1. Introduction

Terrestrial ecosystems store a vast amount of carbon in the vegetation carbon pool (VCP). VCP dynamics have profound implications for potential global carbon sequestration and achieving carbon neutrality targets (IPCC, 2021; Mo et al., 2023; Pan et al., 2024; Yu et al., 2024). In natural ecosystems, VCPs are influenced by abiotic and biotic factors, including climate conditions, soil properties, plant species diversity, and vegetation structural attributes (Hooper et al., 2005; Liang et al., 2016). Furthermore, growing evidence shows that interactions between these factors play a crucial role in shaping VCPs (Wu et al., 2015; Brose and Hillebrand, 2016; Ouyang et al., 2019; Chen et al., 2023a). Therefore, identifying where VCPs are located and the mechanisms that drive their distribution is critical for current and future terrestrial carbon sink estimation and climate change mitigation.

Several studies have demonstrated that biotic factors (plant species diversity and structural attributes) influence VCPs in forest ecosystems (Hua et al., 2022; Ma et al., 2025a; Pigot et al., 2025). According to the diversity-productivity hypothesis, increasing tree species diversity promotes productivity and biomass globally via complementarity and selection effects (Loreau and Hector, 2001; Feng et al., 2022; Chen et al., 2025). Additionally, recent studies have highlighted the importance of vegetation structure in shaping biomass accumulation and distribution (Yang et al., 2024; Zhai et al., 2024). Specifically, moderate stand density and complex canopy structure facilitate efficient utilization of spatial and light resources (Li et al., 2024a). As a result, it has been hypothesized that biomass accumulation is more influenced by structural diversity than by species richness (Chen et al., 2023a). Nevertheless, the combined impacts of species diversity and structural diversity on VCPs remain poorly examined in non-forest woody ecosystems. Elucidating their relative importance in plant communities is essential for accurately predicting terrestrial carbon fixation.

Studies have also shown that the impact of biotic factors on VCPs varies along abiotic gradients. Mountain ecosystems offer ideal systems to examine elevational patterns of VCPs. These ecosystems are characterized by remarkable environmental gradients encompassing climate, soil, and vegetation (Graae et al., 2012; Hagedorn et al., 2019; Rahbek et al., 2019; Hu et al., 2020). Climate and soil properties dominate the availability of resources such as light, nutrients and water, subsequently affecting plant growth, diversity, and structure (Li et al., 2024a), eventually driving VCP elevational patterns (Chen et al., 2023a; Camacho et al., 2025). At high elevations, plant distribution and growth are primarily limited by low temperatures (Su et al., 2023). However, precipitation along an elevation gradient may alleviate limitations driven by temperature and even enhance VCPs (Poudel et al., 2020; Cai et al., 2020; Su et al., 2025). Plant distribution and productivity are also influenced by soil factors such as moisture, nitrogen availability, and pH (Khanal et al., 2024). However, these findings have been predominantly reported from forest ecosystems, with relatively few investigations focusing on non-forest ecosystems, particularly savannas. Consequently, the distribution of VCPs and the factors that drive this distribution in non-forest ecosystems require further exploration.

Savanna ecosystems play an irreplaceable role in the global carbon cycle (Pellegrini et al., 2016; Fei et al., 2017). Due to their extremely arid and hot environments, savanna ecosystems are likely to be more sensitive to global change than forests, grasslands, and other ecosystems, and thus will play a critical role in mitigating future climate change (Godlee et al., 2021). Savanna ecosystems are typically dominated by dwarf shrubs and characterized by a simple canopy structure (Hoffmann et al., 2005; Liu et al., 2022). In contrast, trees in savanna ecosystems exhibit variability in individual height, which creates diverse habitats and provides shady microenvironments for tree seedlings and shrubs. Despite these significant differences, the relative contributions of shrub and tree structural and species diversity to VCPs remain poorly understood. Given the uniqueness of plants adapted to the dry-hot valley environment, assessing the relationships between their VCPs and the factors that drive their formation is highly valuable.

