Ground sample distance is the distance on the ground between the centres of two neighbouring pixels in your image. If your GSD is 2 cm/px, then one pixel covers a 2 cm square of the world.
That single number is the ceiling on everything your survey can ever tell you. A crack 1 cm wide does not exist in a 3 cm/px dataset — not faintly, not as a blur. It falls between the samples. No amount of processing afterwards puts it back, because the information was never captured.
This is why GSD is a planning decision, not a post-processing one. You choose it before takeoff by choosing your altitude.
What actually determines it
Four things, and only one of them is adjustable in flight:
Or, in the form that's easier to use in the field:
- Altitude above ground — the one you control. GSD scales linearly: fly twice as high, get twice the GSD (half the resolution).
- Focal length — a longer lens gives a finer GSD from the same height, at the cost of a narrower footprint per photo.
- Sensor size and resolution — fixed by the aircraft. This is the real difference between a Mini and a Mavic 3 for survey work.
The linearity is worth internalising. It means the trade is completely predictable: every metre of altitude you give up for a finer GSD is a metre you pay for in coverage area, flight time, and battery.
Working backwards from the deliverable
The useful habit is to start from the smallest feature you must be able to identify, then work back to an altitude.
The principle is that a feature has to span several pixels to be reliably identified — and more still if you need to measure it rather than just spot it. A 2 cm crack covers two pixels at 1 cm/px and is visible; at 5 cm/px it falls inside a single pixel and effectively disappears. So detection work wants a GSD comfortably finer than the feature you're hunting, not merely equal to it.
Typical targets people plan to:
- Roof and facade inspection — 0.3–1 cm/px. You are looking for hairline damage.
- Construction progress and volumetrics — 1–3 cm/px. Enough to measure stockpiles and track earthworks.
- Topographic and boundary survey — 2–5 cm/px, often driven by the accuracy standard you're delivering to.
- Agriculture and general orthomosaics — 5–10 cm/px. Crop stress and field patterns don't need centimetres.
GSD is resolution, not accuracy. A 1 cm/px orthomosaic can still be georeferenced two metres off if your positioning is poor. Resolution tells you what you can see; ground control and RTK tell you where it is. They are separate problems and you can fail either one independently.
The part that catches people out: GSD varies across the site
The formula uses altitude above the ground being photographed — not above your takeoff point. On flat ground those are the same number and the distinction never surfaces.
On a slope it matters immediately. Fly a fixed 80 m barometric altitude across a site with 30 m of relief and the high ground sits 50 m below the aircraft while the low ground sits 80 m below. That's a GSD difference of nearly 40% across a single dataset — one end of your survey is meaningfully sharper than the other.
This is exactly the problem terrain follow exists to solve: by holding height above ground rather than height above launch, it holds GSD roughly constant across varied terrain. That has its own limits, which are worth understanding separately.
Setting it in practice
Decide GSD first, derive altitude from it, then check that the resulting altitude is legal and safe for the site. If the altitude your target GSD demands is below what obstacles allow, something has to give — usually a longer lens, or accepting a coarser deliverable. Discovering that conflict at the site, rather than in planning, costs you the trip.
DronePoint calculates GSD from your drone's actual sensor and focal length as you set altitude, so the number is visible while you're planning rather than inferred afterwards.
Next: front and side overlap — the other half of a mapping flight, and the one that decides whether the images stitch at all.
Follow DronePoint
See more of these notes in your Google Search results.
Add DronePoint as a preferred source