4.1 Variation in leaf functional traits and trade-off relationships
We found that the mean values of nine leaf functional traits of 18 dominant tree species in this community were within the ranges of global leaf trait variation (Chen et al., 2016; Zhao et al., 2020), and it was moderate in intraspecific and community-level variations, while significant in interspecific variation, indicating that accurate measurements of multi-source variation of functional traits were significant for deep understanding of the processes of community assembly (Westerband et al., 2021). Compared to that in other vegetation types (Li et al., 2017; Liu et al., 2020; Wang et al., 2021), the intraspecific variation in this community was at a lower level and supported the “a spatial trait variance partitioning hypothesis,” that is, due to the limitation of environmental heterogeneity and individual number, the importance of intraspecific variation was less at the fine scale.
The functional traits of plant leaves are determined by both genetic and environmental factors. They usually show niche differentiation and divergence of ecological strategies through intraspecific variation so as to reduce the intensity of competition (Kang et al., 2017) and to adapt to a broad environmental gradient. In this community, the intraspecific variation in leaf chemical characteristics (LC, LN, LP, and LC:LN) of dominant tree species was higher than that of structural characteristics (LA, SLA, and LDMC). However, the interspecific variations showed the opposite trend except for the related-LP traits, indicating that the variation in leaf chemical traits had higher plasticity in coping with local environmental changes. Zhao (2015) found that the soil and topographic variables in the community had moderate spatial variation, which had significant effects on the species composition and spatial distribution of the community. And as we all know, the availability of nitrogen and phosphorus in soil is the main influencing factor influencing LN and LP contents, respectively (Li et al., 2017). Therefore, to adapt to more complex environmental conditions, dominant species increase their distribution space along the environmental gradient by adjusting the plasticity of leaf chemical traits. Compared to chemical traits, leaf structural traits are mainly limited by genetic factors and are relatively stable and closely related to life forms and species type; therefore, interspecific variation is larger (Ackerly & Reich, 1999). In addition, it greatly differs for species at different phylogenetic stages in leaf size, leaf life span, LN content and photosynthetic ability. Eighteen dominant species in this community belonging to 10 families, including Fagaceae, Theaceae, Anacardiaceae, Pinaceae, and Taxodiaceae (Table 1), exhibited significantly different phylogeny background and thus may help them coexist through the coordination of functional traits. In addition, needle-leaf trees are generally thought to be better adapted to cold or nutrient-poor environments than broadleaf trees (Liu et al., 2020; Liu et al., 2021). With an increase in water stress, plants tend to exhibit xerophytic leaves with a thicker leaf structure (Guerful et al., 2009; Werden et al., 2017). Similarly, our results showed that coniferous species, especially P. massoniana had the highest LT and the lowest LA and SLA in this community, which may be because it was the first pioneer tree species to enter this community; and in order to resist the arid and barren environment, they possessed leaf functional traits such as thicker leaves suitable for storing water and lower LA and SLA to reduce water loss through transpiration. This result was consistent with those of previous studies that showed larger SLA and thinner leaves of broad-leaved trees than those of coniferous trees (Tian et al., 2016). Furthermore, the LN content of evergreen species was significantly lower than that of deciduous species because the generation cost of leaves was related to seasonal variation; leading to different adaptive strategies of evergreen and deciduous trees based on the variation of traits to adverse environments (Liu et al., 2021). In summary, the variations in leaf functional traits in L. glaber–C. glauca evergreen broad-leaved forest community were mainly attributed to the life form and interspecific variation, which was significantly affected by genetic background and taxon, and provided an important prerequisite for community assembly and species coexistence.
Meanwhile, there was a correlation between leaf structure and chemical traits in the community. SLA, LDMC, and LN all reflect adaptation strategies to the environment (Wright et al., 2004; Xun et al., 2020), and there was a significant positive correlation between SLA and LN, reflecting photosynthetic capacity and nutrient turnover at the species and community levels. SLA significantly correlated with LT, LDMC, and LC negatively, while no correlation was observed between LDMC and LT (Table 4 and 5), indicating that LT and LDMC affected SLA in different ways in this community. Simultaneously, the construction of a leaf defense structure requires a large amount of photosynthate, and LC is usually used to compensate for consumption during development (Schulze et al., 1994). Therefore, LC increased with an increase in LDMC and LN. In addition, light is an important factor affecting SLA (Wyka et al., 2012), and LDMC is closely related to water (Saura-Mas et al., 2009). The results showed that LDMC had lower interspecific variation and did not correlate with the chemical traits (except with LC), indicating that subtropical evergreen broad-leaved forests had sufficient hydrothermal conditions and that the major factor affecting community assembly should be light rather than water.

