2.1 Leaf trait data

We collected data on leaf hydraulic efficiency (K_leaf-max) and safety (K_leafP₅₀), stomatal conductance (g_s), and photosynthetic rate (A) from 232 relevant publications (Supplementary material; Table S1), only considering woody plants confined to their natural habitat. We excluded laboratory incubation studies and manipulative experiments conducted under any specific treatment. Graphical data was digitized using Origin Pro 8 SR0 (Origin Lab Corporation). We calculated the ratio of hydraulic supply to demand (K_leaf/g_s) and intrinsic water use efficiency (WUE_i) using the ratio of K_leaf-max/g_s-max and A_max/g_s-max, respectively. We also compiled values of dark respiration rate (R_d-max – μmol m⁻² s⁻¹), leaf area-based nitrogen (N_area – g m⁻²) and phosphorus (P_area – g m⁻²) content, leaf mass per unit area (LMA – g m⁻²), and leaf longevity (LL – months) from the TRY global plant trait database (https://www.try-db.org/TryWeb/Home.php) (Kattge et al., 2011). We used all replicate values to account for variability and increase the robustness of our statistical analysis. Our dataset included 7949 observations of 9 variables related to leaf hydraulic (K_leaf-max, K_leaf P₅₀) gas exchange (A_max, g_s-max), and economic traits (A_max, R_d-max, LMA, LL, N_area, P_area), collected across 3391 species.

2.2 Climatic variables

We obtained the geographical coordinates (i.e. latitude & longitude) for each species' native habitat from the Global Biodiversity Information Facility (GBIF; https://www.gbif.org). Given the abundance of coordinates for each species, we based the species' current native habitat on the occurrence from the most recent year. When multiple records were found in the most recent year, we included all the corresponding coordinates. Using these coordinates, we extracted values for altitude, mean annual precipitation (MAP) and mean annual temperature (MAT) from raster maps downloaded from the WorldClim historical climate database (http://www.worldclim.org) (Fick & Hijmans, 2017) with the help of Arc Map (v.10.8) software. Additionally, we downloaded 30 years of monthly precipitation and temperature data (1970–2000) from the WorldClim database. For the months with ≥4°C mean temperature and precipitation ≥2X mean temperature, we then calculated growing season mean climatic variables, such as temperature, precipitation, and growing season length (Lasky et al., 2012). Annual heat moisture index (HMI) was calculated as (MAT +10) / (MAP/1000) (Roach et al., 2021). We acquired atmospheric pressure data since 1979 from the Multi-Source Weather (MSWX) product, a comprehensive meteorological dataset that has been corrected for biases. This dataset provides global coverage and has a spatial resolution of 0.1° (Beck et al., 2022). The partial pressures of O₂ and CO₂ were calculated by multiplying the fraction of the gas by total atmospheric pressure (Hlastala 2006). Using ArcMap (v.10.8) software, we generated a world map to represent the geographical distribution of species along the altitude gradient (Figure 1).

2.3 Phylogenetic tree

We constructed a phylogenetic tree of 3,391 woody plant species using the phylo.maker function in the V. PhyloMaker package in R (R Core Team, 2022) (Figure S1). Blomberg's K statistic was used to quantify the extent to which related species resemble one another in terms of leaf hydraulic, gas exchange and economic traits.

2.4 Data analysis

A simple linear regression analysis was used to assess both the relationships between environmental variables and altitude and trait–trait relationships. We performed a linear mixed-effects model using the R package “lme4” to evaluate the relationships between leaf hydraulic, gas exchange and economic traits and altitude in each site (i.e. latitude and longitude), with species used as random effects and altitude as a fixed effect. We conducted Spearman's correlation analysis to evaluate the trait-climate associations, and results were visualized using a heatmap. Bivariate relationships of leaf traits with climate were evaluated by performing standardized major axis (SMA) regression analysis in the R package “smatr”. We estimated Blomberg's K to evaluate the phylogenetic signals of leaf traits among woody species using the R package “phytools”. All statistical analysis was completed in R 3.1.0 (https://www.r-project.org/).

