Ecogeographic patterns of macrophyte metacommunities in the Hengduan Mountain region

ABSTRACT Biological communities exhibit multiple distribution patterns at the metacommunity scale, and assessing the major drivers of these patterns is a key issue in community ecology. Here we investigated how the environmental and geographic gradients shape the distribution patterns of macrophytes at the metacommunity level. We applied the framework of the elements of the metacommunity structure (EMS) to identify the distinct types of 48 macrophyte metacommunities in the Hengduan Mountain region (HDMR) of China. We then used a generalized linear model and model selection approaches to determine which variables contributed to the variations of EMS and linear discriminant function analysis to evaluate how well the tested variables predicted metacommunity patterns. We found wide variations in the 3 EMS (i.e., coherence, range turnover, and range boundary clumping): latitude and alpha diversity were most important in determining coherence; nestedness was mostly related to turnover; and sampling depth was significantly associated with boundary clumping. Seven metacommunity types were identified in HDMR, and most metacommunities best fitted the Gleasonian and Clementsian patterns as well as their quasi-structures. Notably, Gleasonian and Q-Gleasonian patterns as well as the other 3 patterns (i.e., Q-nested, evenly spaced, and nested) were for the first time detected for macrophytes. These metacommunity types were best discriminated by nestedness, altitude, and latitude. Our results provide strong evidence of the impact of geographic patterns on macrophyte metacommunities, with the Gleasonian patterns dominant at both ends of the latitude/altitude gradient and Clementsian patterns common near the central part of the gradient.


Introduction
Understanding how local communities are assembled from regional species pools and how the processes vary across spatial scales has been a key issue in community ecology (Márquez and Kolasa 2013).While environmental controls on species coexistence have been studied extensively for individual local communities (Bernard-Verdier et al. 2012, Myers et al. 2013, Fu et al. 2014, Püttker et al. 2015, Göthe et al. 2016), the potential dispersal role originating from a set of interacting local communities (metacommunity) within a region has attracted much attention (Grönroos et al. 2013, Heino et al. 2015a, 2017a, Gianuca et al. 2017, García-Girón et al. 2020).The metacommunity framework has emerged as a fundamental theoretical basis for understanding how the links between dispersal, environmental factors, and species interactions determine the regional coexistence of species within landscapes (Leibold et al. 2004).Therefore, moving beyond the local community by focusing on a set of metacommunities may advance our understanding of the processes shaping a series of metacommunity patterns across a landscape (Henriques-Silva et al. 2013).
As demonstrated by the metacommunity concepts, a suite of spatially interacting communities could be structured in multiple patterns in a large geographic area (Leibold et al. 2004).Originally, 2 distinct metacommunity types are proposed by ecologists: the discrete community boundaries where the species pool is partitioned into a set of species along environmental gradients (i.e., Clementsian gradients; Clements 1936) and the continuous community boundaries where the individual species respond independently to environmental gradients (i.e., Gleasonian gradients; Gleason 1926).In addition, a number of other metacommunity patterns have been identified, such as checkerboard (Diamond 1975), evenly spaced (Tilman 1982), nested (Patterson and Atmar 1986), and randomness (Simberloff 1983).Leibold and Mikkelson (2002) developed a rigorous quantitative approach to distinguish these suggested patterns by hierarchically assessing 3 elements of metacommunity structure (hereafter called EMS)-coherence, range turnover, and range boundary clumping-based on a species incidence matrix.Coherence relates to how species respond to the same environmental gradient, the range turnover reflects how community composition varies along the gradient, and range boundary clumping indicates the degree to which species range boundaries occur together (Leibold andMikkelson 2002, Presley et al. 2010).Using the EMS framework, previous studies have identified a series of idealized metacommunity patterns in both terrestrial and aquatic ecosystems (Presley and Willig 2010, Alahuhta and Heino 2013, Bonthoux and Balent 2015, Heino et al. 2015c, Vieira et al. 2020).However, the best-fit species distribution models for a given metacommunity are largely context dependent and usually show distinct patterns with regard to organisms, spatial scale, and environmental gradients (Tonkin et al. 2016, Erős et al. 2017).For example, Clementsian metacommunities are more likely found for macrophytes while quasi-nested metacommunities are more common for lentic organisms (e.g., fingernail clam, snail, and fish) in drainage basins (Heino et al. 2015c).A set of metacommunities showing variable patterns at smaller spatial scales may aggregate to form a Clementsian structure at larger scales (Heino et al. 2017b).Moreover, most nested metacommunities occur in regions with colder climate and shorter growing season, whereas Clementsian metacommunities are found in regions with warmer climate and longer growing season (Henriques-Silva et al. 2013).Recently, some comparative studies compiling multiple organisms and regions across a large spatial, and even global, scale aimed to find ecological correlates with metacommunty patterns, but no specific generalities emerged (Heino et al. 2015b, García-Girón et al. 2020).Therefore, because context dependency may hinder our attempts to generalize findings regarding species distribution patterns at the metacommunity level, a single study comparing multiple spatial scales and geographic settings would improve our understanding of the geographic determinants of metacommunity patterns.
In this study we investigated how environmental and geographic gradients (e.g., latitude, altitude) shape the distribution patterns of macrophytes at the metacommunity level.We sampled 802 local macrophyte communities in 48 counties of the Hengduan Mountain region (HDMR), China, where 15-77 local communities within a county were defined as a metacommunity.The region is characterized by relatively wide ranges of latitude and altitude and by pronounced geographic barriers formed by a series of parallel mountains and rivers.Following the EMS framework, we assessed the 3 elements (i.e., coherence, turnover, boundary clumping) for each metacommunity and related them to environmental and geographic factors.Because our previous findings identified a profoundly deterministic assembly process (i.e., dominant role of environmental filtering) on species distribution patterns of macrophyte communities across spatial extents (Fu et al. 2019), we hypothesized that nonrandomness patterns (i.e., Clementsian or Gleasonian) caused by species sorting effects would be more common in the studied region (Leibold et al. 2004).Specifically, we explored how these patterns differed according to environmental and geographic drivers.

