nature communications Article https://doi.org/10.1038/s41467-022-32784-1 Leaf water content contributes to global leaf trait relationships Received: 3 March 2022 Zhiqiang Wang 1,2, Heng Huang 3,4 , Han Wang5, Josep Peñuelas 6,7, Jordi Sardans 6,7, Ülo Niinemets8, Karl J. Niklas 9, Yan Li10, Jiangbo Xie10 & Accepted: 15 August 2022 Ian J. Wright11,12 Check for updates Leaf functional traits are important indicators of plant growth and ecosystem dynamics. Despite a wealth of knowledge about leaf trait relationships, a 1234567890():,; mechanistic understanding of how biotic and abiotic factors quantitatively 1234567890():,; influence leaf trait variation and scaling is still incomplete. We propose that leaf water content (LWC) inherently affects other leaf traits, although its role has been largely neglected. Here, we present a modification of a previously validated model based on metabolic theory and use an extensive global leaf trait dataset to test it. Analyses show that mass-based photosynthetic capacity and specific leaf area increase nonlinearly with LWC, as predicted by the model. When the effects of temperature and LWC are controlled, the numer- ical values for the leaf area-mass scaling exponents converge onto 1.0 across plant functional groups, ecosystem types, and latitudinal zones. The data also indicate that leaf water mass is a better predictor of whole-leaf photosynthesis and leaf area than whole-leaf nitrogen and phosphorus masses. Our findings highlight a comprehensive theory that can quantitatively predict some global patterns from the leaf economics spectrum. Leaf functional traits are closely related to plant growth and ecosystem along natural environmental gradients3,13. Considerable variations in dynamics1,2, and strongly influence carbon cycling and energy balance leaf traits, such as SLA (and thus LMA), have been observed across at the ecosystem level3–6. Several studies have explored the scaling plant species, functional groups, and environmental conditions14. Yet, relationships among leaf functional traits1,7–12. However, whether there the underlying mechanisms of these relationships remain elusive. In exists a general model that quantitatively predicts all of the pertinent addition, variations in SLA indicate the variability of the scaling of leaf bivariate leaf trait relationships remains an open question. Among area (AL) with leaf dry mass (ML) which can be described by a gen- these important traits, leaf photosynthesis, specific leaf area (SLA), and eralised power law, i.e. AL = βM L α , where α is the scaling exponent, and leaf dry mass per area (LMA, the inverse of SLA) play critical roles β is the normalisation constant. The overall numerical value of α is because they reflect the potential for plant growth and mirror plant reported to be smaller than 1.0, indicating a phenomenon called ecological strategies in response to environmental changes as well as “diminishing return” because increasing investments in ML fail to 1 Sichuan Zoige Alpine Wetland Ecosystem National Observation and Research Station, Southwest Minzu University, 610041 Chengdu, China. 2Institute of Qinghai-Tibetan Plateau, Southwest Minzu University, 610041 Chengdu, China. 3Department of Environmental Science, Policy, and Management, University of California, Berkeley, Berkeley, CA 94720, USA. 4Department of Biological and Agricultural Engineering, Texas A&M University, College Station, TX 77843, USA. 5Department of Earth System Science, Ministry of Education Key Laboratory for Earth System Modeling, Institute for Global Change Studies, Tsinghua University, 100084 Beijing, China. 6CSIC, Global Ecology Unit, CREAF-CSIC-UAB, Bellaterra, Catalonia 08193, Spain. 7CREAF, Cerdanyola del Vallès, Catalonia 08913, Spain. 8Institute of Agricultural and Environmental Sciences, Estonian University of Life Sciences, Kreutzwaldi 1, 51006 Tartu, Estonia. 9Plant Biology Section, School of Integrative Plant Science, Cornell University, Ithaca, NY 14853, USA. 10State Key Laboratory of Subtropical Silviculture, Zhejiang A&F University, 311300 Hangzhou, China. 11Hawkesbury Institute for the Environment, Western Sydney University, Penrith, NSW 2750, Australia. 12Department of Biological Sciences, Macquarie University, North Ryde, NSW 2109, Australia. e-mail: henghuang@berkeley.edu Nature Communications | (2022)13:5525 1 Article https://doi.org/10.1038/s41467-022-32784-1 produce proportional increases in AL. However, the numerical value of two important traits in the leaf economics spectrum due to their α can vary from <1.0 to >1.0 among or within some groups of essential roles in plant metabolism27 and also to their strong correla- species8,10. Nevertheless, theoretical attempts to explain the observed tions with other leaf traits1,10. Thus, we further test the critical role of variability of α are rare and thus incomplete. LWC by comparing the robustness of leaf water mass versus leaf Tissue water content is another important, yet relatively under- nitrogen (N) and phosphorus (P) mass in predicting leaf photosynth- investigated, plant functional trait. Water plays a fundamental role in esis and leaf area. biochemical reactions and metabolism15, and its availability regulates plant growth and metabolism, thereby influencing the productivity Results and discussion and carbon cycling of ecosystems16–20. At the leaf level, water avail- The theoretical model ability may also directly affect other leaf traits, such as leaf photo- The model presented here builds on a recently developed synthetic rates and SLA (and thus LMA)21. In fact, leaf dry matter metabolic theory based on biochemical kinetics. It describes a non- content (LDMC), which is functionally equivalent to leaf water content linear relationship between plant metabolic rate per unit of dry mass (LWC), has been suggested