The Nonggang National Nature Reserve, located in the southern Guangxi Zhuang Autonomous Region, southwestern China (22°13'56"–22°39'09"N, 106°42'28"–107°04'54"E; Fig. 1), covers an area of 10,080 ha. The region has a tropical monsoon climate, with an average annual temperature of approximately 22 °C. The temperature ranges from above 13 °C in January, the coldest month, to above 28 °C in July, the hottest month. The annual accumulated temperature ranges from 7,400 to 7,800 °C. Annual average rainfall ranges from 1,200 to 1,500 mm and is concentrated between May and September. The bedrock of the area is primarily composed of limestone, and the main soil types are primitive calcareous, brown calcareous, and black calcareous. The topography of the region is characterized by typical karst peak-cluster depression landforms, with depressions formed by clustered peaks sharing a common base and a funnel-shaped landscape. Elevation ranges from 150 to 600 m. Unique landforms, such as peak clusters, forests, depressions, and funnels, result in significant variations in light availability, soil thickness, and moisture. Other micro-landform features formed by numerous rock outcrops—such as stone gullies, stone surfaces, and depressions—contribute to the heterogeneity of small-scale habitats. The reserve contains 1,752 recorded species of vascular plants belonging to 810 genera and 184 families (Huang et al. 2013). Among these, 33 species are rare and endangered in China, and 278 species are endemic to karst habitats (Wang et al. 2014). The dominant tree species in the TKSR include Excentrodendron tonkinense (A. Chev.) H.T. Chang & R.H. Miao, Camellia petelotii (Merr.) Sealy, Cephalomappa sinensis (Chun & F.C. How) Kosterm., Garcinia paucinervis Chun ex F.C. How, and Deutzianthus tonkinensis Gagnep.

Figure 1.

Geographical location of the study area, with photographs showing tropical karst seasonal rainforests along steep karst hillslopes.

During the peak growing season from July to September 2020, forest plots were established at different slope positions based on elevation variations in the karst peak-cluster depressions in the study area (Fig. 1). There were 10 plots, each with an area of 20 × 20 m, at four slope positions — depression (DE), lower slope (LS), middle slope (MS), and upper slope (US)—with a spacing of more than 30 m between plots. A total of 40 plots were established along steep karst hillslopes in the study area. Tree-layer investigations included species names, diameters at breast height (DBH), and heights. Within each plot, two 5 × 5 m subplots were randomly established to investigate the shrub layer, including species names, average heights, coverages, and numbers. Additionally, four 1 × 1 m subplots were randomly established to investigate the herb layer, including species names, numbers, heights, and coverages. All vascular plants in each plot and subplot were inventoried, with species identifications verified by two plant taxonomy experts. Furthermore, latitude and longitude coordinates, elevation, aspect, slope position, and rock outcrop rate were recorded. Three experienced observers visually estimated the slope position and rock outcrop rate, and a handheld GPS (Garmin GPSMAP 60CSx) was used for all other measurements. Slope positions were represented by categorical variables 1, 2, 3, and 4, corresponding to depression, lower slope, middle slope, and upper slope, respectively (Qiu and Zhang 2000). The slope aspect was converted from a range of 0–360° azimuth to a value ranging from 0 to 1 using a previously documented formula (Yu et al. 2013).

