Assessment of the sample size to estimate the weed seedbank in soil

It was found that? in order to estimate the number of weed seeds in soil with a desirable degree of accuracy and confidence level, the sample size depends to large extent on the spatial distribution pattern of the seeds in the soil. The seeds of most weeds found in cultivated soils are distributed according to one of two statistical distributions: Poisson or Negative Binomial. For an acceptable degree of accuracy in estimating the number of seeds in the soil, the sample size must be large. Taking this into account, in this study a practical solution based on a theoretical approach is proposed for the problem of establishing the sample size for a species distributed according to either statistical pattern. In addition, given the sample size, it was possible to determine the estimation error to be expected. This theoretical approach is compared with that proposed in previous studies. Two abaci, based on simple expressions, are provided to determine the sample size, according to the Poisson or the Negative Binomial distribution of the weed species. For Poisson distribution, the sample size is determined (given the desired maximum relative error estimation and the confidence level) as a function of only one parameter: the expected number of seeds per sampling unit m (estimated by the sample mean). For Negative Binomial distribution, the sample size is determined as a function of two parameters: rn and p (estimated according to the relationship between the sample mean and the sample variance). A sample size n not very different from those given in previous studies is obtained, but the lower limit found for the n-values is lower than that found in these studies.