Statistical analyses
The effects of tree species diversity on Ψpd, Ψmd, ΔΨ, Anet, gs, xylem δ2H and δ18O were determined through linear mixed-effects models for each species using the package lmer . The effect of the season (i.e., spring, summer, fall), year (i.e., 2021, 2022), and species diversity (i.e., monospecific/monofunctional/multifunctional/four-species mixtures) were used as fixed effects, and the individual plot was treated as a random effect. Similar models were used to determine differences in soil water δ2H and δ18O. Sampling dates, species diversity, species, and soil depth were used as explanatory factors in the fixed part of the model. Significant differences between depths for each species’ diversity and sampling dates were found, allowing us to use the Bayesian isotope mixing model to determine the water source contribution of trees as described above (Fig. S3). The output of this model was analyzed similarly with linear mixed-effects models for each species. First, the effect of soil depth (i.e., shallow, deep, bedrock), season, year, and species diversity (i.e., monospecific and four-species mixture) were set as fixed effects, and the individual plot as a random effect. Then, we ran similar models for each soil depth and species where the season, year, and species diversity were used as fixed effects and the plots as random effects. To reveal significant differences between species richness for each measurement at each sampling date and each species, post hoc analyses were performed with Tukey’s HSD test, with FDR correction for multiple testing. Linear regressions were used to test the relationships between ΔΨ, Ψpd, gs, xylem δ18O, P/PET, and PW. All analyses were performed using the R v.4.2.2 statistical software (R Development Core Team, Vienna, Austria, 2022). Before performing each model, the homogeneity of variances and the normality of residuals were assessed, and data were log-transformed if necessary.