Using filter-based community assembly models to improve restoration outcomes

1. Ecological filter models derived from community assembly theory can inform restoration planning by highlighting management actions most likely to affect community composition. Despite growing interest in these models, many restoration studies solely manipulate a single assembly filter - the biotic filter by altering interspecific competition - while ignoring abiotic and dispersal filters that may also influence restoration success. 2. To examine how manipulating all three assembly filters (biotic, abiotic, dispersal) affected restoration in an annual-type grassland, we seeded native forbs from the same functional group as a target invader to increase biotic resistance to invasion (biotic filter), cut standing biomass and either removed it or returned it to plots as litter to alter light conditions (abiotic filter), and added native forbs at different seeding rates to alter density of establishing native populations (dispersal filter). We measured grassland restoration success by recording native species and invader cover in plots. 3. The addition of native species with phenological and morphological traits similar to the target invader reduced invasion and increased native populations, but only in litter-free plots when high densities of native seed were added. 4. Seeding two species with functional traits similar to the invader was more effective for achieving restoration goals than seeding just one functionally similar species. As such, trait differences among restoration species, even species belonging to the same functional group, may increase biotic resistance to invasion in restored communities. 5. Litter removal altered native-invader interactions. When litter was left on the plots, added natives did not reduce invader cover. However, when litter was removed, added natives led to declines in invader cover. 6. Increasing native seeding rates led to larger native populations and increased invasion resistance. 7. Synthesis and applications. In this study, simultaneously manipulating biotic, abiotic and dispersal filters was necessary to optimize grassland restoration outcomes. In particular, the biotic filter only contributed to successful restoration outcomes under abiotic and dispersal conditions that were created through management actions specifically targeting these two additional filters. Restoration planning based on filter models should incorporate actions that target all three assembly filters, rather than solely focusing on the biotic filter.