to be a better predictor of ecosystem (Bs, nmol g−1 s−1), such as light-saturated photosynthetic rates and dark above-ground net primary productivity (aNPP) than SLA22, although respiration rates, plant water content (S, g g−1), and temperature aNPP is influenced by many other factors23, suggesting an important (in degrees K)19,20, i.e. role of LWC in the leaf economics spectrum, particularly since pre- vious studies reveal that LWC plays a key role in leaf thermal regulation Bs = g 1 ek 1 S=ðK 1 + SÞ eE=kT ð1Þ and further affects leaf carbon assimilation24–26. Therefore, exploring the quantitative links between LWC and other leaf traits could improve where g1 is a normalisation constant, k1 represents the maximum our understanding of not only trait variation and scaling relationships increase in specific metabolic rates due to changes in water content but also of trait shifts in response to climate change. (i.e. from dehydrated to fully hydrated), K1 represents the water con- In order to improve our understanding of these lineages, we tent when the mean reaction rate of cellular metabolism reaches one- derive a theoretical framework to explore the universal effects of half of its maximum, E is the activation energy, and k is Boltzmann’s temperature and LWC on leaf photosynthesis and SLA. We then use a constant. In this model S is defined on a dry mass basis (i.e. the ratio of comprehensive global leaf trait database compiled from 3427 species plant water mass to plant dry mass) to broaden its range and better across a variety of ecosystems (Fig. 1) to (i) test whether the theoretical reflect the proportional changes in the amount of water in plant model successfully predicts the empirical relationships between leaf tissues20. Full details of the model’s assumptions can be found in the traits (mass-specific light-saturated leaf photosynthetic rate and SLA) Methods and Huang et al.19,20. The model was tested and shown to hold and LWC and temperature and (ii) examine whether the exponent of true for a broad range of species and for whole plants and above- and leaf trait scaling converges onto a canonical value of 1.0 after tem- belowground organs20. Here, we first applied the model to describe the perature and LWC are corrected, as predicted by the model, across quantitative effects of dry mass-based LWC (the ratio of leaf water different plant life forms, ecosystem types, and latitudinal zones. Apart mass to leaf dry mass) and temperature on the light-saturated leaf from LWC, leaf nitrogen (N) and phosphorus (P) are also considered photosynthetic rate per unit of dry mass or leaf photosynthetic a Whittaker biomes b 400 Tundra Precipitation (cm) Boreal forest 300 Temperate seasonal forest Temperate rain forest 200 Tropical rain forest Tropical seasonal forest/savanna 100 Subtropical desert Temperate grassland/desert 0 Woodland/shrubland –10 0 10 20 30 Temperature (°C) Fig. 1 | The distribution of the sampling sites of the studied leaf functional dimensional climate space represented by mean annual temperature and mean traits. a Global map showing the geographic locations of the field sites. b Two annual precipitation superimposed on Whittaker biomes. Nature Communications | (2022)13:5525 2 Article https://doi.org/10.1038/s41467-022-32784-1 capacity (Ps, nmol CO2 g−1 s−1), which can be expressed as examined the effect of LWC on leaf trait scaling. The numerical value of   the scaling exponent for the PL versus ML relationship was 0.95 (Fig. 2a;  P  k  LWC E 95% CI = 0.92 and 0.99, r2 = 0.83). The non-linear relationship between ln P s = ln L = ln g 1 + 1  , ð2Þ ML K 1 + LWC kT Pscor and LWC described by Eq. (3) was supported by the empirical data (Fig. 2b; Supplementary Table 1). The non-linear model (Eq. 3) also had where PL is the light-saturated whole-leaf photosynthetic rate (nmol s−1). a lower Akaike’s Information Criterion score than the simple linear Equation 2 indicates that the log-transformed temperature-corrected Ps model between log-transformed Pscor and LWC (i.e. 793.5 versus (i.e. Pscor) should increase with LWC, following Michaelis-Menten type 1988.8). After LWC and temperature were corrected (see Eq. 4), the hyperbolic response, i.e. numerical value of the scaling exponent became 0.97 (Fig. 2c; 95% CI = 0.94 and 1.01, r2 = 0.85), which was statistically indistinguishable     k  LWC from 1.0 (P > 0.05), as predicted by the model. Likewise, the numerical ln P scor = ln P s eE=kT = ln g 1 + 1 : ð3Þ K 1 + LWC value of the exponent (i.e. α) for the AL versus ML scaling relationship was 1.02 (Fig. 2d; 95% CI = 1.02 and 1.03, r2 = 0.92). Additional analyses Rearranging Eq. (2) shows that the temperature- and LWC- using the pooled dataset showed that log-transformed SLA increased corrected whole-leaf photosynthetic rate (i.e. PLcor) should scale iso- with LWC following Michaelis-Menten dynamics, as predicted by metrically with ML, i.e. Eq. (6) (Fig. 2e; Supplementary Table 1). After the effects of LWC and temperature were accounted for (using Eq. 7), the numerical value of α P Lcor = P L ek 1 LWC=ðK 1 + LWCÞ eE=kT = g 1 M L : ð4Þ became 1.01 (Fig. 2f; 95% CI = 1.00 and 1.01, r2 = 0.95). Thus, both of the scaling exponents numerically converged onto 1.0 once LWC and In this study, we refer to the temperature or water content cor- temperature were corrected. In addition, an inspection of the locally E rection as moving the temperature term (kT ) or water content term weighted smoothing (LOWESS) curves showed that the curvature in k 1 LWC (K + LWC) to the left-hand side of the model, an approach that has been both scaling relationships was reduced after the effects of LWC and 1 commonly used in previous studies28. temperature were corrected, as predicted by the model (Fig. 2). Following previous studies29–31, we assume that PL is proportional The