Research on the degree to which carbon (C) availability limits growth in trees, as well as recent trends in climate change and concurrent increases in drought-related tree mortality, have led to a renewed focus on the physiological mechanisms associated with tree growth responses to current and future climate. This has led to some dispute over the role of stored non-structural C compounds as indicators of a tree's current demands for photosynthate. Much of the uncertainty surrounding this issue could be resolved by developing a better understanding of the potential functions of non-structural C stored within trees. In addition to functioning as a buffer to reconcile temporal asynchrony between C demand and supply, the storage of non-structural C compounds may be under greater regulation than commonly recognized. We propose that in the face of environmental stochasticity, large, long-lived trees may require larger C investments in storage pools as safety margins than previously recognized, and that an important function of these pools may be to maintain hydraulic transport, particularly during episodes of severe stress. If so, survival and long-term growth in trees remain a function of C availability. Given that drought, freeze-thaw events and increasing tree height all impose additional constraints on vascular transport, the common trend of an increase in non-structural carbohydrate concentrations with tree size, drought or cold is consistent with our hypothesis. If the regulated maintenance of relatively large constitutive stored C pools in trees serves to maintain hydraulic integrity, then the minimum thresholds are expected to vary depending on the specific tissues, species, environment, growth form and habit. Much research is needed to elucidate the extent to which allocation of C to storage in trees is a passive vs. an active process, the specific functions of stored C pools, and the factors that drive active C allocation to storage.

Widespread tree mortality associated with drought has been observed on all forested continents and global change is expected to exacerbate vegetation vulnerability. Forest mortality has implications for future biosphere–atmosphere interactions of carbon, water and energy balance, and is poorly represented in dynamic vegetation models. Reducing uncertainty requires improved mortality projections founded on robust physiological processes. However, the proposed mechanisms of droughtinduced mortality, including hydraulic failure and carbon starvation, are unresolved. A growing number of empirical studies have investigated these mechanisms, but data have not been consistently analysed across species and biomes using a standardized physiological framework. Here, we show that xylem hydraulic failure was ubiquitous across multiple tree taxa at drought-induced mortality. All species assessed had 60% or higher loss of xylem hydraulic conductivity, consistent with proposed theoretical and modelled survival thresholds. We found diverse responses in non-structural carbohydrate reserves at mortality, indicating that evidence supporting carbon starvation was not universal. Reduced non-structural carbohydrates were more common for gymnosperms than angiosperms, associated with xylem hydraulic vulnerability, and may have a role in reducing hydraulic function. Our finding that hydraulic failure at drought-induced mortality was persistent across species indicates that substantial improvement in vegetation modelling can be achieved using thresholds in hydraulic function

Recent world‐wide episodes of tree dieback have been attributed to increasing temperatures and associated drought. Because these events are likely to become more common, improved knowledge of their cumulative effects on resilience and the ability to recover pre‐disturbance conditions is important for forest management. Here we propose several indices to examine components of individual tree resilience based on tree ring growth: resistance (inverse of growth reduction during the episode), recovery (growth increase relative to the minimum growth during the episode), resilience (capacity to reach pre‐episode growth levels) and relative resilience (resilience weighted by the damage incurred during the episode). Based on tree ring analyses, we analyzed historical patterns of tree resilience to successive drought‐induced low growth periods in ponderosa pine trees growing in unmanaged, remote forests of the Rocky Mountains. Low‐growth periods registered in tree rings were related to anomalies in the Palmer drought severity index (PDSI) and were attributed to drought. Independently of the impact of a specific event, subsequent growth after a single low‐growth episode was related to the growth prior to the event. Growth performance differed with tree age: young trees were overall more resistant to low‐growth periods, but older trees recovered better from more recent events. Regardless of tree age, recently burned sites exhibited lower post‐episode growth and lower resistance and resilience than unburned ones. We found mixed evidence for the cumulative effect of past low‐growth episodes: overall, greater impacts of a prior event and greater cumulative effects of past low‐growth periods caused a decrease in resistance. However, we did not find a progressive decrease in resilience over time in old trees. Our results highlight the value of using a combination of estimators to evaluate the different components of resilience. Specifically, while tree responses to disturbance depend on past disturbance episodes, the response is context‐specific and depends on the impact the capacity to recover after disturbance. This suggests that recent increases in forest mortality under current climate trends could relate to thresholds on specific components of resilience (resistance, recovery, resilience itself) rather than to an overall loss of resilience over time. Identifying such thresholds and their underlying mechanisms is a promising area of research with important implications for forest management.

