The field experiment was conducted in the subtropical zones of China, which represent the main distribution areas of moso bamboo forests: Northern subtropic zone; including Xinyang (XY), Anji (AJ); Middle subtropic zone; including Changning (CN), Taojiang (TJ); South subtropic zone; including Conghua (CH), Longmen (LM) ( Figure 1 ). Those moso bamboo forests range in latitude from 23°N to 31°N, and in longitude from 105°E to 119°E. The plots ranged from 14.2 to 20.2°C in mean annual temperature (MAT) and from 910 to 1510 mm in mean annual precipitation (MAP). The diameter at breast height of moso bamboo ranged from 7.1 to 10.7 cm, and the tree height ranged from 9.0 to 14.4 m. Those moso bamboo forests range in GPP from 0.98 to 1.69 kg C m^-2 yr^-1, detailed site information of the study areas is shown in Table S1 .

In July of 2019, a total of 24 plots (6 site × 4 plots) were established in subtropical zone. To minimize human interference, the sampling plots were set up in unmanaged, unlogged areas, representing a typical bamboo forest ecosystem. Each sample plot was 20 m × 20 m, and the distance between sample plots was more than 1000 m to avoid spatial pseudoreplication. In total of 120 (6 site × 4 plots × 5 replications) new culms were selected.

Twenty mature and healthy leaves were collected from the middle of the canopy on the south-facing side of each bamboo, and small pieces (1.0 × 0.5 cm) containing the midrib and part of the leaf were immediately cut from the leaves and fixed in FAA fixative (90 ml 75% ethanol + 5 ml formalin + 5 ml glacial acetic acid + 5 ml glycerol) for storage. The stomatal traits of leaves were observed using a scanning electron microscope (XL30-ESEM, Philips, Holland). Three small pieces were randomly selected from the fixed samples, and three photos were randomly taken at 650× and 3500× magnification, respectively, and the number of stomata in each photo was recorded (N_Photo). Stomatal length (SL, µm) and stomatal width (SW, µm) were measured using Image J software (National Institutes of Health, Bethesda, MD, U.S.A.), and a total of 2160 photos were taken, and about 1080 stomata measured. The stomatal density (SD), stomatal size (SS) and stomatal relative area (SRA) of the leaves were calculated as follows ( Table 1 , Figure 2 ):

Leaf samples were gradually dehydrated through a series of ethanol (50-100%) and permeated with paraffin. Leaf samples of approximately 10 μm in size were obtained with a slicer (Leica, RM2016, Germany) and then stained with Safranin O-Fast Green. Then, the anatomical structures were observed and photographed under a light microscope (ECLIPSE E100, NIKON, Japan) and the following 10 leaf anatomical traits were calculated using CaseViewer software 2.3 (3DHistech, Budapest, Hungary) ( Table 1 , Figure 2 ).

After removal of the understory litter, soil samples were collected using a soil auger. A total of 72 soil samples (6 site × 4 plots × 3 replications) were collected from 0-20 cm soil layer. Each replicate was a soil mixture collected in each plot using a five-point sampling method. All soil samples were mixed and sieved through 100 mesh to remove large components, such as visible roots, debris, leaves. We analyzed soil bulk density (SBD), water content (SWC), pH, and concentrations of total carbon (TC), total nitrogen (TN), total phosphorus (TP), ammonia nitrogen (NH₄ ⁺-N), nitrate nitrogen (NO₃ ^–N), soil available phosphorus (SAP), soil organic carbon (SOC), readily oxidizable carbon (ROC). The SBD and SWC were measured using the ring knife, respectively. Soil pH was determined using a pH meter (PHS-3C, INESA Inc., China). The concentrations of TC, TN, TP, NH₄ ⁺-N and NO₃ ^–N were determined, using an elemental analyzer (Costech ECS 4024 CHNSO, Picarro, California, U.S.A.) and automatic chemical analyzer (Smartchem 4024, AMS Group, Italy), respectively. The concentrations of SAP, SOC and ROC contents were measured, using a continuous flow analyzer (AA3, Seal, Germany) and the potassium dichromate oxidation colorimetric method and potassium permanganate oxidation method, respectively ( Table S1 ) (State Forestry Bureau, 1999; Guo et al., 2022).

