1. Introduction

Protecting fragmented habitats is crucial for conservation of regional biodiversity (Fahrig, 2017). Based on inferences from the Island Biogeography Theory, a single large patch in fragmented habitats is often assumed to have higher conservation value than several small patches (Diamond, 1975). However, several studies have revealed that small patches significantly contribute to preserving regional biodiversity and supporting ecosystem functions (Wintle et al., 2019; Gestich et al., 2022). Although recent research indicates that 72% of case studies suggest several small patches contain more species (Fahrig, 2020), this long-standing debate around conserving “single large or several small” (SLOSS) habitat patches continues.

Previous research on the SLOSS debate primarily focused on taxonomic diversity (Fahrig, 2020). However, phylogenetic and functional diversity have become recognized as important indicators of biodiversity and are now widely applied in biodiversity conservation (Dias et al., 2020). Phylogenetic diversity quantifies the history and differences of evolution among species within communities (Faith, 1992; De Pauw et al., 2021), whereas functional diversity measures the variation of species functional traits in communities, bridging species composition and ecosystem functions (Doxa et al., 2020). Importantly, phylogenetic and functional measures of diversity may generate contrasting recommendations for conservation strategies. For example, previous work on bird diversity in urban parks found that when only species richness (i.e., taxonomic diversity) was considered, the best conservation strategy was to protect several small habitat patches; however, when phylogenetic or functional diversity was considered, the best strategy (several small vs. single large) varied across seasons and cities (La Sorte et al., 2023). One proposed approach to resolving the SLOSS debate in biodiversity conservation is to integrate the taxonomic, phylogenetic, and functional dimensions of diversity (Xie et al., 2025).

Previous studies have integrated taxonomic, phylogenetic, and functional diversity into a framework called the diversity-area relationship (DAR) (Mazel et al., 2014; Matthews et al., 2023). The “species-area relationship” (SAR) describes a common pattern of species numbers increasing with sampled area and is central to ecological and biogeographical theories like island biogeography (MacArthur and Wilson, 1967; Rosenzweig, 1995). The SAR, which assumes that species within a community are evolutionarily independent and ecologically equivalent (Dias et al., 2020), has often been used to determine the impact of habitat fragmentation on biodiversity and guide the establishment of nature reserves (Saura, 2021; Liu et al., 2022; Song et al., 2024). Relationships between phylogenetic diversity and area (PDAR) as well as functional diversity and area (FDAR) have been proposed as extensions of the SAR to describe how the phylogenetic composition of species assemblages and trait variation among species change with area (Helmus and Ives, 2012; Smith et al., 2013). Each dimension of diversity provides valuable information and may contribute important properties to protect biological communities. Thus, this integrated framework promises to provide comprehensive representation of biodiversity across a region, providing more precise recommendations for conservation strategies (Leclerc et al., 2022).

The agro-pastoral ecotone in northern China is a classic example of a grassland landscape fragmented by human activity, shaped by centuries of agricultural expansion. Since the 1960s, land policy reforms have driven rapid farmland expansion in regions like the Tabu River Basin in Inner Mongolia, converting continuous natural grasslands into fragments (Yan et al., 2023, 2025). In this study, we determined the best conservation strategies (i.e., single large or several small) for grasslands of the Tabu River Basin. For this purpose, we first analyzed SAR, PDAR, and FDAR in 78 remnant grassland patches, then compared appropriate conservation strategies as indicated by these distinct diversity measures.

2. Materials and methods 2.1. Study area

The study was conducted in the Tabu River Basin, located in the agro-pastoral ecotone of Inner Mongolia, China (Fig. 1). This region has a typical temperate continental semi-arid climate, with annual precipitation ranging from 225 to 322 mm, primarily concentrated from June to September. The annual average temperature ranges from 1.5 to 5.0 ℃, and the elevation varies between 1360 and 1700 m. The dominant vegetation consists of Stipa krylovii and Stipa breviflora, and the soil type is light chestnut soil (Yan et al., 2024). Agricultural expansion has led to the fragmentation of this region, where grasslands are the main natural habitat type and farmlands are the dominant matrix type. The primary crops are economic crops such as potatoes and corn. Owing to extensive agricultural development during the 1960s, many natural grasslands were converted into farmland, dividing the previously continuous natural grasslands into isolated remnant patches. Therefore, the grassland patches in the region have similar fragmentation ages (Chen et al., 2019).

