What is ground sample distance, and why your field disappears into one pixel

Ground sample distance is the thing that decides whether your field shows up on a vegetation index map as its own field, or gets dissolved into whatever three or four neighboring plots happen to share a pixel with it.

GSD, short for ground sample distance, is the distance on the ground that one pixel in a satellite image covers. A sensor with a 10 m GSD means every pixel in the image represents a 10 m by 10 m patch of ground. A sensor with 30 m GSD means every pixel covers a patch closer to the size of a football pitch. Whatever is growing inside that square gets blended into a single average value, no matter how many different crops, planting dates, or management decisions are sitting inside it.

This is a different idea from image resolution in the everyday sense. A satellite pass can be sharp and clear and still have coarse GSD. The sensor samples the earth in big squares, and the number it reports for each square is a single average, not a readout of everything growing inside it.

Why this matters more for a half-hectare plot than a fifty-hectare one

On a commercial operation running center-pivot irrigation over fifty or a hundred hectares, 10-30 m GSD barely registers as a problem. The whole field is one crop at one growth stage, so a pixel grid laid over it is just sampling the same thing over and over. Average a thousand near-identical pixels together and you get a stable number.

Smallholder fields don't work that way. A single wide-swath pixel at 10 m GSD can sit across a maize plot, a bean intercrop next door, a fallow strip, and a stand of banana, all at once. The index value that comes back for that pixel averages all four together. It answers a question nobody asked: "what's the blended vigor of everything in this 100 square meter box," instead of "how is this farmer's maize doing this week."

For an advisory agronomist trying to flag stress before it shows up as yield loss, that blended number can hide the exact thing you need to see. A healthy fallow strip can drag the average up and mask early drought stress in the maize sharing its pixel. A struggling intercrop can drag the average down and make a fine field look like it needs a visit it doesn't.

Source resolution versus field-scale resolution

Most of the multispectral coverage with the cadence and swath to revisit a region weekly comes from sensors in the 10-30 m GSD range. That's the trade-off built into wide-swath platforms: broad, frequent coverage, at a pixel size tuned for large, uniform parcels. Sharper-resolution passes exist, but they don't come with the same weekly cadence or swath width on their own.

Getting a field-scale reading out of that source imagery means fusing the wide-swath pass with a sharper look of the same ground, so the coarse pixel gets broken down to something closer to the actual boundary of the field it's reporting on. That's a processing step, not a different satellite. The source GSD doesn't change. What changes is how finely the signal inside each coarse pixel gets resolved before it reaches you as a number per field.

This is also why "accuracy" isn't really the right question to ask about GSD on its own. A 30 m pixel reports a precise average of whatever is inside it, no more and no less. The mismatch is between what the sensor was built to sample and what a smallholder parcel actually looks like on the ground.

If you're reading index values at a regional or district average and trying to make a per-farmer call off them, that mismatch is the gap worth understanding before the reading, not after a bad recommendation. Field Scale Index is building a weekly field-scale reading by fusing wide-swath passes with sharper looks, field type by field type, for exactly this gap. Join the early-access waitlist if that's the reading you've been missing.