Comparing morphological vs ecophysiological traits
The PCA constructed only from morphological traits separated species according to leaf structure (PC1: C, N, C:N, LDMC, LT) and plant size (PC2: width, height, LA, Root.w, Figure 2a,c). The PCA with only ecophysiological traits separated species based mainly on their gas exchange (PC1: E, WUEintr, WUEinst, GS, R) and water use strategy (PC2: δ13C, δ18O, GS, depth, Figure 2b,d). The second axes of both PCAs (i.e. describing size/ water use strategy) correlated with species’ frequency response to transplantation (χ2(1) = 12.3, p < 0.001 and χ2(1) = 9.1, p = 0.003) and did so as described above for the full trait PCA (i.e. small species with prolific water use decreasing at warmer sites; Figure 1b,c). The second axis of the ecophysiological PCA describing water use strategy (χ2(1) = 4.54, p = 0.03) also predicted species cover change but it was the first axis of the morphological PCA describing leaf structural traits that was associated with cover change (χ2(1) = 7.81, p = 0.005; species with thick, nitrogen-rich leaves reducing in cover, Figure S11c,d). The amount of variation in the cover or frequency response explained by the LMMs did not depend on whether all traits, only morphological traits, or only ecophysiological traits were included (Table S3).
The two PCAs correlated significantly with each other (RV = 0.37, p = 0.01, Figure S14), even though most of the pairwise correlations between morphological and ecophysiological traits were non-significant (Figure S15). Using only size-related morphological traits or only ecophysiological traits resulted in a similar ranking of the winning and losing species (Figure 4), whereas the ranking based on leaf structural traits was different (Table S7).