Etude du stock de semences de mauvaises herbes dans le sol : le problème de l’échantillonnage

National audienceQuantitative study of the buried weed seed bank involves the problem of the number of cores to be taken, and this article seeks to solve this sampling problem. Soil cores were collected following a two-stage sampling (5 x 20 cores in each plot) for 5 types of plots. The cores were treated separately (sieving, germination in glasshouse for 2 months, seedling counts, recovery of dormant seeds by flotation), three types of spatial distribution laws could be established for seed populations, depending on the abundance of the species and not on the species itself ; Poisson distribution, aggregated distribution, normal distribution. The mean number of seeds per core and its confidence interval (calculated separately for each species in each subplot) was used to determine the number of cores to be taken to achieve a precision p = 0.05. From these results, a relation can be established between n, the minimum number of cores to be taken, and the mean number of the seed population to be sampled. In fact, n must be > 100 for a majority of species to achieve a precision p = 0.05, which is very difficult to achieve in practice. For the same precision, 50 100 pour la plupart des espèces et pour une précision de 20 p. 100, ce qui est très difficile à réaliser en pratique. Pour la même précision, 50 < n < 100 convient si la loi de distribution est normale et si l’espèce est abondante (plus de 25 x 106 semences par hectare)