Precision of soil seedbank sampling: How many soil cores?

A prerequisite for a precise estimate of the abundance of weed seeds in soil is prior knowledge of sampling variability. Based on an analysis of EWRS working group trials, it is shown that the sampling variance (s(2)) of seed counts increases with the sampling mean ((x) over bar) and can be predicted with the logarithmic form of Taylor's power law: log(10)(s(2)) = a + b log(10)((x) over bar). This relationship is constant over time for means greater than 0.1 seeds per core within each of the five different sites studied (temporal variability) but differs slightly among sites (geographical variability). An attempt is made to use a general s(2) : (x) over bar relationship to predict the number of samples as a function of precision and density.