The effect of sites on leaf stomatal and anatomical traits of moso bamboo were studied by one-way ANOVA. Mean comparisons were performed using the Duncan’s multiple range test at a probability level of 0.05. All values presented are means ± SD. Then, linear mixed-effect models were used to study the relationship among climate, leaf stomatal and anatomical traits of moso bamboo, and plot nested within site (random effects) (Bartholomew et al., 2022). Spearman correlation coefficients were calculated between soil environmental factors, stomatal and anatomical traits. The trait coordination network of moso bamboo was described by statistically significant correlations between stomatal and anatomical traits (Xie et al., 2022). Nodes in the network represent leaf stomatal and anatomical traits, and edges indicate correlations between traits. Only significant correlations are shown in the network (P < 0.05). Network centrality was characterized by degree (the number of edges connected to a node) and weighted degree (the sum of the edge weights of the adjacent edges of each node) (Messier et al., 2017). All variables were checked for normality and log-transformed when necessary. All statistical analyses were carried out using R-4.0.3 (R Core Team, 2020). The used R packages include ggplot2 v3.3.5 (Wickham, 2016), ggpubr v0.4.0 (Kassambara, 2020), ggsignif v0.6.3 (Ahlmann-Eltze and Patil, 2021), ggpmisc v0.4.5 (Aphalo, 2021), psych v2.1.9 (Revelle, 2017), rcompanion v2.4.15 (Mangiafico and Mangiafico, 2017), igraph v1.2.7 (Csardi, 2013), lme4 1.1-27.1 (Bates, 2010), lmerTest 3.1-3 (Kuznetsova et al., 2017).To explore the causal relationships among environment, leaf stomata and anatomical traits, and GPP, we used structural equation models (SEM) to estimate the path coefficients and variatiton of the dependent variables. We hypothesized that the environment affects leaf stomatal and anatomical traits, and ultimately affects GPP. A causal relationship between the variables was determined based on previous knowledge and preliminary qualitative studies. SEM was analyzed using Amos software ver. 23.0, and the model was evaluated using the maximum likelihood method (Arbuckle, 2014). The model parameters (0≤ χ2/df ≤3, RMSEA < 0.05, GFI > 0.9, CFI > 0.9, P > 0.05) were used to evaluate the fitting effect of the model (Lavorel and Grigulis, 2012; Ali et al., 2019).

Site significantly affected the soil physicochemical properties of moso bamboo ( Table S2 ). For soil physical properties, CN areas had the highest SWC, XY and TJ areas had the lowest, while SBD showed the opposite trend. For soil chemical properties, the content of TC, TN, TP, NH₄ ⁺-N, SOC, ROC and SAP were higher in the CN area, while the soil nutrient content was lower in the XY area ( Table S2 ).

In general, site significantly affected the bamboo leaf stomatal and anatomical traits (P < 0.05) ( Figures 3 , 4 , Table S3 ). In terms of leaf stomatal traits, the northwestern site CN had significantly higher SD and lower SS than the other sites. In contrast, the northern site XY showed the opposite results. In terms of leaf anatomical traits, the northwest site CN had the highest SAP, MPT, SAFVB, and LET, and the northern site XY had the highest SAX, SABC, DBAVB, and SASVB, and both sites had lower SCT. While the central site TJ had the highest SCT and lowest SAX and SAFVB, the other sites had no significant differences in leaf stomatal and anatomical traits (P > 0.05) ( Figures 3 , 4 ).

At the regional scale, the average degree of the trait coordination network was 4.615 (13 nodes, 30 edges), and SS showed the highest centrality in the network ( Figure 5 , Table S4 ). The positive correlations were dominant in the trait coordination network, but all leaf stomatal and anatomical traits associated with SAFVB showed negative correlations. In addition, SAX and MPT were not significantly correlated with other leaf stomatal and anatomical traits ( Figure 5 ).

Overall, leaf stomatal traits were limited by SR, and leaf anatomical traits were mainly affected by MAP and SR ( Table 2 ). SR was significantly negatively correlated with SD, LET, and MPT, and significantly positively correlated with SS and SRA (P < 0.05). MAP was significantly negatively correlated with SAP, SASVB, MPT, and SABC, and significantly positively correlated with SCT (P < 0.05). Moreover, MAT was significantly negatively correlated with DBAVB of moso bamboo (P < 0.05).

Generally, leaf stomatal and anatomical traits of moso bamboo were affected by soil factors ( Figure 6 ). SWC, SBD and TC were the main factors that affected leaf stomatal and anatomical traits. The SS responded most strongly to the soil environment in leaf stomatal traits, while SAFVB and SAP were more sensitive to the soil environment in leaf anatomical traits.

SEM analyses showed a significant positive correlation between environment and leaf traits stomata and anatomical traits ( Figure 7 ). The effects of the environment on leaf stomata and anatomical traits were similar, and standardized path coefficients were 0.730 and 0.626, respectively. SEM revealed a direct positive effect of leaf stomatal traits on GPP (standardised direct effect = 0.553), while anatomical traits had no significant effect on GPP. Furthermore, the environment exerted indirect control on GPP through its positive effect on stomatal traits (standardized indirect effect = 0.404). There was no direct effect of environment on GPP, and leaf stomatal traits may have mediated the direct effect of environment on GPP. Overall, environment explained 53.3% and 39.2% of the variation in leaf stomatal and anatomical traits, respectively; leaf stomatal traits were the key driver explaining regional-scale variation in GPP, explaining 20.8% of GPP variation ( Figure 7 ).

