Tree mode of death and mortality risk factors across Amazon forests

Tropical forests, particularly the Amazon, are crucial for the global carbon cycle, storing vast amounts of carbon and acting as a significant carbon sink. Tree mortality, rather than productivity, significantly influences the spatial distribution of carbon storage and the variation in carbon sink capacity over time. Despite its importance, the mechanisms contributing to tree mortality across the Amazon remain unclear. This lack of understanding hinders accurate representation of this process in Earth-System models, limiting robust projections of the carbon cycle under future climate scenarios. Tree mortality results from the complex interaction of species and tree characteristics with the environment. Physiological failure, potentially caused by senescence, stress (light competition, moisture stress, pathogen attack), or hydraulic failure, often leads to trees dying standing. Structural failure, due to storms and treefalls, results in stem breakage or uprooting. These processes can interact; for instance, long-term physiological stress can increase vulnerability to structural failure. Direct observation of these processes is rare, making long-term forest monitoring plots crucial for understanding large-scale geographical patterns and differentiating potential drivers of mortality. The mode of death (standing vs. broken/uprooted) provides insights into the underlying causes. Spatial patterns in mortality causes are expected to relate to regional variations in climate, forest structure, and dynamics. Previous studies suggest structural failure dominates in fertile Western Amazonia, where trees prioritize growth over wood structure. The Amazon exhibits a strong precipitation seasonality gradient, from consistently wet conditions in the Northwest to a pronounced dry season in the South. Physiological failure is anticipated to be higher in drier regions with a larger proportion of standing dead trees. Individual tree attributes, such as size, influence mortality likelihood. Hydraulic failure during droughts disproportionately affects larger trees. Taller trees with large crowns are more susceptible to lightning strikes. Light competition primarily impacts smaller trees. Stress conditions (light or water shortage) reduce stem growth rate, potentially leading to death. Therefore, relative stem diameter growth rate can indicate physiological stress. While tree size is a known predictor, recent studies highlight individual growth rate's importance, though their combined influence requires further investigation, particularly across large geographical areas. The complexity of assessing mortality causes is amplified by the extreme diversity of Amazonian forests. Species vary greatly in baseline mortality rates and tolerance to different causes. A species' mortality rate is expected to be predicted by its mean growth rate, reflecting a growth-survival trade-off: fast-growing species, with less investment in wood structure, are more susceptible to mechanical damage and have shorter lifecycles, while slower-growing species invest more in defense and structure, exhibiting lower mortality rates. However, strong evidence for this trade-off in adult trees across large geographical areas has been lacking. Tolerance to water stress also varies widely among species, influencing mortality likelihood. This study aims to address these gaps by analyzing long-term forest plot data to assess the mode of tree death and identify its risk factors across Amazonia.

Existing research on tree mortality in Amazonian forests has been largely localized or restricted to small numbers of plots, providing an incomplete picture of the biome-wide patterns and underlying drivers. Studies have highlighted the roles of various factors, including drought, storms, and competition, but a comprehensive, spatially explicit understanding integrating individual tree characteristics and species traits remained elusive. Some studies focused on structural failure as the dominant mode of death in specific regions, linked to overall mortality rates. Others emphasized the importance of individual growth rate as a predictor of mortality, but these findings were often based on limited geographical scope. The theoretical expectation of a growth-survival trade-off, where fast-growing species experience higher mortality, has been demonstrated in some studies, but often limited to saplings or juvenile trees and not consistently supported across different Amazonian regions. Understanding the varying responses of different species to drought and other environmental pressures, especially considering their drought tolerances and biogeographic distributions, has also been an area requiring further investigation at the pan-Amazonian level.

