1. Introduction

Grasslands cover nearly one-third of the Earth's land surface (Lemaire et al., 2014), serving as the main production base for livestock products (meat, milk, etc.) and staple crops (Bengtsson et al., 2019; Bardgett et al., 2021), while also constituting a globally important species bank. However, global climate change and human demand for animal products have threatened grassland ecosystems (Wang et al., 2019). Therefore, understanding the impact of the type and intensity of human management on grassland ecosystems is of critical importance.

Land use and biological resource exploitation are the main factors that drive biodiversity change (Bai et al., 2007; Li et al., 2015, 2017). The intensity and direction of the impact of management types on biodiversity are influenced by geography, vegetation type, and spatial scale. The alpine grasslands of the Qinghai–Tibet Plateau and the temperate grasslands of Northern China are traditional pastures of vast areas. Numerous studies have investigated the biodiversity of the grasslands in these areas and its changes under different management regimes and intensities (Jiang et al., 2003; Zhao et al., 2008; Wang et al., 2019; Li et al., 2021b; Sun et al., 2021). Studies have found that the taxonomic diversity of grassland communities increase under moderate mowing and grazing, which is consistent with "the intermediate disturbance hypothesis" (Connell, 1978; Pei et al., 2008; Baoyin et al., 2014; Zheng et al., 2022, 2023). Moderate grazing directly affects grassland heterogeneity, increasing species dispersal by counteracting the effects of environmental gradients related to topography, soil texture, nutrients or hydrology (Xu et al., 2005). However, even at medium intensity, long-term continuous grazing can lead to grassland degradation, because constant livestock grazing and trampling not only reduce grassland biomass, but also produce a selection effect by weakening graminoids, reducing the phylogenetic diversity of grasslands (Socher et al., 2012; Li et al., 2019; Wang et al., 2019; Zhu et al., 2020). This type of grazing also leads to soil compaction and structural destruction, and eventual reduction of vegetation productivity (Yang et al., 2002; Teague and Dowhower, 2003, 2011; van Klink et al., 2015). Compared to grazing, which is highly selective, the non-selective mowing is more likely to homogenize plant communities (Gossner et al., 2016; Li et al., 2020). Nevertheless, moderate mowing may also stimulate compensatory growth of shorter subdominant plant species by increasing light availability and germination rates, thus, promoting species coexistence (Li et al., 2016; Wang et al., 2020; Zhu et al., 2021b; Yang et al., 2022). In contrast, intensive mowing generally leads to dwarfing of plants, reducing seed yield and, therefore, diversity (Wan et al., 2016). Regional differences in plant species pools, environmental and geographical characteristics will also change the effects of management measures on grassland biodiversity (Egorov et al., 2014).

The southern grasslands of China generally refer to the montane and subalpine meadows and grassy slopes in subtropical humid and semi-humid regions (Zhang et al., 1998; Ren, 1999), mainly distributed in the areas south of the Qinling Mountains and the Huai River, and east of the Qinghai–Tibet Plateau. This type of grasslands covers an area of about 63.41 × 10⁴ km², accounting for 21.8% of the total grassland area in the country (Fang et al., 2018). The southern grasslands generally have a secondary nature and fragmented distribution. The vegetation primarily consists of grass clumps, bushveld, as well as small amounts of lowland and alpine meadows (Ren and Zhang, 2002). Due to its favorable hydrothermal conditions and long frost-free season, the southern grasslands have high coverage, active regeneration, and high grass production, with productivity reaching 4–6 times that of typical temperate grasslands, showing considerable potential for livestock production (Ren, 1999; Liu et al., 2008; Zhang et al., 2021).

