Do tradeoffs govern plant species responses to different global change treatments?

Plants are subject to tradeoffs among growth strategies such that adaptations for optimal growth in one condition can preclude optimal growth in another. Thus, we hypothesized that the response of plant species abundance to one global change treatment would relate inversely to the response to a second treatment, particularly for treatment combinations that accentuate distinct traits. To address this hypothesis, we examined plant species abundances in 39 global change experiments manipulating CO2, nitrogen, phosphorus, water, temperature, or disturbance. Overall, the directional response of a species to one treatment was 13% more likely than expected to oppose its response to a second. This tendency was detectable across the global dataset but held little predictive power for individual treatment combinations or within individual experiments. While tradeoffs in the ability to respond to different global change drivers exert detectable effects globally, other forces may obscure their influence in local communities.

Methods
We analyzed the CoRRE database (Komatsu et al. 2019; corredata.weebly.com), which compiles species abundance information from herbaceous plant communities subjected to experimental manipulations mimicking global change drivers. We included studies from the database that manipulated at least two global change drivers (elevated CO2, nitrogen addition, phosphorus addition, multiple nutrient addition, water addition, droought, warming, and disturbance, where disturbance included burning, mowing, or clipping) for at least three years. For each of these experiments, we estimated each species' mean abundance across all years in the control plots and in each relevant treatment plot (single factor manipulations only). We calculated three different effect size metrics using species mean abundance in treatment plots (t) and control plots (c): 1) E=(t-c)/(t+c), 2) LRR=ln(t/c), 3) PS=(t-c)/c. We also include data on species mean abundance in the control plots. We coded data to indicate whether species were present in at least one treatment plot and one control plot (present), or if they were absent from either all treatment or all control plots.

We also include in this dataset the results of 999 runs of a null model where we randomly reshuffled the treatment assignments among all control and all treatment plots within each experiment. We then calculated species mean abundances and effect sizes as above for the observed data.