This study analyzed over 30 years of data from 189 long-term forest inventory plots within the RAINFOR network, encompassing 124,571 trees (≥10 cm diameter at breast height) and 23,683 tree deaths across Amazonia. The data, obtained from ForestPlots.net, included standardized measurements of tree characteristics and mortality modes. Plots were located in lowland, terra firme, intact forests, with regular monitoring; plots smaller than 0.5 hectares or with census intervals exceeding 10 years were excluded. The average census interval was 2.8 years, and the average plot size was 1.23 hectares. Plot-level mortality rates were calculated as the mean across all censuses, weighted by the census-interval length. Tree mortality rates were calculated using a standard formula, and were also calculated separately for trees that died standing versus those that were broken or uprooted (i.e., mode of death). Four Amazonian regions (Northern, East-Central, Western, and Southern) were defined to assess regional variations. Comparisons among regions employed post hoc Tukey's tests. Species traits (wood density, maximum size, mean growth rate, and drought tolerance – proxied by water-deficit affiliation (WDA)) were obtained from previous studies, using genus or family means where species-level data were unavailable. Individual tree characteristics included size (diameter, D) and relative growth rate (calculated from diameter measurements in the antepenultimate and penultimate censuses). A Cox proportional hazard model was used to analyze the risk of death, including tree-level characteristics and species traits as predictors. To account for the nested structure of trees within plots, plot was included as a random effect. A step-wise AIC selection procedure was implemented to determine the combination of variables that best predicted mortality, identifying the relative importance of tree and species attributes. The analysis was also repeated separately for standing and broken/uprooted deaths to identify the specific risk factors associated with each mode of death. The statistical analyses were performed in R, using packages such as 'survival' for survival analyses, 'stats' for statistical comparisons, and 'multcomp' for post-hoc analyses. Corrections were applied to account for potential biases related to census interval length on the proportion of trees found standing vs broken/uprooted after death. This was achieved by including the census interval length as a covariate in a model relating mode of death to the regional location.

Mortality rates varied significantly across the Amazon, being higher in Western and Southern regions than in Northern and East-Central regions. Across the entire Amazon basin, trees were equally likely to die standing (48.4%) as to be broken or uprooted (51.2%), indicating that catastrophic structural damage is a common cause of death. There was little evidence that the proportion of trees dying from structural failure was strongly related to overall mortality rates; rather, both the rates of standing death and broken/uprooted death tended to increase in regions with higher overall mortality rates. Species traits, particularly species mean growth rate, were stronger predictors of tree mortality than individual tree characteristics. Fast-growing species were at significantly higher risk of death, providing strong empirical support for the growth-longevity trade-off hypothesis across tropical tree species. This trade-off appeared pervasive across adult trees and consistent across forests with different species composition and climate. Individual tree growth rate was a key predictor of mortality across all regions, with slower-growing trees more likely to die. The effect of tree size on mortality varied across regions: in Western and Southern Amazonia, smaller trees were at greater risk, likely due to the higher prevalence of broken/uprooted deaths; conversely, in East-Central Amazonia, larger trees were more at risk. Drought tolerance was a significant predictor of mortality only in Southern Amazonia, the driest region. Wet-affiliated species were at greater risk, suggesting that these forests are already experiencing conditions beyond their adaptive limits. In East-Central Amazonia, drought-tolerant species were at greater risk, potentially due to trade-offs between flood and drought tolerance. Separate analyses for trees that died standing versus those that died broken or uprooted revealed differences in risk factors. Slower-growing trees were at greater risk of standing death, while the risk of being broken or uprooted was less strongly associated with tree-level growth rates.

This study provides the most comprehensive assessment of Neotropical tree mortality to date, demonstrating the importance of both species traits and individual tree characteristics in predicting mortality risk across Amazonian forests. The finding that species traits were stronger predictors than tree-level variables emphasizes the role of species composition in shaping overall mortality rates and suggests that changes in species composition due to climate change or other factors could substantially alter baseline mortality rates. The equal likelihood of trees dying standing versus being broken or uprooted challenges previous assumptions that structural failure was primarily driven by overall mortality rates. The regional variations in mortality risk highlight the need for spatially explicit models of tree mortality that account for factors like climate, species composition, and individual tree growth patterns. The finding that drought tolerance significantly predicted mortality only in the Southern Amazon suggests that these forests are already facing climatic conditions exceeding their adaptive limits, potentially increasing vulnerability to future climate change.

This study provides a comprehensive, pan-Amazonian assessment of tree mortality, revealing the overarching importance of species-level growth rate and the growth-survival trade-off in driving tropical tree mortality. The findings highlight significant regional variation in mortality patterns and risk factors, emphasizing the complexity of predicting future forest dynamics under changing climate conditions. Future research should focus on temporal analyses to understand the interplay between environmental factors and tree mortality, informing the development of more realistic Earth-system models.

While this study represents the most comprehensive analysis of Amazonian tree mortality to date, several limitations exist. The reliance on long-term plot data means that infrequent events, such as large-scale disturbances, may not be fully captured. Species trait data were obtained from other studies and had some missing values; these were addressed through imputation but still could introduce uncertainty. The accuracy of mortality mode classification relies on the careful application of standardized protocols; misclassification of mortality mode remains a possibility. Further work is needed to link these findings to specific environmental drivers with high temporal resolution to more precisely quantify climate effects on mortality.