The southern grasslands of China are scattered in remote mountains. Consequently, these grasslands have only recently been exploited for economic purposes. Accordingly, research on how grassland structure and function are impacted by management type remains inadequate. Furthermore, little is known about how management types and intensity influence biodiversity of southern grasslands. Southern grasslands are generally grazed and mowed. When mowed, seeds of forage species are also sown to enhance grassland productivity. Over-seeding of commonly used forage seeds, such as Dactylis glomerata, Lolium perenne and Trifolium repens, has changed the species composition and dominant population of grassland communities, increasing stress on native flora, plant community dynamics and soil conditions (Xin et al., 2004). Recent studies of the ecology of southern grasslands have examined soil physicochemical properties (Zhu et al., 2021a), microbial community diversity (Du et al., 2014; Yang et al., 2018; Zhou et al., 2023), and the relationship between biodiversity and productivity (Liu et al., 2005, 2018; Haynes et al., 2013). The effects of different management measures and intensity on grassland quality and carrying capacity have also been explored (Wan et al., 2013; Wang et al., 2021). However, the community composition and structure of southern grasslands have undergone rapid changes due to climate change and human interference. Furthermore, the determining factors of grassland biodiversity vary greatly across both space and time. So far, studies of these grasslands have been scarce and mostly focused on the community scale, while little is known about the spatial variation of plant diversity and the impact of natural and human factors at the regional scale. Therefore, it is of great significance to comprehensively study the plant diversity patterns in southern grasslands and the roles of key processes, especially management type and intensity, that drive community assembly.

The Wumeng Mountains, located in the northeastern Yunnan, are characterized by a broad distribution of grassland landscapes across the flat mountain tops (Ren et al., 2018). This region possesses a tradition of animal husbandry that has been largely intensified by modern grassland management (Liu, 2016; Wu, 2019). This study focused on the southern grasslands distributed in the Wumeng Mountains, comparing plant diversity of grasslands under different management types (protected, grazed, mowed grasslands), and estimated the effects of different influencing factors on grassland species diversity. We specifically aim to answer two questions: (1) Are the differences significant in plant diversity and composition among grassland communities under different management types? (2) Does the impact of different management types surpass the influence of geographical and environmental factors on the biodiversity of grasslands in this region?

2. Materials and methods 2.1. Study area

Northeastern Yunnan is situated within the Chuan-Dian Block of the Sichuan Basin. The region is characterized by the southwest–northeast oriented mountain ranges known as the Wumeng Mountains, which reach elevations ranging from 2500 to 4000 m. Landforms of northeastern Yunnan represent the "residual denudational surface" feature from the Qinghai–Tibet Plateau (He and Pu, 2014), and include high-altitude, low-relief, and gentle terrains of mountain-plains. Northeastern Yunnan is located in the transitional zone between the eastern and western subtropical regions of China, and experiences distinct monsoon climatic patterns characterized by simultaneous occurrence of rainfall and heat. The annual average temperature in northeastern Yunnan ranges from 11.2 ℃ to 21.3 ℃. The coldest month is January, with temperatures ranging from 1 ℃ to 12 ℃, while the warmest month is July, with temperatures ranging from 20 ℃ to 27 ℃. The annual average rainfall ranges between 660 and 1230 mm.

Northeastern Yunnan comprises six types of grasslands: sparse woodland grasslands, mountain shrub grasslands, mountain grasslands, alpine meadows, marsh meadows, and abandoned secondary grasslands (Fan, 2009). Apart from the dry-hot river valley savanna bushveld along the shore of the Jinsha River at low elevations, these grasslands primarily consist of temperate and sub-frigid grasslands situated at elevations above 2000 m on the upper surfaces of mountain landforms, along with grass slopes that have been transformed from forests and shrubs due to human interference at medium to low elevations (Hu et al., 2011; He and Pu, 2014). Northeastern Yunnan has a tradition of livestock farming in mountainous areas. Currently, this region has a total of nearly 167, 000 ha of natural and artificial grasslands, which provide abundant resources for animal husbandry. The rapid economic development in past decades has prominently increased the demand for meat, eggs, and milk, providing new incentives for increased livestock farming in southern grasslands (Fan, 2009).