2.2. Data collection 2.2.1. Patches selected

We downloaded 2-m resolution Gaofen-1 satellite data from the China Satellite Data Center (http://rs.ceode.ac.cn/) and used the visual interpretation method in ArcGIS v.10.3 to extract 78 remnant grassland patches surrounded by farmland (Fig. 1). All chosen patches featured undulating terrain and were located at higher elevations, making them unsuitable for agricultural cultivation, suggesting that these patches had likely never been cultivated. We also checked these patches with historical land cover data and found that the selected patches had remained grasslands from 1990 to 2019 (Yang and Huang, 2021). Patch area was calculated in ArcGIS v.10.3. The 78 remnant grassland patches varied in size from 0.32 to 1451.59 ha.

2.2.2. Vegetation survey

Field surveys were conducted from late July to mid-August 2019. A total of 78 remnant grassland patches were investigated using the transect method. Patch sizes varied, thus, the length and number of transects varied accordingly. Transects were established along the longest axis of each patch. For smaller patches, one transect was established; for larger patches, 2 to 3 transects were established. Along each transect, a 1 m × 1 m quadrat was placed every 10 m. In the final count, each patch had between 10 and 48 quadrats.

During the vegetation sampling process, the longitude and latitude of each patch were recorded using GPS. Then, plant species from each quadrat were documented and sampled. In addition to the transect survey, we conducted a comprehensive inventory of plant species within each patch by recording those species not occurring in the transects (Takkis et al., 2018). All species recorded in our study were vascular plants. The names of species were verified and corrected on The Plant List website (http://theplantlist.org/).

2.2.3. Functional trait data collection

We selected seven plant functional traits: plant height, leaf area, leaf carbon content (leaf C), leaf nitrogen content (leaf N), leaf phosphorus content (leaf P), specific leaf area, and leaf dry matter content (leaf dry matter). Plant height and leaf area are fundamental metrics for assessing the ability of plants for light capture (Wright et al., 2004). Leaf C, leaf N, and leaf P are used as indicators to assess the nutritional status of plants (Jin et al., 2024). Specific leaf area and leaf dry matter reflect the ability of plants to utilize resources (Udayakumar and Sekar, 2021). In each quadrat, we measured these seven plant traits in 15 healthy, randomly-selected individuals of each species. Rare species had limited samples, so their traits were difficult to measure. Therefore, we supplemented corresponding trait data using the TRY database (Kattge et al., 2020).

Plant functional traits were measured based on Pérez-Harguindeguy et al. (2016). Plant height (cm) was measured as the natural vertical height of the plant. Leaf area (cm²) was measured using Li-3000 leaf area meter. Leaf C (%) and leaf N (%) were measured using a C/N elemental analyzer. Leaf P (mg/g) was measured using the phosphomolybdenum blue colorimetric method. Specific leaf area (cm²/g) was calculated by dividing the leaf area by its dry weight. Fresh leaves were oven-dried at 60 ℃ until reaching a constant weight to determine their dry weight. Leaf dry matter (%) was calculated by dividing the leaf dry weight by the leaf saturated fresh weight. In a dark setting, fresh leaves were immersed in water for 24 h. Prior to weighing the leaves to determine their saturated fresh weight, the surface water on leaves was gently dried with a cloth.

2.3. Data analysis 2.3.1. Calculation of taxonomic, phylogenetic, and functional diversity

We used species richness (SR) as the taxonomic diversity index. The phylogenetic and functional diversity indexes were both calculated based on summing branch lengths (PD, FD), which are equivalent to species richness for taxonomic diversity (Dias et al., 2020; Matthews et al., 2023). Thus, all indexes we used were dimensionally equivalent and comparable.

SR was calculated as the total number of species in each patch. PD was determined by first constructing the phylogenetic relationships of vascular plants surveyed in the field using “U.PhyloMaker” package in R v.4.4.1 (Jin and Qian, 2023), then by summing branch lengths as in Faith (1992). FD was determined by first constructing a species-trait matrix. We then used the “dist” and “hclust” functions in R v.4.4.1 to calculate pair-wise species distance matrices based on the Euclidean Distance. Functional tree was produced using the “Unweighted Pair Group Method with Arithmetic Mean”. Finally, FD was calculated based on branch lengths of the functional tree (Petchey and Gaston, 2006; Mammola et al., 2021).

