The Qinghai-Tibetan Plateau boasts a unique natural environment and spatial differentiation, which is attributed to the reduction of atmospheric circulation and the distinct topography of the plateau (Li et al., 2022). The distinctive geographical combination of water and thermal conditions created ideal experimental sites for this research. The research sites were primarily situated in the eastern Qinghai-Tibetan Plateau, across the Yunnan, Sichuan, and Gansu provinces of China (25.72 °N - 33.67 °N, 98.49 °E - 104.82 °E). The mean annual precipitation (MAP) at the research sites ranged from 525 mm to 1240 mm, and the mean annual temperature (MAT) ranged from -4 °C to 21 °C. Additionally, the sites contained a large elevation drop (860 m - 4200 m) and an array of vegetation types, including alpine shrub, subalpine coniferous forest, subalpine coniferous and broad-leaved mixed forest, dry valley shrub, and dry-hot valley shrub (Du et al., 2020). These natural experimental conditions provided an excellent basis for our investigation into the leaf traits of woody plants ( Figure 1 ; Table S1 ).

Field surveys and sampling were conducted from 2018 to 2020 during the peak period for plant growth (July to September). The 50 broadly representative woody sample sites on the eastern Qinghai-Tibetan Plateau were selected and 3 - 5 plots were set up (forest, 20 m × 20 m; shrub, 10 m × 10 m) at each site. To minimize human interference, the plots were set up within natural reserves, and we recorded the longitude and latitude of each plot using a global positioning system (Thales Navigation, Santa Barbara, CA, USA). Five to seven individuals of each species with 15 - 20 g leaves per individual within a plot were collected as one sample for measuring leaf traits. Expanded, mature, and sun-exposed leaves of trees and shrubs were collected. Totally, the 771 samples were collected from 110 species across the 50 sites, including three plant functional types (PFTs) (e.g., needle-leaved evergreens, NE; broad-leaved evergreens, BE; and broad-leaved deciduous, BD).

The study measured six leaf traits, including leaf thickness (LT, mm), leaf dry matter content (LDMC, mg/g), leaf dry mass per area (LMA, mg/cm²), nitrogen content per unit area (N_area, mg/cm²), nitrogen content per unit mass (N_mass, mg/g), and carbon:nitrogen ratio (C/N) (Rawat et al., 2021). LT was measured using a vernier caliper with an accuracy of 0.01 mm, and the leaf fresh weight (LFW) was weighed using an electronic balance with an accuracy of 0.001 g. The leaf area (LA) was calculated using image analysis software (image J v.1.8.0) after scanning the leaves using a scanner (Seiko Epson Co., Nagano, Japan). After recording all the measurements, the leaf samples were baked in an oven at 75°C for 72 hours to a constant weight, and the leaf dry weight (LDW) was weighed. N_mass, N_area, and C/N were analyzed after the leaf samples were ground into fine powder using a steel ball mixing mill MM200 (Retsch GmbH, Haan, Germany) (Cornelissen et al., 2003; Pérez-Harguindeguy et al., 2013). LMA and LDMC were calculated using the following equations:

Meteorological data for the study was obtained from China Ecosystem Research Network. The dataset was generated using data from 2400 weather stations of the China Meteorological Administration spanning from 1980 to 2020 and had a spatial resolution of 1 × 1 km (https://data.cma.cn). MAT and MAP values for each plot were extracted using the interpolation software of ANUSPLIN (v.4.36) (Hutchinson and Xu, 2013) from climate dataset. This allowed us to obtain accurate and representative climatic conditions of the study area.

The data analyzed in this study include species-mean trait values for each species sampled at each plot (Dong et al., 2020; Wang et al., 2022). Linear mixed models were used to explore whether the variation of leaf traits was independent of PFTs identity. In this analysis, species were used as fixed effects and PFTs were used as random effects (package Ime4, Ismeans). LSD’s post-hoc test was then performed to compare the differences in leaf traits among three PFTs. Principal component analysis (PCA) was utilized to explore differences in major functional dimensions among three PFTs. Linear models and Pearson’s correlation were applied to determine relationships between leaf traits. Linear models were used to explore whether the effect of environmental variability on leaf traits was independent of PFTs. At the same time, linear mixed models with PFTs as random effects and environmental factors as fixed effects were used to compare and verify the results of the linear model. Because there was no obvious difference between the results of the two models, the results of the linear model were finally selected for analysis (Warton et al., 2006). To quantify the relative contributions of PFTs, MAT, and MAP on leaf traits, partial general linear models analyses were used with leaf traits as dependent variables and PFTs, MAT, and MAP as predictors. The partial regressions was used to divide the variation in response variables explained by predictive variables into independent components (PFTs, MAT, MAP) and joint components (PFTs - MAT, PFTs - MAP, MAT - MAP, PFTs - MAT - MAP) (Zhan et al., 2018). Additionally, linear mixed models and linear models were conducted using R v.3.6.1 (R Core Team, 2019), the other analyses were performed using SPSS v.23.0 (Field, 2013), the graphs were performed using Origin v.2021 (Chen, 2006).