Consistent with our second hypothesis, we found that the environment (climate and soil) significantly affected the leaf stomatal traits of moso bamboo ( Figures 6 , 7 , Table 2 ). SR was significantly negatively correlated with SD, and positively correlated with SS and SRA, indicating that bamboo could adapt to the external environment by regulating stomatal traits ( Table 2 ). In other case, the SD of sedge was positively correlated with temperature and precipitation in the Eurasian Arctic, while the SS was the opposite (Stenström et al., 2002). The studies have shown that SD and SS were negatively correlated with precipitation and temperature, respectively, this difference may be related to genetic factors, leaf age, leaf vein density, altitude, etc. (Batos et al., 2010; Bertolino et al., 2019; Du et al., 2021; Amitrano et al., 2022). In this study, soil nutrients and moisture affected leaf stomatal traits of moso bamboo, and the SS of moso bamboo was more sensitive to soil variables ( Figure 6 ). A previous study showed that SD and SRA were negatively correlated with soil nutrients (C, N and P) and pH, while SS showed the opposite trend (Liu et al., 2017). Sun et al. (2021) found that SS and SRA were significantly negatively correlated with C, N, and P, while SD showed the opposite trend. The effects of different environmental variables on leaf stomatal traits may be the result of their combined effects (Larcher et al., 2015; Du et al., 2021). SEM showed that leaf stomatal traits were the key drivers explaining the regional-scale variation of GPP ( Figure 7 ). It was shown that stomatal density was positively correlated with ecosystem productivity, explaining 51% of its variation (Wang et al., 2015). To improve the prediction accuracy of the model, leaf stomatal traits should be included as new ecological parameters in the model for predicting ecosystem function.

Consistently as with the stomatal, traits, the environment (climate and soil) also significantly affected the leaf anatomical traits of moso bamboo ( Figures 6 , 7 , Table 2 ). General, MAT, MAP, and SR were significantly negatively correlated with leaf anatomical traits ( Table 2 ). It was shown that leaf anatomical traits were significantly influenced by climate, especially MAT and MAP, which explained 33-72% of the large-scale total variation (He et al., 2017). Previous study have shown that grasses tend to increase vessel and vascular bundle diameters to resist drought stress (Wang et al., 2011; Wyka et al., 2019). This difference may be that the suitable zone of moso bamboo has not reached the level of drought stress, and its well-developed bamboo whip system has a strong water transport and storage function. In our study, the leaf anatomical traits of moso bamboo were affected by SWC, TC, SOC, ROC, SAP ( Figure 6 ). It was found that leaf anatomical traits were influenced by soil moisture and nutrients in different habitats, and their total variation on leaf anatomical traits was about 20%, respectively, and leaf anatomical traits were more influenced by TN than TP (Wu et al., 2022). However, Lin et al. (2021) found that variation in leaf anatomical traits of Cyclobalanopsis was mainly controlled by soil pH. SEM showed that anatomical traits had no significant effect on GPP ( Figure 7 ). In contrast, at the community scale, leaf anatomical traits of dicotyledonous plants were closely related to GPP (He et al., 2017). Therefore, the relationship between leaf anatomical traits of monocotyledons and ecosystem function may need more data to test this hypothesis.

We found that there may be a trade-off between leaf stomata and anatomical traits, and the traits were mainly synergistic ( Figure 5 ). Qi et al. (2016) found that graminaceous plants adapt to environmental changes through the coordination between traits and have a strong overall adaptation to the environment. Previous studies have shown that high photosynthesis requires stomatal opening for gas exchange and transpiration for water loss, and plants may improve water transfer by regulating leaf anatomical traits to increase water use efficiency (Lawson and Blatt, 2014; Schneider et al., 2017). In accordance with our first hypothesis, leaf stomatal trait (SS) was the highest centrality value in moso bamboo, indicating that stomatal traits may play a dominant role in the rapid development of new bamboos ( Figure 5 ). Previous studies have shown that SS shows the highest centrality value in humid area, while in regional scale, semi-humid and arid areas it is the stomatal opening ratio. Plants can adjust SD and SS in response to external environmental stresses, but when they reach a physiological threshold, the stomatal opening ratio may regulate changes in the trait coordination network along environmental gradients (Bertolino et al., 2019; Xie et al., 2022). Moreover, the leaf economic spectrum indicates growth condition-related trade-offs among leaf traits within a species, and thus leaf vein traits and other leaf traits should also be considered to better understand the acclimation and adaptation of plant photosynthetic mechanisms to environmental changes (Sack et al., 2013).