Leaf drought tolerance cannot be inferred from classic leaf traits in a tropical rainforest

Plants are enormously diverse in their traits and ecological adaptation, even within given ecosystems, such as tropical rainforests. Accounting for this diversity in vegetation models poses serious challenges. Global plant functional trait databases have highlighted general trait correlations across species that have considerably advanced this research program. However, it remains unclear whether trait correlations found globally hold within communities, and whether they extend to drought tolerance traits. For 134 individual plants spanning a range of sizes and life forms (tree, liana, understorey species) within an Amazonian forest, we measured leaf drought tolerance (leaf water potential at turgor loss point, pi(tlp)), together with 17 leaf traits related to various functions, including leaf economics traits and nutrient composition (leaf mass per area, LMA; and concentrations of C, N, P, K, Ca and Mg per leaf mass and area), leaf area, water-use efficiency (carbon isotope ratio), and time-integrated stomatal conductance and carbon assimilation rate per leaf mass and area. We tested trait coordination and the ability to estimate pi(tlp) from the other traits through model selection. Performance and transferability of the best predictive model were assessed through cross-validation. Here pi(tlp) was positively correlated with leaf area, and with N, P and K concentrations per leaf mass, but not with LMA or any other studied trait. Five axes were needed to account for >80% of trait variation, but only three of them explained more variance than expected at random. The best model explained only 30% of the variation in pi(tlp), and out-sample predictive performance was variable across life forms or canopy strata, suggesting a limited transferability of the model. Synthesis. We found a weak correlation among leaf drought tolerance and other leaf traits within a forest community. We conclude that higher trait dimensionality than assumed under the leaf economics spectrum may operate among leaves within plant communities, with important implications for species coexistence and responses to changing environmental conditions, and also for the representation of community diversity in vegetation models.