2.3 Measurement of species richness, species composition and community stability
We calculated species number in a quadrat as species richness. To distinguish differences in species composition among different fencing duration or shrub cover treatments, non-metric multidimensional scaling (NMDS) based on Bray-Curtis distance was run. We used the envfit function from vegan to test the community composition difference between treamtments of different fencing duration or shrub cover, with treatment type as a categorical factor variable. Significance of these differences was tested using Analysis of Similarity (Anosim) by calculating 999 random permutations. Theses anlyses were performed in the ‘vegan’ package in R (R Core Development Team).
M-Gordon stability index, a popular international method, is used for determining community stability (Hou et al., 2019). It uses the number and frequency of all species in a plant community to establish a mathematical model, and calculates the reference point of the community. Then, the stability of the plant community was calculated according to the reciprocal of the distance between the reference point and the stable point. More details for calculation methods of M-Gordon stability index see Supplementary material. The closer the reference point to the stable point, the higher the stability of plant community (Hou et al., 2019).
2.4 Measurement of phylogenetic structure and trait phylogenetic conservatism
Before quantifying phylogenetic structure and phylogenetic conservatism, we constructed a phylogenetic tree for our study species. All the 19 species investigated in this study were identified based on Angiosperm Phylogeny Group III system (The Angiosperm Phylogeny Group, 2009) using the Plant List database (http://www.theplantlist.org/). And then, the completed phylogenetic tree with branch lengths (Figure. S1) was directly assembled within the online tool Phylomatic (http:// phylodiversity.net/phylomatic/) based on the ultrametric phylogeny of Zanne et al. (2014). Next, a phylogenetic tree was randomly solved by ‘multi2di’ function, and then we ultrametrized the tree using ‘compute.brlen’ function from ‘ape’ package in R (R Core Development Team).
To characterize community phylogenetic diversity and structure, we calculated Faith’s phylogenetic diversity (PD) (Faith 1992), mean nearest taxon distance (MNTD) and mean pairwise distance (MPD) (Webb, 2000). PD is defined as the total branch length among all taxa in a comnunity. MPD generally reflects phylogenetic structure across the entire tree, whereas MNTD is sensitive to patterns among closely related species towards the tips of the phylogeny (Blaimer et al., 2015). In additon, we calucaluted the standardized effect sizes (SES, differences of observed values versus null communities) of MPD and MNTD (SESMPD and SESMNTD) by weighting species abundance and their associated quantiles (p value) to assess statistical significance.
To further strengthen the conclusion, we analyzed the SES with two null models: 1) randomizing the community data matrix (‘independent swap’) and 2) shuffling taxon labels randomly across the phylogeny (‘taxa labels’), this algorithm randomizes the relatedness of species to one another, but it maintains species occurrence frequency and abundance. SESMPD and SESMNTD are equivalent to net relatedness index (NRI) and nearest taxon index (NTI) multiplied by – 1. Phylogenetic clustering is indicated by negative SES values and low quantiles (p <0.05), corresponding to -1.96, and overdispersion by positive SES values and high quantiles (p >0.95), corresponding to 1.96. Non-significant positive or negative values indicate observed values close to the median of the random dispersion.
To quantify trait phylogenetic conservatism, we employed a widely used Blomberg’s K statistic with 999 iterations (Blomberg et al., 2003). Blomberg’s K statistic is assessed by using the Brownian motion (BM) of trait evolution, and is considered as one of the most powerful approahces (Pavoine and Ricotta, 2013). K-value close to 1, greater than 1 and less than 1 indicate that the trait distribution perfectly matches BM expectation of trait evolution on the phylogeny, is more conserved and is less conserved than BM expecatation, respectively (Kraft et al., 2007). K-values between 0.4 and 0.7 are hereby in the following considered to represent intermediate, whereas values below 0.4 express low levels of conservatism. We assessed the statistical significance of K by comparing observed variances of phylogenetic independent contrasts to a null model generated by shuffling tips randomly across the phylogeny. Analyses of phylogenetic structure and trait phylogenetic conservatism were performed in the ‘picante’ package in R (R Core Development Team).