Yunnan Province is one of the most biodiverse regions in China. In the dry-hot valleys of Yunnan (e.g., Yuanjiang Nature Reserve), savannas are a distinctive vegetation type that harbors a rich diversity of species adapted to arid habitats (Zhu et al., 2024). These species are globally rare and represent China's most typical and representative forms of savanna ecosystems (Li et al., 2022). In the Yuanjiang Nature Reserve, savanna ecosystems act as a significant carbon sink (Fei et al., 2017), and their carbon storage savanna can exceed the average values of global savanna ecosystems (Jin et al., 2017). Here, we explored the elevational distribution of VCPs in the Yuanjiang dry-hot valley savannas and determined the intrinsic factors that have driven this distribution. We propose the following hypotheses: (1) Both plant diversity and structural diversity positively affect the VCP, with structural diversity being the more influential driver; (2) Elevation shapes the VCP patterns by regulating both abiotic factors and structural diversity (Fig. 1B and Table S2).

2. Materials and methods 2.1. Study area and field survey

Our study was conducted in the Yuanjiang Nature Reserve in Yunnan Province, China (101°21′24″–102°21′12″E, 23°19′12″–26°46′12″N), a region characterized by the typical sparse shrub vegetation of Southwest China (Wu et al., 1987). The mean annual temperature (MAT) in the Yuanjiang dry-hot valley is 23.8 ℃, with an average temperature of 28.7 ℃ in the hottest month and 16 ℃ in the coldest month. The mean annual precipitation (MAP) is 732.8 mm and potential annual evaporation is 1750 mm (data from the Yuanjiang Savanna Ecological Station, Xishuangbanna Tropical Botanical Garden, Chinese Academy of Sciences) (Li et al., 2024b). Due to limited precipitation and high evapotranspiration, the vegetation is dominated by drought-tolerant trees, shrubs, and herbs. Moreover, the vegetation exhibits a distinct elevational distribution, transitioning from a mixture of herbs and shrubs dominating at higher elevations to trees as the dominant species at lower elevations (Zhu et al., 2020). Thus, this region is ideal for studying the elevational patterns of VCPs in savannas.

Field vegetation surveys and soil sampling were conducted at four elevations: low elevation (415.8–460.9 m), middle elevation (868.2–877.1 m), middle–high elevation (1223.7–1238.1 m), and high elevation (1627.5–1644.1 m). In 2021 and 2022, we established a total of 80 sample plots, with 32, 16, 16, and 16 plots at each elevation, respectively. The size of each plot was 10 m × 10 m (Fig. 1A; Table 1). We recorded basic information for each plot, including latitude, longitude, and elevation. Within each plot, we also documented all species and measured diameter at breast height (1.3 m) (DBH), height, and ground diameter of all woody plants (trees and shrubs) with a ground diameter greater than 1 cm. The dominant species across elevations were as follows. (1) Low elevation: Huberantha cerasoides, Lannea coromandelica, and Tarenna depauperata; (2) middle elevation: Phyllanthus emblica, Bauhinia brachycarpa, and Heteropogon contortus; (3) middle–high elevation: Quercus cocciferoides, Campylotropis henryi, and Pistacia weinmanniifolia; (4) high elevation: Q. variabilis, Osteomeles schwerinae, and Carissa spinarum.

After completing plant surveys, we used the five-point sampling method to collect soil samples from each plot. Surface soil samples (0–10 cm) were collected at the four corners and center of each plot. The five soil subsamples were thoroughly mixed to form a composite sample and then were transported to the laboratory, where we measured soil water content (SWC), pH, soil organic carbon (SOC), nitrate nitrogen (NO₃^--N), ammonia nitrogen (NH₄⁺-N), and total N (TN).

