Ecological uniqueness of species assemblages and their determinants in forest communities

Beta diversity can be partitioned into the contributions of individual sampling units to overall beta diversity, which are comparative indicators of the ecological uniqueness of species assemblages in the sampling units. Yet, what determines ecological uniqueness has rarely been examined. Here, we investigated the determinants of ecological uniqueness in species assemblages in forest communities.


| INTRODUC TI ON
Understanding the mechanisms that determine the spatial distribution of species and species diversity is a central theme in ecology (Chave, 2004;Chesson, 2000;Hutchinson, 1961;Ricklefs, 1990).
For community ecologists, the interest to investigate beta diversity is that it can provide fundamental insights into the processes that determine the spatial patterns of species assemblages (Legendre & De Cáceres, 2013;Myers et al., 2013). Legendre and De Cáceres (2013) suggested that the total beta diversity can be partitioned into local contributions to beta diversity (LCBD), which are comparative indicators of the ecological uniqueness of species assemblages in sampling units (hereafter abbreviated as ecological uniqueness). Calculations of LCBD are suitable for quantifying which sites contribute more (or less) to beta diversity than the mean and, thereby, for evaluating the ecological uniqueness of species assemblages at each site in a community (Sor et al., 2018).
A high degree of ecological uniqueness (i.e. high LCBD values) may also be linked to unusual habitat conditions and the presence of newcomers or exotic species (Vilmi et al., 2017). Previous studies have indicated that biodiversity conservation programmes that apply this practical and heuristic approach show a high potential for better planning and monitoring to achieve their conservation goals (Legendre & De Cáceres, 2013;Legendre & Gauthier, 2014;Tonkin et al., 2016;Vilmi et al., 2017). Some recent studies have applied LCBD approaches to partition total beta diversity. The majority of existing studies, however, mainly focus on freshwater ecosystems and the organisms therein, such as stream insects (Heino & Grönroos, 2017), urban pond insects , fish in flood-pulse systems (Kong et al., 2017), lake zooplankton (Lopes et al., 2014), organisms in bomb crater ponds (Vad et al., 2017) and diatoms in streams and lakes (Vilmi et al., 2017).
Very few studies have focused on other organisms and ecological systems (e.g. forest ecosystems; Legendre & Gauthier, 2014;Qiao et al., 2015;Yao et al., 2019). Validating the applicability of their findings in an ecosystem other than an aquatic one will indeed be welcome. Moreover, studies that investigate general patterns with regard to ecological uniqueness and their determinants in different ecosystems and organism groups are urgently needed.
Patterns in ecological uniqueness can be correlated with the local environmental conditions at sampling sites (e.g. characteristics of the topography and soil); thus, they can be examined using site-based approaches (Heino & Grönroos, 2017;Legendre & De Cáceres, 2013). The relationships between ecological uniqueness and local environmental conditions have previously been studied in different ecosystems and have yielded unclear results. Sor et al. (2018) and da Silva et al. (2018) found that variations in ecological uniqueness are well-explained by local environmental variables in macroinvertebrate and dung beetle communities, respectively, whereas Heino and Grönroos (2017) and Tonkin et al. (2016) found that ecological uniqueness is weakly related to environmental factors in stream invertebrate assemblages. In forest ecosystems, the relationship between the ecological uniqueness of species assemblages and the local environmental conditions has not been fully verified, and relevant results are quite scarce (but see Legendre & Gauthier, 2014;Qiao et al., 2015). Previous studies have shown that community characteristics (e.g. species richness, abundance and the basal area of each quadrat) also affect ecological uniqueness (da Silva et al., 2018;Sor et al., 2018;Vilmi et al., 2017). In particular, the relationship between ecological uniqueness and species richness is of great significance for biodiversity conservation or restoration (Legendre & De Cáceres, 2013). Similarly, there is a lack of consensus on the relationship between ecological uniqueness and species richness. Existing studies have shown that the ecological uniquenessspecies richness relationship (EUSRR) can be negative (e.g. Legendre & De Cáceres, 2013;da Silva et al., 2018) or positive (e.g. Kong et al., 2017). It is assumed that the proportions of rare and common species in the assemblages may determine the degree to which the EUSRR is negative, positive or non-significant (da Silva et al., 2018). A negative EUSRR seems to occur when the proportion of common species is dominant, whereas it is positive when rare species are dominant in a community (Qiao et al., 2015;da Silva et al., 2018). Moreover, the generality of the EUSRR needs to be further verified in other ecosystems and organism groups (e.g. forest ecosystems). However, regardless of whether the EUSRR is positive or negative, the degree of ecological uniqueness may relate to species-rich or species-poor sites. Hence, linking environmental characteristics to sites with high ecological uniqueness may help identify which type of environment should be given more attention from a conservation context (Legendre & Gauthier, 2014;da Silva et al., 2018).
The forest dynamics plots (FDPs) in the present study include the tropical rainforests in southern China, subtropical evergreen and deciduous broadleaved mixed forests in central China, warm temperate deciduous forests found along the geographical dividing line between North and South China and temperate coniferous forests in north-western China. Here, we investigate the patterns of ecological uniqueness in species assemblages as well as their determinants, and test whether these patterns vary across tropical, subtropical and temperate forests. Specifically, the aim of this study was to test whether the degree of ecological uniqueness in species assemblages can be predicted by community characteristics and/or local environmental conditions. Additionally, we provided insight into whether the local environmental factors that affect ecological uniqueness can be linked to the factors that affect the variations in species composition (i.e. beta diversity).

