A grid is a square law and the count runs away
Forty acres is 161,874 square metres, which as a square block is 402 m on a side. At 100 m spacing that is five gaps across and five along, so six points by six — 36 control points, with 8 checkpoints on top at 20 percent. At 8 minutes a point that is 5 hours 52 minutes of point work before anyone walks anywhere.
Halve the spacing to 50 m and it is 11 by 11, or 121 points. Not double: roughly quadruple, because the grid is two dimensional. That relationship is the single most useful thing to hold in mind when someone suggests tightening the control a bit, and the page prints the three spacings side by side so it is visible rather than assumed.
The walk is real and it is not short
Sweeping the default grid row by row is 4.02 km of walking, plus a bit for the checkpoints. At 60 m a minute that is 74 minutes on top of the point work — call it seven and a quarter hours of ground time for the block. And 60 m a minute is a fair pace on a mown field; across ploughed earth, standing crop or slope it is nothing like that.
The geometry also assumes straight lines between points, which almost no real block allows. Hedges, ditches, gates, livestock and the fact that the shortest route is rarely the walkable one all add to it. Doubling the walking figure is a reasonable sanity check for a site you have not stood in.
A target has to be big enough to be seen
Placing the centre of a target in the imagery is limited by how many pixels it covers. At 2 cm per pixel, wanting 12 pixels across the target means the target is 24 cm — call it a 10 inch square. At 1 cm per pixel the same 12 pixels only needs 12 cm, but the flight to get 1 cm per pixel is four times the images and four times the processing.
The floor under all of it is the pixel itself. A centre that cannot be placed better than a pixel or two is a centre placed to within a few centimetres, however good the receiver standing over it was. Spending on ground equipment past that point buys nothing the imagery can express.
Checkpoints are the only honest number
Every point that goes into the solve is a point the solve is fitting. The residuals it reports afterwards say how well it fitted the data it was handed, which is not the same question as how well it predicts anywhere else. Points held back and surveyed the same way answer the second question, and there is no other way to answer it.
How many control points a job actually needs is not on this page and will not be. That is set by the accuracy the deliverable is held to, the specification whoever is paying wrote, the terrain and the method — and it is signed off by the surveyor responsible, not by arithmetic. Spacing goes in as a number you supply, and everything here is what that number costs.
Questions people ask
How many ground control points does a survey need?
This page will not tell you, because the answer belongs to the specification the job is being held to and to the surveyor responsible for signing it off — it depends on the accuracy required, the terrain, the method and the client. What the page does is turn a spacing you supply into a count, a target size, a walk and a number of hours, so the cost of any given answer is visible before you commit to it.
What does halving the GCP spacing cost?
Roughly four times the points, because a grid is two dimensional. Over a 40 acre square block, 100 m spacing gives 36 points and 50 m gives 121. At 8 minutes a point that is the difference between six hours of point work and twenty, before any walking, which is why the page prints half, current and double side by side.
How big should a GCP target be?
Big enough to cover the number of pixels you need to place its centre confidently, which is the ground sample distance times that pixel count. At 2 cm per pixel and 12 pixels, that is a 24 cm target. The pixel size is the floor under the whole survey: no receiver on the ground improves a centre that cannot be resolved better than a pixel or two in the imagery.
Why hold back checkpoints instead of using every point?
Because a solve fits the points it is given, so its own reported residuals measure the fit rather than the prediction. Points surveyed identically but kept out of the solve are the only independent check available, and 20 percent of the control count is the placeholder in the field — the number itself belongs to your specification, not to this page.
Is the walking time realistic?
It is a floor. The figure comes from sweeping the grid row by row in straight lines at the pace you enter, which gives 4.02 km and 74 minutes for the default block at 60 m a minute. Real ground has hedges, ditches, gates, crop and slope, and none of them are in the geometry. Doubling it is a fair sanity check for a block you have not walked.