2.2. Carbon pool estimation

We used biomass allometric equations from a dataset developed by Zhou et al. (2018) to evaluate biomass. This dataset, comprising ~900 equations, has been widely used in biomass assessment in Chinese forests (Tang et al., 2018; Zhang et al., 2021). According to the equation for subtropical evergreen broadleaf forests (Zhou et al., 2018), we calculated tree biomass using individual DBH and height. We also assessed shrub biomass using the shrub-specific biomass equation for subtropical regions in China (Xie et al., 2018). The specific biomass equations are provided in Table S1. VCP was defined as vegetation biomass multiplied by the carbon concentration (0.4816) of woody plants (Ma et al., 2018).

2.3. Explanatory variables

We categorized potential explanatory variables for elevational patterns into three groups: environmental variables, plant species diversity, and community structure. Environmental variables included elevation, MAT, MAP, and soil properties. Plant species diversity was represented by species richness, the Shannon–Wiener index, and Pielou's evenness index. Species richness refers to the number of species in each plot. The Shannon–Wiener index is a comprehensive index that consists of the variance of species richness and relative abundance in a community (Eq. (1)). Pielou's evenness index represents the evenness of species distribution in a community (Eq. (2)).

H′=−∑i=1spiln(pi)

(1)

J′=H′ln⁡(S)

(2)

where H′ represents the Shannon diversity index, J′ represents the Pielou's evenness index, S represents species richness, and p_i is the proportion of species i (p_i = N_i/N, N_i = abundance of species i, N = total abundance in a plot).

Community structure included stand density and structural diversity. We calculated stand density as the number of woody individuals in each plot. For structural diversity, we quantified DBH diversity and height diversity as the coefficients of variation of DBH (CVD) (Eq. (3)) and height (CVH) (Eq. (4)), respectively. CVD and CVH represent horizontal and vertical variability in vegetation structure and spatial resource utilization, respectively (Yi et al., 2021; Li et al., 2026). The formulas are as follows:

CVD=100*SDDBHMeanDBH

(3)

CVH=100*SDHMeanH

(4)

where SD_DBH and SD_H are the standard deviations of DBH and height in each plot, respectively. Mean_DBH and Mean_H are the mean values of DBH and height in each plot, respectively. A high CV value denotes high structural diversity for a community.

2.4. Data analysis

Prior to all analyses, we tested normality of all variables and log-transformed those that followed logarithmic distribution. To evaluate significant differences in soil properties at different elevations, we used one-way analysis of variance (ANOVA) followed by Tukey's post-hoc test for multiple comparisons. To assess how the relationships between explanatory variables and VCP vary with growth forms, we categorized plants into three groups: woody plants (trees + shrubs), trees, and shrubs. To identify the factors correlated with VCP, we conducted Pearson correlation analysis between VCP and variables, including environments (elevation, climate (MAT, MAP), soil), plant diversity (species richness, Shannon diversity index, Pielou's evenness index), and community structures (CVH, CVD, density). We performed ordinary linear regressions to explore how the woody plant, tree, and shrub VCPs change with the explanatory factors.

To assess the relative importance of variables in each group for VCP, we used the randomForest package and calculated the percentage increase in mean square error (%lncMSE) for each variable. For the first three variables in each group identified by random forest analyses, we performed partial variance partitioning analysis to quantify their total and interactive contributions to VCP variation. All the above analyses were conducted in R 4.3.2 (R Core Team, 2023).

Finally, we used structural equation models (SEMs) to explore the multivariate effects of elevation, soil, plant diversity, and structure on VCP. The SEMs were fitted using maximum likelihood estimation. The chi-squared (χ²) test, root mean square error of approximation (RMSEA), and goodness-of-fit index (GFI) were used to evaluate the model fit. Generally, a model was considered acceptable if P > 0.05, RMSEA < 0.05, and GFI > 0.90 simultaneously. SEM analyses were performed with Amos 21.0 (Amos Development Corporation, Chicago, IL, USA).

3. Results 3.1. Elevation patterns of VCPs and explanatory variables

The study plots covered a temperature range greater than 5 ℃. Soil properties showed complex elevational variation (Table 1). Specifically, both MAP and soil moisture increased at higher elevations, whereas pH followed a unimodal pattern, peaking at an elevation of 800 m. In contrast, SOC displayed an inverse trend, reaching its minimum value at 1200 m. NH₄⁺-N showed increased at higher elevations, whereas NO₃^--N decreased with elevation (Fig. S1).