2.2. Sampling design for vegetation investigation

From July to September of both 2020 and 2021 (peak growing season), we investigated typical temperate montane grassland communities in Ganshan Village of Yongshan County, Dazhai Village of Daguang County, Dashanbao Nature Reserve, and Huize Caohai (Fig. 1). We established sampling plots in grasslands of different management types (protected, grazed, mowed grasslands), along an elevational gradient and at different slope aspects and positions. At each sampling plot, we investigated plant communities in three 1 m × 1 m duplicate quadrats, resulting in a total of 179 surveyed sampling plots. We also collected grassland community data from 23 additional sampling plots from the Peng et al. (2006) and 12 sampling plots from Yang et al. (2022).

For each sampling plot, we interviewed landowners about grassland management types (including protected, grazed, and mowed grasslands), fertilization and artificial reseeding, recorded the geographical coordinates, elevation, slope position, aspect and angle, as well as other relevant habitat factors of the plot, and defined slope position levels (position) based on the terrain, i.e., ridge tops = 5, slope shoulders = 4, middle of the slope = 3, lower slope = 2, foot of the slope or depression areas = 1. We then recorded all plant species names within each quadrat, their abundance, coverage ratio (%), average height (cm), etc., and estimated the total vegetation coverage ratio (%). The species abundance level was determined according to the Drude method: cop. – Copiosae (cop.3 – extremely abundant, cop.2 – very abundant, many individuals of this species, cop.1 – abundant); sp. – Sparsae (plants are found occasionally, scattered, in small numbers), sol. – Solitariae (plants are rare or single); un – unicum. Soil samples were collected from each plot at a depth of 0–20 cm beneath the surface using the five-point sampling method. Following thorough mixing, these samples were brought back to the laboratory and naturally air-dried for a month before experiments measuring soil physical and chemical properties. We also searched literature for data of southern grasslands in the study region, including habitat factors (coordinates, elevation), land use practices, and community information (species abundance, coverage, height). In total (i.e. our field studies and data from previous studies), our study consists of 214 grassland plots, including 75 protected, 100 grazed, and 39 mowed (Table S1).

2.3. Environmental variables 2.3.1. Climate data

To calculate the long-term average values of climate indicators for the entire region, we utilized meteorological data from 184 county-level meteorological stations, covering all of Yunnan Province as well as its neighboring counties in Xizang, Sichuan, Guizhou, and Guangxi provinces, over a 30-year period (1986–2015). Meteorological observation data were obtained from the official website of the China Meteorological Administration (http://data.cma.cn). Asplin software's Thin Plate Spline was utilized in combination with a digital elevation map of 30 m resolution to interpolate 100 m resolution climate factor grid data for all of Yunnan. Subsequently, utilizing the latitude and longitude coordinates of each sampling plot, we extracted climate factor values using ArcGIS, with a specific focus on monthly precipitation and temperature. Additionally, we computed the mean annual precipitation (MAP) and mean annual temperature (MAT) based on extracted data.

2.3.2. Soil data

Field-collected soil samples were ground after air-drying to prepare test soil samples. Prior to measuring mechanical properties and soil pH, soil samples were passed through a 2 mm sieve. Total soil carbon, total nitrogen, and total phosphorus were measured after samples were passed through a 100-mesh sieve. Total soil carbon and total nitrogen content were determined using a total organic carbon analyzer (vario TOC select, Germany) through the dry combustion method. Soil pH values were measured by the Sartorius PB-10 pH meter (soil: water ratio of 1:2.5). The mechanical composition of soil samples was evaluated using a BT-9300H laser particle size analyzer (bettersize, China). Total phosphorus was measured by the alkali fusion – Mo–Sb Anti spectrophotometric method (HJ 632–2011).

2.4. Statistical analyses 2.4.1. Spatial pattern of α diversity

In this study, we used species richness (SR) as a measure of taxonomic α diversity, which represents the number of species in each plot. Relationships between species richness and geographical and environmental factors (latitude, elevation, MAP, MAT) were evaluated using linear and polynomial regressions. Akaike information criterion (AIC) was used to compare the goodness-of-fit of the regression models. To test differences in species richness among functional groups, we applied the Kruskal–Wallis test for multiple comparisons. The species–area curve was implemented utilizing the "vegan" package in R 4.2.1 (Baselga and Orme, 2012).