Within each plot, soil samples were collected at depths of 0–20 cm using the five-point sampling method (Lu et al. 2023). The collected soil samples were mixed and returned to the laboratory. After air-drying and sieving, the soil samples were subjected to chemical analyses to determine various soil parameters. The measured indicators included soil pH, exchangeable calcium (ECa), exchangeable magnesium (EMg), available phosphorus (AP), available potassium (AK), ammonium nitrogen (NH₄⁺-N), nitrate nitrogen (NO₃--N), organic matter (SOM), total nitrogen (TN), total phosphorus (TP), and total potassium (TK). Soil pH was determined in a suspension with a 1:2.5 soil-to-water ratio. ECa and EMg were measured using atomic absorption spectrophotometry (AA-7000, Shimadzu, Kyoto, Japan). NH₄⁺-N was quantified using the indophenol blue colorimetric method, and NO₃--N was measured using the cadmium reduction method followed by a sulfanilamide–NAD reaction. SOM content was determined using the traditional potassium dichromate oxidation method. TN was measured using the semi-micro Kjeldahl method. AP and TP levels were determined using the molybdenum–antimony colorimetric method. AK and TK levels were measured using the sodium hydroxide fusion flame photometry method. The surface soil water content (SWC) of each plot was measured using a portable three-parameter soil tester (WET-2, Delta-T Devices Ltd., United Kingdom) following the five-point sampling method (Lu et al. 2023). These methods for measuring soil properties were described in detail by Lu (1999) and Bao (2000). Environmental variables (mean ± SD) in four slope positions are presented in Suppl. material 1: table S1.

A phylogenetic tree of vascular plants was constructed using the V.PhyloMaker package (Jin and Qian 2019), incorporating evolutionary branch lengths based on the “GBOTB.extended.tre” framework following the methodological approach of scenario S3. This scenario employs average distances to constrain phylogenetic tip positions, thereby reducing bias introduced by polytomies. The “GBOTB.extended.tre” tree integrates the evolutionary branches of seed plants from GBOTB (Smith and Brown 2018) and ferns from the updated, revised, and expanded phylogenetic system of Zanne et al. (2014).

The Faith Diversity Index (PD) was used to measure phylogenetic diversity (Faith 1992). The net relatedness index (NRI) and nearest taxon index (NTI) were used to determine the phylogenetic structure. NRI and NTI represent the standardized values of the mean pairwise distance (MPD) and mean nearest taxon distance (MNTD), respectively. The NRI primarily describes the overall phylogenetic relationships among species in a community, whereas the NTI focuses on phylogenetic relationships at the terminal branches of the phylogenetic tree. Compared with NRI, NTI places greater emphasis on describing the phylogenetic distances between neighboring species (Webb et al. 2002). MPD and MNTD were calculated using the comdist and comdistnt functions in the picante package (Kembel et al. 2008). The actual values for each community were compared using an unconstrained null model (Kembel and Hubbell 2006), in which the terminal branches of the phylogenetic tree were randomly permuted 999 times for all species in the community. The differences were then calculated, and the NRI and NTI values for each community were obtained (Webb et al. 2002; Swenson et al. 2007).

PD = ∑L_b (4)

$$\mathrm{NRI}=-1\times \frac{MP{D}_{\text{observed}}-MP{D}_{\text{randomized}}}{sdMP{D}_{\text{randomized}}}$$ (5)

$$\mathrm{NTI}=-1\times \frac{MNT{D}_{\text{observed}}-MNT{D}_{\text{randomized}}}{sdMNT{D}_{\text{randomized}}}$$ (6)

where L_b represents the branch length connecting the species on the tree. MPDobserved and MNTDobserved are the observed values of the average pairwise phylogenetic distance and mean nearest taxon distance, respectively, for each plot. MPDrandomized and MNTDrandomized are the average pairwise phylogenetic distance and mean nearest taxon distance, respectively, obtained by randomly generating null models for the community. sdMPDrandomized and sdMNTDrandomized are the standard deviations of the average pairwise phylogenetic distance and mean nearest taxon distance obtained from the randomly generated null models for the community, respectively. If the NRI (or NTI) > 0, the phylogenetic relationships among individuals in the community are closer than those expected from a randomly generated null model, suggesting phylogenetic clustering. Conversely, if the NRI (or NTI) < 0, it suggests phylogenetic overdispersion, indicating that the phylogenetic relationships among individuals in the community are more divergent than expected from the null model (Webb et al. 2002). If the NRI (or NTI) < −1.96, it indicates significant phylogenetic evenness compared with the null model, whereas an NRI (or NTI) > 1.96 indicates significant phylogenetic clustering compared with the null model (Vamosi et al. 2009).