Although the diversity metrics we used share an identical mathematical framework that makes them directly comparable, PD and FD are highly correlated with SR. Therefore, to partial out the influence of SR, null models were used to generate standardized effect sizes of PD and FD (sesPD, sesFD) (Dias et al., 2020). We created a variant of the “taxa. labels” null model (999 iterations) to calculate the sesPD and sesFD by using the “ses.pd” function of the “picante” package in R v.4.4.1. This null model shuffled taxa labels across the tips of the tree (across all taxa included in the tree) (Matthews et al., 2023). The formula of ses is as follows:

ses=(pd.obs−pd.rand.mean)/pd.rand.sd

where ses is standardized effect size, pd.obs is observed PD/FD in community, pd.rand.mean is mean PD/FD in null communities, pd.rand.sd is the standard deviation of PD/FD in null communities (Webb et al., 2008).

A larger ses value indicates greater dispersion, whereas a smaller ses value indicates higher clustering (Feng et al., 2024). These indexes may be related to ecological processes that influence the distribution of phylogenetic and functional diversity in landscapes (Weiher and Keddy, 1995; Mayfield and Levine, 2010). Consequently, they provide valuable insights into how different aspects of diversity shift with area, regardless of species richness. Additionally, we examined the relationships between SR, sesPD, and sesFD using Pearson correlation analysis by “Hmisc” package in R v.4.4.1.

2.3.2. Fitting of diversity-area relationship

The “sars” package was used to fit the diversity-area relationship (DAR) in R v.4.4.1. This R package, which provides 20 linear and non-linear regression models, was developed for fitting, evaluating, and comparing species-area relationship models (Matthews et al., 2019) (Table S2). We selected the best-fit model based on the lowest AICc, and presented R² as a descriptive goodness-of-fit measure. Because many models do not support values less than or equal to 0, we transformed the sesPD and sesFD values by subtracting the overall lowest value from each one and then adding 1 (Dias et al., 2020; Yuan et al., 2024). This transformation of sesPD/sesFD did not affect the curve shapes or relative comparisons (Fig. S1; Tables S6 and S7).

2.3.3. SLOSS inferences

Based on SR, sesPD, and sesFD, we used an improved version of the species accumulation curves method by Quinn and Harrison (1988) for SLOSS inference. For the specific methods, we referred to La Sorte et al. (2023): firstly, we mapped each point from the large-to-small accumulation curve to its corresponding position on the small-to-large accumulation curve, connecting them with vertical lines. Next, we calculated the total length of the vertical lines where the small-to-large accumulation curve was positioned above the large-to-small accumulation curve. This sum was then divided by the combined length of all vertical lines, regardless of whether they were above or below the small-to-large accumulation curve (as shown in Fig. 2). This ratio is used to estimate the area proportion between the two curves, where a ratio close to 1.0 indicates “SS > SL”, a ratio near 0.50 indicates “SL = SS”, and a ratio close to 0.0 indicates “SL > SS” (La Sorte et al., 2023). The ratio was calculated in R v.4.4.1. Prior to calculating the ratio, we used the “dplyr” package for data processing and “ggplot2” package for plotting the accumulation curves in R v.4.4.1. By not calculating from the origin, this method improves the integral ratio approach introduced by Quinn and Harrison (1988) and avoids the bias towards “SS > SL” that can occur (Fahrig, 2020).

3. Results 3.1. Relationships between taxonomic, phylogenetic, and functional diversity

Our survey of 78 patches in the Tabu River Basin identified 152 species of plants in 39 families and 96 genera (Table S1). The study area is predominantly characterized by species from the families Asteraceae, Fabaceae, and Poaceae, as well as the genera Artemisia, Astragalus, and Potentilla. SR was positively correlated with sesPD (r = 0.30, p < 0.01; Fig. 3A), but negatively correlated with sesFD (r = −0.27, p < 0.05; Fig. 3B). sesPD was negatively correlated with sesFD (r = −0.28, p < 0.05; Fig. 3C).

3.2. Diversity-area relationship of three dimensions

Based on the fitting results of “sars”, we sorted the AICc values of all models of SAR, PDAR, and FDAR in ascending order (Table S3-S5). The model with the smallest AICc value was selected as the best-fit model. The best-fit model for SAR was the Persistence Function 2 (AICc = 487.33, R² = 0.62; Fig. 4A and Table S3); the best-fit model for PDAR was the Rational (AICc = 147.52, R² = 0.15; Fig. 4B and Table S4); and the best-fit model for FDAR was the Rational (AICc = 237.20, R² = 0.12; Fig. 4C and Table S5).