results presented here show that temperature and LWC to AL, i.e. P L / AL because leaf area directly determines the light quantitatively correlate with other leaf traits, such as Ps and SLA (or interception capacity1. Therefore, the quantitative relationship LMA), as predicted by the model. The increases in Ps and SLA attenuate between SLA and LWC and temperature can be described as with increasing LWC (Fig. 2b, e), indicating that leaf water availability sets a constraint on the maximum Ps and SLA that leaves can reach. It   AL  k  LWC E has long been recognised that SLA is closely correlated with leaf lnðSLAÞ = ln = ln g 2 + 1  , ð5Þ ML K 1 + LWC kT growth rate and metabolic activity14,36,37. Therefore, it is reasonable to also expect that SLA, as well as Ps, will be quantitatively affected by where g2 is another normalisation constant. Given that leaf area is a LWC, which can change as a function of developmental status (such as direct indicator of leaf photosynthetic capacity and that both traits leaf maturation and the accumulation of lignified tissues) and tran- reflect the long-term adaptation of plants to environmental change3,13, siently as a function of evapotranspiration. LWC is also a reflection of k1 in Eqs. (2) and (5) represent the maximum increase in mass-specific species-specific adaptation to environmental conditions in different leaf photosynthetic capacity due to changes in water content. Since biomes. Nevertheless, our model, as well as the empirical data used to the temperature has a direct effect on the metabolic rates32 and test it, reveal a broad and statistically robust correlation between cri- productivity of ecosystems33, it is reasonable to assume that tical leaf functional traits and leaf tissue water content. However, the temperature can affect SLA globally34,35. Therefore, in this context T observed variations in Pscor (Fig. 2b) suggest that in addition to tem- denotes the mean growing-season temperature (in degrees K), as SLA perature and LWC, other factors (e.g. plant phylogeny and soil fertility) might be more responsive to long-term changes in temperature. may also affect leaf photosynthetic capacity, which is not accounted Equation (5) predicts that SLA should increase with both LWC and the for in our model and should be critically examined in future research. growing-season temperature. Rearranging Eq. (5) yields Leaf area-mass scaling among different groups     k  LWC The numerical value of α varied across different plant growth forms, ln SLAcor = ln SLAeE=kT = ln g 2 + 1 , ð6Þ K 1 + LWC ecosystems, and latitudinal zones (Fig. 3 and Supplementary Table 2). In particular, the leaf area versus mass scaling relationship showed a which predicts that the log-transformed temperature-corrected SLA clear pattern along a latitudinal gradient (Fig. 3c and Supplementary (i.e. SLAcor) should increase with LWC following Michaelis-Menten Table 2). The numerical value of α decreased from 1.10 in boreal dynamics. Likewise, by moving Kk 1+LWC E and kT to the left-hand side and regions (95% CI = 1.08 and 1.12, r2 = 0.91) to 1.00 in temperate regions 1 LWC ML to the right-hand side of Eq. (5), we observe that the temperature- (95% CI = 0.99 and 1.01, r2 = 0.91), and to 0.94 in tropical regions (95% and LWC-corrected leaf area (i.e. ALcor) should scale isometrically with CI = 0.92 and 0.95, r2 = 0.91). However, after correcting for the effects leaf mass, i.e. of LWC and temperature, as predicted, α converged onto 1.0 across all different groupings, and the r2 values of the scaling relationships also ALcor = AL ek 1 LWC=ðK 1 + LWCÞ eE=kT = g 2 M L : ð7Þ increased (Fig. 3 and Supplementary Table 2). Our analyses show that LWC also affects the numerical values of We note that the scaling of PLcor and ALcor with respect to ML will the exponents of leaf trait scaling relationships, which helps to explain reveal how LWC mediates the scaling exponent of leaf trait why different works sometimes report significant differences in the relationships. exponents governing these relationships8,10. In particular, the numer- ical values of the scaling exponent governing the leaf area versus mass Effects of LWC and temperature on leaf trait scaling scaling relationship differ among different plant growth forms, eco- The numerical value of the exponent for the PL versus AL scaling systems, and latitudinal zones (Fig. 3 and Supplementary Table 2), relationship calculated from the empirical data was 0.99 (Supple- indicating that no invariant “scaling exponent” (i.e. α) holds true for mentary Fig. 1; 95% CI = 0.95 and 1.02, r2 = 0.87), strongly supporting the leaf area-mass scaling relationship. For example, in our dataset, α is the model assumption that PL scales isometrically with AL. We then significantly smaller than 1.0 in tropical regions (i.e. in keeping with a Nature Communications | (2022)13:5525 3 Article https://doi.org/10.1038/s41467-022-32784-1 Fig. 2 | The quantitative effects of dry mass-based leaf water content (LWC) on (ML, g). e Non-linear fit to the relationship between temperature-corrected specific leaf photosynthesis and SLA. a Scaling of leaf photosynthetic rate (PL, nmol s−1) leaf area (SLAcor, cm2 g−1) and LWC based on Eq. (6). f Scaling of temperature- and with leaf dry mass (ML, g). b Non-linear fit to the relationship between temperature- LWC-corrected leaf area (ALcor, cm2) with leaf dry mass (ML, g). Data with LWC corrected mass-specific leaf photosynthetic rate (Pscor, nmol g−1 s−1) and LWC (g g−1) greater than 25 were not shown in panel e for a better visualisation. LOWESS curves based on Eq. (3). c Scaling of temperature- and LWC-corrected leaf photosynthetic (blue lines) and 95% confidence intervals are shown. rate (PLcor, nmol s−1) with ML (g). d Scaling of leaf area (AL, cm2) with leaf dry mass Fig. 3 | The exponents of leaf