Environmental Drivers of Algal Metacommunity Structure in in River-Connected Lakes


 The metacommunity approach provide insights into how the biological communities are assembled along the environmental variations. The current study presents the importance of water quality on the metacommunity structure of algal communities in six, river-connected lakes using long-term (8 years) monitoring datasets. Elements of metacommunity structure were analyzed to evaluate whether water quality structured the metacommunity across biogeographic regions in the riverine ecosystem. The algal community in all lakes was found to exhibit Clementsian or quasi-Clementsian properties, indicating that the communities responded to the environmental gradient. Reciprocal averaging clearly classified the lakes into three clusters according to the geographical region in river flow (upstream, midstream, and downstream). The dispersal patterns of algal species, including Aulacoseira, Cyclotella, Stephanodiscus, and Chlamydomonas across the regions also supported the spatial-based classification results. Although conductivity, chemical oxygen demand, and biological oxygen demand were found to be important variables (loading > |0.5|) of the entire algal community assembly, temperature was a critical factor in water quality associated with community assembly in each geographical area. These results support the notion that the structure of algal communities is strongly associated with water quality, but the relative importance of variables in structuring algal communities differed by geological regions.


Introduction
The metacommunity concept is an important approach for community ecology because it allows both local (e.g., nutrient, biotic interaction) and regional (e.g., dispersal) factors that contribute to community assembly to be identi ed 1 . Interest in community assembly is increasing because the local community is constantly reassembled in response to changes in the local environment, and the diversity and functionality of the community is controlled by the spatial distribution and interaction of species in the community 2,3 .
To understand the role of the community assembly in the eld, focusing on the community level (such as the metacommunity concept), rather than the species level, can provide new insights to associate the environmental factors with the community 4 .
Elements of metacommunity structure (EMS) is a useful analysis tool that evaluates the assembly process of the community and determines the effects of environmental factors on the community assembly by assessing community patterns 5,6 . The EMS calculates three elements (coherence, turnover and boundary clumping) to identify the idealized metacommunity pattern (e.g. checkerboard, random, evenly spaced, Gleasonian, or Clementsian pattern). The patterns facilitate the search for general rules determining metacommunity structure 6 . EMS approaches have been applied to terrestrial and aquatic systems for various organisms.
Most studies on sh communities have been focused in freshwater, and a few of studies have dealt with insects and zoo plankton [7][8][9][10][11] . The algal community have largely been neglected when applying the EMS approach. Although the abundance of algae species is known to be affected by water quality, such as by the phosphorus and nitrogen levels in the water, it has not been proven whether the assembly of algal communities is also affected by water quality.
The incidence of algal species is in uenced by complex relationships between biological and environmental factors such as species dispersion, competition, water quality, and topography 12,13 . Although independent biological and environmental factors have been identi ed for incidence of a single algal species in laboratory conditions, the study of factors affecting community was limited in freshwater 14 . Relationships between these factors and the algal community have been found to vary across regions and spatial scales 15 . No studies have focused on the river-connected lakes, which is an aquatic system in which species disperse naturally and share a large number of algal species. Algal communities may vary depending on local environmental factors, compared to geographically distant lakes 13,14 . If the phenotype (e.g. algal blooms, order, and biotoxins) varies for each region depending on the algal community structures, it is essential to elucidate community assembly in the same riverine system.
The goal of this study was to understand how environmental variables in uenced algal community assembly in six river-connected lakes in South Korea. The algal community was examined for the idealized metacommunity structures at each lake using long-term (8 years) monitoring datasets and how environmental factors (water qualities) in each of the EMS analyses were associated with algal community assembly in the lakes. Our present study provides new comparative information about the response of algal community to local environmental factors although the metacommunity structures are largely invariable at different biogeographic scale in riverconnected lakes.

Materials And Methods
Algal community sampling and data acquisition Field sampling were performed at six lakes of the North Han River, Paldang (PD), Cheongpyeong (CP), Uiam (UM), Chuncheon (CC), Soyang (SY), and Hwacheon (HC) (Fig. 1). In detail, we monthly collected water samples spring to autumn (March to November) from 2008 to 2016 (excluding 2014) at six lakes in the North Han River. All samples were taken at a depth of 1 m below the surface at the middle of the lake using a Van Dorn sampler (Horizontal water sampler, iStech, Korea). While sample transportation, the water sample was kept at 2 L vinyl containers (BT1550-2000, Korea) and stored in an icebox.
The nine water quality parameters of samples such as temperature, conductivity, pH, BOD (biological oxygen demand), COD (chemical

Element Of Metacommunity Structure Analysis
EMS analysis consists of three components: coherence, turnover, and boundary clumping. Through EMS analysis, idealized metacommunity models are determined 5 , or quasi-models 6 . Coherence is assessed by counting the number of embedded absences in the ordinated matrix and comparing this to a null distribution. We identi ed metacommunity structure with presence/absence algal matrix based on R1( xed-xed) model as null model. Turnover is calculated from the number of species replacing each other from site to site 5 . Boundary clumping is evaluated by comparing the observed distribution of range boundaries with an expected equiprobable distribution 5,6 . To identify metacommunity structure at sampling sites, the coherence, turnover, and boundary clumping were computed in R 16 , using the 'metacommunity' function in the 'metacom' package (version 1.5.2). The metacommunity structure was determined using the p-value and z-score. When all three components of coherence, turnover, and boundary clumping had signi cant pvalues, 12 metacommunity structures were identi ed by z-score and Morisita index. Firstly, the metacommunity structures were classi ed by coherence z-score into a checkboard (greater than 1.96), random (-1.96 to 1.96), and nest or gradient type (less than − 1.96). The nest or gradient type was separated according to whether the turnover z-score was positive (nest type) or negative (gradient type). If the turnover z-score value was between − 1.96 to 1.96, the metacommunity structure became a 'quasi-structure' (i.e., quasinested, quasi-Clemensian). Lastly, Morisita index (I) separated gradient metacommunity structures as Clemensian (I > 1), Gleasonian (nonsigni cant), and evenly spaced (I < 1). The metacommuntiy order of each sample was calculated by reciprocal averaging to ordinate the site-by-species matrix. Then, we ranked the site score following order of the samples in the overall metacommunity structure.