Study area
The HDMR, located at the eastern edge of the Qinghai-Tibetan plateau, China, is a core region of the Sino-Himalayan floristic region, one of Earth's 34 biodiversity hotspots, where 6 parallel mountain chains and rivers stretch from north to south and the traffic from west to east is blocked (Fig. 1).HDMR has a total area of 500 000 km 2 and includes the eastern Tibet Autonomous Region, the western Sichuan Province, and the northwestern Yunnan Province, which includes ∼99 counties.The generalized geographic range of HDMR is 22-33°N and 97-103°E (Gan 2007).The elevation ranges from 386 to 7143 m and is higher in the northwest than the southeast.The topography declines from northwest to southeast.In addition, HDMR covers both sides of the "Hu Line," proposed by the Chinese geographer Hu Huanyong, and today is regarded as the Chinese population density distribution line (Fig. 1).On the northwestern side of the Hu Line, population density is sparse and economic development relatively slow while on the southeastern side the population is dense and economic development is strong.

Field sampling
Field surveys of macrophyte communities were conducted during summer (Jun-Aug) from 2014 to 2018.In this study, a suite of local communities (n ≥ 15, sites within or near a river or a lake or temporarily connected ponds) originating from a single county were defined as a metacommunity.We sampled 802 local macrophyte communities consisting of 48 metacommunities (Fig. 1, Table 1) across 48 of 99 counties in HDMR and estimated the abundance of macrophytes using 3-5 randomly positioned quadrats (0.25 m 2 ) at each site.All species occurring in the plots were identified and recorded.At each site, we measured sampling depth and water pH and recorded the geographic coordinates (i.e., latitude and longitude) and altitude using a portable GPS.