The result indicated that there were significant differences in leaf traits among different PFTs (p<0.01) ( Table 1 ). LT, LDMC, LMA, C/N, and N_area of NE were significantly higher than those of BE and BD, and showed the pattern of NE>BE>BD. In contrast, N_mass showed the opposite trend, with NE<BE<BD. Principal components analysis (PCA) of the data showed that the multivariate space occupied by three PFTs was distinct ( Figure 2 ). The first and second PC axes explained 74.8% and 12.8%, respectively. PC1 was strongly positively related to LT, LDMC, LMA, C/N, and N_area, and negatively related to N_mass ( Figure 2 ).

Significant correlations were observed between leaf traits, which varied among three PFTs ( Figure 3 ; Table 2 ). In most cases, the correlations between leaf traits of three PFTs were similar. For instance, a positive correlation was observed among C/N-LDMC ( Figure 3A ), LMA-LT ( Figure 3C ), C/N-LMA ( Figure 3D ), LDMC-LMA ( Figure 3E ), N_area-LT ( Figure 3I ), and C/N-LT ( Figure 3J ), while a negative correlation was observed between LMA-N_mass ( Figure 3F ), LDMC-N_mass ( Figure 3H ), and N_mass-LT ( Figure 3L ). However, in only a few cases, the slope relations were significantly different among three PFTs ( Figures 3B, G, K ; Table 2 ). The common slope test showed that the scaling index varied significantly among the PFTs, indicated that the scaling relationship was dependent on the PFTs associated with most of the bivariate traits examined ( Figure 3 ; Table 2 ).

The analysis revealed effects of elevation, MAT, and MAP on leaf traits, and these effects varied across PFTs ( Figure 4 ). While leaf traits of BE and BD exhibited similar changes with increasing elevation and MAT, NE showed a different pattern ( Figures 4A, B, D, E, G, H, J, K, M, N ). Leaf traits of three PFTs exhibited similar changes with increasing MAP, but it is not significant in most cases ( Figures 4C, F, I, L, O ).

The variations in leaf traits were mainly explained by PFTs, with a higher proportion than that of MAT and MAP ( Figure 5 ). The explained fractions of LT, LDMC, LMA, C/N, N_mass, and N_area by PFTs were as high as 62.2%, 18.2%, 63.3%, 41.4%, 27.3%, and 51.5%. On the other hand, MAT and MAP had relatively lower explanatory power for variations of leaf traits, but had a higher proportion for LDMC and N_mass. Compared with MAP, MAT had relatively high explanatory power ( Figures 5B, E ).

The result showed that the mean values of leaf traits differed among the three PFTs, with NE plants exhibiting greater values of LT, LDMC, LMA, C/N, and N_area compared to BE plants and BD plants ( Table 1 ). It can be explained by NE plants’ ability to resist hostile environments, which is significantly stronger than BE plants and BD plants (Fyllas et al., 2020). The thickness of leaves can help plants avoid damage from strong light and low temperature, and it provided a protective substrate for plants to survive in hostile environment (Körner et al., 1986; Qi et al., 2014). Meanwhile, the leaf with higher LDMC had smaller intercellular space and the more gas diffusion resistance from mesophyll cells, these led to lower photosynthetic carbon assimilation capacity (Hamdani et al., 2019; Maccagni and Willi, 2022). Furthermore, according to the leaf economic theory, higher LMA indicated that NE plants adopted more conservative survival strategy, while BD plants adopted acquired survival strategy (Maracahipes et al., 2018; Fyllas et al., 2020; Rawat et al., 2021; Liu et al., 2022a). This was also confirmed in a study of woody plants from the temperate forest in the Himalayan region of India, where plants adopted conservation or acquisition strategies to adapt to environmental change (Rawat et al., 2021). The C/N and N_area of NE plants and BE plants were significantly higher than BD plants, while N_mass was lower ( Table 1 ). This suggested that the nitrogen invested in the photosynthetic system of evergreen leaves (NE and BE) was relatively lower, and the remaining nitrogen was invested in non-photosynthetic systems such as cell wall proteins, lipids, and amino acids. More nitrogen was used in the construction of leaf tissue structure to make leaves tougher which enhanced the ability of evergreen plants to resist stress (Mediavilla et al., 2012; Ghimire et al., 2017; Byeon et al., 2021). On the other hand, BD plants were completely on the contrary, and they increased the nitrogen investment in the photosynthetic system so that plants with shorter leaf life could assimilate as much carbon dioxide as possible at a limited time to ensure growth and development. This indicated that there were differences in the survival strategies among three PFTs (Berveiller et al., 2007; Bai et al., 2015; Weng et al., 2017). These differences in survival strategies among the three PFTs were further confirmed by the principal component analysis ( Figure 2 ), which showed differences in the multivariate space occupied by three PFTs. The smaller intersection of NE plants and BD plants indicated that they adopted different survival strategies, while BE plants were somewhere in between. The component of the first PC axis was related to the resource capture and utilization strategies of plants. It was emphasized that LMA was a better indicator of this strategy, and the dimension of resource utilization seemed to explain the more differences among PFTs (Markesteijn et al., 2011; Lasky et al., 2014). Overall, this study provided valuable insights into the survival mechanisms of plants with different functional types, which could aid in the development of effective conservation and management strategies in the future.

