Soil seed bank dynamics in hayfield succession

1. Changes in the species composition of the soil seed bank were determined in a dry and a more species-rich wet chronosequence. Each sequence represented a 25-year hayfield succession following cessation of fertilizer application in the Drentse A Nature Reserve (NL), where the vegetation has been studied since 1972. 2. The number of seeds of many late successional species showed a significant increase during succession with only two characteristic late species present in the seed bank of the early stage of each series. Most of the early successional species showed a decrease in the number of seeds, which paralleled their relative frequency within the vegetation. Overall, seed density and species diversity of the seed bank was higher in the wet than the dry chronosequence. 3. Mean values for seed longevity were calculated for each successional phase in each series using a published seed bank data base. When based simply on the presence of species, the seed longevity index decreased significantly during the wet-series succession in both surface and deeper seed banks and in the vegetation. When based on the relative frequency of species present, the index decreased significantly in both series but only in the surface seed bank and the vegetation. 4. Comparison with data for an ancient undisturbed wet hay meadow in England suggests that seed longevity of hayfield species is generally low. Similarity between seed bank and established vegetation was low in both the undisturbed British meadow (only 47%) and in the relatively young Dutch study sites (c. 50%). 5. The soil seed bank is unlikely to determine hayfield succession in the Drentse A Nature Reserve, since the composition of the soil seed bank tended to follow that of the above-ground vegetation. Increases in plant species richness following the cessation of fertilizer application, the main goal of restoration management, therefore depend to a large extent on an influx of the seeds of most species from outside the site.