Elements of metacommunity structure
To test which metacommunity structure best fitted the macrophyte species distribution, we used the EMS methodology described by Leibold and Mikkelson (2002) and Presley et al. (2010).Metacommunity structure was assessed by hierarchically evaluating the coherence, turnover, and boundary clumping elements of a site using a species incidence matrix, ordinated using reciprocal averaging (i.e., correspondence analysis).
First, coherence was calculated as the number of embedded absences in an ordinated incidence matrix.The significance of coherence was then tested by comparing the observed values to a null distribution of 999 simulated matrices.A significant negative coherence (i.e., more absences than expected by chance) indicates the existence of a checkerboard metacommunity type with strong interspecific competition, whereas nonsignificant coherence is related to randomness of species distribution with regard to gradient.A significant positive coherence (i.e., less absence than expected by chance) suggests a common gradient shapes the species distribution within a metacommunity (Leibold and Mikkelson 2002).The metacommunity type was further specified by assessing the turnover and boundary clumping patterns.
Second, turnover was calculated as the number of times that 1 species replaces (Rep) another between 2 sites in an ordinated incidence matrix.The significance of turnover was then tested by comparing the observed number of replacements to those calculated from the null distribution.A significantly lower turnover than expected indicates a nestedness pattern; a significantly higher turnover than expected suggests evenly spaced Gleasonian or Clementsian patterns identified with further analysis of boundary clumping.
Third, boundary clumping was evaluated as the Morisita's index (MI; Morisita 1971) and compared to a null expectation of 1 using a χ 2 test.A randomly distributed range boundary (i.e., Gleasonian) is expected when MI Table 1.The EMS results for each metacommunity, based on the fixed-proportional (r1) null model and the first reciprocal averaging axis.The interpretation follows Presley et al. (2010).Abs = number of embedded absences; Coh Z = Z-value of coherence, Rep = number of species replacements, Tur Z = Z-value of turnover, Q = Quasi.L&M = Leibold and Mikkelson (2002).In addition, quasi-structures of a metacommunity appear when significant positive coherence and nonsignificant turnover exist.Thus, nonsignificant negative turnover suggests quasi-nestedness, and nonsignificant positive turnover suggests quasi-Gleasonian, quasi-Clementsian, or quasi-evenly spaced gradients, which can be separated by boundary clumping (Presley et al. 2010).

Site
To detect the significance of coherence and turnover, we applied a fixed-proportional null model (r1) to create 999 random matrices with fixed species richness for each site (rows).We calculated the Z score for coherence and the turnover for each metacommunity as follows: Z score = (I obs -I null )/I sd , where I obs is the observed index values for coherence (Abs) and turnover (Rep), I null is the mean index values calculated from 999 null models, and I sd is the standard deviation of the index values calculated from 999 null models.We checked symmetry of null distribution before calculating the standardized effect size (SES) value and used a log-transformation Z score of the test statistic when the null distribution was skewed (Botta-Dukát 2018).Z scores allow comparisons among metacommunities and can thus subsequently be used in comparative analyses.Z scores higher than 1.97 and lower than −1.97 indicate significance at the 0.05 level.EMS analysis was conducted using functions of the metacom package.