Assuming that the data were normally distributed and satisfied the assumption of homogeneity of variances, one-way ANOVA was used to compare differences in environmental variables among the different slope positions. For multiple comparisons of plant diversity between slope positions, the least significant difference (LSD) method was applied if variances were equal; however, if variances were not equal, the Games–Howell test was used. Pearson correlation analysis was conducted to examine relationships between species, phylogenetic diversity, and environmental variables. Factors with r > 0.6 were removed. The remaining factors were then included in multiple stepwise regression analyses, along with diversity indices and phylogenetic structures, to identify the main factors influencing species and phylogenetic diversity. The stepwise regression procedure was based on the Akaike Information Criterion (AIC), in which variables were added or removed by selecting the model with the minimum AIC value. Residual analysis confirmed that the model satisfied the assumptions of normality, linearity, and homoscedasticity, thereby validating the use of a regression model based on the Gaussian distribution. The factors retained in the regression equations were quantified using the rdacca.hp package (Lai et al. 2022) in R 4.2.2 to determine their explanatory power for the diversity indices through hierarchical partitioning analysis. One-way ANOVA and Pearson correlation analysis were performed using SPSS 26.0 (SPSS Inc., Chicago, IL, USA).

Species diversity in the herb, shrub, and tree layers initially decreased and then increased with slope position (Fig. 2). In the herb layer, species richness in the middle slope was significantly lower than that in the lower and upper slopes (p < 0.05), and the Pielou index showed an increasing trend, with values for both the middle and upper slopes significantly higher than those in the lower slopes (p < 0.05). There were no significant differences in species diversity in the shrub layer among the different slope positions. In the tree layer, species richness initially decreased and then increased with slope position, whereas the other indices showed an overall increasing trend, with no significant differences among slope positions.

Figure 2.

Changes in species diversity at different slope positions. Different lowercase letters indicate a significant difference between communities at different slope positions (p < 0.05). DE, depression; LS, lower slope; MS, middle slope; US, upper slope.

Phylogenetic diversity in the herb, shrub, and tree layers initially decreased and then increased with slope position (Fig. 3). Specifically, phylogenetic diversity in the lower and upper slopes was significantly higher than that in the middle slope in the herb layer (p < 0.05). In contrast, there were no significant differences in phylogenetic diversity among slope positions in the shrub and tree layers. Across the three layers, phylogenetic diversity was highest in the tree layer, followed by the herb and shrub layers.

Figure 3.

Phylogenetic diversity at different slope positions. Different lowercase letters indicate a significant difference between communities at the various slope positions (p < 0.05). DE, depression; LS, lower slope; MS, middle slope; US, upper slope.

The NRI and NTI in the herb, shrub, and tree layers showed inconsistent patterns in community species phylogenetic relationships across the slope positions (Fig. 4). In the herb layer, both NRI and NTI indicated phylogenetic clustering across the four slope positions. In the shrub layer, both NRI and NTI exhibited phylogenetic divergence on the lower and upper slopes, whereas the middle and depression slopes showed a coexistence of phylogenetic divergence and clustering, with significant differences in NRI among the slope positions (p < 0.05). In the tree layer, both NRI and NTI exhibited an overall pattern of phylogenetic clustering on the lower and upper slopes. In contrast, the depression and middle slopes showed a coexistence of phylogenetic divergence and clustering. Overall, the herb and tree layers showed phylogenetic clustering, whereas the shrub layer exhibited phylogenetic divergence.

Figure 4.

Changes in phylogenetic structures at different slope positions. Different lowercase letters indicate significant differences between communities at the various slope positions (p < 0.05). DE, depression; LS, lower slope; MS, middle slope; US, upper slope.

Correlation analysis revealed that species diversity in the herb layer was significantly and negatively correlated with NO₃--N, SOM, and SWC, and significantly and positively correlated with SP, SA, and SD (p < 0.05). The phylogenetic structure was significantly and positively correlated with AK, TP, and BA (p < 0.05). Phylogenetic diversity was significantly and negatively correlated with NO₃--N and SOM contents (p < 0.05; Fig. 5).

