Experiments under carefully controlled conditions have been criticized as having limited applicability because they do not adequately reflect the heterogeneity of natural environments. This heterogeneity, arising from multiple sources, can affect the magnitude and direction of ecological interactions. We evaluated the importance and magnitude of spatiotemporal variability relating to plant establishment in a field restoration experiment in western Washington. Our experiment was conducted at four sites (two each in North and South Puget Sound) and in four seeding years (2009-2012). Sites were abandoned agricultural fields initially dominated by non-native grasses and agricultural weeds. We killed the extant vegetation using three site preparation treatments: glyphosate-burn-glyphosate (burned), glyphosate-mow-till-solarize (solarized), or two years of glyphosate applications as needed (herbicide). Plots were broadcast seeded with one of three seed mixes (grass-rich, mixed, forb-rich) in autumn, keeping the seeding mixture as consistent as possible among sites and years (mean: 24 sown species at a density of 707 seeds m-2). Establishment (species richness and density of seeded species) was measured during the spring after seeding on 1756 quadrats distributed amongst 349 plots, and was summarized as proportions of the sown richness and sown density.
Results/Conclusions
Two analyses were conducted. First, variance was partitioned to temporal (seeding year) and spatial sources (region, site, plot, quadrat). Region accounted for much of the variation in establishment success (42%); counterintuitively, establishment was much higher in the drier region (58% vs. 33% of sown species detected). Variation among quadrats, within plots, accounted for a third of the variation in establishment success; this degree of small-scale heterogeneity was unexpected given the apparent homogeneity of the sites and our intensive site preparation. Second, site preparation treatments and seed mixes were compared at each site. Seed mix had little effect on establishment, while site preparation effects were more complex. Site preparation treatments were similarly effective in the North Sound, whereas herbicide was most effective in the South Sound. These results highlight the importance of replicating ecological experiments both spatially and temporally. Variation at both regional and micro-site scales should be expected, can inform our understanding of ecological interactions, and needs to be incorporated into planned management actions. Edaphic or other microsite factors have a significant role in determining establishment success, and require further research. In addition, species-specific responses to treatments need to be evaluated.