Statistical analysis
Local variables (e.g., sampling depth and water pH) and geographic variables (e.g., altitude and latitude) were calculated as polygon centroids across sites within each metacommunity.We calculated spatial extent within each metacommunity using the Euclidean distance between sites.In addition, we classified HDMR into 2 parts (northwest and southeast) according to the Hu Line, which is considered a regional variable.All variables were log/square root transformed prior to analysis if the distribution was not normal.We tested the correlations among the 6 variables using the Spearman method.
We were also interested in testing how alpha, beta, and gamma aspects of biodiversity measures were related to the metacommunity structure.Alpha diversity was calculated as the mean species richness across the sites within a metacommunity and gamma diversity as the sum of species occurring in a metacommunity.Total beta diversity (Sørensen coefficient) and its 2 components-turnover (Simpson coefficient) and nestedness (nestedness coefficient)-were calculated using the functions beta.multi for multiple-site indices in the R package betapart (Baselga and Orme 2012).
To determine which variables contributed to explaining the geographic patterns of macrophyte metacommunities in HDMR, we used 2 complementary approaches.For the 3 EMS, we first used a generalized linear model (GLM) to assess the effects of environmental variables and diversity measures (predictors), applying separately the Z scores of coherence, the Z scores of turnover, or the index of boundary clumping as response variables.The variance inflation factors (VIF) of all tested variables were <6.1, indicating no problem with collinearity among the predictor variables.Next, we applied the model selection approaches using the second-order Akaike's information criterion (ΔAIC < 2) to select the best models with the most important explanatory variables for the beta diversity metrics.The sum of Akaike weights including all models was calculated to estimate the relative importance of explanatory variables.The model selection and model averaging were conducted using functions of the MuMIn package (Bartoń 2018).We applied the same statistical approaches to assess the relative importance of environmental variables (predictors) for determining the variations of the alpha, beta, and gamma aspects of biodiversity measures.
To distinguish metacommunity types, we used linear discriminant function analysis (DFA) to evaluate how well the 6 environmental variables and the 3 diversity measures maximized the differences in the structure of the 48 macrophyte metacommunities using all the detected categorical metacommunity types as response variable.DFA was conducted using the function lda in the R package MASS (Ripley et al. 2013).We also used stepwise selection of predictor variables to determine the most important predictors in separating the metacommunity types using the function greedy.wilks in the R package klaR (Weihs et al. 2005).
All statistical tests were performed using R 3.51 software (Core 2013), figures were constructed using the R package ggplot2 (Wickham 2009), and significance was tested at the 0.05 level.

Results
We found 63 macrophyte species in HDMR during the study periods (38 submersed, 9 floating-leaved, and 16 emergent macrophytes; Supplemental Table S1) and significant variations in diversity measures (i.e., alpha, beta, gamma) among the 48 metacommunities (Supplemental Table S2).Gamma diversity ranged from 6 to 40 species, richness from 2 to 16 species, and total beta diversity (Sorenson dissimilarity index) from 0.09 to 0.71.For the beta diversity measures, turnover components (Simpson dissimilarity index) accounted for 74% (range: 33-91%) of the total beta diversity and dominated most of the metacommunities.
Significant correlations were found among the 6 tested environmental variables (Supplemental Fig. S1), except between spatial extent and sampling depth.That is, the northwestern parts of HDMR are often located at higher latitudes and tend to be more geographically isolated, with shallower habitats and lower water pH.
Latitude was the most important variable determining alpha diversity, which was lower at higher latitudes (Table 2).Hu Line, followed by spatial extent (marginally), was the most important variable affecting total beta diversity and its turnover component, whereas spatial extent accounted for most of the variation of the nestedness component (Table 2).No variables were significantly associated with gamma diversity (Table 2).
For the 3 EMS, we found that latitude and alpha diversity were most important in determining the variations of Z scores of coherence (Table 3).We used nested coefficient as a proxy of beta diversity for predicting metacommunity structure because it was negatively correlated with the Sørensen coefficient (r = −0.31,p < 0.05).Nestedness was the most important variable relating Z scores of turnover, whereas sampling depth was marginally associated with the boundary clumping index (Table 3).
All 48 metacommunities exhibited significantly greater coherence values than the null expectation (Table 3, Fig. 2).Most metacommunities (n = 30) showed no significant turnover patterns, with 17 positive turnovers and 1 negative turnover.The boundary clumping Table 2. Relative importance (RI) of explanatory variables for all model compilations and standardised coefficients (β) obtained from model averaging over all combinations of model terms.Models were calculated for the 5 diversity measures respectively.For RI, 1.00 indicates that the variable is selected in all models, whereas 0 means the variable is not selected in any of the models.β indicates the directions between the 3 elements and the environmental variable.If a given variable (not shown in the table) was not included in the most important models (AICc < 2.0), the direction of influence was obtained from a full model including all the variable candidates.The most important predictors of each metric are given in bold, and the marginally important predictors of each metric are shown in italics.index was not significantly different from 1 for 27 metacommunities, >1 for 20 metacommunities, and <1 for 1 metacommunity (Table 3, p < 0.05).According to the EMS framework proposed by Presley et al. (2010), we found 7 metacommunity patterns (Table 3): 12 Clementsian, 6 Gleasonian, 1 evenly spaced, 1 nested, 8 Q-Clementsian, 17 Q-Gleasonian, and 3 Q-nested patterns.Comparing the metacommunity types between the 2 parts of HDMR relative to the Hu Line, we found that evenly spaced and Gleasonian metacommunities were only present in the northwestern part while nested metacommunities only occurred in the southeastern part.Q-Gleasonian metacommunities constituted the highest proportions of both parts of HDMR (Fig. 3).