3.1 Changes in climatic variables along an altitudinal gradient

Mean annual precipitation (MAP), growing season precipitation (P_gs) and the annual heat moisture index (HMI) showed statistically significant negative correlations with altitude (Figure 2; R² = 0.01–0.08, p < 0.001). In addition, mean annual temperature (MAT) and growing season temperature (T_gs) decreased linearly with altitude (Figure 2; R² = 0.33–0.34, p < 0.001) and explained about 34% of variation by altitude. Among all climatic variables used in our analysis, altitude displayed a negative correlation (Figure 2; R² = 0.90–0.91, p < 0.001) and explained the greatest proportion of variance (about 90%) in atmospheric pressure (AP), partial pressure of O₂ (PPO₂) and partial pressure of CO₂ (PPCO₂). We also observed a substantial explanatory power of altitude on growing season length (GSL), with GSL decreasing as altitude increased (Figure 3; R² = 0.34, p < 0.001).

3.2 Changing patterns of leaf hydraulic, gas exchange and economic traits with increasing altitude

To examine global patterns of leaf hydraulic traits in relation to altitude, we performed analysis using a linear mixed-effects model, with sites and species as random effects, and showed that leaf hydraulic efficiency (K_leaf-max), ranging between 0.98 to 100 mmol m⁻² s⁻¹ MPa⁻¹, declined linearly with increasing altitude (Figure 4a; R² = 0.12, p = 0.01). Similarly, we observed a negative correlation between leaf water potential at 50% loss of K_leaf (K_leaf P₅₀) and altitude, with values ranging from −0.09 to −8.35 (Figure 4b; R² = 0.10, p < 0.001). Thus, species adapted to higher altitudes tended to exhibit lower K_leaf-max and more negative K_leaf P₅₀ compared to species native to lower altitudes, demonstrating a hydraulic efficiency-safety trade-off relationship (Figure 4d; R² = 0.23, p < 0.001).  Furthermore, linear mixed-effects models of stomatal conductance (g_s-max) with altitude showed decreasing trends (Figure 4c; p < 0.001) in which fixed effects alone explained a modest portion of variation (marginal R² = 0.06), whereas both random and fixed effects explained most of the variation (conditional R² = 0.94).

At high altitudes, the significant decrease in K_leaf-max values also led to a lower K_leaf/g_s ratio (Figure 5a; R² = 0.14–0.15, p < 0.001). In contrast, water use efficiency (WUE_i) displayed a positive correlation with increasing altitude (Figure 5b; marginal R² = 0.28, conditional R² = 0.96, p < 0.001), and species at high altitudes had mean values of 98.79 for WUE_i, compared to the 57.86 of species at lower altitudes. We also observed upregulation of WUE_i in species with a shorter growing season (Figure S2 R² = 0.05, p < 0.001). Both photosynthesis (A_max) and dark respiration (R_d) rates declined with increasing altitude and, together, random and fixed effects explained most of the variation (Figure 6; marginal R² = 0.03–0.08, conditional R² = 0.98–0.99, p < 0.001). In addition, with increasing altitude, leaf mass area (LMA) and leaf longevity (LL) increased linearly, and area-based nitrogen (N) and phosphorus (P) content decreased linearly (Figure 7; marginal R² = 0.01–0.03, conditional R² = 0.01–0.99, p < 0.001). The lower A_max in high-altitude species was associated with lower N and P contents (Figure S3; R² = 0.07–0.15, p < 0.001).

3.3 Relationships between leaf traits and climate variables

Most of the environmental variables had either a positive or a negative influence on leaf hydraulic, gas exchange and economic traits, as shown by the Pearson's correlations displayed in the heatmap (Figure 8). Following standardized major axis (SMA) regression analysis, AP, PPO₂ and PPCO₂ explained the majority of variation in leaf traits, while other variables, such as MAP, MAT, P_gs, T_gs and HMI, accounted for smaller proportion of variation (Table S2).

3.4 Phylogenetics

Leaf hydraulic safety (K_leaf P₅₀) displayed a significant phylogenetic signal among woody plant species, as indicated by Blomberg's K > 1, whereas all the other studied leaf traits showed weak phylogenetic control, with Blomberg's K < 1 (Table 1; Figure S1).