area-mass scaling with and without temperature ecosystem types (n = 97, 1367, 1285, 1271, and 114 for deserts, forests, grasslands, and LWC corrections among different groups. a Comparison of scaling expo- tundra, and wetlands, respectively). c Comparison of scaling exponents among nents among plant growth forms (n = 1688, 491, 1097, and 832 for forbs, grami- different latitudinal zones (n = 1111, 2113, and 910 for tropical, temperate, and boreal noids, shrubs, and trees, respectively). b Comparison of scaling exponents among zones, respectively). Error bars indicate 95% confidence intervals. “diminishing returns” relationship in leaf area with respect to latitudinal gradient may be associated with the different strategies to increasing leaf mass), close to 1.0 in temperate regions (i.e. a break- cope with variations in water availability, e.g. the high evapo- even relationship), and significantly larger than 1.0 in boreal regions transpiration rates in tropical regions may constrain increases in leaf (i.e. an “increasing returns” relationship). This shift in α along a area with increasing leaf mass, therefore resulting in diminishing Nature Communications | (2022)13:5525 4 Article https://doi.org/10.1038/s41467-022-32784-1 Fig. 4 | Leaf water is a robust predictor of leaf photosynthesis and area. The regression lines (black lines), the LOWESS curves (blue lines) and 95% confidence scaling of whole-leaf photosynthesis (PL, nmol s−1, a–c) and leaf area (AL, cm2, d–f) intervals (grey areas) are shown. with leaf nitrogen (N, g), phosphorus (P, g), and water mass (MW, g). The SMA returns, whereas, in boreal regions, the reduced water stress may Leaf water mass is a robust predictor of leaf photosynthesis enable plants to maximise leaf area to achieve relatively high photo- and area synthetic capacities. Despite the variability in the numerical values of We used the China Plant Trait Database38 (see Methods) to explore scaling exponents across different groups, the degree of curvature in whether whole-leaf N and P mass (g) or water mass (MW, g) show these scaling relationships is reduced, and exponents converge onto stronger scaling relationships with PL and AL. This dataset was speci- unity after the effects of LWC and temperature are accounted for fically used for this purpose because it includes paired data required (Fig. 3), as predicted by the model (Eq. 7). This finding indicates for this comparison. The numerical values of the scaling exponents of that the variations in the exponent can, at least partially, be ascribed the relationships of PL versus leaf N and P were 0.94 (Fig. 4a; 95% to the effects of LWC on SLA and the relative rate of increase in CI = 0.90 and 0.98, r2 = 0.80) and 0.76 (Fig. 4b; 95% CI = 0.72 and 0.82, leaf area versus leaf mass. It is noteworthy that the numerical value of r2 = 0.71), respectively. Likewise, the exponents for the scaling of AL the scaling exponent for the PL versus ML relationship is also very versus leaf N and P were 0.95 (Fig. 4d; 95% CI = 0.92 and 0.99, r2 = 0.86) close to 1.0, indicating the relatively weak effects of temperature and and 0.81 (Fig. 4e; 95% CI = 0.77 and 0.86, r2 = 0.78), respectively. LWC on leaf photosynthesis-mass scaling. This may be partially However, PL and AL scaled as the 0.95 power (Fig. 4c; 95% CI = 0.91 and attributed to the relatively limited number of data with concurrent 0.99, r2 = 0.81) and as the 0.97 power of MW (Fig. 4f; 95% CI = 0.94 and LWC measurements and the adaptation of leaf traits to long-term 0.99, r2 = 0.91), respectively. Thus, MW provided a higher explanatory temperature changes (see below for detailed discussion). Never- power in predicting PL and AL with an exponent closer to 1.0 than either theless, more measurements on LWC and leaf photosynthetic rates are leaf N or P. Moreover, an inspection of the LOWESS curves (blue lines needed to further test how LWC mediates leaf photosynthesis-mass in the graphs of Fig. 4) showed that the scaling of PL and AL with respect scaling. to both leaf N and P had a stronger degree of curvature than the scaling Given that LWC was weakly correlated with ML (Supplementary of PL and AL with respect to MW. Thus, water appears to have a higher Fig. 2, log-log slope = −0.01, 95% CI = −0.02 and 0, r2 < 0.01, P = 0.17), it hierarchical status than nutrients in terrestrial ecosystems where is reasonable to assume that LWC is independent of ML in order almost all biological activities, including nutrient uptake, depend on to derive the change in temperature-corrected  PL and AL with respect water availability15. to differences  in ML i.e. ∂ P L eE=kT =∂M L = g 1 ek 1 LWC=ðK 1 + LWCÞ and Leaf N and P are considered important leaf traits in the leaf eco- ∂ AL eE=kT =∂M L = g 2 ek 1 LWC=ðK 1 + LWCÞ . These equations predict that nomics spectrum because of their close correlations with other leaf increases in temperature-corrected PL and AL with respect to ML should traits, such as PL and AL1,9,10. However, the numerical values of scaling become faster with increasing LWC when LWC is relatively low, and exponents are known to vary significantly across different plant achieve a relatively constant rate (plateau) when LWC is saturated, groups9, and no mechanistic model has been proposed to explain this such that k 1  LWC= K 1 + LWC ≈1 (Supplementary Fig. 3). Therefore, variation. In this study, we found that the scaling relationships of PL LWC plays an essential role in regulating the dynamics of whole-leaf and AL, with respect to leaf N and P, exhibited curvature, whereas using photosynthesis and area. MW resulted in much more robust and linear scaling relationships Nature Communications | (2022)13:5525 5 Article https://doi.org/10.1038/s41467-022-32784-1 Fig. 5 | The three-dimensional leaf trait relationships. The relationships of log- (g