2.4.2. Constructing phylogenetic tree and assessing the phylogenetic community structure

We used 374 plant species recorded in 214 survey plots to generate a species-level phylogenetic tree (Fig. S1) using the "phylo.maker" function implemented in the R (v.4.2.1) package, "U.PhyloMaker" (Jin and Qian, 2023). The package uses the latest dated phylogeny for plants, GBOTB (Smith and Brown, 2018), as a backbone for vascular plants. This study uses the Net related index (NRI) (Webb et al., 2002) to represent the phylogenetic structure of communities. NRI is standardized by the average phylogenetic distance (MPD) between species through the independent swap model. This calculation was performed 999 times to assess the dispersion level of the average phylogenetic distance among all species in the community compared to the null model, which represents the average state of the community's phylogenetic structure (Webb et al., 2002). The NRI was thus calculated as follows:

(1)

We created the null model communities by shuffling species labels across the entire phylogeny, thus randomizing phylogenetic relationships among species, with 999 iterations. The value of NRI can be positive, negative or no different from zero, which indicates phylogenetic clustering, overdispersion and randomness, respectively. The Wilcoxon signed-rank test was used to test whether the NRIs were significantly different from zero. The Kruskal–Wallis test was used to assess differences in NRI between management types.

Structural equation modeling (SEM) is a widely used technique for evaluating the direct and indirect effects among variables (Shipley, 2000; Grace and Keeley, 2006). Before the analysis, we screened for the multicollinearity among the environmental factors. We calculated Spearman rank correlation between environmental factors and excluded environmental factors with correlation coefficients greater than 0.7 (Dormann et al., 2013). Soil total carbon was excluded from the later analysis. SEM was employed to assess the relative importance of climate, topography, soil, and other factors on NRI, as well as to examine the relationships between these factors. All analyses were performed in R 4.2.1 (R Core Team, 2011). The picante package was utilized for NRI calculation, the piecewiseSEM package was employed for the construction and analysis of SEM, and the ggplot2 package was used for data visualization.

2.4.3. β diversity indices

The Sorensen index is used to quantify and compare the taxonomic and phylogenetic β diversity of communities. As an indicator of species composition changes in communities, the Sorenson diversity index measures two fundamental community processes, namely species addition or reduction (nestedness index) and species replacement (turnover index) (Baselga, 2010). The functions for these factors are as follows:

(2)

(3)

(4)

where a is the number of species shared by two locations, b is the specific number of species at one location, and c is the specific number of species at another location. In calculating phylogenetic β diversity, a shared species is replaced by a shared branch length (Leprieur et al., 2012; Qian et al., 2021). The taxonomic and phylogenetic β diversity values were computed utilizing the "betapart" package in R 4.2.1 (Baselga and Orme, 2012).

2.4.4. Mantel test and partial Mantel test

To decompose the influence of spatial distance and environmental differences on grassland β diversity variation, we employed Mantel tests, which estimated the matrix correlations between species composition differences and both environmental and spatial distances (Legendre and Legendre, 1998). Additionally, to address the non-independence of distance data, we conducted partial Mantel tests to examine the correlations between the two matrix factors while controlling for a third matrix factor (Legendre, 2000). A geographic distance matrix was constructed by calculating the distance between each pair of sampling plots using their coordinates. Similarly, the Euclidean distance between each pair of sampling plots was computed for all environmental variables, or specifically for a particular factor. All analyses were performed using the corresponding software packages in R 4.2.1 (R Core Team, 2011).

