Management Filters and Species Traits: Weed Community Assembly in Long-Term Organic and Conventional Systems

Abstract

Community assembly theory provides a useful framework to assess the response of weed communities to agricultural management systems and to improve the predictive power of weed science. Under this framework, weed community assembly is constrained by abiotic and biotic “filters” that act on species traits to determine community composition. We used an assembly approach to investigate the response of weed seed banks to 25 yr of management-related filtering in three different row-crop management systems in southeastern Pennsylvania: organic manure-based, organic legume-based, and conventional. Weed seed banks were sampled in April of 2005 and 2006 and quantified by direct germination in a greenhouse. We also assessed the filtering effects of weed management practices and relationships between assembled seed bank and emergent weed communities by allowing or excluding weed control practices within each management system and measuring emergent weed community response. Germinable weed seed bank densities and species richness in the final year of the study were over 40% and 15% higher, respectively, in the organic systems relative to the conventional system. Seed bank community structure in the organic systems was different from the conventional system, and the relationships between assembled seed banks and the emergent flora varied. Primary tillage, weed control, timing of planting, and fertility management appeared to be the main filters that differentiated weed seed banks in the three systems. Weed life history, emergence periodicity, seed size, and responsiveness to soil fertility and hydrology appeared to be the most important functional traits determining how weed species responded to management-related filters. Our results suggest that management systems can exert strong filtering effects that can persist over relatively long (greater than one growing season) time scales. Legacy effects of community-level filtering might be more important than previously assumed, and should be incorporated into predictive models of weed community assembly.