Most structural diversity variables decreased with elevation, except for shrub DBH diversity (coefficient of variation of shrub diameter at breast height) (Fig. S2). Species diversity of woody plants and shrubs increased with elevation, whereas tree species diversity decreased with elevation (Fig. S1). Total VCP of woody plants showed no significant elevational trend (Fig. 2A), although distinct patterns emerged when analyzing tree and shrub components separately (Fig. 2A1 and A2). Specifically, tree VCP decreased with elevation (R² = 0.11, P < 0.01), whereas shrub VCP increased (R² = 0.08, P < 0.01).

3.2. Relationships between VCP and explanatory variables

Statistical analyses revealed divergent responses of VCPs to environmental variables. While climatic factors showed no significant relationships with woody VCP, they had significant relationships with both tree and shrub VCPs (Figs. 2B, C, B1, C1, B2, C2, and S1). Specifically, shrub VCP decreased with MAT (R² = 0.08, P < 0.01; Fig. 2B2) and increased with MAP (R² = 0.08, P < 0.05; Fig. 2C2). Tree VCP significantly increased with MAT (R² = 0.10, P < 0.01, Fig. 2B1), but was not significantly changed with MAP. Tree VCP was positively correlated with both SOC (R² = 0.19, P < 0.001; Fig. 2D1) and NO₃^--N (R² = 0.11, P < 0.01; Fig. 2F1), but negatively correlated with soil pH (R² = 0.11, P < 0.01; Fig. 2E1). Conversely, shrub VCP was negatively correlated with NO₃^--N (R² = 0.09, P < 0.01; Fig. 2F2).

Both species diversity (e.g., species richness, Shannon diversity, Fig. 2G, H, G1, and H1) and structural diversity (Fig. 2J, K, J1, and K1) were positively correlated with woody and tree VCPs. Moreover, the relationships between structural diversity and VCP were lifeform-specific patterns. Tree VCP increased with coefficient of variation of tree height (R² = 0.24, P < 0.001, Fig. 3A) and that of tree diameter at breast height (R² = 0.39, P < 0.001, Fig. 3B), while shrub VCP increased with coefficient of variation of shrub diameter at breast height (R² = 0.20, P < 0.001, Fig. 3D) and density (R² = 0.23, P < 0.001, Fig. 2L2).

3.3. Relative importance of driving factors based on random forest modeling and partitioning of variance

All abiotic and biotic factors were categorized into three groups (environment, species diversity, and structure) to quantify their relative contributions to woody, tree and shrub VCP variations by using random forest algorithms (Fig. 4). Elevation was the most influential environmental factor for all three VCPs (Fig. 4A, D, and G). Secondary drivers exhibited lifeform-specific patterns: SOC ranked second for both woody and tree VCPs, whereas soil NO₃^--N was a stronger predictor for shrub VCP. Among species diversity metrics, Shannon index for trees proved to be the best predictor for woody and tree VCPs (Fig. 4B and E), while Pielou's evenness index of shrubs contributed the most to shrub VCP (Fig. 4H). For structural variables, CVD accounted for the greatest variance in both woody and tree VCPs, followed by the coefficient of variation of tree height and that of tree diameter at breast height, respectively (Fig. 4C and F). Shrub density and coefficient of variation of shrub diameter at breast height were the key predictor for shrub VCP (Fig. 4I).

Variance partitioning analysis quantified the contributions of environmental, species diversity, and structural variables to VCP variation (Fig. 5). The total explained variance reached 78.2% for woody plants, 73.1% for tress, and 88.8% for shrubs, respectively. Specifically, structural variables contributed the most, while environmental variables accounted for relatively small proportions (shrubs: 7.1%; trees: 16.8%; woody plants: 17.5%). Furthermore, the individual effects of structure were higher than those of plant diversity and environment; the latter mainly acted through interactive effects.

