Severe droughts have the potential to reduce forest productivity and trigger tree mortality. Most trees face several drought events during their life and therefore resilience to dry conditions may be crucial to long-term survival. We assessed how growth resilience to severe droughts, including its components resistance and recovery, is related to the ability to survive future droughts by using a tree-ring database of surviving and now-dead trees from 118 sites (22 species, >3,500 trees). We found that, across the variety of regions and species sampled, trees that died during water shortages were less resilient to previous non-lethal droughts, relative to coexisting surviving trees of the same species. In angiosperms, drought-related mortality risk is associated with lower resistance (low capacity to reduce impact of the initial drought), while it is related to reduced recovery (low capacity to attain pre-drought growth rates) in gymnosperms. The different resilience strategies in these two taxonomic groups open new avenues to improve our understanding and prediction of drought-induced mortality.

Observations of forest mortality are increasing globally, but relatively little is known regarding the underlying mechanisms driving these events. Tree rings carry physiological signatures that may be used as a tool for retrospective analyses. We capitalized on a local soil water drainage event in 1982 that resulted in increased mortality within a stand of oak trees (Quercus robur), to examine the underlying physiological patterns associated with survival and death in response to soil water limitations. Pre-dawn water potentials showed more negative values for trees in the process of dying compared with those that survived. We used tree rings formed over the 123 years prior to mortality to estimate productivity from basal area increment (BAI, mm(2)), multiple xylem hydraulic parameters via anatomical measurements and crown-level gas exchange via carbon isotope discrimination (Δ, ‰). Oaks that died had significantly higher BAI values than trees that survived until the drainage event, after which the BAI of trees that died declined dramatically. Hydraulic diameter and conductivity of vessels in trees that died were higher than in surviving trees until the last 5 years prior to mortality, at which time both groups had similar values. Trees that died had consistently lower Δ values than trees that survived. Therefore, tree mortality in this stand was associated with physiological differences prior to the onset of soil water reduction. We propose that trees that died may have been hydraulically underbuilt for dry conditions, which predisposes them to severe hydraulic constraints and subsequent mortality. Measurements of above-ground/below-ground dry mass partitioning will be critical to future tests of this hypothesis. Based on these results, it is probable that pedunculate oak trees will experience greater future mortality if climate changes cause more severe droughts than the trees have experienced previously.