Spatial heterogeneity confounded ozone-exposure experiment in semi-natural grassland

Interpretation of observations from manipulative experiments is often complicated by a multitude of uncontrolled processes operating at various spatial and temporal scales. As such processes may differ among experimental plots there is a risk that effects of experimental treatments are confounded. Here we report on a free-air ozone-exposure experiment in permanent semi-natural grassland that suggested strong ozone effects on community productivity after 5 years. We tested ozone effects and investigated the potential of confounding due to changes in nutrient management. Repeated-measure ANOVA revealed mainly negative temporal trends for frequency of abundant productive plant species. Constrained ordination additionally showed converging trajectories of species compositions for ozone and control treatments with time. Yields sampled prior to the start of the experiment and soil nitrogen concentrations revealed that spatial heterogeneity in the soil nutrient status was not accounted for by the random allocation of treatments to plots with a bias towards less productive patches in the elevated-ozone plots. Re-analysis of yield data using repeated-measure ANOVA with a covariable to account for productivity prior to the start of fumigation revealed effects on the temporal changes in total yield and yield of legumes that cannot be separated between ozone and pre-treatment nutrient status. Changes in species composition favour an ecological interpretation with spatial heterogeneity as the major cause of different yield declines. Although elevated ozone may cause subtle physiological changes with longer term implications, our new results suggest that species-rich mature grassland such as the one studied at Le Mouret may be less sensitive to elevated ozone than previously assumed. In this experiment a confounded design was hidden at the start by transitory effects of a prior change in nutrient treatments.