3.4. Driving factor impact on VCP based on structural equation modeling

SEMs revealed that elevation indirectly affected woody plant, tree, and shrub VCPs by mediating soil properties, which further regulated species diversity and structure (Fig. 6 and Tables S3–S5). The reduction in soil nutrients at increasing elevation indirectly reduced species diversity and community structure, thereby negatively affecting VCP. Community structure had a direct, positive effect on all three VCPs. Both structural diversity and species diversity had similarly positive effects on the VCP of woody plants (Fig. 6A). Although SOC did not directly affect woody VCP, it had an indirect effect by promoting species diversity and structural diversity.

SEM pattern of the tree VCP were similar to that of the woody plant VCP. Notably, Shannon index for trees promoted tree VCP by increasing variation of tree height (Fig. 6B). SOC had a direct impact on the tree VCP and an indirect impact via Shannon index for trees. The latter was the most positive influential factor. In contrast, the impact of environmental factors on shrub VCP differed substantially: soil NO₃^--N negatively affected shrub VCP (Fig. 6C). Given the decreasing elevational pattern of NO₃^--N, shrub VCP was the only carbon pool that was positively affected by elevation. In addition, the community structure of shrubs had the strongest effect on its VCP (Fig. S4). Overall, these results suggest that plant diversity (mainly Shannon index for trees) can enhance total and tree VCPs by increasing structural diversity (Figs. 6 and S3).

4. Discussion 4.1. Both structural and species diversity are strong drivers of VCP

Consistent with our first hypothesis, both species diversity and structural diversity positively influenced VCP, with structural diversity having a stronger effect (Figs. 2 and 5). These results confirm the existence of diversity-carbon relationships in the hot-dry valley savanna ecosystems of Southwest China, driven primarily by the complementarity effect. Furthermore, the stronger influence of structural diversity on total VCP underscores the importance of ecological niche differentiation in the uptake and utilization of spatial light resource (Ali et al., 2019), which may facilitate biomass production and carbon storage (Zhang and Chen, 2015; Yuan et al., 2016; Chen et al., 2023a). This aligns with previous findings in forest ecosystems (Zhang and Chen, 2015; Ali et al., 2019; Chen et al., 2023a). Notably, the VCP was mediated not only by structural diversity but also by tree density (Fig. 4C and F), indicating that stand structural attributes are key regulators of biomass carbon accumulation.

Notably, the contributions of tree structure (e.g., coefficient of variation of tree height and that of tree diameter at breast height) to both woody and tree VCPs were greater than those of shrubs (Fig. 4C and F), which we partially attribute to trees' large size, canopy dominance, and capacity to create suitable habitats for shrubs (Ma et al., 2025b). Variations in tree size enhance the spatial and light resource utilization, thereby boosting tree carbon sequestration (Van Pelt et al., 2016; LaRue et al., 2023). Moreover, large trees create shaded understory habitats and heterogeneous light environments, which regulate hydraulic processes by reducing shrub transpiration while improving water-use efficiency (Valladares et al., 2016). Consequently, trees play an irreplaceable role in savanna ecosystems and should be prioritized for conservation amid future climate change. These findings align with previous research emphasizing the dominant mediating role of tree structural diversity in species diversity–VCP relationship in mixed plantations (Mensah et al., 2020). Overall, our results demonstrate that structural diversity is a fundamental mechanism governing the relationships between biodiversity and ecosystem functioning, extending beyond forests to savanna ecosystems, where structure-related complementarity effects appear particularly critical.

In contrast, shrub VCP was largely influenced by horizontal structural traits (e.g., shrub density, the coefficient of variation in shrub diameter; Fig. 4I), reflecting adaptive strategies such as lateral branching and prostrate growth that enhance survival and carbon storage despite limited vertical development (Götmark et al., 2016; Tumber-Dávila et al., 2022). The strong effect of density, coupled with the reduced importance of species richness, supports the role of competitive exclusion and environmental filtering in shaping shrub community and carbon accumulation (Li et al., 2019; Liu and Zheng, 2024). Strong environmental filtering leads to functional trait convergence among coexisting shrub species, an adaptation to nutrient scarcity and climatic extremes (Zhao et al., 2024). This convergence may diminish the role of species richness relative to structure. Overall, these findings demonstrate that establishing communities with high tree structural diversity and shrub density enhances carbon sequestration in savanna ecosystems.

