Pixel size vs. field size: is your plot too small for satellite monitoring?

An agronomist pulls up a vegetation index map, zooms into a client's plot, and sees one blurry square covering three different fields with three different planting dates. The number on screen is an average of all of them. That's the pixel problem, and it's worth understanding before you hand a grower a reading you can't defend.

What GSD means at the field edge

Ground sample distance (GSD) is the size of the square on the ground that one pixel represents. A sensor with 10 m GSD gives you a pixel that's 10 m by 10 m, roughly a tenth of a hectare, before any resampling happens. Most of the wide-swath multispectral satellites doing the weekly passes agronomists rely on sit somewhere in the 10 to 30 m range at native resolution.

That's fine for a 40-hectare block with uniform rows. It's a different story for a half-hectare smallholder plot wedged between a neighbor's maize and a drainage line. At 10 m GSD, a plot that's 50 m by 100 m is only five pixels across its short side. At 30 m, it's less than two. Any pixel that straddles the boundary is picking up reflectance from whatever's next door, the field edge, the access track, sometimes bare soil from last season's fallow strip.

A quick check: count the pixels across the narrowest dimension

Before you trust an index reading for a plot, measure its narrowest side and divide by the sensor's GSD. If the answer is less than roughly four or five, you don't have a clean interior pixel. You have edge contamination on both sides and maybe one pixel in the middle that's still partly mixed.

Walk a plot boundary with a GPS or pull the shapefile and overlay the sensor's pixel grid if your toolchain supports it. You'll usually find the same pattern on small holdings: the index value for the "field" is really an average of the crop, the bund, a footpath, and a corner of someone else's plot. The pixel simply doesn't know where the property line sits, a geometry limit built into the grid rather than a fault in the sensor itself.

Shape matters as much as area. A square half-hectare plot has a better interior-to-edge ratio than a long thin strip of the same size, which is common on subdivided smallholder land. Irregular or elongated boundaries lose more of their area to mixed pixels than the raw hectare count suggests.

What a mixed pixel does to the number you report back

A mixed pixel doesn't fail loudly. It gives you a plausible-looking NDVI or EVI value that's just wrong, pulled toward whatever's in the neighboring cells. If the field next door is stressed and yours isn't, the index for your client's plot reads lower than reality. If the opposite is true, you over-report vigor right before a problem shows up. Either way, the advisory call built on that number is built on an average the grower never actually planted.

This is the gap between a regional or district-level reading, which averages out fine over hundreds of hectares, and a per-field reading on a plot that's smaller than the sensor's native pixel. Wide-swath resolution was built for scale, not for the boundary of a single smallholder's field.

Where fusion changes the math

Fusing the wide-swath pass with a sharper look narrows the effective pixel without waiting for a dedicated high-resolution tasking order every week. It doesn't make the source imagery something it isn't, but it lets a field that would otherwise collapse into two or three mixed cells get resolved closer to its actual boundary.

That's the gap Field Scale Index is building toward: a weekly field-scale index for plots that would otherwise get smeared into a handful of mixed pixels, built out field type by field type rather than promised as a finished national layer on day one.

If you've got a plot this size sitting in your monitoring list right now, it's worth checking whether Field Scale Index already covers that field type, and getting on the early-access list while the pipeline is being built out.