Data analysis
We used mixed effects models to test g1separately for each canopy with the following independent variables: site, species, warming, rainfall reduction, and up to 4-way interactions; with measurement campaign (e.g., year and campaign during that year) as a random effect. To separate the direct effect of warming temperature from the indirect effect of warming treatment on soil moisture
we used soil VWC categories and tested the effect of experimental treatment when soil VWC was high in contrast to when it was low. We ran those tests separately for each canopy with independent variables: warming, rainfall reduction (for open only), and soil VWC category up to 3-way interactions; with measurement campaign set up as a random effect. In addition, we tested the effect of warming and rainfall reduction ong1 for high soil VWC independently from other soil VWC categories. Moreover, we used soil VWC as a covariate in combination with fixed variables (as outlined above) and campaign measurement and site set as random variables. We tested whether different cohorts behaved differently in response to environmental drivers and found no evidence for this, so did not further consider those in analyses (but see Figure S3). We also constructed additional mixed effects models to analyze the effect of warming and drought on higher groupings of the species following their biome association, drought tolerance, and phylogenetic affiliation (for details about mixed effects models for species and their respective groupings [drought tolerance, biome, phylogeny] see Tables 3 and S2). All statistical analyses were carried out in JMP statistical software (JMP 14.2, SAS Institute).