4.1 Modulation of leaf hydraulic traits by altitude

We compared global patterns of leaf hydraulic efficiency and safety along an altitudinal gradient and found that high-altitude species tend to exhibit lower hydraulic efficiency (K_leaf-max) but their greater hydraulic safety (K_leaf P₅₀) gives them a competitive advantage (Figure 4). Thus, our findings support the hypothesis that species prioritize the construction of safer water transport systems (a more negative K_leaf P₅₀) over hydraulic efficiency to adapt to high-altitude environments (Liu et al., 2019; Tyree & Sperry, 1989). Indeed, it has been shown that freeze–thaw events trigger anatomical changes (i.e. thickening of cell walls and modifications of xylem vessels and vein density) that help plants withstand lower temperatures at higher altitudes by effectively preventing the formation and spread of xylem embolism (Mayr et al., 2003; Petit et al., 2011). These traits enable plants to resist hydraulic impairment by constructing safer water transport systems, maintaining leaf hydraulic function, and consequently, sustaining photosynthetic activity (Mayr et al., 2006; Wang et al., 2018; Waseem et al., 2021, 2024; Scoffoni et al., 2016; Yao et al., 2021a).

Moreover, our study revealed the phylogenetic signal in K_leaf P₅₀ (Table 1), indicating that both evolutionary relationships and environmental pressures are involved in shaping the plasticity of K_leaf P₅₀ among woody species. Pieces of evidence suggest that environmental conditions exert selective pressure while evolutionary history shapes the genetic and structural limits within which these adaptations can occur (Liu et al., 2022; Shahzad et al., 2024; Zhang et al., 2011). Anyway, these patterns indicate that hydraulic safety play a significant role in shaping species distribution along altitude gradients.

Although K_leaf-max increased with altitude among five Rhododendron species (Taneda et al., 2016), our global study of diverse species with greater environmental heterogeneity indicated that K_leaf-max decreased with increasing altitude. The decrease of K_leaf-max with increasing altitude has also been observed in two shrub species (Yao et al., 2024). It has been suggested that higher VLA increases K_leaf-max by shortening the post-venous hydraulic pathway, thereby facilitating more efficient water transportation (Scoffoni et al., 2016), and that altitude-related trends of VLA have shown lower VLA with increasing altitude, tightly related to low K_leaf-max (Yao et al., 2024). Our data show that species growing at higher altitudes tend to have a lower K_leaf/g_s ratio than those at lower altitudes (Figure 5a). At lower altitudes, a higher K_leaf/g_s ratio would help alleviate the increased evaporative demand caused by high vapour pressure deficit (VPD) (Will et al., 2013). Thus, evaporative cooling helps reduce leaf temperature (Brodribb et al., 2007; Scoffoni et al., 2011, 2016; Yao et al., 2021). In contrast, a higher K_leaf/g_s ratio might not be necessary at higher altitudes due to the low VPD in cooler temperatures (Will et al., 2013).

Previous global studies have reported a general hydraulic efficiency-safety trade-off among grasses, shrubs and angiosperm species (Ocheltree et al., 2016; Yan et al., 2020; Yao et al., 2021b). Our findings indicate that species inhabiting high-altitude environments exhibit lower K_leaf-max and more negative K_leaf P₅₀, whereas species in lower altitudes tend to have higher K_leaf-max and less negative K_leaf P₅₀ (Figure 4). This global study strongly supports the hydraulic efficiency-safety trade-off framework along the altitudinal gradient.

4.2 Modulation of gas exchange and leaf economic traits by altitude

The optimization of water use efficiency largely depends on constructing a safe and effective water transport system as well as possessing a coordinated signal pathway to finely adjust stomatal regulation (Brodribb et al., 2017; Yang et al., 2021). Species growing at high altitudes, which experience shorter growing seasons, are characterized by higher water use efficiency compared to those at lower altitudes (Figure 5b, S2). This increased efficiency may be associated with greater sensitivity of stomatal closure to low temperatures and precipitation or to reduced atmospheric pressure or low hydraulic supply in high-altitude environments (Leakey et al., 2019). The increased stomatal sensitivity at high altitudes reduces the rate of evaporation and, in turn, enhances the efficiency of water usage (Picotte et al., 2007; Zou et al., 2010).

