Abstract
Tree death is a key process for our understanding of how forests are and will respond to global change. The extensive forests across the southern Amazonia edge-the driest, warmest and most fragmented of the Amazon regions-provide a window onto what the future of large parts of Amazonia may look like. Understanding tree mortality and its drivers here is essential to anticipate the process across other parts of the basin. Using 10 years of data from a widespread network of long-term forest plots, we assessed how trees die (standing, broken or uprooted) and used generalised mixed-effect models to explore the contribution of plot-, species- and tree-level factors to the likelihood of tree death. Most trees died from stem breakage (54%); a smaller proportion died standing (41%), while very few were uprooted (5%). The mortality rate for standing dead trees was greatest in forests subject to the most intense dry seasons. While trees with the crown more exposed to light were more prone to death from mechanical damage, trees less exposed were more susceptible to death from drought. At the species level, mortality rates were lowest for those species with the greatest wood density. At the individual tree level, physical damage to the crown via branch breakage was the strongest predictor of tree death. Synthesis. Wind- and water deficit-driven disturbances are the main causes of tree death in southern Amazonia edge which is concerning considering the predicted increase in seasonality for Amazonia, especially at the edge. Tree mortality here is greater than any in other Amazonian region, thus any increase in mortality here may represent a tipping point for these forests.