Freezing temperatures and water availability shape the distribution of evolutionary lineages of angiosperm trees in the Americas.

Global patterns of species and evolutionary diversity in plants are primarily determined by a temperature gradient, but precipitation gradients may be more important within the tropics, where plant species richness is positively associated with the amount of rainfall. The impact of precipitation on the distribution of evolutionary diversity, however, is largely unexplored. Here we detail how evolutionary diversity varies along precipitation gradients by bringing together a comprehensive database on the composition of angiosperm tree communities across lowland tropical South America (2,025 inventories from wet to arid biomes), and a new, large-scale phylogenetic hypothesis for the genera that occur in these ecosystems. We find a marked reduction in the evolutionary diversity of communities at low precipitation. However, unlike species richness, evolutionary diversity does not continually increase with rainfall. Rather, our results show that the greatest evolutionary diversity is found in intermediate precipitation regimes, and that there is a decline in evolutionary diversity above 1,490 mm of mean annual rainfall. If conservation is to prioritise evolutionary diversity, areas of intermediate precipitation that are found in the South American ‘arc of deforestation’, but which have been neglected in the design of protected area networks in the tropics, merit increased conservation attention.

Predicting drought-induced mortality (DIM) of woody plants remains a key research challenge under climate change. Here, we integrate information on the edaphoclimatic niches, phylogeny and hydraulic traits of species to model the hydraulic risk of woody plants globally. We combine these models with species distribution records to estimate the hydraulic risk faced by local woody plant species assemblages. Thus, we produce global maps of hydraulic risk and test for its relationship with observed DIM. Our results show that local assemblages modelled as having higher hydraulic risk present a higher probability of DIM. Metrics characterizing this hydraulic risk improve DIM predictions globally, relative to models accounting only for edaphoclimatic predictors or broad functional groupings. The methodology we present here allows mapping of functional trait distributions and elucidation of global macro-evolutionary and biogeographical patterns, improving our ability to predict potential global change impacts on vegetation.