g−1). The dots projected on the floor and walls indicate bivariate correlations. The transformed Ps (nmol g−1 s−1, a) and SLA (cm2 g−1, b) with temperature (°C) and LWC planes represent the theoretical predicted relationships based on Eqs. (2) and (5). (Fig. 4). These analyses indicate that leaf water availability may be a ecosystem productivity are weakly correlated with growing-season air better indicator of PL and AL compared to leaf N and P, probably temperature33,43. This observed weak effect of temperature may be because (i) water serves as an essential biochemical reactant, solvent, probably attributed to species-specific short-term phenotypic accli- and nutrient carrier in plants15,20, which indicates that leaf water mation and long-term adaptation of plant photosynthetic traits to availability plays a more fundamental role in determining leaf growth changes in ambient temperature43,44. However, as discussed and metabolism than leaf N and P; and (ii) LWC is more sensitive to elsewhere33,43, we used long-term mean air temperature for the SLA changes in endogenous (e.g. developmental age) and exogenous fac- analyses, which may not effectively reflect the leaf temperature at tors (e.g. drought stress) thereby providing a better reflection of plant which leaf traits were measured. Future investigations are needed to growth status and changes in other leaf traits15,39. Leaf N shows a more critically examine the effect of leaf temperature on leaf trait variations robust scaling relationship with PL than P (Fig. 4a, b), which is rea- and trait scaling relationships. sonable because N is a major component of key enzymes involved in Although the effect of precipitation was not directly quantified in photosynthesis, such as Rubisco. our model, it might be partially accounted for by the effect of LWC because the plant water content is closely correlated with a variety of The important role of LWC in the leaf economics spectrum hydrological processes such as precipitation and evapotranspiration, Although LWC is a relatively understudied leaf trait, its important role soil water status, and plant growth16,20. Our analyses show that both in the leaf economics spectrum is indicated by the strong correlations LWC and SLAcor are overall weakly correlated with mean annual pre- between LDMC, a trait that is functionally related to LWC, and other cipitation (Supplementary Fig. 4), indicating that precipitation alone traits, such as SLA40. Recent work reveals that LDMC also affects the might not be able to fully explain leaf trait variation. In contrast, LWC ability of leaves to store and exchange heat in response to changes in shows a strong non-linear correlation with other leaf traits as predicted surface energy fluxes, which further influences leaf carbon by our model and as shown by using a global dataset, demonstrating assimilation26. In addition, LDMC is predicted to be an effective indi- that LWC is an integrative trait that plays a critical role in mediating leaf cator of plant resource use41 and ecosystem net primary production22, trait variability. which corroborates the crucial role of LWC in regulating leaf growth In this study, we combined a theoretical model and a global leaf and metabolism and, therefore, influencing productivity at the whole- trait dataset to quantify the effects of LWC and temperature on leaf plant and at the ecosystem levels. Although leaf functional traits have trait scaling relationships (Fig. 5). The theoretical framework presented been extensively explored in ecological studies1,7–12, leaf trait variation here provides a mechanistically deeper insight into the integrated across species or ecosystems, as well as the quantitative relationships functional traits of leaves and into how future changes in global cli- among leaf traits, are not well understood due to the lack of a general mate may affect leaf traits such as Ps and SLA (and LMA). Empirical model that captures the fundamental mechanism governing leaf trait observations and global climate simulations predict an overall rise in variation. Our findings indicate that LWC is an important leaf trait that temperature and especially an intensification of drought stress in shows non-linear quantitative relationships with other important leaf many regions of the world45,46. These climatic changes will undoubt- traits, i.e. Ps and SLA (or LMA), and directly regulates the leaf trait edly influence leaf traits, including Ps and SLA, directly through the scaling relationships. Although temporal variations in LWC might exist effect of temperature or indirectly through the drought-induced in response to instantaneous changes in microclimate conditions, changes in leaf structure and leaf water status. Our study highlights the previous work indicates that variation in LDMC (and thus LWC) is importance of LWC as a simple but powerful trait that can be easily relatively conserved across seasons42. Further investigations should be measured, and that quantitatively correlates with other important leaf conducted to better understand the temporal (e.g. seasonal and traits. Consequently, LWC should be included in the leaf economics yearly) dynamics of LWC and its link to the dynamics of other leaf traits spectrum to better understand the trait variations and scaling across different species and ecosystems. relationships. Relatively weak effects of temperature and precipitation Methods It is worth noting that the energy activation E in the Boltzmann factor is The model assumptions observed to be very close to 0 for both the leaf photosynthesis-mass In this study, we extended a previously published model19,20 to derive scaling (−0.13, 95% CI = −0.40 and 0.15) and leaf area-mass scaling the model for leaf trait relationships. The previously validated model relationships (0.03, 95% CI = 0.01 and 