Figure 5.

Pearson correlation analysis between diversity metrics and environmental factors. H_S, H_D, H_H, H_J, H_NRI, H_NTI, and H_PD indicate the species richness, Simpson index, Shannon–Wiener index, Pielou index, net relatedness index, nearest taxon index, and phylogenetic diversity of the herb layer, respectively; S_S, S_D, S_H, S_J, S_NRI, S_NTI, and S_PD indicate the species richness, Simpson index, Shannon–Wiener index, Pielou index, net relatedness index, nearest taxon index, and phylogenetic diversity of the shrub layer, respectively; T_S, T_D, T_H, T_J, T_NRI, T_NTI, and T_PD indicate the species richness, Simpson index, Shannon–Wiener index, Pielou index, net relatedness index, nearest taxon index, and phylogenetic diversity of the tree layer, respectively. SP, slope position; SA, slope aspect; SD, slope degree; ROR, rock outcrop rate; ECa, exchangeable calcium; EMg, exchangeable magnesium; AP, available phosphorus; AK, available potassium; NH₄⁺-N, ammonium nitrogen; NO₃--N, nitrate nitrogen; pH, soil pH; TK, total potassium; TN, total nitrogen; TP, total phosphorus; SOM, soil organic matter; SWC, soil water content; CD, canopy density; BA, basal area of woody species. * p <0.05; ** p <0.05.

In the shrub layer, species diversity was significantly and negatively correlated with TP and BA and significantly and positively correlated with SA and pH (p < 0.05). The phylogenetic structure was significantly and positively correlated with NH₄⁺-N, NO₃--N, pH, and CD (p < 0.05). Phylogenetic diversity was significantly and positively correlated with SA and negatively correlated with TP (p < 0.05; Fig. 5).

In the tree layer, species diversity was significantly and positively correlated with SP and CD and significantly and negatively correlated with soil TK, TP, and BA (p < 0.05). The phylogenetic structure was significantly negatively correlated with SP, SD, and SWC, and significantly positively correlated with SD, ROR, ECa, EMg, TK, TN, TP, SOM, and SWC (p < 0.05). Both NRI and NTI showed different responses to SD and SWC. Phylogenetic diversity was significantly negatively correlated with soil TK, TP, and BA (p < 0.05; Fig. 5).

In the herb layer, the environmental variables retained in the regression equation collectively explained 24.5%, 13.6%, 19.6%, and 22.1% of the variance in species richness, Simpson index, Shannon–Wiener index, and Pielou index, respectively (Table 1). Among these, SA, SP, and SOM had high explanatory powers for species diversity (Fig. 6). The retained environmental variables in the regression equation collectively explained 14.0% and 16.8% of the variance in NRI and NTI, respectively (Table 1), with BA, ROR, and soil pH contributing significantly (Fig. 6). The environmental factors retained in the regression equation collectively explained 14.6% of the variance in the PD index, with SOM as the main influencing factor (Fig. 6).

Figure 6.

Contribution rates of the environmental variables to the explained variation in species and phylogenetic diversity and structure using hierarchical partitioning analysis. H_S, H_D, H_H, H_J, H_NRI, H_NTI, and H_PD indicate the species richness, Simpson index, Shannon–Wiener index, Pielou index, net relatedness index, nearest taxon index, and phylogenetic diversity of the herb layer, respectively; S_S, S_D, S_H, S_J, S_NRI, S_NTI, and S_PD indicate the species richness, Simpson index, Shannon–Wiener index, Pielou index, net relatedness index, nearest taxon index, and phylogenetic diversity of the shrub layer, respectively; T_S, T_D, T_H, T_J, T_NRI, T_NTI, and T_PD indicate the species richness, Simpson index, Shannon–Wiener index, Pielou index, net relatedness index, nearest taxon index, and phylogenetic diversity of the tree layer, respectively. SP, slope position; SA, slope aspect; ROR, rock outcrop rate; ECa, exchangeable calcium; EMg, exchangeable magnesium; AP, available phosphorus; NH₄⁺-N, ammonium nitrogen; NO₃--N, nitrate nitrogen; pH, soil pH; SOM, soil organic matter; CD, canopy density; BA, basal area of woody species.

