Understanding variation in tree mortality rates across the tropics

Old growth tropical forests are an important carbon store, and currently function as a carbon sink. However, this sink capacity may already be declining and its future is highly uncertain, as patterns of tree mortality, one of the key processes shaping the future of tropical forest carbon, are still poorly understood. Accurate estimates of tropical tree mortality require substantial monitoring efforts across time and space, and analytical techniques that can account for variation at multiple scales. Here, I therefore develop Bayesian models to explore how tree mortality rates vary across the tropics and within Amazonia. Firstly, I derive recommendations for the sample size of plot monitoring networks to confidently detect short- and/or long-term changes in tree mortality rates. A key result is that forests with high baseline mortality rates require smaller plot sampling networks to detect a given change, compared to forests with lower rates. Secondly, using observations from an extensive long-term pan-tropical monitoring network, I derive mortality rate distributions at different, nested spatial scales: at the level of individual plots, biogeographical regions and continents. The results show that stem-based mortality rate distributions are best described at the scale of biogeographical regions: for example, forests in North Australia have lower mean mortality rates but suffer occasional larger- scale disturbances, whereas forests in western Amazonia have higher mean mortality rates but few large-scale disturbances. Finally, I test whether the long-term increase in tree mortality rates in Amazonian forests is related to long-term trends in cumulative water deficit. The results confirm a long-term increase in tree mortality rates, but indicate that this trend is not primarily driven by increasing drought stress. Overall, these findings are important for designing efficient national strategies for monitoring the impact of climate change on forests, calibrating vegetation models and predicting the future of tropical carbon under climate change.