Predictors
Here, we excluded the 2 cases of evenly spaced and nested patterns from the following results because only one case of each type was found.The DFA analysis showed that Q-nested followed by Q-Gleasonian and Gleasonian patterns of metacommunity types were relatively well predicted to their original source groups, whereas the Clementsian and Q-Clementsian metacommunity types were poorly predicted to the correct groups (Table 4).The DFA with stepwise selection of predictor variables showed that nestedness (Wilks.lambda = 0.50, F = 10.46,p < 0.001), altitude (Wilks.lambda = 0.36, F = 6.66, p < 0.001), and latitude (Wilks.lambda= 0.30, F = 4.92, p < 0.001) remained significant variables discriminating the observed 5 metacommunity types.Gleasonian metacommunities had the highest altitude and latitude and the lowest nestedness.Clementsian and Q-Clementsian metacommunities occurred at middle altitude and latitude (Supplemental Fig. S2).Q-Gleasonian metacommunities had the lowest altitude and Q-nested metacommunities the lowest latitude and highest nestedness (Supplemental Fig. S2, Table 4).

Variations of EMS in HDMR
We applied the EMS framework to investigate macrophyte metacommunity distributions and variations in assemblage structure in HDMR to identify the geographic patterns of macrophyte metacommunities.We found that the 3 EMS (i.e., coherence, turnover, and boundary clumping) varied widely.All 48 metacommunities showed significant positive coherence, indicating that the species distribution (from a single metacommunity) was responding similarly to the same underlying environmental gradient, and species sorting (i.e., niche dynamics) exerted a dominant control on metacommunity structure in HDMR.Further, latitude was the most important abiotic predictor for determining the coherence patterns, with more positive coherence (more negative Z-value of coherence) at higher latitude (Table 3; β = −0.38),suggesting that higher latitude macrophyte species exhibited a more consistent response to the environmental gradient.In addition, the metacommunities occurring at lower latitudes had a higher species richness (alpha diversity) and a more positive coherence than at higher latitudes, which might largely reflect a warmer climate at lower latitude favoring the growth of more macrophyte species.Likewise, macrophyte communities tend to be richer in subtropical than temperate and cold regions (Murphy et al. 2019).These results indicate that latitude might not only affect coherence patterns directly but also has important indirect effects on coherence patterns through changes in the local alpha diversity of the metacommunities.Moreover, the greater sampling depth resulted in more clumped range boundaries of metacommunities, as indicated by the larger and more positive Morisita's dispersion index.Water depth is one of the most important abiotic factors influencing macrophyte species distributions and often shows a significant zonation pattern along a gradient (Spence 1982).Thus, a clumped species range could be expected in metacommunities with deeper water.