Numerous studies have examined the impact of climatic factors, such as precipitation and temperature, on water use efficiency. These studies have shown that species in low precipitation environments increase their water use efficiency (Ponce-Campos et al., 2013; Ogaya and Peñuelas, 2003; Xiao et al., 2013). Similarly, the increased water-use efficiency at high altitudes enables plants to allocate limited water resources more effectively to photosynthetic processes. This adaptation helps mitigate the impact of hydraulic supply limitations and compensates for the short growing season, allowing plants to accumulate more assimilated carbon and enhance their tolerance to extreme alpine environments (Bresson et al., 2009; Ogaya & Peñuelas, 2003). This provides a competitive advantage for plants at high altitudes by maintaining carbon fixation under lower temperatures.

The interrelationship of leaf economic traits has been reported previously both at local and larger scales (Reich et al., 1997, 1999; Wright et al., 2001; Wright & Westoby, 2003; Wright et al., 2004). Here, we have generalized the modulation of these traits by altitude on a global scale. The results showed that species at lower altitude exhibited high foliar N and P content, A_max, R_d-max, coupled with lower LMA and LL. Conversely, species at higher altitude exhibited lower A_max, R_d-max, and foliar N and P contents, but with higher LMA and LL (Figures 6 and 7). From lower to higher altitudes, the observed trait-altitude interactions indicate a continuum of adaptation to temperature, precipitation, and resource constraints along an altitudinal gradient. Thus, species adapted to high altitude environments tend to exhibit a conservative leaf economic trait strategy than those in lower altitude environments at global scale. Species inhabiting high-altitude environments tend to have leaves requiring greater foliar construction costs (high LMA) to cope with stressors such as reduced precipitation, low temperature, strong radiation, and wind (Midolo et al., 2019; Poorter et al., 2009; Zhang et al., 2020).

There are several key aspects in which high LMA can be seen as a strategy of adaptation for high altitude-prone species. Firstly, high LMA with increasing altitude tends to improve the strength or elastic modulus of leaves (Nadal et al., 2023), thus to protect them from damage caused by low temperatures or wind at high altitudes (Lutz, 2010). Secondly, this can be regarded as a strategy to maintain carbohydrate production in stressful environments (Poorter et al., 2009). Finally, structural reinforcement with high LMA protects species from herbivory and other physical hazards (Reich et al., 1997; Wright & Cannon, 2001), and can thereby increase leaf longevity (LL). Furthermore, in our study, the increase in LMA also resulted in a corresponding increase in LL at high altitude environments (Figure 7 a,b), thus suggesting a strong correlation between LMA and LL (Wright et al., 2004, 2005; Zhang et al., 2010), and longer LL can optimize their nitrogen use efficiency by extending N mean residence time, especially where soil decomposition, mineralization, and nutrient absorption rates may be restricted by low temperatures (Escudero et al., 1992; Eckstein et al., 1999; Körner & Paulsen, 2004; Sullivan et al., 2015).

In our global study, species at high-altitude sites displayed lower values of A_max and R_d-max that might be due to the decline in the enzymatic activity of Rubisco caused by colder temperatures and smaller CO₂ concentrations in the liquid phase of mesophyll cells (Terashima et al., 1995). Another plausible explanation for this lower gas exchange is the correlation of foliar N and P with A_max and R_d-max. The bulk of nitrogen found in the photosynthetic complex contributes to higher A_max, and to a large extent, phloem loading of photosynthate contributes to the dark respiration rate (Field 1986; Lambers et al., 1998), but species at higher altitudes have lower N. Furthermore, because bioenergetic molecules like ATP and NADPH play a crucial role in photosynthesis and respiration, these metabolic processes (A_max, R_d-max) correlate tightly with phosphorus (Atwell 1999), but species at higher altitudes have lower P. Other factors, such as lower atmospheric pressure and lower partial pressure of O₂ and CO₂, may also reduce photosynthetic efficiency in high-altitude environments (Friend & Woodward, 1990; Takeishi et al., 2013; Viveros et al., 2024).

Based on the shift of traits, species at lower altitudes exhibited an acquisitive strategy that takes advantage of abundant resources. This is because higher temperatures at lower altitudes stimulate microbial activity and make more resources available for these species to exploit (Raich & Schlesinger,  1992). Conversely, species at higher altitudes exhibited a conservative strategy exists to cope in less fertile ecosystems. Low temperatures at high altitudes hamper soil biological activity, thus further limiting soil decomposition and mineralization and reducing the availability of soil nutrients (N, P) as altitude increases (Sundqvist et al., 2013; Vincent et al., 2014).