Wood density predicts mortality threshold for diverse trees

In the past decades, drought-induced forest die-off has been recorded on every forested biome, exerting great effects on biodiversity and ecosystem functions (Allen et al., 2010; Hartmann et al., 2018; Brodribb et al., 2020). A general understanding of forest vulnerability to damage due to climate is lacking, because of our limited understanding of the variation in lethal water potential (Ψlethal) of trees. Achieving this critical goal requires a reliable proxy for Ψlethal that can be used to characterize many species in a forest community. Tree water potential (Ψ), which shapes tree hydraulic conductance and water uptake from the soil, is a robust and direct indicator of physiological stress (Steppe, 2018). The Ψlethal is a critical threshold of Ψ beyond which trees are unable to recover from drought, even after a year of rewatering (Brodribb & Cochard, 2009; Choat, 2013). In the face of drought, along with decreasing soil water availability, tree Ψ gradually drops (i.e. becomes more negative), inducing xylem cavitation and loss of xylem conductivity (Sperry et al., 2002). When tree Ψ falls below the Ψlethal, cavitation spreads widely within xylem conduits, and eventually, trees die from excessive dehydration of cells (McDowell et al., 2008; Körner, 2019). A recent data synthesis showed that trees dehydrated to Ψlethal experienced a > 60% loss of xylem conductivity, suggesting a ubiquitous hydraulic failure in drought-induced tree mortality (Adams et al., 2017). However, due to technical problems associated with measuring tree Ψ in plants undergoing lethal tissue damage, the Ψlethal leading to tree mortality has not been quantified widely across species, despite its importance in understanding and predicting the response of vegetation to drought (Choat et al., 2018; Blackman et al., 2019; McDowell et al., 2019). The determination of the Ψlethal of trees is particularly time- and labor-consuming. Quantification of Ψlethal is typically done by monitoring the recovery of potted plants after exposure to different degrees of water stress. This approach may take months to years for the potted experiments (Brodribb & Cochard, 2009; Kursar et al., 2009), and even years to decades for the studies examining natural drought events (Breshears et al., 2009; McDowell et al., 2016). Added to this is the challenge of identifying tree death during water stress as opposed to seasonal or transient drought-induced leaf deciduousness (Wolfe et al., 2016). Therefore, we need to explore a reliable proxy for Ψlethal to understand forest vulnerability to drought. Based on the previous few studies, useful associations between Ψlethal and xylem vulnerability to drought-induced cavitation have been observed (Brodribb & Cochard, 2009; Brodribb et al., 2010; Urli et al., 2013). If these associations emerge as a constitutive feature of plants, then the vulnerability curve data, which are now available for hundreds of species (Maherali et al., 2004; Choat et al., 2012; McCulloh et al., 2019), could provide important insights into tree mortality in drought. However, two drawbacks currently limit the broad application of xylem vulnerability as a proxy for Ψlethal; first, the relationship between Ψlethal and xylem vulnerability is yet to be generalized across diverse species; and second, the vulnerability technique is laborious and technically challenging. In this respect wood density (dry weight per fresh volume of sapwood) may offer an alternative. Wood density is thought to be an essential index associated with drought tolerance (Hacke et al., 2001; Greenwood et al., 2017), yet the correlation between the Ψlethal and this easily measured and widely available trait has never been tested before. In the present study, Ψlethal values of 59 tree species were compiled from published journal articles. We aimed to investigate (1) the variation in Ψlethal at the tree species level across biomes and (2) the correlations between the Ψlethal and the functional traits believed to be related to drought tolerance. We hypothesized that Ψlethal was lower (more negative) for tree species with lower P50 and P88 (Ψ at 50% and 88% loss of xylem conductivity, respectively), and higher wood density. First, Ψlethal values of 59 tree species were extracted from journal articles. Where possible, trait values of P50, P88, and wood density were obtained from the same data source or the same location for all species. For species that location-specific trait values were not found, the mean values of all available data were obtained via searching published literature or datasets, ignoring intra-specific variation in traits. The final dataset included Ψlethal, P50, P88, and wood density across 59 tree species from three biomes, i.e. tropical, subtropical, and temperate forests. Specifically, there were 14 gymnosperms from seven genera, three families, and 45 angiosperms from 36 genera, 21 families (Supporting Information Table S1). Across all tree species assessed, the Ψlethal ranged widely from −1.5 MPa (Populus balsamifera, an angiosperm) to −14.7 MPa (Callitris columellaris, a gymnosperm) (Fig. 1). The Ψlethal was significantly lower for gymnosperms (mean value: −8.2 MPa) than angiosperms (mean value: −6.0 MPa) (Student's t-test: t = 2.33, P < 0.05), in line with previous findings that gymnosperms are more drought-tolerant than angiosperms (Maherali et al., 2004; Choat et al., 2012). There was no significant difference in the Ψlethal between temperate and subtropical gymnosperms (Student's t-test: t = −1.83, P = 0.09), while angiosperms from temperate forests had a significantly less negative Ψlethal (−4.4 MPa) than those from tropical (−6.7 MPa) or subtropical forests (−7.0 MPa) (ANOVA: F = 4.32, P < 0.05; Fig. 1). The differences in angiosperms among biomes might result from differences in climatic seasonality (Liu et al., 2021), with most subtropical and tropical angiosperms compiled here being from seasonally dry forests. For example, the angiosperm tree species having the most negative Ψlethal were from the Barro Colorado Island of Panama, which has a pronounced 4-month dry season (Kursar et al., 2009). As the subtropical and tropical angiosperms are all evergreens, they are exposed to low soil water availability during the period of the dry season, leading to probable evolutionary selection for higher drought resistance. By contrast, all temperate angiosperm trees compiled here were deciduous species (Table S1), which usually do not experience a regularly dry period