Towards an advanced assessment of the hydrological vulnerability of forests to climate change-induced drought



Preserving the integrity of the hydraulic system is crucial for survival of trees under dry conditions (Bréda et al., 2006). Maintaining a hydraulic safety margin (HSM) between the water potential (WP) associated with stomatal closure and the WP associated with cavitation can therefore be an important drought resistance trait in many tree species. Based on a broad survey of HSMs across tree species, it was recently concluded that all forest biomes are similarly and strongly threatened by climate change-induced drought (Choat et al., 2012). We submit that while the prevalence of narrow HSMs across tree species and climates is remarkable, additional factors must be considered to assess vulnerability of trees and forests to drought, that is, the degree to which they are susceptible to, and unable to cope with, adverse impacts of drought (IPCC, 2007). In fact, such prevalence of narrow HSM is likely a basis, and a trigger, for the development of a range of strategies to cope with this situation in light of the ever-imminent drought stress. Furthermore, settling for an apparently small HSM alone may slow down research into critical aspects of water-use strategies in trees. For example, the accumulating evidence in recent years for rapid recovery from loss of hydraulic conductivity (Brodersen & McElrone, 2013, and references cited therein) indicates a lesser importance of the HSM and any temporary loss of conductivity than suggested. Here, we briefly review key issues in tree strategies to cope with drought that are critical to consider in addition to HSM, if we are to improve predictions of drought related vulnerability of forests.

Mechanisms of drought-induced tree mortality

Climate predictions indicate drying trends and higher frequency of drought events in many regions worldwide (Alpert et al., 2006; Burke et al., 2006; IPCC, 2007; Zhang et al., 2007), strongly affecting biogeochemical and eco-physiological processes in terrestrial ecosystems (Reichstein et al., 2013). The most alarming consequence of climate change to forests is increased drought- and heat-induced tree die-off in multiple locations across the globe (Adams et al., 2009; Allen et al., 2010; and references cited therein). The following main underlying mortality mechanisms were proposed: (1) carbon starvation in response to stomatal closure under prolonged drought stress, thus leading to lower photosynthetic activity; (2) increased attacks by biotic agents, such as insect outbreaks, promoted by a reduced ability of plant defense under drought; and (3) cavitation-induced hydraulic failure (McDowell et al., 2008; Adams et al., 2009). The latter occurs where stem water transport is interrupted by air bubbles that form when (WP) decreases below a species-specific threshold. Some tree-mortality events were attributed to a single mechanism, for example, to carbon starvation in Pinus edulis (Adams et al., 2009), to insect outbreak in Abies concolor (Ferrell et al., 1994), and to hydraulic failure in conifers of the southern hemisphere (Brodribb & Cochard, 2009). In other situations, a combination of mechanisms might be causing tree die-off in a drier climate (McDowell et al., 2011).

The potential and limits of hydraulic safety margins of trees

Water transport in trees is usually under large negative pressures and, consequently, trees live under the threat of xylem embolism (Tyree & Sperry, 1988; Holbrook et al., 1995; Zwieniecki & Holbrook, 1998) . Embolism breaks the continuity of water columns, thereby reducing hydraulic conductance and water supply to transpiring leaves, with major implications for tree functioning and survival. Twenty-five years ago, the question was posed, whether woody plants operate near the point of catastrophic xylem dysfunction caused by dynamic water stress (Tyree & Sperry, 1988). To answer this question, plant scientists have been defining HSMs in various ways, for example, by the difference in WP between stomatal closure and the point of 50% loss of conductivity (WP50), and by the gap between minimum leaf WP under natural conditions and WP50 (Meinzer et al., 2009). Sperry (2004) suggested that narrow HSMs allow for higher xylem functioning, and hence were advantageous in the evolution of higher plants. In woody plants, such low margins are more characteristic of species with high diurnal water storage capacity and low wood density (Meinzer et al., 2009), and are more frequent in pioneer and deciduous tree species than in shade-tolerant and evergreen trees (Markesteijn et al., 2011)