3. Results 3.1. Taxonomic α diversity of grassland communities under different management types

The species richness of grassland plants in different functional groups varied with latitude and elevation. Specifically, species richness significantly decreased with increasing latitude (26.23°–28.23° N). Species richness of legumes and forbs also decreased across latitude, although that of graminoids did not; in contrast, species richness of sedges increased with latitude. Grassland species richness best fit a unimodal curve along an elevational pattern (1384–3943 m), with the peak occurring at about 3000 m. Forbs exhibited the same trend. However, species richness of sedges decreased along the same elevational gradient, and that of graminoids remained stable (Fig. 2a, b).

The species richness of grassland plants was negatively correlated with MAP (876–1472 mm), including all functional groups except sedges. Grassland species richness exhibited a unimodal relationship with MAT (1.79–12.78 ℃). The species richness of sedges was positively correlated with MAT, however, species richness of other functional groups showed no significant correlation (Fig. 2c, d).

Species richness was the highest in grazed grasslands, and lowest in mowed grasslands (Fig. 3a). α diversity was highest in forbs. The mean value of forbs' α diversity was significantly higher in grazed grasslands than in both protected and mowed grasslands (Fig. 3b). The second highest α diversity was found in graminoids, with values significantly higher in protected grasslands than in grazed and mowed grasslands. The α diversity was lowest in sedges and legumes, which showed no significant difference among the three types of grasslands.

3.2. Phylogenetic structure and drivers of grassland community assembly

The NRI of protected grasslands was not significantly different from zero, with a random phylogenetic structure, whereas the NRI of grazed and mowed grasslands was significantly greater than zero, possessing a clustered phylogenetic structure (Fig. 4a). The phylogenetic structure of grassland communities was primarily influenced by MAT, MAP, management types, and soil pH, with soil total nitrogen and slope position having no significant effect. Specifically, clustered communities occurred in areas with higher MATs and more intensive management (protected → grazed → mowed); at increased MAPs, communities were random or overdispersed. In karst regions, the phylogenetic structure of grassland communities tended to be clustered, reflecting habitat filtering in poor soil conditions (low soil N content, high soil pH).

3.3. Taxonomic and phylogenetic β diversity of grasslands under different management types

The taxonomic β diversity (Sorensen index) of grassland communities was mainly composed of turnover components, with nestedness playing a minor role. Difference in species composition of the communities significantly reduced from protected to grazed to mowed grassland (from 0.85 to 0.66), mainly due to a significant decrease in turnover components (from 0.80 to 0.58). Nestedness components increased slightly (from 0.04 to 0.08) (Fig. 5a).

Similarly, the phylogenetic β diversity of grassland communities under different management types was consistent with the taxonomic β diversity, mainly composed of species turnover, with a low proportion of nestedness. Difference in phylogenetic β diversity of the communities also decreased (from 0.49 to 0.33) from protected to grazed to mowed grassland, with both turnover (from 0.34 to 0.25) and nestedness components (from 0.15 to 0.08) significantly decreasing (Fig. 5b).

3.4. The drivers of β diversity in grassland communities under different management types

Partial Mantel tests showed that, in protected and grazed grasslands, the taxonomic and phylogenetic β diversity between communities was more correlated with geographical distance than environmental distance, indicating the dominant influence of spatial isolation (Table 1). However, in mowed grassland, β diversity was significantly related to environmental distance, but not with spatial distance. Protected grasslands were mainly affected by climate and topographical distance. Although grazing and mowing grasslands were also affected by topographic distance, climate distance had a weaker effect.

Further correlation analysis was conducted between spatial distance and the turnover component of the taxonomic and phylogenetic β diversity in different management types of grasslands. Regression coefficients showed that turnover gradually decreased from protected, grazed to mowed grasslands (taxonomic: 0.005 vs 0.004 vs 0.002; phylogenetic: 0.003 vs 0.002 vs 0.001) (Fig. 6).

