Weed seed banks in arable fields under contrasting pesticide regimes

Viable weed seed banks on a clay soil were estimated for five years from 11 arable fields receiving one of three levels of pesticide application. Low seed densities, averaging 2000 m^-2, were recorded, but dicotyledonous weed seed numbers usually increased following oilseed rape crops, reaching 26 000 m^-2 in one field. Field-to-field variability was large. Samples could be grouped using multivariate techniques into those dominated by grass weed seeds and those containing mainly dicotyledonous weed seeds. Within TWINSPAN classifications, most fields maintained similar seed banks during the study period, indicating relative stability of communities. Differences present at the outset of the experiment, including variation between fields in the same treatment area, were maintained for the five years. Patterns were not associated with straw disposal or spring herbicide application. Oilseed rape crops did not cause qualitative change in seed bank communities, but allowed certain species, notably Stellaria media and Sonchus asper, to increase seed numbers. Smaller and less diverse seed banks in the area receiving most herbicide were not statistically significant. Seed banks gave poor predictions of following weed populations.