Plant community composition may be altered due to global change, with consequences for ecosystem function. As such, predicting the trajectory and magnitude of community responses to these altered environmental conditions is imperative. The methods typically employed in ecological studies focus on mean differences in species richness in response to resource manipulations, thus masking the inherent complexity of many ecological systems. However, variability among replicates within a treatment can be informative of the trajectory and predictability of community responses to global change drivers. Here, we present a synthesis of 100 experiments manipulating resource availability in herbaceous systems to determine how plant communities respond to these global change drivers. We separate our analysis into two components, the first examining mean changes in community composition with experimental global change manipulation treatments, and the second examining in variance among replicate communities within treatments. Finally, we compare these plant community changes to aboveground net primary productivity (ANPP) responses.
Global change treatments significantly altered plant community composition relative to control plots. This shift in community composition treatments increased with the duration of the experiment and was stronger when multiple global change drivers were simultaneously manipulated (e.g., combined drought and N additions) than when one driver was manipulated alone. This mean change in plant community composition was greater than changes in richness alone and was robust across a wide variety of ecosystems. In addition to shifts in community composition between treatments, replicate plots within treatments were also observed to vary in community composition. The dispersion among replicates within treatments varied widely relative to the control plots, with some treatments resulting in plant community convergence, others resulting in divergence, and some showing no change in dispersion among replicates. Finally, experiments where the plant community shifted with experimental treatments exhibited corresponding shifts in ANPP; further, the predictability of ANPP increased when replicate plant communities converged in response to treatments and decreased when they diverged. Overall, our results illustrate the importance of examining both changes in mean and variance in community composition when examining ecosystem responses to global change.