K E Y W O R D S
beta diversity, biodiversity conservation, ecological uniqueness of species assemblages, local environmental conditions, species richness 2 | ME THODS

| Study areas and FDPs
The study was conducted in four 6-ha FDPs, which we refer to as Hnjfl, Hbmlz, Gsxls and Xjk FDPs across tropical, subtropical and temperate forests. The Hnjfl FDP is located in the Jianfengling National Nature Reserve (18°23′-18°50′N, 108°36′-109°05′E) in the southwestern region of Hainan Island, China. A 60-ha (1,000 × 600 m) FDP within a vast area of continuous old-growth tropical montane rainforest (TMRF) in the core area of the reserve was established from March 2009 to November 2012 (Xu et al., 2015;Zang et al., 2019).
In the present study, we used data collected from a continuous 6-ha

| Experimental design and data collection
All the FDPs in the present study were established according to the standards set by the Centre for Tropical Forest Science (http://www. ctfs.si.edu/), the main features of which are summarized in Table 1.
The census methodology was identical for all FDPs: all woody stems with a diameter at breast height (dbh) of 1 cm or larger were spatially mapped, measured and identified to the species level, and tagged.
Each FDP was divided into 150 (20 m × 20 m) subplots, hereafter called quadrats. Data on specific topographic and soil variables were also collected for each quadrat. Four topographic variables (mean elevation, convexity, slope and aspect) were calculated for each quadrat following the recommendation of Harms et al. (2001) and Yamakura et al. (1995). In brief, the mean elevation is the mean elevation value at the four corners of each subplot. The slope is the mean inclination angle of the four triangular planes composed of any three quadrat corners. The slope aspect was from 0° to 360°, measured in degrees from north, indicating the azimuth. The convexity was calculated as the mean elevation of the focal quadrat minus the mean elevation of its eight surrounding quadrats (Ding et al., 2019). At the central point of each subplot, a soil sample (0-20 cm in depth) was collected. The soil samples were air-dried and transported to the soil laboratory for chemical analysis. Six soil environmental and nutrient variables, including soil pH, the amount of organic matter (OM) and their total amounts, as well as the available amounts of nitrogen (N) and phosphorus (P), were measured.

| Beta diversity and the ecological uniqueness of species assemblages
We applied the index proposed by Legendre et al. (2005) and Legendre and De Cáceres (2013) to compute beta diversity (BT Total ) as the variance of the community data (i.e. variation in species composition among all quadrats within each FDP): where Y = [y ij ] tr is the Hellinger transformed cell-by-species data table (each y ij element represents the number of individuals of species j in quadrat i). The total sum of squares, SS(Y), is the sum of all species and all quadrats of the squared deviations from the species means. n is the number of quadrats. Following Legendre and De Cáceres (2013), the local contributions to beta diversity of a sampling unit i (LCBD i ) represent the relative contributions of that sampling unit to the total beta diversity. LCBD i indicates how exceptional the composition of site i is compared with the centroid of all points, which would represent a theoretical site with the average species composition of all the sampling units (Yao et al., 2019). The ecological uniqueness of each quadrat i (LCBD i ) in terms of community composition can be calculated as.