4.2 Phylogenetic effects on leaf functional traits

The evolution is close to Brownian motion; that is, species with similar phylogenetic positions have similar characteristics and have certain evolutionary conservation. Species with similar functional traits are often phylogenetically similar (Losos, 2008). When a strong phylogenetic signal is detected in functional traits, environmental filters are probably selected for phylogenetically close species, causing phylogenetic clustering (Amaral et al., 2021). Here, only a number of leaf structural traits (LA and LT) showed strong and significant phylogenetic signals, indicating that LA and LT were closely related to phylogenetic history and showed strong phylogenetic conservation; that is, the more phylogenetically close species were more similar to LA and LT. All the dominant tree species had considerably high values of LA, except the deciduous tree species (Q. fabri and A. kurziivar. Kurzii ). All evergreen tree species belonging to the same family and genus had more similar traits, especially for the two most dominant species (L. glaber and C. glauca ) in the community, both belonging to Fagaceae (Table 2 and Fig. 1).
The distributions of chemical traits (K < 1, p> 0.05) (Fig. 1) were not consistent with the phylogenetic relationships, indicating that the phylogenetic signals of these traits were random or divergent. Therefore, compared to the leaf chemical traits, the formation and development of structural traits were more affected by genetic differences, which is consistent with the results of Cao et al. (2013). In other words, the phylogenetic signal test based on functional traits showed a lack of consistency between the leaf functional trait patterns and phylogenetic patterns in this community, and no specific trend or relationship between them was observed. The phylogenetic relationships of L. glaber–C. glauca evergreen broad-leaved forest community were inconsistent with the changes in functional traits with the historical processes. This observation was supported by the work of Cheng et al. (2019) on the construction mechanism of tropical cloud forest communities.
Numerous studies have indicated that phylogeny has a significant effect on the functional trait composition and that the relationships among traits are generally weakened after removing phylogeny (Cadotte et al., 2019; Wang et al., 2020; Liu et al., 2021). This study also found that the traits of coexisting species in the community had a phylogenetic structure; however, only a few leaf traits (LA and LT) showed strongly phylogenetic signals. But the relationship among traits after removing the influence of the phylogeny remained a little changed at community level or even significantly enhanced at species level. This indicated that in the local community, environment had a greater impact on the spatial distribution of individuals, and plants follow the “realism” strategy to adapt to the environment, meaning, plants would adjust these trait-off relationships according to their habitats to archive the best survival state, which was less limited by the evolutionary history. It was clear that the functional community structure was not consistent with the phylogenetic structure. Combined analysis of phylogenetic and functional trait structures will more accurately infer the main ecological processes driving species coexistence.

4.3 Community assembly mechanisms of integrated phylogenetic leaf functional traits

Many studies have attempted to distinguish between determinative and stochastic processes by partitioning the variation of species composition into environmental and spatial components (Legendre et al., 2009; Chang et al., 2013; Qiao et al., 2015). However, species niches are determined by their functional traits, which further influence their distribution along environmental gradients (McGill et al., 2006). Thus, the effects of determinative processes on community assembly are expected to be underestimated based on species identity, which does not consider the functional properties of species. However, we found that functional traits (12.35%) did not improve the interpretation rate of niche-based processes by considering only species identity (28.10%, Zhao et al., 2015) and phylogenetic structure (29.80%) (Fig. 4). This observation was similar to that by Jiang et al. (2018) for temperate deciduous broad-leaved Korean pine forests, in which functional traits could not better reveal ecological processes compared with that by species.
However, integrating functional traits with phylogeny can greatly improve the ability to infer determinative and stochastic processes, and trait- and phylogenetic-based approaches are powerful ways to detect community assembly processes (Li et al., 2019; Amaral et al., 2021). We found that the interpretation rate of community assembly based on functional traits (63.38%) was higher than that based on phylogeny (47.96 %). However, the pure space variable could significantly explain higher functional traits than that by environmental variables (Fig. 4), indicating that the neutral stochastic process played a leading role in the construction of community functional traits. Compared to the community functional trait composition, the environment had a greater contribution to the spatial variation in the phylogenetic structure (Fig. 4). This indicates that the phylogenetic structure of the community was aggregated (mainly affected by habitat filtration); however, leaf functional trait composition showed a dispersion pattern (mainly affected by stochastic processes). It also verified the opinion that species identity is a more holistic concept and could better depict multiple traits of a plant, while leaf functional traits could depict one or certain facets of a plant (Jiang et al., 2018). At the same time, among all edaphic and environmental variables, only altitude, aspect, and TK reflecting light and water utilization of plants had a significant effect on community functional composition (Fig. 3), which was related to the fact that leaves were key organs of plant photosynthesis and mainly exercise the functions of photosynthesis and nutrient turnover. Therefore, we need to consider more functional traits (such as height and wood density) to provide more detailed information to improve the interpretation rate of community functional trait composition.