Nonstructural carbon (NSC) provides the carbon and energy for plant growth and survival. In woody plants, fundamental questions about NSC remain unresolved: Is NSC storage an active or passive process? Do older NSC reserves remain accessible to the plant? How is NSC depletion related to mortality risk? Herein we review conceptual and mathematical models of NSC dynamics, recent observations and experiments at the organismal scale, and advances in plant physiology that have provided a better understanding of the dynamics of woody plant NSC. Plants preferentially use new carbon but can access decade-old carbon when the plant is stressed or physically damaged. In addition to serving as a carbon and energy source, NSC plays important roles in phloem transport, osmoregulation, and cold tolerance, but how plants regulate these competing roles and NSC depletion remains elusive. Moving forward requires greater synthesis of models and data and integration across scales from -omics to ecology. 667

The mechanisms of carbon starvation: how, when, or does it even occur at all?Recent observations of increasing vegetation mortality events appear to be a result of changing climate, in particular, an increase in the frequency, length and intensity of droughts (e.g. Allen et al., 2010). The threat of widespread increases in future mortality has rekindled interest in the mechanisms of plant mortality and survival because we do not yet understand them well enough to confidently model future vegetation dynamics (Sitch et al., 2008). In this issue of New Phytologist, provide a viewpoint on the 'carbon (C) starvation hypothesis ' (McDowell et al., 2008). Their viewpoint is invaluable for stimulating our field to explicitly refine our definitions and identify the key experiments needed to understand mechanisms of vegetation survival and mortality. Two important conclusions of their paper were that mortality can occur at nonzero carbohydrate levels and that careful experiments focused on the explicit mechanisms of C starvation, as well as on partitioning the roles of hydraulic failure and C starvation, are needed to understand the physiological underpinnings of how plants die. We applaud these conclusions, and agree that hasty acceptance of any hypothesis before adequate testing is foolish. In this commentary, we highlight some of the valuable ideas from Sala et al. and provide additional comments that we hope will prompt careful future tests on the mechanisms of plant mortality.When the C-starvation hypothesis was proposed (McDowell et al., 2008), it represented an attempt to summarize and interpret the existing literature on vegetation mortality, of which there was a wealth of indirect studies, but a paucity of true, mechanistic tests. The original formulation of the hypothesis suggested that stomatal closure minimizes hydraulic failure during drought, causing photosynthetic C uptake to decline to low levels, thereby promoting carbon starvation as carbohydrate demand continues for maintenance of metabolism and defense. The plant either starves outright, or succumbs to attack by insects or pathogens, whichever occurs first. By contrast, failure to maintain xylem water tension lower than its cavitation threshold results in embolisms, which, if unrepaired, can eventually lead to widespread hydraulic failure, desiccation and mortality. We hoped that the C-starvation and hydraulic failure hypotheses would generate discussion and new ideas; and 'The paucity of studies that quantified mortality forces scientists to use data from nonmortality studies to develop hypotheses … we do this at the risk of confusing stress responses with mortality mechanisms.' , as summarized by Sala et al., active discussion is taking place. A primary conclusion from the discussion is that we need clarification of the various mechanisms by which C starvation can occur, if it occurs at all.Plants maintain metabolism through respiratory processes that consume carbohydrates, and in doing so their C budgets must obey the law of conservation of energ...