The partial Mantel test results of grassland taxonomic and phylogenetic β diversity under different management types showed (Fig. 7) that the drivers of grassland taxonomic and phylogenetic β diversity were similar under different management types. In protected grasslands, both taxonomic and phylogenetic β diversity were driven by slope position and slope. Taxonomic β diversity was also significantly affected by MAP, whereas phylogenetic β diversity was affected by MAT and total phosphorus in soil. The β diversity in grazed grasslands was mainly driven by difference in slope position, while β diversity of mowed grasslands was mainly determined by slope position and angle, as well as soil total nitrogen.

4. Discussion 4.1. Management type affects α diversity and phylogenetic structure of grasslands

Spatial patterns of plant species diversity reflect the community responses to environmental changes (Bergamini et al., 2001). This study found that species diversity of temperate moist grasslands in northeastern Yunnan exhibited a unimodal distribution as elevation increased. At lower elevations, where environmental conditions were comparatively favorable, some competitively dominant species gained an advantage in communities (Wu et al., 2019), such as in wetland meadows. At the highest end of the elevational gradient, cold stress limits species richness and favors more cold-tolerant species (Wang et al., 2004). In contrast, mid-elevation habitats represent the major distribution area of southern grasslands. At this elevation, the favorable combination of water and heat conditions results in a higher species richness (Sa et al., 2012).

Human land use changes are a significant driver of the changes in biodiversity (Pereira et al., 2012) and ecosystem functions (Allan et al., 2015; Chillo et al., 2017). Compared to protected grasslands, moderate grazing in grassland communities appears to have reduced the competitiveness of dominant species, providing opportunities for new species to colonize, and thus promoting species coexistence and biodiversity (Merdas et al., 2021). Previous studies have indicated that ecological differentiation is stronger among forbs, which can maintain high levels of species richness, than among graminoid species, for which ecological niches are closer (Brambila et al., 2020). The impact of grazing on species diversity has also been found to vary due to productivity (Lezama et al., 2014). We found that moderate grazing promotes species diversity, which is consistent with the "intermediate disturbance hypothesis" (Connell, 1978). However, in dry years or in low productivity grasslands, the resilience of plant communities to grazing disturbances has been found to be greatly reduced, and the deteriorating habitat after local species loss has been shown to limit species re-colonization, resulting in community degradation and species richness reduction (Osem et al., 2002; Frank, 2005; Li et al., 2021a, Li et al., 2021b).

As the most intensive type of grasslands management, mowing aims to maintain the hay output of specific species, often by increasing grassland productivity through fertilization and reseeding of dominant species after mowing (Harpole et al., 2016), and by giving rapid-growing resource-acquiring species a competitive advantage over slower-growing, resource-conservative species (DeMalach et al., 2017; Li et al., 2022) through light limitation (Hautier et al., 2009), thereby reducing grassland biodiversity (Socher et al., 2012; Leps, 2014). Our interviews with local residents found that artificial grasslands in the mountainous areas of northeastern Yunnan were often mowed 2–3 times a year, supplemented by fertilization and reseeding of dominant grass species. This intensive human management, which maintained the primary dominance and biomass of desirable forage species such as Dactylis glomerata, Lolium perenne, and Trifolium repens while inhibiting the growth of other species, resulted in a reduction in the species diversity of mowed grasslands.

Ecological niche conservatism indicates that closely related species have similar responses to disturbances (Li et al., 2014). Therefore, the clustered phylogenetic structure of communities after disturbance implies an increase in interspecific competition (Chollet et al., 2018). Compared with protected grasslands, we found grazing and mowing increased the survival of some disturbance-tolerant species, which were often more closely related, leading to a clustered phylogenetic structure. Previous studies have found that the phylogenetic structures of northern and alpine grasslands also became clustered under grazing pressure (Li et al., 2019; Zhu et al., 2020; Sun et al., 2021). In contrast, functional types and species richness are higher in grassland communities located in habitats with lower temperatures, higher levels of precipitation, rich soil conditions, and less selective grazing (i.e., livestock grazing) and mowing.

