Drought-induced cavitation resistance varies considerably between tree species and forest ecosystems (Maherali et al., 2004; Delzon et al., 2010) and is closely linked to survival under severe drought in both conifers (Brodribb & Cochard, 2009; Brodribb et al., 2010) and angiosperms (Kursar et al., 2009; Anderegg et al., 2012; Barigah et al., 2013; Urli et al., 2013). Choat et al. (2012) recently reported that most trees operate very close to their threshold of cavitation, leaving them potentially vulnerable to drought-induced mortality in a warmer/drier world (Engelbrecht, 2012). Indeed, species growing in dry environments are more resistant to drought-induced cavitation (more negative water potential at 50% cavitation, P50) but experience a more negative minimum water potential (Pmin) than those growing in wet environments. The so-called hydraulic safety margin, the difference between the level of water stress experienced by a species in the field (Pmin) and the level of water stress leading to hydraulic failure, is, therefore, remarkably narrow, whatever the forest species and biome considered (Choat et al., 2012). This pattern provides clues to the global drought-induced mortality currently observed, even in very wet environments, such as tropical forests (Allen et al., 2010).
Klein et al. (2014) play down the functional significance of the hydraulic safety margin in the vulnerability of forests to drought, pointing out the important role played by additional mechanisms, such as the ability of trees to repair embolism. While it is obvious that drought-induced forest dieback is a complex process involving a number of biotic and abiotic factors, we would like to draw the attention of scientists to the state of evidence for embolism repair, thereby guiding research on tree drought resistance into the most relevant and fruitful directions.
A misunderstanding concerning xylem embolism
Recent advances in tree hydraulics have demonstrated that, contrary to what was previously believed, embolism and repair may be far from routine in trees (Cochard et al., 2013; Wheeler et al., 2013). Indeed, several studies had previously concluded that trees seemed to be highly vulnerable to cavitation (Salleo et al., 1996), with Pmin values much more negative than P50, leading to a high level of xylem cavitation and an apparently negative safety margin, but with recovery on a daily basis. Methodological problems that have only recently been discovered put this alternative conception of tree functioning under question.
Daily cavitation is not routine in trees
Daily patterns of embolism formation and apparent recovery in well watered trees (Canny, 1997; Zwieniecki & Holbrook, 1998; Zufferey et al., 2011) are in some cases known to be biased by sampling procedures. Canny (1997) used cryo-scanning electron microscopy (SEM) to provide evidence of the occurrence of considerable daily variation in cavitation, but Cochard et al. (2000) refuted this conclusion by demonstrating that cavitation occurred in stretched sap upon freezing of the sample. Similarly, Zwieniecki & Holbrook (1998) detected large daily fluctuations in xylem hydraulic conductance that were presumably caused by cavitation and suggested that embolism removal occurred in two tree species (Fraxinus americana and Acer rubrum) despite the existence of tension within the xylem. Wheeler et al. (2013) have also recently demonstrated in the same two species that the formation of embolism following the cutting of branches under water is overestimated when xylem tensions are large. These authors concluded that the idea that many plants are prone to embolism under typical midday water potentials should be reconsidered. Without further work that explicitly avoids such artifacts, we have no unequivocal evidence for daily cavitation (Rockwell et al., 2014). Independent observations in intact seedlings by use of high-resolution computed tomography are urgently needed to unravel this issue in trees as it has been done in grapevine (McElrone et al., 2012; Brodersen et al., 2013).
Issues concerning highly vulnerable species
Cochard et al. (2013) conducted an extensive literature survey on the methods used to estimate cavitation resistance and concluded that the long-vessel species reported to be highly vulnerable to cavitation had actually been evaluated with biased techniques. New methods have been introduced based on air injection (Cochard et al., 1992) or centrifugation (Alder et al., 1997; Cochard, 2002) for the assessment of cavitation resistance, and these attractive methodologies are now widely used. The in situ flow centrifuge technique (Cavitron) has been shown to overestimate vulnerability to cavitation in species with very long vessels (Cochard et al., 2010; Martin-St Paul et al., 2014; Wang et al., 2014) due to open vessel artifact. We concluded that this method as well as the air injection technique (Ennajeh et al., 2011) are not appropriate for use in the measurement of cavitation resistance in species with long vessels. By contrast, other studies assessing the validity of the static centrifuge technique (in which flow measurements are made outside the centrifuge) in large vessels of ring-porous trees (Christman et al., 2012; Sperry et al., 2012; Tobin et al., 2013) did not find support for the open vessel artifact hypothesis. However, a recent study has shown that all centrifuge techniques, independent of the rotor design, are prone to the open vessel artifact (Torres-Ruiz et al., 2014). Direct observations of vessel function using high-resolution computed tomography have also confirmed the remarkable ability of long vessels to resist cavitation (Torres-Ruiz et al., 2014). We therefore recommend interpretation of vulnerability curves showing rapid decline in conductance at high pressure (between 0 and −1 MPa) with caution, before a clear understanding of the open vessel artifact can be obtained.
