Consistent seed bank spatial structure across semi-natural habitats determines plot sampling

Question: Seed bank sampling remains a critical bottleneck to the quality of studies investigating community patterns in the seed bank. The main cause is a large knowledge gap in two aspects critical to sampling, i. e. spatial autocorrelation and species-area relations. The central question of this study is howthe seed bank of a single plot should be sampled, in order to obtain more precise estimates on plot seed bank characteristics, without resorting to a disproportionate investment of available resources. Similar seed bank samples may then enable better plot-based statistical inference of ecological patterns in the seed bank in community ecology studies. Location: Semi-natural habitats in Flanders (Belgium) and northern France. Methods: We investigated the fine-scale spatial structure of individual seed banking species across 12 2.1 m 9 2.1 m plots in three widespread habitats: temperate forest, grassland and heathland. Soil core samples (128) were collected in each plot, using a combined systematic (64) and random design (64). This enabled both geostatistical analyses of the fine-scale spatial structure of individual species-plot combinations as well as the calculation of sampled-based species rarefaction curves. Results: Fine-scale (i. e. within plot) spatial seed bank structure was detected in all plots in each habitat, in at least one or usually more plant species. Over half of the species records displayed significant spatial structure -visible as a random distribution of seed clusters -with medium to strong spatial dependence between point observations of a species of ca. 30 cm. Species rarefaction curves did not attain an asymptote at the actual sampling intensity of 128 samples. Seven out of 12 extrapolated species rarefaction curves did reach an asymptote in less than 384 samples. Conclusions: Using these consistent results in spatial structure and species-area relations across habitats, we present a method of how researchers can develop a tailor-made seed bank design to accommodate their individual needs, abiding by simple predefined boundaries. When the tailored design samples ca. 3% of a plot surface area along a systematic grid with a mesh width of at least 30 cm, these studies will potentially significantly increase the comparability among future seed bank community studies in semi-natural habitats.