TA B L E 1 Locations, climatic conditions and overall statistics of the forest dynamics plots (FDP) ordered by latitude
where s ij is the square of the difference between the y ij value and the mean value of the corresponding jth species; the LCBD indices are scaled to add up to one. We used the function 'beta.div' in the {adespatial} R package (Dray et al., 2018), available on CRAN (https:// CRAN.R-proje ct.org/packa ge=adesp atial), to calculate the BD Total and LCBD indices. LCBD indices can be tested for significance using the same function and R package (Legendre & De Cáceres, 2013).

| Predictors of ecological uniqueness in species assemblages
To determine whether the ecological uniqueness of species assemblages in quadrats is related to local environmental conditions, or results from community characteristics, we modelled the degree of ecological uniqueness (i.e. LCBD indices) as a function of local environmental variables (i.e. topographic and soil variables) and community characteristics (i.e. species richness, abundance, and F I G U R E 1 Maps of the ecological uniqueness of species assemblages in quadrats at the study forest dynamics plots (FDPs). The size of the circles is proportional to the degree of ecological uniqueness in terms of community composition (i.e. LCBD values). Black circles indicate significant LCDB indices at the 0.05 significance level after Holm correction for multiple testing (Legendre & De Cáceres, 2013). Grey lines are contours with intervals of 4 m total basal area), using beta regression with the logit link function (Cribari-Neto & Zeileis, 2010). We implemented the best-subset selection for beta regression based on genetic algorithms using the function 'kofnGA' in the package {kofnGA} (Wolters, 2015). For all continuous explanatory variables, the values were z-transformed by subtracting the mean value of the variable (across all 150 quadrats) and then dividing by one standard deviation. This allows for a direct comparison of the effect of explanatory variables (Gelman & Hill, 2006). Finally, to explore whether the proportions of rare species in the assemblages determined the EUSRR, we first calculated the Spearman rank correlation coefficient to measure the EUSRR. Then, the relationship between the Spearman rank correlation coefficient and the percentage of rare species (%) was investigated. Here, we defined rarity using species occupancy (i.e. the
The quadrats that had significantly higher ecological uniqueness (i.e. LCBDs denoted by black circles in Figure 1) showed a certain degree of aggregated distribution in Hnjfl, Hbmlz and Gsxls FDPs. The degree of ecological uniqueness of eight quadrats (black circles) was several orders of magnitude larger than that of the others in the Xjk FDP ( Figure 1d). The total variation explained by the first canonical axes of the environmental variables decreased from tropical through subtropical to temperate forests (Table 2). Additionally, the first canonical axis of the environmental variables was strongly correlated with convexity in Hnjfl (R 2 adj = .56), mean elevation in Hbmlz (R 2 adj = .58) and Gsxls (R 2 adj = .68). The edaphic variables showed a weak correlation with the first canonical axis in Hnjfl but showed moderate correlation in Hbmlz (Table 2). All environmental variables were weakly correlated with the first canonical axis in Xjk. These results indicated that local environmental factors that affect ecological uniqueness and beta diversity were not completely consistent. is well-explained by local environmental variables in macroinvertebrate and dung beetle communities, respectively. However, our results contrast with those reported by Heino and Grönroos (2017) and , who found that ecological uniqueness is weakly related to environmental factors in stream insect and urban pond insect assemblages, respectively. This discrepancy may have partially resulted from the different biological groups and ecosystems that were the focus of these studies. As few previous studies have focused on forest ecosystems, our results shorten the gap in the correlation between ecological uniqueness and local environmental conditions in forest ecosystems.