Observations of increased drought-induced tree die-off on the one hand, and narrow HSMs in many tree species on the other hand are potentially related. However, the existence of additional factors makes this relationship far more complex than a simple cause and effect. First, tree die-off can occur in spite of a large HSM. For example, P. edulis suffered a large-scale mortality episode in the semi-arid woodlands of southwest United States (Breshears et al., 2009), despite an HSM of 2.61 MPa, much higher than the meta-analysis average HSM of 0.47 MPa (Choat et al., 2012). A combination of continuous hydraulic stress, lack of carbon assimilation, and attack by biological agents may be held accountable for this massive tree die-off (McDowell et al., 2011). Second, xylem cavitation is not always fatal: trees survived extreme droughts even when continuously operating very close to the HSM (e.g. Quercus alba during the 1995 heat wave in southeast United States; Baldocchi, 1997) or within the HSM (e.g. Swartzia racemosa under rainfall exclusion; Fisher et al., 2006; and Pinus halepensis under soil drought; Klein et al., 2011). Third, in other cases WP was kept above the HSM, in spite of extreme drought conditions (e.g. multiple tree species during the 2003 extreme drought in central Europe; Ciais et al., 2005; Leuzinger et al., 2005; Granier et al., 2007; and Quercus petraea under soil drought; Kuster et al., 2012). These observations suggest that HSMs cannot be easily used as predictors of forest vulnerability to drought.

Strategies of drought avoidance

Avoiding drought stress is one way to survive under dry conditions. Tree drought-stress avoidance strategies range from the leaf scale to the whole-tree level, and from immediate responses to seasonal changes. In many tree species, when exposed to very low WP, stomata close to protect the xylem from developing embolism (Cruiziat et al., 2002; Brodribb & Holbrook, 2004; Yang et al., 2012). In parallel, photosynthetic rate decreases, though a residual and often sufficient carbon gain is maintained without risking excessively low WP (Larcher et al., 1981; Kuster et al., 2012; Klein et al., 2013; Wolf et al., 2013). Gas exchange can also be maintained by homeostatic adjustments of, for example, the ratio of leaf area to sapwood area (McDowell et al., 2002, 2006; Mencuccini, 2003). Alternatively, carbon supply can be facilitated by carbon mobilization between storage and sinks (Guehl et al., 1993; Canham et al., 1999; Körner, 2003; Sala et al., 2010). To further avoid drought stress, trees may shift growth phenology towards moist periods (Grünzweig et al., 2003; Rotenberg & Yakir, 2010; Klein et al., 2012a), adjust root architecture, mycorrhizal colonization and root distribution in the soil profile (Bréda et al., 2006), and adapt xylem vessel anatomy (Eilmann et al., 2009, 2011). Trees might also adapt leaf-shedding phenology in dry periods according to the sensitivity of leaves to drought (Méndez-Alonzo et al., 2012), which is part of an adjustment strategy of the canopy size to water shortage. At the canopy level, self-shading and hydraulic lift are also important drought avoiding strategies. Such stress avoidance strategies prevent progressively lower WP. In turn, they allow trees that have narrow HSMs to hold off from the ‘danger zone’ and, in the long term, facilitate their survival under drought irrespective of the HSM.

The prevalence of recovery from embolism

The capacity of xylem refilling to reverse embolism provides another route of survival under drought episodes, which overcomes the narrow HSM. Evidence is accumulating from manipulation experiments and field measurements on the ability of trees to survive low WP and high levels of hydraulic conductivity loss, and xylem embolism (Cochard, 2006; Taneda & Sperry, 2008; Klein et al., 2011). This suggests efficient recovery mechanisms, which are more common and routine than previously assumed (Brodersen & McElrone, 2013). Empiric evidence for the reversibility of conductivity loss and embolism on a diurnal timescale has become available for over a decade (Tyree et al., 1999). Recently, sub-diurnal cycles of embolism and refilling were also shown and visualized in the laboratory for grapevine (Brodersen et al., 2010; Zufferey et al., 2011), bamboo (Yang et al., 2012), and poplar (Secchi & Zwieniecki, 2011). Routine embolism repair was shown in the field for a desert woody shrub (Schenk & Espino, 2011). Hydraulic conductance can be also recovered by growth of new xylem vessels (Cochard et al., 2001; Ameglio et al., 2002; Eilmann et al., 2010). These mechanisms enable activity and survival at or beyond the HSM, inferring it a less important role. Nevertheless there are constrains to recovery from embolism: First, xylem refilling is limited by the threshold of water extraction by roots. Second, accumulating deterioration of the plant hydraulic system was reported as a consequence of repeated cavitation and its repair (Hacke et al., 2001; Anderegg et al., 2013). Third, in some tree species, the evidence for refilling must be verified with the most current methodology (Wheeler et al., 2013). Fourth, in various tree species, no recovery has been demonstrated (Brodersen & McElrone, 2013). In fact, such inter-specific variation is yet another indication that similarity in forest vulnerability to drought cannot be expected.