0.05), which is consistent was developed to quantify the effects of body mass and temperature with previous work showing that both individual tree growth and on plant metabolic rates based on the following assumptions. Nature Communications | (2022)13:5525 6 Article https://doi.org/10.1038/s41467-022-32784-1 First, cellular metabolism is limited by the water availability of plant compare differences in the numerical values of scaling exponents tissues, given the fundamental role of water in metabolism (e.g. as a among groups. Species belonging to geophytes, pteridophytes, and biochemical reactant and nutrient carrier)15. In addition, the effect of vines were only used for pooled data analysis due to their limited tissue water content on the mean rate of metabolic chemical reactions number of measurements. To investigate whether the numerical follows the Michaelis-Menten type hyperbolic response. Finally, the values of scaling exponents exhibited a latitudinal pattern, the data rate of change in cellular metabolism is proportional to the rate of were segregated into three latitudinal zones (tropical 0–25°, tempe- change in the critical metabolic chemical reaction. More details on the rate 25–50°, and boreal >50°). The bivariate scaling relationships model derivation are provided in Huang et al.19,20. among leaf traits for the pooled data and for each species group were fitted using standard major axis (SMA) regression, except for the Data synthesis scaling relationship between LWC and leaf mass, which was fitted using To test the predictions of our model, we merged the TRY plant trait ordinary least squares (OLS) regression because of the expectation database47, the Tundra Trait Team (TTT) database48, the BROT plant that the slope should be small and close to zero. These regression functional trait database49, and the GLOPNET database1, and a database analyses were performed using the lmodel2 function in the R package compiled from 12 additional peer-reviewed papers (see Supplemen- lmodel250. The leaf trait scaling relationships were also fitted by locally tary Data sources). We focused on the six key traits most commonly weighted sums of squares (LOWESS) smoothing lines to reveal the reported across the dataset: leaf area, leaf fresh and dry mass, SLA, and curvature across leaf size. The non-linear relationships of specific leaf LWC or LDMC. Data were included using two stringent criteria: (1) the photosynthetic rate and SLA with LWC and/or temperature T (Eqs. 2, 3, data for leaf dry mass, leaf area, LWC or LDMC was concurrently 5 and 6) were fitted to data by applying non-linear regression protocols measured and reported simultaneously; (2) only data from natural using the nls function from the R package car (ver. 2.0–25). In the non- ecosystems at non-polluted sites were used (this excluded data from linear regression analysis, we treated 1/kT as a whole to estimate the fertilised plants or plants grown in greenhouses). We also excluded activation energy E in Boltzmann’s factor. To obtain more robust data for which the identification of species was lacking. In total, the estimates of model parameters, we applied the bootstrapping resam- resulting dataset contained more than 17,000 trait measurements for a pling approach for non-linear regression analysis with 1000 re- total of 3427 species distributed globally across different ecosys- samplings using the function nlsBoot in the R package nlstools51. The tems (Fig. 1). three-dimensional plot was created using the scatter3D function in the Given the relatively limited data available for instantaneous leaf R package plot3D52. All analyses were conducted using the statistical photosynthetic rates with paired measurements of LWC, we used a software R53. specific dataset from the TRY plant trait database (i.e. the China Plant Trait Database38), which consists of paired leaf trait measurements Reporting summary (including leaf light-saturated photosynthesis rate, leaf area, LDMC, Further information on research design is available in the Nature and leaf N and P) from 1215 plant species across a variety of ecosys- Research Reporting Summary linked to this article. tems, to test the effect of leaf N and P mass versus leaf water mass on leaf photosynthesis and leaf area. We removed records with missing Data availability values of the selected leaf traits, resulting in a total of 404 measure- The TRY plant trait database is publicly available at https://www.try-db. ments. Leaf mass-specific photosynthetic rate (nmol g−1 s−1) was con- org. The TTT database is publicly available at https://github.com/ verted from area-specific photosynthetic rate (umol m−2 s−1) by TundraTraitTeam/TraitHub. The BROT plant functional trait database multiplying SLA (m2 kg−1). In this study, we used dry mass-based LWC is publicly available at https://www.uv.es/jgpausas/brot.htm. The (i.e. the ratio of leaf water mass to leaf dry mass, g g−1), which was GLOPNET database is publicly available at https://www.nature.com/ calculated from the fresh mass-based leaf water content or LDMC articles/nature02403 and also available from I.J.W. (ian.