|| Baker, J.D., Wilson, S.D., Christian, J.M., Li, X., Ambrose, L.G., Waddington, J.
||Semiarid ecosystems such as grasslands are characterized by high temporal variability in abiotic factors, which has led to suggestions that management actions may be more effective in some years than others. Here we examine this hypothesis in the context of grassland restoration, which faces two major obstacles: the contingency of native grass establishment on unpredictable precipitation, and competition from introduced species. We established replicated restoration experiments over three years at two sites in the northern Great Plains in order to examine the extent to which the success of several restoration strategies varied between sites and among years. We worked in 50-yr-old stands of crested wheatgrass (Agropyron cristatum), an introduced perennial grass that has been planted on > 10 x 10(6) ha in western North America. Establishment of nativegrasses was highly contingent on local conditions, varying fourfold among years and threefold between sites. Survivorship also varied greatly and increased significantly with summer precipitation. No consistent differences were found between drilling and broadcasting in their effects on. establishment, but survivorship was nearly threefold higher in broadcast plots. Plots without seed added, or with native hay added, had almost. no seedlings of native grasses. In contrast, broadcasting the residue remaining after cleaning seeds from native hay produced the highest seedling densities of any treatment. Competition from A. cristatum was significantly and consistently reduced through annual application of a generalist herbicide (glyphosate), which increased native grass establishment and survivorship and the richness and total cover of native species. Herbicide decreased standing crop and increased soil moisture and available nitrogen. A. cristatum was controlled without suppressing native vegetation, both by spraying in early spring, which selectively killed the cool-season A. cristatum, and by application with a wick, which selectively killed the taller A. cristatum. A. cristatum persisted over four years, however, in spite of annual herbicide application. A. cristatum cover in control plots increased significantly with summer precipitation. In summary, broadcasting and drilling differed little in their effects on establishment, but broadcasting increased survivorship and will allow the emergence of plant-induced heterogeneity. Competition from introduced species can be reduced but not eliminated by continuing herbicide application. Lastly, the positive relationships between precipitation and both A. cristatum and native seedling survivorship suggest that management should focus on controlling A. cristatum during dry years and on introducing native species during wet years.