In the shrub layer, the environmental variables retained in the regression equation collectively explained 15.1%, 16.9%, 17.0%, and 20.9% of the variance in species richness, Simpson index, Shannon–Wiener index, and Pielou index, respectively (Table 1). Among these, BA, SA, and soil pH had high explanatory powers for species diversity (Fig. 6). The retained environmental variables in the regression equation collectively explained 36.3% and 35.6% of the variance in NRI and NTI, respectively (Table 1), with soil pH, NH₄⁺-N, and CD exerting strong influences (Fig. 6). The environmental factors retained in the regression equation collectively explained 21.5% of the variance in the PD index, with SA and soil NO₃--Nhaving strong impacts (Fig. 6).

In the tree layer, the environmental variables retained in the regression equation collectively explained 30.1%, 32.7%, 32.1%, and 29.3% of the variance in species richness, Simpson index, Shannon–Wiener index, and Pielou index, respectively (Table 1), with BA, CD, and SP showing the highest explanatory power for species diversity (Fig. 6). The retained environmental variables in the regression equation collectively explained 41.1% and 24.7% of the variance in NRI and NTI, respectively (Table 1). SP, soil EMg, ECa, and pH also had high explanatory powers (Fig. 6). The environmental factors retained in the regression equation collectively explained 32.0% of the variance in the PD index, with SP, ROR, and BA making significant contributions (Fig. 6).

Plant community distributions result from the combined effects of various environmental factors (climate, soil, topography, and biology) operating at multiple scales. In this study, noticeable differences were observed in topographic, soil, and biological factors along the slope gradient (see Suppl. material 1: table S1), resulting in habitat heterogeneity at different slope positions, which subsequently affected species diversity in the plant communities. Terrain influences species distribution and diversity by spatially redistributing resources such as soil nutrients, light, heat, and water (Du et al. 2015). Slopes integrate various environmental factors, including light, soil thickness, moisture, and nutrient conditions, which in turn significantly influence plant diversity (Liu et al. 2018). In this study, as slope increased, the species diversity of the herb and shrub layers first decreased and then increased. Species richness in the tree layer showed a similar trend, whereas other diversity indices, such as the Simpson, Shannon–Wiener, and Pielou indices, generally showed a gradual increase (Fig. 2). The elevation of the karst peak-cluster depression is relatively low (<600 m), and there are substantial topographic differences in soil and water availability on the mountainside (Zhang et al. 2021). Conditions in the depression are favorable for plant community species diversity. In contrast, the mountainside contains various microhabitats (rock cliffs, gullies, and open soil) that provide living spaces for plants with different preferences (Zhang et al. 2013). Huang et al. (2016) suggested that conditions on the mountainside may be harsh due to factors such as water scarcity, thin soil, and strong solar radiation, which can make it difficult for any one species to become dominant, thus promoting higher species diversity.

Soil is an essential component of terrestrial ecosystems and is responsible for many ecological processes; therefore, it is a critical factor in research on the renewal and succession of plant communities (Daunoras et al. 2024). In karst areas, plant roots are widely distributed in surface soil, the quality of which determines its favorability for plant growth (Yan et al. 2022). Numerous studies have demonstrated that the physicochemical properties of soil have a significant impact on plant diversity (Sheng et al. 2015; Guo et al. 2018; Gastauer et al. 2020). In this study, the diversity of the herb, shrub, and tree layers was influenced by multiple soil factors. The species diversity of the herb layer was significantly negatively correlated with soil NO₃--N, SOM, and SWC. However, in the shrub layer, it was negatively correlated with soil TP but positively correlated with pH. In contrast, diversity in the tree layer was negatively correlated with soil TK and TP. Although nitrogen and phosphorus are widely recognized as the primary limiting nutrients for terrestrial plants globally (Borer et al. 2015), our study site in the tropical karst region aligns with findings identifying phosphorus and potassium as key limiting elements (Tan et al. 2019). This difference highlights the uniqueness of the karst ecosystem and suggests that potassium limitation may replace nitrogen as a key constraint alongside phosphorus in this environment. Borer et al. (2015) detected widespread phosphorus limitation at the tree species level in tropical forests, suggesting that intense competition for soil phosphorus may be a common phenomenon at the species scale. Han et al. (2021) indicated that soil potassium is a major nutrient-limiting factor in the subtropical karst areas of Guangxi, China. Our study provides further evidence that soil phosphorus and potassium are key factors limiting plant species diversity in karst areas.