- reported in the global leaf trait dataset. wright@mq.edu.au) upon request. The China Plant Trait Database is publicly available at https://doi.org/10.1002/ecy.2091. The WorldClim Data analysis 2.1 database is publicly available at https://www.worldclim.org/. For leaf photosynthetic rates, the in situ temperature at which the rates were measured was used for model fitting. For SLA, the monthly Code availability average temperature data (with a spatial resolution of ~1 km2) at each The analysis was conducted using existing R packages, which are site were retrieved from the WorldClim 2.1 database (http://www. described in the “Methods”. The R codes for the data analysis are worldclim.org/). Following the previous work1, we defined “growing available from the corresponding author upon request. season months” as months with a mean temperature >4.99 °C. The growing-season temperature at each site was calculated as the average References temperature across growing-season months. To reduce errors intro- 1. Wright, I. J. et al. The worldwide leaf economics spectrum. Nature duced by transient temperature changes, we used long-term tem- 428, 821–827 (2004). perature data to calculate growing-season temperatures. This 2. Craine, J. M. Resource Strategies of Wild Plants. (Princeton Univ. “averaging” approach should be appropriate because of the great Press, 2009). number of field sites included in this study. The mean annual pre- 3. Lavorel, S. & Garnier, E. Predicting changes in community compo- cipitation (mm) at each site was also retrieved from the WorldClim 2.1 sition and ecosystem functioning from plant traits: revisiting the database to explore its correlations with LWC and SLA. Holy Grail. Funct. Ecol. 16, 545–556 (2002). Given the differences in leaf sample sizes across sites, leaf trait 4. Díaz, S. et al. The plant traits that drive ecosystems: evidence from values were averaged for each species at each sampling site to obtain three continents. J. Veg. Sci. 15, 295–304 (2004). species mean trait values. We note that we did not average the trait 5. Pearson, R. G. et al. Shifts in Arctic vegetation and associated values for the same species across sites because some species exist in feedbacks under climate change. Nat. Clim. Change 3, different ecosystems and latitudinal zones, which adds difficulty to the 673–677 (2013). comparison analyses among ecosystem types and latitudinal gra- 6. Bjorkman, A. D. et al. Plant functional trait change across a warming dients. Species were categorised into one of the four plant growth tundra biome. Nature 562, 57–62 (2018a). forms (trees, shrubs, graminoids, and forbs) and into one of the five 7. Reich, P. B. et al. Generality of leaf trait relationships: a test across ecosystem types (forest, grassland, desert, tundra, and wetland) to six biomes. Ecology 80, 1955–1969 (1999). Nature Communications | (2022)13:5525 7 Article https://doi.org/10.1038/s41467-022-32784-1 8. Milla, R. & Reich, P. B. The scaling of leaf area and mass: the cost of 32. Gillooly, J. F., Brown, J. H., West, G. B., Savage, V. M. & Charnov, E. L. light interception increases with leaf size. Proc. R. Soc. B Biol. Sci. Effects of size and temperature on metabolic rate. Science 293, 274, 2109–2114 (2007). 2248–2251 (2002). 9. Wright, I. J. et al. Assessing the generality of global leaf trait rela- 33. Michaletz, S. T., Cheng, D. L., Kerkhoff, A. J. & Enquist, B. J. Con- tionships. N. Phytol. 166, 485–496 (2005). vergence of terrestrial plant production across global climate gra- 10. Niklas, K. J. et al. “Diminishing returns” in the scaling of functional dients. Nature 512, 39–43 (2014). leaf traits across and within species groups. Proc. Natl Acad. Sci. 34. Niinemets, Ü. Global-scale climatic controls of leaf dry mass per USA 104, 8891–8896 (2007). area, density, and thickness in trees and shrubs. Ecology 82, 11. Osnas, J. L. D., Lichstein, J. W., Reich, P. B. & Pacala, S. W. Global leaf 453–469 (2001). trait relationships: mass, area, and the leaf economics spectrum. 35. Wang, H. et al. Leaf economics explained by optimality principles. Science 340, 741–744 (2013). bioRxiv https://doi.org/10.1101/2021.0207430028 (2021). 12. Keenan, T. F. & Niinemets, Ü. Global leaf trait estimates biased due 36. Shipley, B. Structured interspecific determinants of specific leaf to plasticity in the shade. Nat. Plants 3, 16201 (2016). area in 34 species of herbaceous angiosperms. Funct. Ecol. 9, 13. Westoby, M., Falster, D. S., Moles, A. T., Vesk, P. A. & Wright, I. J. 312–319 (1995). Plant ecological strategies: some leading dimensions of variation 37. Meziane, D. & Shipley, B. Direct and indirect relationships between species. Ann. Rev. Ecol. S. 33, 125–159 (2002). between specific leaf area, leaf nitrogen and leaf gas exchange. 14. Poorter, H., Niinemets, Ü., Poorter, L., Wright, I. J. & Villar, R. Causes Effects of irradiance and nutrient supply. Ann. Bot. 88, and consequences of variation in leaf mass per area (LMA): a meta- 915–927 (2001). analysis. N. Phytol. 182, 565–588 (2009). 38. Wang, H. et al. The China Plant Trait Database: towards a compre- 15. Sibly, R. M., Brown, J. H. & Kodric-Brown, A. Metabolic Ecology: A hensive regional compilation of functional traits for land plants. Scaling Approach. (Wiley-Blackwell Press, 2012). Ecology 99, 500 (2018). 16. Huxman, T. E. et al. Convergence across biomes to a common rain- 39. Wang, R., He, N., Li, S., Xu, L. & Li, M. Spatial variation and use efficiency. Nature 429, 651–654 (2004). mechanisms of leaf water content in grassland plants at the biome 17. Breshears, D. D. et al. Tree die-off in response to global change-type scale: evidence from three comparative transects. Sci. Rep. 11, drought: mortality insights from a decade of plant water potential 9281 (2021). measurements. Front. Ecol. Environ. 7, 185–189 (2009). 40. Garnier, E. et al. Consistency of species ranking based on functional 18. Anderson-Teixeira, K. J., Delong, J. P., Fox, A. M., Brese, D. A. & leaf traits. N. Phytol. 152, 69–83 (2001). Litvak, M. E. Differential responses of production and respiration to 41. Garnier, E., Shipley, B., Roumet, C. & Laurent, G. A standardized temperature and moisture drive the carbon balance across a cli- protocol for the determination of specific leaf area and leaf dry matic gradient in New Mexico. Glob. Change Biol. 17, matter content. Funct. Ecol. 15, 688–695 (2001). 410–424 (2011). 