4.2. Grazing and mowing cause taxonomic and phylogenetic homogenization of grassland communities

Beta diversity can be decomposed into turnover and nestedness components. The nestedness of species between sites indicates the extent to which species in sites with lower species richness form a subset of species in sites with higher species richness (Baselga, 2010; Mori et al., 2018). Spatial turnover occurs when some species are replaced by other species due to environmental filtering, spatial or historical limitations (Qian et al., 2005). Phylogenetics can reflect the genetic relationships and evolutionary history among species. The turnover component reflects the replacement of the evolutionary spectrum, while the nestedness component reflects the loss or increase of spectra at different sites (Leprieur et al., 2012; Xu et al., 2023). In the southern grasslands studied, both taxonomic and phylogenetic β diversity of communities under different management types were dominated by turnover components. In grazed and mowed grasslands, both the taxonomic and phylogenetic β diversity of grasslands decreased, and the phylogenetic community structure became more similar, with mowing having a greater impact. The taxonomic dissimilarity of grasslands under different management types was generally greater than their phylogenetic dissimilarity, indicating that the species have a close phylogenetic relationship. The increase in taxonomic α diversity and the decrease in taxonomic and phylogenetic β diversity in grazed grasslands suggests the impact of grazing, which suppresses the dominance of the graminoids, increases invasive forbs and grazing-resistant grass species, thus reducing spatial variation in species diversity and composition (Lezama et al., 2014). Mowing decreased the taxonomic α diversity and also reduced taxonomic and phylogenetic β diversity. This suggests that intense management such as fertilization and reseeding cause the extinction of rare species, resulting in the homogenization of grassland species composition under mowing management, with retained species having closer phylogenetic relationships (Yang et al., 2019).

4.3. Management type alters the relative contributions of geographical and environmental factors on grassland β diversity

Environmental and spatial distances are the major drivers of β diversity (D'Amen et al., 2018). These drivers of diversity stem from the underlying mechanisms described by niche and neutral theories. Niche theory emphasizes the importance of contemporary environmental conditions, including abiotic factors (e.g., climate and soil properties) and biotic factors (Chase and Leibold, 2002; Tang et al., 2012; Ulrich et al., 2014). Niche theory also suggests that diversity patterns are mainly determined by environmental filters (Chesson, 2000; Chase and Leibold, 2002), whereas neutral theory highlights the role of spatial processes in shaping species composition (e.g., drift and dispersal limitation; Hubbell, 2001). We found that the relative contributions of environmental difference and spatial distance to grassland β diversity were influenced by different management types. Our results indicate that both environmental and spatial distances significantly impact the taxonomic and phylogenetic β diversity of protected grasslands, which indicates that environmental filtering and dispersal limitation are not mutually exclusive but rather jointly regulate β diversity in protected and grazed grasslands, with dispersal limitation playing a stronger role. This finding may be attributed to the fact that these grasslands are generally located on separate mountain tops or dissected plateau at elevations of more than 2000 m, where distribution is scattered (Ren and Zhang, 2002). Grassland species are mainly apomictic organisms (e.g., tillering) with limited propagation distance (Ott and Hartnett, 2011). Even for species reliant on seed dispersal, their range is minimal (Baer and Maron, 2018; Pinto et al., 2014). Research has shown that the seed dispersal distance of grassland species is generally 20–50 m (Donath et al., 2003), not exceeding 1 km (Fahrig, 2013). Grazing and mowing, which act as habitat filters (Laliberte et al., 2013; Salgado-Luarte et al., 2019), gradually diminish the impact of spatial distance. For example, livestock grazing has been shown to accelerate internal seed spread through soil, fur, and feces (Kapas et al., 2020; Michaels et al., 2021). In mowed grasslands subject to intensive artificial management, including artificial seeding and fertilization, geographic distance has been shown to have little impact, whereas environmental distance has a greater impact. These findings imply that niche processes predominantly shape β diversity in mowed grasslands (Wang et al., 2021).

