Unifying intra- and inter-specific variation in tropical tree mortality

Tree death is a fundamental process driving population dynamics, nutrient cycling, and evolution within plant communities. While past research has identified factors influencing tree mortality across a variety of scales, these distinct drivers are yet to be integrated within a unified predictive framework. In this study, we use a cross-validated Bayesian framework coupled with classic survival analysis techniques to derive instantaneous mortality functions for 203 tropical rainforest tree species at Barro Colorado Island (BCI) Panama. Specifically, we develop mortality functions that not only integrate individual, species, and temporal effects, but also partition the contributions of growth-dependent and growth-independent effects on the overall instantaneous mortality rate. We show that functions that separate mortality rates into growth-dependent and growth-independent hazards, use stem diameter growth rather than basal-area growth, and attribute the effect of wood density to growth-independent mortality outperform alternative formulations. Moreover, we show that the effect of wood density - a prominent trait known to influence tree mortality - explains only 22% of the total variability observed among species. Lastly, our analysis show that growth-dependent processes are the predominant contributor to rates of tree mortality at BCI. Combined, this study provides a framework for predicting individual-level mortality in highly diverse tropical forests. It also highlights how little we know about the causes of species-level and temporal plot-scale effects needed to effectively predict tree mortality.