The probability of soil and atmospheric drought

A realistic assessment of forest-tree vulnerability to drought must also consider the site-specific conditions that interact with its water-use strategy to determine the tree's fate. Environmental drivers of leaf WP are the WP of soil and atmosphere, which are primarily influenced by precipitation, evapotranspiration, and vapor pressure deficit (VPD). These parameters not only vary with time, but also largely differ among biomes. Annual trends of daily VPD were measured in eight flux-tower sites representing various biomes across the Americas (data adopted from the BIGFOOT dataset; Turner et al., 2012). Notably, VPD > 4 kPa was recorded in only two biomes (desert grassland and tall grass prairie), whereas at the boreal, temperate, and tropical forest sites, VPD was never above 2 kPa. To some extent, the effect of high VPD on tree function should vary among biomes, as tree species in those biomes that are less prone to atmospheric drought are less adapted to high VPD. Yet WP is regulated by stomatal closure, and when VPD increases stomatal conductance tends to converge across species (Oren et al., 1999). In this way, the effect of high VPD should be rather similar among trees in dry and humid biomes. Tree vulnerability to drought is hence mostly determined by the probability that VPD reaches (or remains below) a critical threshold. While VPD maxima may change from one year to the next or in the course of climate change, it is assumed that not all forest biomes are similarly threatened by atmospheric drought.

In the soil, water availability is often studied locally without reference to a standard scale, and hence ‘soil drought’ can have different meanings in different studies (Vicca et al., 2012). This is because water availability for transpiration depends on the local soil and plant hydraulic properties, that is, soil water retention curves and the WP threshold for root uptake (Sinclair, 2005). Soil water retention changes with soil type, which is mostly independent of the forest biome type, and should therefore introduce additional variation among biomes to their susceptibility to soil drought. Knowledge of soil water retention and the WP threshold for root uptake for a specific soil structure and tree species allows for the calculation of the actual amount of transpirable soil water content (tSWC). It has recently been demonstrated that changes in tSWC predicted to a high degree episodes of tree mortality in a semi-arid pine forest (Klein et al., 2012b). Quantification of tSWC (and its decay) offers a general framework for the extent of soil drought, and thereby allows comparisons among different forest ecosystems in terms of their hydraulic vulnerability.

Limitations on forest water-use

An ecosystem approach might be necessary to evaluate the limitations on forest water-use (i.e. hydrological limitations) and to improve our understanding of forest–climate interactions. Forest drought vulnerability was assessed for the Amazon rainforest and the Alaska boreal forest (Phillips et al., 2009; Welp et al., 2007, respectively), but those studies focused mainly on carbon pools and fluxes rather than on hydrological vulnerability. A recent analysis of the water balance in various biomes showed that water-use in most forests worldwide is smaller than expected from potential evapotranspiration, and hence was below the demand and supply limits, that is, well within the hydrological range (Williams et al., 2012). This implies first that water-use in forests is adapted to the current variation in climate. Second, a system-based approach needs to be applied for determining safety margins at the forest scale and, in turn, identifying forests that may be most vulnerable to climate change.

Conclusions and perspectives for a vulnerability assessment

Mechanisms of drought avoidance, xylem refilling at low WP, and fundamental differences among biomes in the probability of drought events suggest that similarity in HSMs in trees of various origins does not imply similarity in forest vulnerability to drought. An advanced assessment of the hydrological vulnerability at the forest scale also needs to apply a system-based approach and integrate the often opposing water-use strategies of different tree species, for example, into the framework of dynamic vegetation models (Van der Molen et al., 2011; Reichstein et al., 2013). More large-scale field manipulation experiments, studies along regional or global water-availability gradients, and monitoring of impacts of extreme events are required. Considering the complex nature of such multi-scale analysis, a realistic vulnerability assessment can yield a highly heterogeneous forest map, with various levels of drought vulnerability depending on local site conditions, tree species, and their interaction.