4.2. Elevation influences VCP through abiotic and biotic factors

Our study found no significant elevational pattern of total VCP across these savanna ecosystems. However, contrasting patterns emerged between the two growth forms: tree VCP decreased, whereas shrub VCP increased with elevation (Fig. 2A, A1, and A2). This finding differs from the common single-peak pattern (Su et al., 2025). Additionally, the elevational pattern of shrub VCP is consistent with results from previous studies (Dyola et al., 2022; Wu et al., 2022). These findings deepen our insights into the elevational pattern of VCP in savanna ecosystems.

Elevation influenced VCP by regulating soil properties (mainly SOC and NO₃^--N; Fig. 6A-C), which supports our second hypothesis. SOC serves as a proxy for soil organic matter (SOM) and an integrated indicator of ecosystem properties such as soil fertility, biodiversity, and overall health (Schmidt et al., 2011; Liptzin et al., 2022). Higher SOC is closely associated with soil fertility and water retention (Spohn et al., 2023), and we speculate that this may enhance plant resistance to drought and high temperature stress (Laliberté et al., 2014), thereby promoting growth and carbon storage. Nitrogen (N) is a critical element for plant growth and productivity (Chen et al., 2023b). It is well-established that trees preferentially take up NH₄⁺-N, whereas shrubs prefer to utilize NO₃^--N (Mao et al., 2025). In fact, we found that both NO₃^--N and NH₄⁺-N were important for VCPs, with NO₃^--N having a stronger effect (Figs. 4 and 6). Previous research has also shown that soil N availability promotes carbon accumulation (Li et al., 2023). Under harsh conditions (high temperature, low precipitation), environmental filtering fosters niche differentiation and plasticity in water and nutrient use among plants, thereby reducing sensitivity to climatic variability (Cisneros et al., 2021). Furthermore, plant growth and survival strongly depend on soil properties to maintain fertility supply (Schmidt et al., 2011). This dependence is reinforced by a recent savanna study highlighting the critical effect of soil factor on plant diversity and biomass (He et al., 2024).

In contrast to numerous studies identifying climate as a crucial determinant of VCP (Uribe et al., 2023; Fatunsin and Naka, 2025; Gao et al., 2025), our study found that both woody and tree VCPs were more strongly affected by soil properties than by climate (Fig. 4A, D and G). At increasing elevation, temperature decreases and precipitation increases (Table 1), leading to reduced evaporation and higher soil moisture. At middle and high elevations, there is a lack of soil water supply from underground water and runoff. The latter only exist at lower elevations (~400 m), where soil moisture reaches its minimum. Thus, climatic conditions may become more favorable at higher elevations. If so, it is expected that the tree and total VCPs would increase with elevation. However, our results did not support this view (Fig. 2A, A1, and A2). Instead, high-elevation sites are characterized by strong winds, high radiation, steep slopes, and erosion conditions (Wei et al., 2024) that disproportionately limit the growth of deep-rooted trees compared to shrubs. In contrast, low-elevation sites, despite high temperature and evaporation, possessed favorable edaphic conditions (gentle topography, higher soil fertility, and better water retention) that might facilitate dominant tree growth and carbon fixation under relatively adequate moisture (Fei et al., 2017). Consequently, climate alone had a limited influence on VCP in the present study. These findings suggest that elevation integrates multiple co-varying stressors (e.g., high temperature, high radiation, and low fertility) (Körner, 2021; Wei et al., 2024), thereby mediating and reshaping the apparent impact of climate on VCP patterns.