Metacommunity types in HDMR
The Q-Gleasonian and Clementsian metacommunity types prevailed in our study systems, followed by Gleasonian and Q-Clementsian, whereas Q-nested, evenly spaced, and nested metacommunity types were less common.The findings of Clementsian and Q-Clementsian structures were consistent with the results of previous studies of macrophytes conducted within 16 regions globally, whereas the Clementsian pattern usually characterizes the dominant metacommunity types for macrophytes across regions or basins (García-Girón et al. 2020).Further, a Clementsian (or sometimes the quasi-Clementsian) pattern has also been reported for other aquatic biotas, including diatoms, bacteria, invertebrates, damselflies, fishes, and microcrustaceans (Heino et al. 2015b).The dominance of Clementsian and Q-Clementsian patterns indicates that the macrophyte metacommunity in some regions of HDMR is composed of discrete community boundaries that show similar responses to the environment (i.e., coherent species distributions) and replace as groups across space.In this case, 2 or more species groups respond to the same ecological gradient and have similar range boundaries (Leibold andMikkelson 2002, Presley et al. 2010), reinforcing the role of the environmental filter  in structuring the macrophyte metacommunity.By comparison, our recent studies have also identified a significant functional convergence of macrophyte communities and highlighted a dominant role of deterministic assembly processes governing the spatial functional turnover of macrophyte communities across this region (Fu et al. 2017), lending further support to our results.However, Gleasonian and Q-Gleasonian patterns as well as the other 3 patterns (i.e., Q-nested, evenly spaced, and nested) have not been detected for macrophytes in previous studies (Heino et al. 2015c, García-Girón et al. 2020).Instead, these highly variable observed patterns between macrophyte communities are consistent with previous comparative analyses of lake biota within small regions (Heino et al. 2015c), possibly reflecting the relatively low environmental heterogeneity within basins, where the relative control of environments and spatial dispersion may vary greatly with spatial extent and geographical region (Alahuhta andHeino 2013, Fu et al. 2019).Notably, the dominance of Gleasonian and Q-Gleasonian patterns in HDMR suggests that most macrophyte metacommunities are composed of species responding individualistically to the environment, a response also evidenced for metacommunities of other aquatic biotas (e.g., bacteria, diatoms, zooplankton, fish, bryophytes, and invertebrates) in aquatic ecosystems worldwide (e.g., Finland, China, United States; Dallas and Drake 2014, Heino et al. 2015b, Erős et al. 2017, He et al. 2020).

Determinants of metacommunity patterns in HDMR
We found that latitude and altitude were the most significant variables driving the variations of metacommunity patterns in HDMR, especially for discriminating the Gleasonian and Clementsian as well as their quasi-structure patterns along the gradient.Coincidently, Heino et al. (2015b) found that mostly Gleasonian and Clementsian metacommunities of aquatic biotas dominated along the geographic gradient, while Gleasonian metacommunities dominated in the southernmost drainage basin (66°N), and Clementsian metacommunities in the northernmost drainage basin (70°N) of 3 Finnish drainage basins.By contrast, results from a comparative analysis of biologically highly different organismal groups (bacteria, algae, macrophytes, invertebrates, and fish) revealed no clear relationships between geographic gradients (1300 km) and metacommunity types in the freshwater realm (Heino et al. 2015c).More recently, a comparative study based on macrophyte data from 16 regions worldwide (∼10 000 km latitudinal gradient) proved the existence of strongly context-dependent metacommunity-environment relationships, indicating that no single mechanism accounts for the variability of macrophyte metacommunity patterns (García-Girón et al. 2020).Notably, the failure of comparative studies to detect the geographic patterns of metacommunities might be due not only to the relatively larger spatial scale that highlights the environmental heterogeneity within a metacommunity, predominantly leading to Clementsian structures, but also the differences in sampling time of biotas and metacommunity-level environmental variables (Heino et al. 2015c).Thus, a potential caveat is that we sampled macrophytes and environmental variables in different years from 2014 to 2018, although we selected the same season (summer growth season) in all years because temporal variation in environmental conditions is known to affect the composition of freshwater plant communities at local and regional scales (Wojciechowski et al. 2017, Diniz et al. 2021).
In this study, Gleasonian and Q-Gleasonian patterns often dominated at the 2 ends of the geographic gradients (altitude and latitude) while Clementsian and Q-Clementsian metacommunities prevailed at the middle of the geographic gradients.In HDMR, habitats at higher altitude and latitude are characterized by colder and severe winters, longer ice cover periods, and shorter growing seasons than habitats at lower altitude and latitude, creating homogeneous, temperaturelimited conditions for macrophyte growth.Likewise, habitats at lower altitude and latitude, such as Gongshan, Fugong, and Tonghai, mostly exhibit temporarily connected ponds and small shallow lakes undergoing seasonal drought and are usually more homogeneous than the large deep lakes, such as Erhai, Luguhu, Fuxianhu, Napahai, located at the middle geographic gradients.Furthermore, the local alpha diversity was relatively low for the Gleasonian and Q-Gleasonian metacommunities, indicating that most habitat specialists did not inhabit these unfavorable habitats.Thus, the tolerant species must finish their life history sooner and exhibit individual responses to the harsh and homogeneous environments, producing Gleasonian metacommunity structures.By contrast, Clementsian and Q-Clementsian metacommunities had higher local alpha diversity and inhabited a relatively deeper area, in agreement with a more heterogeneous and benign habitat allowing more species to maximize their exploitation of functional niches (Fu et al. 2020), thus promoting coexistence, which could potentially increase the effects of environmental filtering and species sorting and lead to clumped species replacement structures (Heino et al. 2015b, Fu et al. 2019).