The best-fit model indicated that SR generally increased with area (Fig. 5A). sesPD also increased with area (Fig. 5B), which indicates that phylogeny tends to disperse as area increases. sesFD, however, decreased with area (Fig. 5C), indicating that functional traits tend to cluster as area increases.

3.3. SLOSS inferences of diversity in three dimensions

Taxonomic diversity indexes indicated that the best conservation strategy for the Tabu River Basin is several small patches (SS > SL) (Fig. 6A). In contrast, phylogenetic and functional diversity indexes indicated that the best conservation strategy for the study area is a single large patch (SL > SS) (Fig. 6B and C).

4. Discussion 4.1. Taxonomic, phylogenetic, and functional diversity had different relationships with area

In this study, we asked whether Tabu River Basin grasslands were best conserved by protecting a single large patch or several small patches of habitat. To answer this question, we first determined how taxonomic diversity, phylogenetic diversity, and functional diversity changed with the habitat patch area. We found that patch area significantly affected species richness (SR), standardized effect sizes of phylogenetic diversity (sesPD), and standardized effect sizes of functional diversity (sesFD), but in distinct ways.

In the Tabu River Basin, SR generally increased with patch area. This finding is consistent with the Island Biogeography Theory (MacArthur and Wilson, 1967), which posits that larger areas provide more resources and habitats, reduce the risk of extinction, and attract more species to colonize. sesPD also increased with patch area, suggesting that large patches not only host more species but also favor the coexistence of phylogenetically distant lineages. This may be because greater environmental heterogeneity in large patches provides opportunities for the colonization of species with different evolutionary strategies (Miao et al., 2022; Qian et al., 2024). Furthermore, large patches are more likely to preserve species with long evolutionary histories, thereby enhancing phylogenetic dispersion. sesFD decreased with area, which indicates that functional traits tend to cluster as area increases. These findings show that the expansion of patch area increases taxonomic and phylogenetic diversity without a proportional increase in functional divergence.

One plausible explanation for our findings is that increased numbers of plant within the patch area intensifies competition among species (Tian et al., 2024). During this competitive process, plants with similar resource acquisition strategies and utilization efficiencies are more likely to coexist, leading to the clustering of functional traits (Mammola et al., 2021). The clustering of functional traits often indicates that species communities share core adaptive strategies, such as drought resistance, cold resistance, or efficient resource utilization (Cornwell and Ackerly, 2009). Thus, plants in the fragmented grassland may select the core functional traits to adapt to the environment of large patches rather than promote the diversification of functional traits. These community assembly processes jointly explain the significant negative correlations of sesFD with SR and sesPD. Specifically, because SR and sesPD both increased with patch area, whereas sesFD decreased, their contrasting responses to the same driver (patch area) ultimately led to the observed negative correlations.

Our study also suggests that the diversity-area relationship in the Tabu River Basin, as determined by taxonomic, phylogenetic, and functional diversity metrics, is subject to a small-island effect. This is consistent with a previous study that found remnant grassland patches in the Tabu River Basin were subject to a small-island effect interval between 2.0 and 12.0 ha (Yan et al., 2023). Our findings also suggest the need to consider the critical threshold for patch area (i.e., the influence range of the small-island effect) in ecological protection planning. For instance, we found that SR initially decreased but then increased below around 10 ha, then rose rapidly above it. sesPD showed a slow increase below around 10 ha but a rapid rise above it. sesFD decreased sharply below around 10 ha, and as the area exceeded around 10 ha, it slowed down and stabilized. These findings lead us to recommend that the area of nature reserves should exceed the range susceptible to the small-island effect, as this may effectively promote the recovery of diversity and maintain ecosystem stability and function.

4.2. Taxonomic, phylogenetic, and functional diversity measures recommend distinct conservation strategies

In recent years, most research regarding the SLOSS debate has supported the conclusion that it is more important to protect several small patches than a single large patch (SS > SL) (Fahrig, 2020). However, our study found that the distinct measures of diversity (i.e., taxonomic, phylogenetic, or functional) produce conflicting recommendations for conservation strategies.

