How much of an area must a drone survey fly?
It depends on two numbers: how many moose the area holds and how unevenly they stand. We have written down the rule that gives the answer, checked it against SLU's report and put it where anyone can read it and recompute it.
de Meijere, D. (2026). Coverage from the count. Preprint, Zenodo. doi.org/10.5281/zenodo.23725802. 21 pages, free to share (CC BY 4.0). The code that computes every figure in the paper comes with it.
Why "20 percent" is not an answer
SLU counted every moose in two areas and then had a computer run thousands of plot surveys from those counts, with different shares of the land flown (§2.3.3). In Alster, in Värmland, flying 20 % of the area (§3.2) was enough for the density to reach adequate precision. Outside Fredrika, in Västerbotten, it took 50–60 % (§3.3). Same method, same requirement, nearly three times the share.
The difference is how the moose stand. In Alster they stood moderately clumped (§4.2). In Fredrika they stood in clumps, so that a few plots held most of the animals and the rest were empty. The report measures that with a number called VMR (the glossary (in Swedish) explains it): 3.0 in Alster and 4.7 in Fredrika (§3.2 and §3.3), where 1 means scattered entirely at random. The report's conclusion is that coverage has to be assessed for each area (§5). Our paper answers how.
The rule, in words
Precision is bought with moose seen, not with land flown. Two things decide how sure a density estimate is: how many moose the survey sees, and how much the count varies from plot to plot. Written down, it is one line of standard sampling theory, and it gives the coverage directly:
coverage ÷ (1 − coverage) ≥ 15.4 × VMR ÷ number of moose on the land
With the report's own numbers the rule gives 19.0 % for Alster and 55.9 % for Fredrika, where the report found 20 % and 50–60 % by simulation. Nothing is adjusted to fit.
What the rule says, and what sounds wrong the first time you hear it, is that it is the number of moose seen that counts. In a large area a survey needs to see about 46 moose at Alster's clumping, 3.0 times more than if the moose stood at random, however large the land. What share of the land that takes follows from how many moose are on it: 197 in Alster and 57 in Fredrika (§3.2 and §3.3).
What it means for your land
- Large areas. At a given clumping the hectares that must be flown level off while the area keeps growing, because it is a number of moose that has to be seen and not a share of the land. A fixed percentage overpays on large areas.
- Small areas. With few moose the share becomes large, and below a certain size a plot survey is the wrong method. With Fredrika's 57 moose the rule asks for more than half the land.
- Before you fly. The clumping on your land is measured with a first round of plots. A survey sized directly on that measurement falls short more often than not, because the measurement lands below the true need as readily as above it. So we size on an upper bound for the clumping, and the paper says how much extra that costs and how often it still is not enough.
What the paper does not show
It builds on the report's printed tables, not on where the moose stood. The positions are not published, so the report's simulations cannot be rerun.
The interval a survey prints must be corrected for the share of the land that was flown, or it comes out wider than the rule sizes for: 1.51 times at Fredrika's coverage. And an interval that truly holds the real density 95 times in 100 is wider than the one the rule aims at. The coverage the rule gives is therefore a floor, not a ceiling.
How to set the upper bound on clumping where the moose stand in tight groups is open. The paper is a preprint and has not been reviewed by a journal. The report's lead author has read an earlier version and answered our questions, but has not checked every figure.
Read, recompute, cite
- The paper on Zenodo Coverage from the count, 21 pages, with the code attached. Cite it as: de Meijere, D. (2026). Coverage from the count (version 1.0). Zenodo. doi.org/10.5281/zenodo.23725802.
- The code Python with no dependencies. One command recomputes every figure in the paper and stops if anything differs.
Read more about moose surveys
- The method behind the moose survey What we build on, and where we have taken it further.
- What SLU's report says What the report tested, what it found and what it leaves open.
- How many sample plots are needed (in Swedish) The arithmetic behind the number, and why a fixed rule misses.
- What a moose survey costs (in Swedish) Why the price follows the number of moose you need to see, not the area.
Contact
Want to know what the rule gives for your land? Send the area's boundary and roughly how many moose you think are on it, and we will run the numbers.