Conclusions
We identified 7 metacommunity types in HDMR, most of which best fitted the Gleasonian and Clementsian as well as their quasi-structures, even though previous studies have only detected Clementsian and Q-Clementsian structures for macrophytes at regional scale.This finding greatly increases our understanding of macrophyte metacommunity structures in a series of narrowly isolated biogeographic regions and allows us to identify general ecological patterns (i.e., the EMS analysis).Our results provide strong evidence of the impact of geographic patterns on macrophyte metacommunities, with Gleasonian and Q-Gleasonian dominating the 2 ends of the altitude and latitude gradients and Clementsian and Q-Clementsian most common in the middle gradients.

Figure 1 .
Figure 1.Geographic location of the Hengduan Mountain region.The hatched line indicates the Hu Line, which empirically divides the population distribution as well as economic development in China into 2 parts (i.e., low population density and poor; high population density and prosperous).Colors represent their canonical scores along a discriminant analysis separating macrophyte metacommunity patterns using local and regional variables together, and letters represent the original results from the EMS analysis.

Figure 2 .
Figure 2. The 48 metacommunity types plotted in the space of the Z scores of coherence and turnover.Bubble size denotes the index of boundary clumping.The dashed line indicates Z scores of −1.96 and 1.96.

Figure 3 .
Figure 3.The proportion of metacommunity types observed in the southeast (E) and northwest (W) part of the Hengduan Mountain region.

Table 3 .
Relative importance (RI) of explanatory variables for all model compilations and standardised coefficients (β) obtained from model averaging over all combinations of model terms.Models were calculated for the 3 elements of metacommunity: coherence, nestedness, and the coherence, turnover and boundary clumping, respectively.For RI, 1.00 indicates the variable is selected in all models, whereas 0 means the variable is not selected in any of the models.β indicates the directions between the 3 elements and the environmental variable.If a given variable (not shown in the table) was not included in the most important models (AICc < 2.0), the direction of influence was obtained from a full model including all the variable candidates.The most important predictors of each metric are given in bold, and the marginally important predictors of each metric are shown in italics.

Table 4 .
Summary of average values for the metacommunity characteristics.Also, shown are correct classifications (%) from discriminant function analysis based on the 3 significant predictors: altitude, latitude, and nestedness.n is the number of metacommunities.HL indicates Hu Line.