Assessing the method’s reliability on a collembola population.

<p>(a) Total measured collembolan biosurface (mm²) depending on the number of pictures in the stack. For low densities, two or three pictures are sufficient to have a good and reliable measurement. For higher densities, four or five pictures are needed. (b) Size distribution of the same collembola population. Four different sets of pictures taken by two different users were analysed. The size classes are 0.02 mm width. We fitted a generalised additive model that did not show any significant difference between the pictures nor between the sets. (c–d) Reliability of the automatic counting method. (c) The estimated biosurface measured in 10 boxes and the number of individuals automatically counted (circles). Results of the manual counting are shown (grey arrows) together with a linear regression between the biosurface and the number of individuals calculated for the five lowest density boxes. (d) Comparison of the number of individuals for automatic and manual counting methods.