Environmental factors greatly affect grasslands β diversity. Specifically, we found that β diversity of protected grasslands was influenced by MAP, MAT, slope position, and soil total phosphorus. These environmental factors vary in communities due to spatial distribution, e.g., variations in water and temperature result from the scattered distribution of the grasslands among different locations. Topography also redistributes environmental resources such as water and soil nutrients, further increasing the environmental heterogeneity of grasslands. For example, in flat areas, southern grasslands form wetland swamp communities dominated by aquatic plants like Eleocharis ovata and Persicaria muricata; on steep slopes, grasslands were characterized by temperate meadows dominated by plants such as Arundinella hookeri and Festuca ovina, resulting in a mosaic pattern of vegetation.

Livestock grazing reduces environmental heterogeneity. Previous studies have found that livestock grazing reduced potential spatial gradients of physical drivers like soil quality and nutrients (Golodets et al., 2011). This occurs through treading and excretion, foraging on non-dominant rare plant species, and facilitating colonization by widespread species. We found that topographical distance (especially slope position distances) played a key role in the β diversity of grasslands under different management types, grassland communities in different topographic conditions developed distinct community structures, such as the community of Potentilla lineata + Festuca ovina on high slopes and the community of Jacobaea analoga + Rumex nepalensis on low slopes. Mowing grasslands also influences environmental distanced such as total soil nitrogen. Nitrogen addition decreases the root length and biomass of forbs, weakening their water absorption capacity and ensuring the dominance of graminoids (Bai et al., 2015; Liu et al., 2021). However, nitrogen addition could also result in soil acidification, leading to a decline in community species diversity (Stevens et al., 2010; Lu et al., 2023). Previous studies have shown that the intensity of mowing and the amount of fertilizer applied affected soil nutrients (Socher et al., 2012; Chiste et al., 2018), altering soil heterogeneity. This may explain decreased taxonomic and phylogenetic β diversity of mowed grasslands. Compared with the uneven spatial and temporal patterns of grazing, mowing has a uniform effect even on a large scale, leading to a stronger homogenization of vegetation through the continued removal of biomass within a short time (Gossner et al., 2016; Li et al., 2021a, Li et al., 2021b).

In conclusion, we found that, compared to protected grasslands, grazing increased whereas mowing decreased taxonomic α diversity of southern grassland communities. Mowed grasslands were more phylogenetically clustered. Both the taxonomic and phylogenetic β diversity of grasslands experienced a significant decrease in grazed and mowed areas, with mowing exerting a stronger impact. The taxonomy and phylogeny β diversity in mowed and grazed grasslands were dominated by species turnover. Both taxonomic and phylogenetic β diversity of protected and grazed grasslands were influenced by environmental filtering and dispersal limitation, with a more pronounced effect of dispersal limitation. However, mowing and related treatment exerted a more intense filtering effect on the grassland community structure, as reflected in the changes in community composition. Our study is the first to show the impact of management types on community composition and structure at a regional scale in grasslands of northeastern Yunnan. Our findings also indicate that conserving biodiversity in the southern grasslands of China may require a network of reserves.

Acknowledgment

We appreciate the Dashanbao National Natural reserve and the Central Yunnan Plateau Observatory of the Ministry of Natural Resources on biodiversity & Critical Zones for supporting our field investigations. We also thank Mr. Liu, Mr. Ding from Maolin and Gaoqiao village for their assistance during our research. The study is sponsored by the Strategic pilot project (type A) of the Chinese Academy of Sciences (Project No. XDA26020203-K207002210014), and the Second Tibetan Plateau Scientific Expedition and Research Program (STEP) (2019QZKK0402).

CRediT authorship contribution statement

Jianghua Duan: Writing – original draft, Visualization, Software, Investigation, Data curation. Liu Yang: Software, Investigation. Ting Tang: Software, Investigation, Data curation. Jiesheng Rao: Validation, Investigation. Wencong Liu: Software, Investigation. Xi Chen: Validation, Investigation. Rong Li: Supervision, Methodology. Zehao Shen: Writing – review & editing, Supervision, Methodology, Investigation, Data curation.

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.2024.04.005.