42. Raouda, A. H. K., Michel, D., Pierre, T. J., Sylvain, P. & Pablo, C. 19. Huang, H. et al. A general model for seed and seedling respiratory Variation in leaf traits through seasons and N-availability levels and metabolism. Am. Nat. 195, 534–546 (2020). its consequences for ranking grassland species. J. Veg. Sci. 16, 20. Huang, H. et al. Water content quantitatively affects metabolic rates 391–398 (2005). over the course of plant ontogeny. N. Phytol. 228, 1524–1534 (2020). 43. Michaletz, S. T. Evaluating the kinetic basis of plant growth from 21. Vendramini, F. et al. Leaf traits as indicators of resource-use strat- organs to ecosystems. N. Phytol. 219, 37–44 (2018). egy in floras with succulent species. N. Phytol. 154, 147–157 (2002). 44. Helliker, B. R. & Richter, S. L. Subtropical to boreal convergence of 22. Smart, S. M. et al. Leaf dry matter content is better at predicting tree-leaf temperatures. Nature 454, 511–514 (2008). above-ground net primary production than specific leaf area. Funct. 45. Dai, A. G. Increasing drought under global warming in observations Ecol. 31, 1336–1344 (2017). and models. Nat. Clim. Change 3, 52–58 (2013). 23. Zhang, Y., Zhou, S., Gentine, P. & Xiao, X. M. Can vegetation optical 46. Feng, S. & Fu, Q. Expansion of global drylands under a warming depth reflect changes in leaf water potential during soil moisture climate. Atmos. Chem. Phys. 13, 10081–10094 (2013). dry-down events? Remote Sens. Environ. 234, 111451 (2019). 47. Kattge, J. et al. TRY plant trait database-enhanced coverage and 24. Leigh, A. et al. Do thick leaves avoid thermal damage in critically low open access. Glob. Change Biol. 26, 119–188 (2020). wind speeds? N. Phytol. 194, 477–487 (2012). 48. Bjorkman, A. D. et al. Tundra Trait Team: a database of plant traits 25. Zwieniecki, M. Stomatal control and leaf thermal and hydraulic spanning the tundra biome. Glob. Ecol. Biogeogr. 27, capacitances under rapid environmental fluctuations. PLoS ONE 8, 1402–1411 (2018). e54231 (2013). 49. Tavşanoğlu, Ç. & Pausas, J. G. A functional trait database for Med- 26. Michaletz, S. T. et al. Plant thermoregulation, energetics, trait‐ iterranean Basin plants. Sci. Data 5, 180135 (2018). environment interactions, and carbon economics. Trends Ecol. 50. Legendre, P. lmodel2: Model II regression. R package v.1.7-3. Evol. 30, 714–724 (2015). [WWWdocument]. https://CRAN.R-project.org/package= 27. Sterner, R. W. & Elser, J. J. Ecological Stoichiometry-the Biology of lmodel2 (2018). Elements from Molecules to the Biosphere. (Princeton Univ. 51. Baty, F. et al. The toolbox for nonlinear regression in R: the package Press, 2002). nlstools. J. Stat. Softw. 66, 1–21 (2015). 28. Brown, J. H., Gillooly, J. F., Allen, A. P., Savage, V. M. & West, G. B. 52. Soetaert, K. plot3D: Plotting Multi-Dimensional Data. R package Toward a metabolic theory of ecology. Ecology 85, version 1.4. https://CRAN.R-project.org/package=plot3D. (2016). 1771–1789 (2004). 53. R Core Team. R: A Language and Environment for Statistical Com- 29. West, G. B., Brown, J. H. & Enquist, B. J. A general model for the puting. R Foundation for Statistical Computing, Vienna, Austria. structure and allometry of plant vascular systems. Nature 400, https://www.R-project.org/. (2020). 664–667 (1999). 30. Niklas, K. J. & Enquist, B. J. On the vegetative biomass partitioning of seed plant leaves, stems, and roots. Am. Nat. 159, 482–497 (2002). Acknowledgements 31. Enquist, B. J. et al. A general integrative model for scaling plant We are grateful to the scientists who contributed the valuable data for growth, carbon flux, and functional trait spectra. Nature 449, this study. This work was supported by National Natural Science Foun- 218–222 (2007). dation of China (32022052, 31971495), International Science and Nature Communications | (2022)13:5525 8 Article https://doi.org/10.1038/s41467-022-32784-1 Technology Cooperation Project of Qinghai province of China (2022-HZ- Peer review information Nature Communications thanks the anon- 817), the Spanish Government (grants PID2019−110521GB-I00 and ymous reviewer(s) for their contribution to the peer review of this work. PID2020115770RB-I), Fundación Ramon Areces (grant ELEMENTAL-CLI- MATE), Catalan Government (grants SGR 2017−1005 and AGAUR- Reprints and permission information is available at 2020PANDE00117), and European Research Council (Synergy grant http://www.nature.com/reprints ERC-SyG-2013-610028, IMBALANCE-P). Publisher’s note Springer Nature remains neutral with regard to jur- Author contributions isdictional claims in published maps and institutional affiliations. Z.W., H.H., and I.W. conceived the study. H.H. and Z.W. developed the model. Z.W., H.H., J.X., Y.L., and H.W. collected data and performed the Open Access This article is licensed under a Creative Commons analysis. J.P., J.S., Ü.N., and I.W. contributed to new analysis. Z.W. and Attribution 4.0 International License, which permits use, sharing, H.H. wrote the first draft, and all authors contributed to the final version adaptation, distribution and reproduction in any medium or format, as of the manuscript. long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if Competing interests changes were made. The images or other third party material in this The authors declare no competing interests. article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not Additional information included in the article’s Creative Commons license and your intended Supplementary information The online version contains supplementary use is not permitted by statutory regulation or exceeds the permitted material available at use, you will need to obtain permission directly from the copyright https://doi.org/10.1038/s41467-022-32784-1. holder. To view a copy of this license, visit http://creativecommons.org/ licenses/by/4.0/. Correspondence and requests for materials should be addressed to Heng Huang. © The Author(s) 2022 Nature Communications | (2022)13:5525 9