| D ISCUSS I ON
It is notable that the local environmental variables that affect ecological uniqueness vary significantly among forest plots ( Figure 2). A striking finding from our study is that the effect of environmental factors on ecological uniqueness gradually weakens from tropical (Hnjfl) to subtropical to temperate (Xjk) forests. This result suggests that environmental control is more important in shaping the community assembly of tropical forests (Figures S1 and S2). The environmental conditions are more homogeneous in Xjk than in Hnjfl (Ding et al., 2019). It is supposed that species are more climatically tolerant in high latitudes than in low latitudes. Lower climatic tolerance may further lead to narrower niche breadths in tropical forests than in temperate forests (Chen et al., 2016). We thus hypothesize that the species distribution in Hnjfl is more sensitive to heterogeneous environmental conditions than that in Xjk. According

F I G U R E 3
The Spearman correlation coefficient between ecological uniqueness and species richness (EUSRR) plotted against the percentage of rare species. Spearman's rank correlation coefficient was used to measure the EUSRR. Percentage of rare species (%) was measured as the proportion of species occurring in fewer than 40% of the 20 m × 20 m quadrats to the environmental filtering hypothesis, there is a negative relationship between habitat heterogeneity and latitude (Pianka, 1966) and the higher beta diversity at lower latitudes is caused by stronger environmental filtering (Qian & Ricklefs, 2007;Ricklefs, 1977).
Therefore, different processes may shape ecological communities, producing different relationships between ecological uniqueness and local environmental conditions. We found that ecological uniqueness did not show the same pattern as beta diversity regarding local environmental variables (Table 2 and Figure 2). An important finding from our study is that the local environmental factors that affect ecological uniqueness and beta diversity were not completely consistent, which illustrates that the local environmental factors that affect ecological uniqueness are not necessarily linked to factors that affect the variation in species composition (i.e. beta diversity). Hence, in practice, focusing simultaneously on variations in species composition (i.e. beta diversity) and ecological uniqueness with regard to local environmental conditions is likely an appropriate approach to study forest community assembly.
Our results revealed that the ecological uniqueness of forest quadrats is better explained by various community characteristics, such as species richness, abundance and basal area, which is consistent with the findings of most previous studies da Silva et al., 2018;Sor et al., 2018;Vilmi et al., 2017 Kong et al. (2017), who investigated spatiotemporal patterns in fish assemblages. We, therefore, infer that the relationship between ecological uniqueness and richness is related to the ecosystems and biological groups targeted.
Furthermore, we have provided direct evidence that the strength and direction of EUSRR were strongly related to the percentage of rare species in a community ( Figure 3). Therefore, our results support the assumption that the proportions of rare and common species in the species assemblages may determine whether the EUSRR is negative, non-significant or positive (da Silva et al., 2018). It is supposed that species-rich quadrats exhibit low ecological uniqueness (i.e. negative EUSRR) owing to the greater chance of sharing species with other quadrats (Maloufi et al., 2016) when a community harbours a comparatively high proportion of common species.
Recent studies on freshwater ecosystems have indicated that the relationship between ecological uniqueness and species richness is of great significance for biodiversity conservation or restoration (Legendre & De Cáceres, 2013;da Silva et al., 2018;Sor et al., 2018;Vilmi et al., 2017). From a conservation biology viewpoint, for instance, a negative EUSRR suggests that it is not enough to preserve sites with high species richness because high-richness sites are not necessarily the most unique ecologically. Moreover, species-poor sites may harbour rare or exotic species; thus, these sites contribute strongly to overall beta diversity and are worth studying in more detail in terms of biodiversity conservation. In practice, limited resources do not allow us to conserve all sites. Our results show that ecological uniqueness and the relationship between species richness and ecological uniqueness could be used to prioritize biodiversity conservation.

ACK N OWLED G EM ENTS
We gratefully acknowledge the numerous scientists, graduate and undergraduate students, and research technicians for their contributions to this research. This work was supported by the Fundamental