Taxonomic diversity indicated that the best strategy for conserving the Tabu River Basin grasslands is to protect several small patches (SS). Many studies showed that this conservation strategy had been observed not only in grasslands but also in forests and other ecosystems, indicating that habitat fragmentation may not reduce taxonomic diversity (Godefroid and Koedam, 2003; MacDonald et al., 2018; Saravia et al., 2025). The “SS” conservation strategy of taxonomic diversity may be because the habitat heterogeneity represented by several small patches is higher than that of a single large patch, thus providing unique habitats for more species (Seibold et al., 2017; Fahrig, 2020). Specifically, several small patches are distributed dispersedly and may cover more environmental gradients, thereby further increasing species richness. In particular, the study area in our research is located in the agro-pastoral ecotone, which itself is an important component of the spatially heterogeneous landscape (Vu Ho et al., 2023).

Phylogenetic and functional diversity indicated that the best strategy for conserving grasslands in the Tabu River Basin is to protect a single large patch (SL). Although several small patches support higher taxonomic diversity, these species may belong to relatively closely related lineages and exhibit redundancy in functional traits. Consequently, the overall increases in phylogenetic diversity and functional diversity are limited. Conserving a single large patch may provide a suitable environment for species with different evolutionary lineages and traits. Large habitat patches are more stable and possess greater resistance to disturbances (Herrera et al., 2017). Moreover, a single large patch provides superior connectivity compared to several small patches, promoting species migration and gene flow (Klinga et al., 2019; Zhang et al., 2025). Together, these factors will better maintain phylogenetic and functional diversity. It is worth noting that in studies of other taxa (e.g., birds) phylogenetic and functional diversity have been shown to be better conserved by several small patches (Xie et al., 2025). Therefore, conservation strategies for phylogenetic and functional diversity may vary across different taxa.

These findings remind us that biodiversity conservation is not only a matter of area size but also a strategic choice of how to balance the protection of different biodiversity dimensions. Our results are consistent with the conclusions drawn from many other SLOSS studies, indicating that taxonomic diversity is best conserved by the “SS” strategy rather than the “SL” strategy proposed by Diamond (1975). However, even so, we believe that in the ecological context of the agro-pastoral ecotone in northern China, the “SL” strategy may remain the most cost-effective biodiversity protection plan, as conserving a single large patch in the Tabu River Basin would protect both phylogenetic and functional diversity of grasslands. Therefore, we recommend adopting the “SL” strategy for conservation of agro-pastoral ecotones.

5. Conclusions

Our study of remnant grassland patches in the Tabu River Basin of the Inner Mongolia agro-pastoral ecotone indicates that taxonomic, phylogenetic, and functional measures of diversity have distinct diversity-area relationships and require distinct conservation strategies (i.e., “SL” vs. “SS”). We found that species richness and standardized effect size of phylogenetic diversity increase with area, whereas standardized effect size of functional diversity decreased. In addition, measures of taxonomic diversity indicate that the best conservation strategy for Tabu River Basin grasslands is to protect several small patches, whereas phylogenetic and functional diversity indicate that the best strategy is to conserve a single large patch. Thus, our study suggests that biodiversity conservation in fragmented landscapes requires comprehensive consideration of multiple diversity dimensions. For the protection of taxonomic diversity, we recommend protecting several small patches, whereas for the protection of phylogenetic and functional diversity, larger continuous patches should be prioritized. These findings are crucial for formulating effective protection strategies for biodiversity, especially in the establishment of nature reserves.

Acknowledgements

This study was supported by the Natural Science Foundation of Inner Mongolia Key Project (2023ZD24), the Erdos City Major Science and Technology Special Project (ZD20232305), the Inner Mongolia Autonomous Region Science and Technology Plan Project (2025KYPT0012), the Inner Mongolia Autonomous Region Science and Technology Plan Project (2022ZD007), the Inner Mongolia Autonomous Region Education Department Project (NMGIRT2409), the Inner Mongolia First-Class Disciplines Scientific Research Special Project (YLXKZX-ND-047). No permits were required for fieldwork in this research.

CRediT authorship contribution statement

Jia-Wei Yu: Methodology, Validation, Formal analysis, Investigation, Data curation, Writing-original draft, Writing-review & editing, Visualization. Yong-Zhi Yan: Methodology, Validation, Formal analysis, Investigation, Data curation, Writing-original draft, Writing-review & editing, Visualization. Qing Zhang: Conceptualization, Methodology, Investigation, Writing-review & editing, Supervision, Project administration, Funding acquisition.

Data availability statement

Upon reasonable request, the data are available from the corresponding author.

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.2025.10.006.