Plants interact with their local environment, leading to changes in community species composition and distribution, which, in turn, affect community species diversity (Gaiarsa et al. 2025). In this study, species diversity in the shrub and tree layers was significantly negatively correlated with BA, whereas CD primarily affected species diversity in the tree layer. You et al. (2016) found that basal area was the most significant factor influencing the composition of shrub layer species in the community. An increased basal area, reflecting a higher presence of large-diameter trees, suppresses the growth of small-diameter plants, leading to negative interactions at a small scale. As the forest naturally thins, community species diversity decreases accordingly (Chen et al. 2023). Canopy closure is a key factor in studying understory plant diversity (Jules et al. 2008; You et al. 2016) because it limits light availability, thereby mitigating plant competition for resources within the community. Consequently, competition has a less significant impact on the tree layer, resulting in reduced interspecific competition, which increases the diversity of tree layer species (Jules et al. 2008; Chen et al. 2023). This study reveals that biological, topographic, and edaphic factors have important effects on the distribution patterns of community species diversity across three forest strata (herb, shrub, and tree layers) in the TKSR. The divergent responses of these plant layers to both biotic and abiotic environmental factors result in measurable variations in species diversity along slope gradients. Nevertheless, the coupled effects of these factors on the pattern of community species diversity require further exploration.

High phylogenetic diversity suggests that spatial habitats can support the coexistence of species with relatively high levels of evolutionary diversity and variation, as well as longer evolutionary histories (Zeng et al. 2018). In this study, the phylogenetic diversity of the herb, shrub, and tree layers initially decreased and then increased with increasing slope position, exhibiting significant differences in phylogenetic diversity across the slope gradient in the herb layer. Plant communities in the depressions and upper slopes exhibited higher species diversity, corresponding to longer evolutionary histories and higher levels of evolutionary diversity and variation. Conversely, plant communities on the lower and middle slopes showed opposite trends. Xiao et al. (2019) suggested that the different phylogenetic diversity responses of species with different life forms to slope variation are related to their adaptive strategies to different environments. The species richness of woody plant communities is primarily determined by climatic factors such as low winter temperatures (Watanabe et al. 2024), whereas microenvironmental factors such as light intensity, soil moisture, and nutrient status in the understory are the main determinants of herbaceous plant community composition (Reich et al. 2012). This may explain why the phylogenetic diversity of the herb layer changed significantly with an increase in slope position in the peak-cluster depression. In contrast, the phylogenetic diversity of the shrub and tree layers did not show a significant trend. In this study, SA and SOM both had significant effects on the phylogenetic diversity of the herb layer; SA, NO₃--N, and BA significantly affected the phylogenetic diversity of the shrub layer; and SP, ROR, BA, and CD significantly influenced the phylogenetic diversity of the tree layer. Generally, sunny slopes with favorable light conditions promote plant growth, whereas relatively low levels of soil moisture on shaded slopes can inhibit plant growth. Thus, slope aspect may either promote or restrict community species richness (Liu et al. 2018; Zhang et al. 2020). In this study, communities on sunny slopes had greater species richness and higher phylogenetic diversity than those on shaded slopes in the herb and shrub layers, possibly due to the relatively favorable habitat conditions on sunny slopes, which enabled plants to reproduce more rapidly. Species continuously evolve and diversify to occupy suitable ecological niches, thereby increasing the phylogenetic diversity of their communities. Previous studies have indicated that species with different life forms respond differently to the same habitat conditions (Ding et al. 2015), and similar results were obtained in this study. Phylogenetic diversity research can help determine which communities should be prioritized for conservation within a region, as communities with higher phylogenetic diversity have greater evolutionary potential and a stronger ability to cope with risks associated with global change (Devictor et al. 2010). The results indicate that communities in depressions and on upper slopes have higher conservation value because of their greater species and phylogenetic diversity.