Problematic evidence of xylem refilling under tension
In the 1990s, rapid xylem refilling in stems has been reported under negative pressure in a few species (Salleo et al., 1996, 2004; Holbrook & Zwieniecki, 1999; Tyree et al., 1999). Direct observations of embolism repair under apparently negative pressures have also recently been reported in grapevine (Brodersen et al., 2010). However, as appropriately pointed out by Sperry (2013), xylem pressure was not precisely measured in the stem in which refilling was observed, but on stressed distal leaves that may have not been hydraulically connected or equilibrated with the vascular system. We still need direct measurements of xylem pressure at the site of refilling based on more appropriate techniques, such as the use of a stem psychrometer or xylem pressure probe. All together, these recent results highlighting methodological artifacts call for much greater care in assessing the presence of refilling capable of reversing drought-induced embolism in trees, and its prevalence may have been greatly overestimated (Cochard & Delzon, 2013).
These findings suggest that cavitation may occur only when xylem pressure falls below a threshold pressure that trees experience only after prolonged episodes of drought and that refilling under tension cannot keep a plant from runaway cavitation. Drought-induced xylem cavitation is thus a symptom of distress. Because the prevalence of the low-cavitation resistance and high-repair capacity paradigm could result from erroneous observations, we propose the high-cavitation-resistance paradigm as the default framework for understanding tree hydraulics and water relations (Fig. 1). Accordingly, tree survival would be unlikely to be dependent on xylem refilling when xylem water potential drops beyond the hydraulic safety margin, as suggested by Klein et al. (2014).
The functional and ecological significance of the hydraulic safety margin
Hydraulic safety margins can be defined in different ways (see Meinzer et al., 2009). From an ecological perspective, it is relevant to define a metric relative to a threshold for tree mortality (see McDowell, 2011, for a review). In conifers, the lethal level of cavitation was found to be close to 50% of embolized conduits in seedling stems (Brodribb & Cochard, 2009; Brodribb et al., 2010), the P50 safety margin (Pmin − P50) being the most relevant metric. In a pine-juniper woodland, Plaut et al. (2012) recently demonstrated in adult trees that the species less prone to drought-induced mortality (Juniperus monosperma) exhibited a greater safety margin at the whole pant level compared to the vulnerable Pinus edulis. Data for angiosperms have recently become available (Anderegg et al., 2012; Barigah et al., 2013; Urli et al., 2013), and have shown that the point of no return is closer to 88% cavitation in the stem. A more conservative lethal safety margin, calculated as (Pmin − P88), may be more appropriate for angiosperms (Choat, 2013). Even though more work is needed to characterize hydraulic failure threshold in different trees species, we suggest that there is a remarkable convergence in lethal hydraulic safety margin across all species and all biomes (combine P50 safety margin in Fig. 2a for conifers and P88 safety margin in Fig. 2b for angiosperms in Choat et al., 2012). This hydraulic safety margin for mortality is overall positive (+2 MPa), suggesting that the point of no return is crossed only after an exceptionally intense and deadly episode of drought. Trees can recover partially or totally from the deleterious effects of water stress until they reach this lethal threshold. However the mechanism underlying this recovery is not necessarily xylem refilling, and is more likely to involve an ability to form new functional xylem conduits (Brodribb et al., 2010).
Use of the species minimum water potential (instead of the water potential associated with stomatal closure) makes the estimate of hydraulic safety margin robust across ecosystems and biomes. Indeed, Pmin is altered by soil, but also by atmospheric water stress, taking the variability of both soil types and vapor pressure deficit across biomes into account. Thus, although not all forest biomes experience similar atmospheric and soil droughts, these differences are fully integrated into the hydraulic safety margin through the use of Pmin. However, we now need to quantify the variation of safety margin within species, as the populations growing at species' warm/dry range margins would be expected to have a narrower safety margin. We hypothesize that these populations experience a more negative Pmin (drier conditions), whereas it has been reported that resistance to cavitation does not vary between populations across the entire range of a species distribution (Martínez-Vilalta et al., 2009; Lamy et al., 2011, 2013).