouse or a run of plastic tunnels shows up as a distinct shape with real edges, not a blur. Below that resolution you're guessing at boundaries, and a guess on a 2-hectare glasshouse can be off by enough area to misjudge the covering film or shade-cloth order that goes with it.

RGB alone gets you the outline most of the time. NIR earns its keep on the edge cases: a plastic structure with heavy condensation or algae film reads differently in RGB than a glass roof catching direct sun, and NIR reflectance helps separate the two when the visual boundary is ambiguous. It also helps flag structures that look occupied versus ones sitting empty, which matters if you're trying to size active growing area rather than total roofed area.

The actual measurement step is polygon digitizing: tracing the structure's footprint and letting the software compute the enclosed area. For a single greenhouse that's a five-minute job. For a region with several hundred structures spread across adjacent properties, it's the part that eats a planner's week, and it's where most of the error creeps in.

Where the counts go wrong

A few things trip up floor-area estimates even with good imagery:

Adjacent structures read as one block. Gutter-connected ranges and tightly packed tunnel rows can merge into a single polygon if the imagery doesn't resolve the gap between them. That inflates the footprint for one structure and zeroes out its neighbor.

Shadow eats the edge. Low sun angle on a glasshouse ridge throws a shadow line that gets traced as the structure boundary. On a north-south oriented range this can shave or add several percent to the measured footprint depending on pass time.

Degraded plastic looks like bare ground. Older poly film, especially after a few seasons of UV exposure, loses the reflective signature that makes a tunnel pop against surrounding soil or turf. These structures get missed in anything short of a careful pass, which matters if your territory includes older smallholder operations alongside newer commercial ranges.

One pass tells you nothing about change. A structure present in last year's imagery might be re-covered, expanded, or torn down by the time you're placing input orders this season. Protected cropping footprint sizing is a standing question that needs a refresh on some cadence. Plastic structures in particular get added, removed, and re-covered inside a single growing year, which makes last year's count a poor stand-in for this year's order volume.

What this means for sizing a territory

Field survey catches what's near a road. It misses what's behind a windbreak or on the far side of a property that doesn't show up on a drive-through. Satellite imagery flips that: it catches everything in the frame, at the cost of needing a careful eye on the edge cases above. For input suppliers trying to size a market, the structures you miss are the orders you miss, and for planners the gap between a field count and an imagery-based count is usually bigger than people expect going in.

This is the exact problem Greenhouse Mapping builds its annual inventory around: every structure in a region, classified by covering type and sized by footprint, refreshed on a yearly cycle so the numbers don't go stale between growing seasons.

If you're trying to put a real number on protected-cropping area across a region rather than guess from what's visible off the highway, that's worth a look.