Community phylogenetic structures result from the interactions among ecological factors, environmental filtering, and historical evolution, and they can reveal the primary ecological processes influencing species diversity (Zhang et al. 2016). Based on the assumption of phylogenetic niche conservation, environmental filtering leads to the coexistence of species with similar physiological tolerances, resulting in an aggregated relationship among phylogenetic lineages within the community. Interspecific competition enables distantly related species to coexist within a community, resulting in divergent relationships among phylogenetic lineages (Webb et al. 2002). In this study, the overall phylogenetic structures of the herb and tree layer communities tended to be aggregated, whereas that of the shrub layer community tended to be divergent. This suggests that the herb and tree layer communities were primarily composed of closely related species and that environmental filtering was the main ecological process influencing community assembly. In contrast, the species in the shrub layer community were more distantly related, and the primary ecological process influencing their community assembly was competitive exclusion. This may be associated with spatial scale, as interspecific competition often occurs when resources within a habitat are limited. In this study, the relatively small sampling area of the shrub layer may have exacerbated competition among species owing to restricted space and resources, leading to the coexistence of different species occupying different niches (Zhu et al. 2018). On a larger scale, environmental filtering leads to the coexistence of species with similar resource utilization, resulting in a more aggregated phylogenetic structure within the community. Studies have shown that closely related species are unlikely to coexist at small scales, leading to significant phylogenetic divergence within the community (Silva and Batalha 2009). As spatial scale increases, phylogenetic structures in communities gradually shift from divergent to aggregated (Jin et al. 2020), a trend that is particularly evident in tropical rainforests (Hubbell 2006). Furthermore, strong environmental filtering in karst habitats results in various phylogenetic structural patterns within plant communities, which are influenced by multiple environmental factors at different levels. Specifically, phylogenetic aggregation in the herb layer was associated with increases in soil AK, TP, and BA. In the shrub layer, aggregation was driven by elevated soil NH₄⁺-N, NO₃--N, pH, and BA. Conversely, the tree layer responded to increases in soil ECa, EMg, TK, TN, TP, SOM, and ROR. However, an increase in SD promoted an overall aggregation trend in the phylogenetic relationships of tree layer species, whereas SOM had the opposite effect, increasing the phylogenetic distance between adjacent species and affecting the phylogenetic structure of the tree layer. The phylogenetic structure of plant communities is closely related to the response of plants to their environment, which in turn affects the ecological processes that dominate community assembly.