Correlations among leaf traits are essential to understand plant strategies and functional trade-offs, and these correlations are often used to infer from one trait to another in dynamic vegetation models (Lohbeck et al., 2015; Sakschewski et al., 2015). There were some similarities in the correlations among leaf traits in three PFTs ( Figure 3 ; Table 2 ). LMA had significant positive correlations with LT, LDMC, C/N, and N_area, and a negative correlation with N_mass. However, recent studies showed that correlations among many traits might be diverse in different PFTs (Fyllas et al., 2020). For example, C/N ratio was positively correlated with N_area in the BE plants and BD plants, but there was a negative correlation in NE plants. The growth environment and ecological adaptation of coniferous plants and broad-leaved plants were different, which may be the reason for the different correlation between their C/N ratio and N_area (Ackerly and Reich, 1999; Faber et al., 2022). Coniferous plants typically thrived in nutritionally deficient soils, and they had evolved to adapt to these conditions by maintaining a high C/N ratio to more efficient use of limited nitrogen resources (Faber et al., 2022; Liu et al., 2022b). In contrast, broad-leaved plants generally grew in areas with more nutrient-rich soil, allowing them to obtain greater amounts of nitrogen during their growth and allocated it towards both growth and metabolic processes (Faber et al., 2022; Liu et al., 2022b). These differences indicated that general research phenomena might not be applicable to evaluate the correlation between leaf traits at regional or global scales in parametric dynamic vegetation models. Therefore, we suggest that PFTs-specific parameters should be developed to better represent the relationships among leaf traits that were generally embedded in such models.

The plasticity of leaf phenotype plays a crucial role in plant survival across different environments (Maire et al., 2012; Wright et al., 2017). However, it was discrepant that the sensitivity of leaves to environmental change among different PFTs (Wright et al., 2004; Kikuzawa et al., 2013). The study found that the leaf traits among three PFTs exhibited significant variation with elevation and MAT changes ( Figure 4 ). As the temperature gradually decreased with increasing elevation, there was a significant collinearity between elevation and temperature (Fu and Sun, 2022). Compared with MAP, MAT was the primary environmental factor driving variation in leaf traits along elevation. With increasing elevation or decreasing temperature, the LT and N_area of NE plants and BE plants gradually increased, while BD plants gradually decreased. To adapt to environmental change, evergreen plants typically maintained a higher leaf thickness and internal nitrogen content at high elevations to support longer lifespans and higher photosynthetic efficiency. However, deciduous plants adopted the opposite approach to reduce nutrient and energy loss. These two PFTs took different response measures to adapt to environmental change (Niinemets, 2001; Scafaro et al., 2017; Togashi et al., 2018; Liu et al., 2022a). In addition, BD plants paid more attention to the acquisition of environmental resources in the short growing season, and weakened the investment of leaf tissue structure in environmental adaptation (Chen et al., 2013; Fyllas et al., 2020). The thinner leaves were beneficial to the gas exchange and transpiration of plants to promote the photosynthetic efficiency and enhance the carbon fixation capacity when the water was sufficient (Sun et al., 2006; Bunce, 2022). Meanwhile, with the increase of elevation, LDMC, LMA, and C/N of BE plants and BD plants showed downward trend, while N_mass showed upward trend. However, NE plants exhibited the opposite trend, possibly due to its tendency to adopt conservative survival strategies to adapt to environmental change, and this ensured their survival in hostile environments by enhancing leaf structure and reducing photosynthesis investment (Lasky et al., 2014). The plant with large leaf tended to prefer acquisitive survival strategies, and BD plants were the most typical representative (Fyllas et al., 2020). With the increase of MAP, the six leaf traits did not show obvious change trend, these might be because MAP was not the main environmental factor causing variation in leaf traits of woody plants in subalpine environment (Zhang et al., 2022).

PFTs, MAP, and MAT accounted for a significant portion of the biogeographic variations in leaf traits ( Figure 5 ). Apart from LDMC, PFTs had a higher explanation for the variation in leaf traits. PFTs were the major factor leading to the differences in leaf traits (Akram et al., 2022). It was consistent with the results of a meta-analysis from global data conducted by Siefert et al. (2015). Compared with the MAP, MAT could explain more changes in leaf traits in subalpine environments. This finding indicated that the temperature was an important environmental factor affecting the variation in leaf traits. PFTs and climate were important drivers of variation in leaf traits, but PFTs were more critical in shaping biogeographic patterns of leaf traits, as demonstrated by previous studies from regional to global scales (Niinemets, 2001; Ordoñez et al., 2009; Zhan et al., 2018). Nevertheless, it had been proved that both soil conditions and plant phylogenetic background also affect leaf traits. Thus, further research is necessary to distinguish the relationship between the three dimensions of heredity, soil, and climate, and to explore the source of leaf traits variation caused by environmental change.