In this study, the positive correlation between SOC and species diversity suggests that higher soil fertility supports the development of more diverse plant communities (Fig. 6A and B). It is generally believed that increased resource availability would reduce community diversity through competitive exclusion (Harpole et al., 2016; Ratcliffe et al., 2017; Muehleisen et al., 2022). However, at low-elevation plots, climatic stresses (high temperature and evaporation) limit species distribution. Concurrently, these conditions enhance microbial activity, thereby accelerating SOC decomposition and nutrient release (Table 1; Ma et al., 2025b). Thus, gentler slopes and richer soil nutrients may help plants mitigate these adverse climatic conditions, thereby enabling species coexistence (Table 1 and Fig. 2D–I). Furthermore, under these condition, certain large tree species might have a high dominance, thereby improving structural diversity and light-use efficiency, ultimately enhancing VCP as observed (Fig. 6). Therefore, abiotic factors (climate and soil) influence VCP primarily indirectly by regulating structural diversity, consistent with findings from tropical mountainous regions (Noulèkoun et al., 2021). Overall, our research on savanna ecosystems in Southwest China indicate that variations in soil properties along the elevational gradient shape both species diversity and structural diversity, thereby influencing the VCP accumulation.

4.3. Limitations

Our study has several limitations that may affect the interpretation of VCP elevation patterns and their underlying mechanisms. These include collinearity between elevation and climate variables, the exclusion of herbaceous plants from our investigation, and a lack of long-term soil microclimate data. Despite these limitations, our findings broaden our understanding of the relationship between biodiversity and ecosystem function in woody communities. Future work should incorporate the combined effects of elevation and climate on VCP to better address climate impacts, quantify the contribution of the herbaceous layer to VCP, and investigate plant-soil-microclimate feedbacks through long-term monitoring.

5. Conclusion

Our study provides new insights into the elevational patterns of savanna VCPs and the underlying mechanisms that drive these patterns in the dry-hot valley region of Southwest China. Total VCP showed no significant elevational trend, whereas its two components (tree and shrub VCP) exhibited contrasting patterns. Total VCP was more strongly influenced by structural diversity than by species diversity. Furthermore, the contribution of trees substantially exceeded that of shrubs, indicating that tree structure-mediated complementarity plays a major role in carbon sequestration in this savanna ecosystem. In contrast to the structural drivers of the tree VCP, the shrub VCP was primarily driven by its density and structural diversity. This demonstrates the unique role of trees and shrubs in shaping VCPs in savanna ecosystems. Our findings emphasize that managing tree and shrub species composition and density to foster high species and structural diversity is key to enhancing carbon storage, supporting biodiversity conservation, and advancing vegetation restoration in stress-prone habitats in the context of climate change.

Acknowledgments

This work was supported by the Key Research and Development Program of Yunnan Province (No. 202303AC100009), National Natural Science Foundation of China (32571918, 32201426, 32260300), Science and Technology Talents and Platform Program of Yunnan Province (No. 202505AA350008), Yunnan Province Field Scientific Observation and Research Station for Vertical Band of Composite Ecosystem in Baima Snow Mountain, Yunnan, China (202205AM070005) and the Scientific Research Fund Project of Yunnan Education Department and Yunnan University (2025Y0078).

CRediT authorship contribution statement

Wan-Chen Li: Writing-original draft, Visualization, Validation, Data curation, Software, Investigation. Qin Huang: Writing-review & editing. Ru-Jing Yang: Writing-review & editing, Investigation. Zhi-Yan Peng: Writing-review & editing. Qiong Cai: Writing-review & editing. Wen-Jing Fang: Writing-review & editing, Funding acquisition. Wen-Jun Liu: Writing-review & editing, Funding acquisition. Su-Hui Ma: Writing-review & editing, Methodology, Formal analysis, Resources, Funding acquisition. Ya-Jun Chen: Resources, Project administration. Zhi-Ming Zhang: Resources, Funding acquisition.

Data accessibility statement

Data on the vegetation carbon pool, community structure variables, and plant diversity indices across the four elevations investigated in this research have been shared as Appendix A.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.pld.2025.12.005.