The phylogenetic structures of the herb, shrub, and tree layers exhibited inconsistent patterns along the slope gradient. The herb layers across the different slopes exhibited aggregated phylogenetic structures. In addition to the influence of microhabitat conditions, the phylogenetic structure of the herb layer may be related to the life histories of herbaceous plants (Qian et al. 2014). Herbaceous plants have short life cycles and strong dispersal abilities, can rapidly occupy suitable habitats, exhibit large population fluctuations, and easily form aggregated phylogenetic structures (Niu et al. 2011). The dispersal of tree seeds may also influence the phylogenetic structure of the herb layer community. Herb layers include woody plant seedlings, and the limited dispersal of seeds from parent trees in the forest leads to a clustered pattern of seedlings, resulting in an aggregated phylogenetic structure in the herb layer community (Niu et al. 2011). The phylogenetic structure of the shrub layer exhibited divergence in the depression and upper slope and aggregation in the middle slope communities. In contrast, the lower slope communities showed inconsistent NRI and NTI values for the phylogenetic relationships of community species, with NRI indicating divergence and NTI indicating aggregation. In the tree-layer communities, the NRI and NTI of the depression and middle slopes exhibited inconsistent phylogenetic structures. In the depression, NRI showed divergence, whereas NTI showed aggregation. Conversely, the opposite pattern was observed in the middle slope communities, while the phylogenetic structures of the upper and lower slope communities were aggregated. In high-altitude areas, due to environmental stress, species within communities undergo convergent evolution, resulting in a trend toward aggregated phylogenetic structures (Kluge and Kessler 2011; You et al. 2013). Similarly, research has shown that the phylogenetic structures of evergreen broad-leaved forest communities exhibit aggregation in low-altitude areas and divergence in high-altitude areas (Huang et al. 2010). In this study, the harsh habitat conditions on the upper slope may have caused strong environmental stress, promoting convergent evolution in plants and resulting in aggregated phylogenetic structures in the tree-layer communities. Conversely, competition for limited resources among plants may have caused the phylogenetic structure of shrub layer plants to diverge. The inconsistent changes in community NRI and NTI values are related to their different emphases, indicating that distinct ecological processes influence the assembly of herb, shrub, and tree layer communities. The different ways in which communities respond to the environment affect the relative importance of ecological processes in community assembly. The results indicate that deterministic processes (environmental filtering and competitive exclusion) have significant impacts on community assembly in the TKSR, while some communities may also be influenced by neutral processes (NRI or NTI < −1.96) (Vamosi et al. 2009).

Karst landscapes are crucial biodiversity reservoirs, particularly in Asia, hosting a wealth of endemic species (Clements et al. 2006). The steep karst regions of southwestern China are recognized as biodiversity hotspots, notable for their remarkable plant species richness and significant conservation value (Guo et al. 2018; Geekiyanage et al. 2019). Protecting biodiversity and rehabilitating degraded vegetation in these areas remain pivotal ecological priorities for the Chinese government and researchers (Chen et al. 2024). Conservation strategies should incorporate phylogenetic diversity into assessment frameworks, complementing traditional species diversity metrics to reflect evolutionary relationships (Tietje et al. 2023). Consequently, areas with high and unique phylogenetic diversity should be prioritized for conservation efforts (Zhang et al. 2024). Our study demonstrates that depressions and upper slopes exhibit greater species and phylogenetic diversity than other slope positions, highlighting their significant conservation value for plant diversity. Karst depressions, characterized by deep, fertile soils, are especially susceptible to anthropogenic conversion. In southwestern China, this has resulted in substantial natural forest loss and degradation, primarily due to their conversion into agricultural and plantation lands (Yuan et al. 2024). Therefore, conservation efforts for forests in karst depressions should be prioritized. Our findings provide critical insights for planning forest reserves and developing conservation strategies in karst regions.

The authors have declared that no competing interests exist.

No ethical statement was reported.

No use of AI was reported.

This work was supported by the Guangxi Natural Science Foundation (2021GXNSFFA196005, 2021GXNSFAA196024, 2022GXNSFBA035633, 2022GXNSFBA035461); the Special Funding for Guangxi Bagui Young Top Talents Program (to Zhonghua Zhang); and the National Natural Science Foundation of China (31960275, 31760128).

Qingling Pang: Conceptualization; data curation; methodology; writing—original draft. Cong Hu: Funding acquisition; supervision; writing—original draft. Chaohao Xu: supervision; writing—original draft. Chaofang Zhong: supervision; writing—original draft. Zhonghua Zhang: Conceptualization; funding acquisition; supervision; visualization; writing—review and editing. Gang Hu: Project administration; supervision; visualization; writing—review and editing.

Qingling Pang https://orcid.org/0009-0002-1168-7402

Cong Hu https://orcid.org/0000-0001-7507-4520

Chaohao Xu https://orcid.org/0000-0002-0437-5390

Chaofang Zhong https://orcid.org/0000-0002-3509-3628

Zhonghua Zhang https://orcid.org/0000-0003-2094-698X

Gang Hu https://orcid.org/0000-0002-0662-811X

All data that support the findings of this study are available upon request.