during the humid growing season. Hence the Ψlethal of temperate deciduous angiosperms was less negative than that of subtropical/tropical evergreen trees. These findings imply that climate, especially rainfall seasonality, matters in shaping plant drought tolerance, and more investigations are needed to draw a complete picture of Ψlethal across biomes at a global scale. The Ψlethal was linearly correlated with P50 for gymnosperms (R2 = 0.75, P < 0.001; Fig. 2a), with a slope not significantly different from 1 (standardized major axis slope test: P = 0.43), suggesting that P50 represents a good predictor for Ψlethal of gymnosperms. The result raises a question: how could trees die with 50% conductance remaining? The answer may lie in leaves (or needles), which are commonly observed to be less resistant to cavitation than stems within a plant (Choat, 2013). Indeed, the Ψlethal of four gymnosperms were close to stem P50 but more importantly, the Ψ at 95% loss of leaf conductivity, at which point a lethal situation of runaway cavitation is likely to completely disconnect leaves from the stem, rapidly causing leaf death (Brodribb & Cochard, 2009). Regarding angiosperms, Ψlethal was always lower than P50 and close to or even lower than P88 (Fig. 2a,b), suggesting that angiosperms die closer to P88. The result explains why so many angiosperms operated at a negative P50 safety margin, i.e. the lowest Ψ under natural conditions was lower than P50 (Choat et al., 2012). It also strengthens the idea that a P88 safety margin provides a more meaningful metric describing functional safety for angiosperms species (Urli et al., 2013). Highly significant negative correlations were found between Ψlethal and wood density in both angiosperms (R2 = 0.50, P < 0.001) and gymnosperms (R2 = 0.43, P < 0.01) (Fig. 2c). Wood properties are correlated with water storage and transport, with higher wood density also being associated with higher resistance to drought-induced xylem cavitation (Hacke et al., 2001; Hoffmann et al., 2011). A recent study showed that in a tropical forest, tree species with higher wood density had lower xylem osmotic potential at full turgor and lower xylem turgor loss point (De Guzman et al., 2021). It has also been reported that trees with denser wood across the globe had lower leaf turgor loss point (Fu & Meinzer, 2019), providing clear physiological evidence that trees with high wood density are more able to retain water and thus survive under lower Ψ. On a global scale, tree species having higher wood density tend to have lower mortality rates during drought (Nardini et al., 2013; Greenwood et al., 2017). Taken together, the evidence here suggests that wood density, an easily measured functional trait, could be considered as a robust indicator of Ψlethal at the species level for both gymnosperms and angiosperms. It should be noted that Ψlethal was lower for gymnosperms than angiosperms in a given wood density (Fig. 2c), which is similar to the relationship between P50 and wood density reported by Hacke et al. (2001). This difference is likely to reflect contrasting functional characteristics of the xylem types found in gymnosperms and angiosperms. Of particular relevance is the distinct anatomy of inter-conduit pits in the xylem of gymnosperms (torus-margo pits) vs most angiosperms (simple bordered pits). Given the importance of pit anatomy in determining the cavitation threshold of xylem (Lens et al., 2011), it is not surprising to see a higher resistance to cavitation in gymnosperms with the same wood density as angiosperms. The time that takes a tree to reach the Ψlethal determines the tree survival during droughts, therefore models that aim to predict tree mortality under drought must include Ψlethal as a critical parameter (Blackman et al., 2019; Brodribb et al., 2020). Our results provide insights into the variations of Ψlethal across species and biomes, and the important conclusion that wood density can provide a meaningful proxy of the Ψlethal. Although our results provide an important first step, more work is needed to expand the sample diversity and confirm the generality of the patterns observed here. Drought-induced tree mortality is widespread across species and biomes, yet our dataset only presented a small number of tree species from three major biomes. Besides, our dataset primarily came from potted experiments of seedlings or saplings, with mature trees in the field poorly represented. The Ψlethal of trees for multiple life stages was available only for Pinus edulis, showing similar Ψlethal at different life stages (Table S1). Nonetheless, as the relationship between cavitation traits and wood density has been shown not to change with tree age (Rosner et al., 2014), the observed relationship between Ψlethal and wood density may also hold regardless of life stage. In sum, our synthesis of data showed that tree Ψlethal ranged widely from −1.5 to −14.7 MPa across species and biomes, and the Ψlethal was significantly lower for gymnosperms than angiosperms. Tree Ψlethal of gymnosperms was close to P50, while for angiosperms, tree Ψlethal was closer to P88. Our results also suggest that the Ψlethal, which is quite hard to measure, can be predicted using wood density, an easily measured and widely available functional trait. These findings should advance our understanding and prediction of tree mortality in a changing climate. The authors thank the editor and reviewers for their constructive suggestions and comments on an earlier version of this manuscript. This work was supported by the National Natural Science Foundation of China (31825005 and 31800336), the Guangdong Basic and Applied Basic Research Foundation (2020A1515010688), the Key Special Project for Introduced Talents Team of Southern Marine Science and Engineering Guangdong Laboratory (Guangzhou; GML2019ZD0408), and the Institution of South China Sea Ecology and Environmental Engineering, Chinese Academy of Sciences (ISEE2018YB01). The authors declare no competing interests. XL and QY conceived the idea, XL analyzed the data and drafted the manuscript, XL, QY, HL and TJB contributed substantially to revisions. All data used in this study are available in Table S1. Table S1 Full dataset and data sources in this study. Please note: Wiley Blackwell are not responsible for the content or functionality of any Supporting Information supplied by the authors. Any queries (other than missing material) should be directed to the New Phytologist Central Office. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.