Changes in tree functioning and structure driven by drought may eventually lead to tree death in case of long-lasting, frequent, and/or intense stress, or when drought interacts with external disturbance agents such as pests. Despite our better understanding of carbon and water dynamics before mortality thanks to the increasing number of drought-simulation experiments, our ability to predict drought-induced mortality in the field is still limited. Most studies thus far focus on saplings and on few stress factors only; consequently their outputs are not necessarily valid for the multifactorial and long-term processes that usually occur in mature forests.
Instead, focusing on the temporal changes in tree-ring properties is pertinent as this information is available for numerous individuals, sites, and species, and for almost the entire tree life span at an annual resolution. They can reflect direct- and lag-effects of drought on individual physiological functions, productivity, and mortality risk.
We compiled a new global tree ring-width database from sites where both dead and living trees were sampled (more than 7000 trees from 190 sites; 36 species), and compared the growth patterns (growth rate, inter-annual variability, and lag-1 temporal autocorrelation) between trees that died and those that survived a given mortality event. For a subset of these sites, we also analyzed the temporal changes in δ13C and in δ18O to get more insights on the physiological mechanisms.
Results/Conclusions
Considering all mortality sources, tree mortality is preceded by a decrease in radial growth in 85% of the mortality events. We found that long-term decrease in growth rates and increase in inter-annual growth variability (but not in temporal autocorrelation) may be used as proxies of mortality risk, particularly for gymnosperm species. For angiosperms, growth-based metrics tended to be less powerful indicators of impending tree death. Trees that experienced high competition intensity showed longer and stronger growth reductions than the ones that died following bark beetles attacks. The dual-isotope approach revealed that dying trees could either show higher or lower δ13C and δ18O than surviving ones, indicating various physiological pathways before drought-induced mortality.
Changes in tree-ring width and isotopic signals before mortality are nearly universal, but their magnitude are species- and site-specific. This variability provides valuable information on the nature of the mortality process. Low water use efficiency and abrupt changes in growth shortly before death potentially reveal hydraulic failure, and long-term growth declines associated with low photosynthetic rates or low stomatal conductance may indicate carbon depletion.