Evaluating the adequacy of sampling germinable soil seed banks in semi-arid systems

Quantification of soil seed is an important tool for understanding vegetation dynamics. However, determining adequate sample sizes and sampling regimes is problematic. The literature reports highly variable sample sizes and regimes without any generally applicable standards appearing to emerge. This problem is exaggerated by the heterogeneous and patchy nature of arid and semi-arid environments. Species accumulation curves are commonly used in floristic sampling to plan sampling effort and to evaluate its adequacy. However, the precision of the sample cannot be quantified and it is difficult to determine the number of species missed. We suggest an alternate approach in which a species richness estimator is compared to the actual number of species found. When applied to data collected in a semi-arid region of South-West Queensland, this method gave a good indication of sampling adequacy at large sample sizes. We suggest that this post hoc comparison of the predicted species richness (with coefficient of variation) to the actual number of species found would be useful in all studies reporting seed bank composition. This would allow the reader to evaluate the adequacy of the sampling effort, when judged against the research aims. We also investigated the deployment of sampling effort by examining structured subsets of our data. The results suggest that sampling effort may be optimized by taking fewer samples per transect in any one site, but selecting more transects and sites.