Evidence for local‐scale community assembly processes from long‐term observations of biodiversity in a grassland chronosequence

QUESTIONS: Habitat restorations offer the opportunity to observe community development over time to infer processes that drive assembly. We used long‐term data from diverse prairie plantings to test predictions of alternate models of community assembly. LOCATION: Prairie plantings in former agricultural fields in Northfield, MN, USA. METHODS: We studied the plant community in adjacent prairie plantings across a 13‐year chronosequence. Permanent plots were surveyed repeatedly over a 16‐year period when the prairies were from 5 to 21 years old. RESULTS: Marked differences in plant composition and diversity among planting blocks obscured temporal inferences across the chronosequence; we found no age effect on any biodiversity parameter within the same sample year. However, strong age effects were evident when longitudinal changes were examined within planting blocks. Changes in local‐scale species richness and Shannon diversity were highly dependent on initial conditions. Low‐diversity plots gained species and high‐diversity plots lost species, resulting in convergence of taxonomic diversity. Phylogenetic diversity increased significantly with age. Processes that could reduce coexistence of related species include competitive exclusion and shared enemies. The species that gained or lost abundance were randomly distributed across the phylogeny overall. Legume species (Fabaceae) consistently increased over time, but grasses (Poaceae) and composites (Asteraceae) had a mix of species that increased or decreased. The species that increased were more likely to be late‐successional prairie specialists; declining species were more likely to be habitat generalists. CONCLUSIONS: Our results show that the trajectory of plant community succession in prairie plantings is highly variable, and depends on initial species diversity and phylogenetic relatedness. Such variation weakens the space‐for‐time substitution assumption in chronosequence studies; caution is necessary in interpreting chronosequence data from a single time sample. Restoration plantings over multiple sites and years are necessary to maximize landscape‐level diversity.