Statistical analysis
The number of observed algal genus during sampling was calculated as richness. Statistical analysis was conducted using R (version 3.6.1). All variables were checked for normality with the Shapiro-Wilks normality test. If the data obeyed the normality test, ANOVA were performed to compare richness, site score distribution and nine water quality parameters of six lakes. Otherwise, the Kruskal-Wallis test was used. The hierarchical clustering was calculated by using the presence/absence matrix and Euclidean distance of six lakes from the 'hclust' function in R. To nd key algal genus three groups which separated by the site score (High: SY and HC; Middle: CC, UM and CP; Low: PD), the R package 'random forest' (version 4.6-14) was used for a random forest classi cation 17 . The classi cation model was designed with 131 trees with 1,000 permutations using sampling data and was validated by the confusion matrix method. From the classi cation models, six key indicator genera were selected with a top 10% mean decrease in Gini. The 'cca' function of the 'vegan' package (version 2.5-6) was used to implement canonical correspondence analysis (CCA) 18 to assess which environmental variables were associated with site score distribution and the underlying metacommunity structure of the North Han River. The correlation analysis was performed between CCA 1 and water quality parameters using the Pearson or Spearman method on normal or nonnormal datasets, respectively.

Elementary Of Metacommunity Structure
Across all datasets, EMS analysis revealed positive coherence, positive turn over and large values (> 1) of boundary clumping with Clemensian structure (Table 1), with ranges of algal species contributing most to these patterns (Fig. 3a). The Clemensian structure indicates that the community was assembled by environmental gradient. Most of individual lakes also exhibited the Clemensian structure as a best-t pattern of metacommunity structure (Table 1, Fig S1). In contrast, PD exhibited a quasi-Clemensian structure due to its non-signi cance in turnover. Even though all lakes were analyzed by season, they were identi ed as Clemensian or quasi-Clemensian as the same as the abovementioned results (Table S1). These results support the algal community were strongly associated with environmental factors regardless of variation in biogeographic units or seasons. The Clemensian structure across all datasets had three compartment-by-site scores, which were determined by an EMS ordination procedure for ordering algal species and sites (Fig. 3b). CP, UM, and CC, located midstream, were between 100 and 300 in rank of site scores, and signi cantly different from upstream (SY and HC) and downstream (PD) (P < 0.05). These results are differentiated from the beta-diversity results, which were divided into two groups. Although algal species dispersed through river hydrologic connections and six lakes shared considerable numbers of algal species, the occurrence of some species was unique geographically. The indicator analysis con rmed that the abundances of key indicator species were clearly varied depending on the location of river networks (Fig. 4). The distribution of Aulacoseira, Cyclotella, and Stephanodiscus increased proportionally from upstream to downstream, while Chlamydomonas decreased. The distribution of Asterococcus was unique to the upstream group. These results support the contention that metacommunity analysis is capable of analyzing the assembly of algal communities in detail at the community level as well as at the species level. Lakes located downstream could be classi ed into two additional clusters (down: PD, and mid: CP, UM, and CC) according to the assembly characteristics.

Environmental Drivers
The water quality of all lakes is summarized in Table S2. Although clear regional differences were not observed in algal diversity and EMS results, the water quality of the downstream (PD) and upstream (SY and HC) groups was statistically different, except for pH. The midstream (CP, UM, and CC) varied regionally within the range of water quality values between upstream and downstream, while COD and TN gradually increased from downstream to upstream. Seasonal or annual temperatures and precipitation, which are known to be important for the algal growth, could not be seen as signi cant differences between lakes (Table S3 and S4).
To identify the environmental drivers of algal community, the association between water quality and site scores generated from EMS analysis was evaluated. The CCA axes were de ned by reciprocal averaging, which is the same ordination method used to identify the main gradient of species distribution in the EMS framework 4 . Site score and CCA axes were highly correlated (Spearman's ρ = 0.92, P < 0.001), indicating that axes from both analyses represent variations in the same latent environmental gradients. The variation in temperature was most highly associated with the CCA1 axes along with metacommunities structure in each location (Table 2), whereas temperature was not important in all lakes (loading = -0.061). Temperature was positively related to the EMS ordination axes of upstream (loading = 0.893), but negatively related to axes of downstream (loading = -0.801) and midstream (loading = -0.547). Moreover, other environmental variables which were highly associated with EMS ordination axes (i.e. loading < 0.500 or > 0.500) were signi cantly different by location. Conductivity and COD were the environmental variables highly associated with ordination axes for upstream, but pH or BOD were associated downstream and midstream. Compared to the individual analyses of lakes, these results tended to be similar (Table S5).

Discussion
We used EMS analysis combined with CCA to identify the relationship between water quality and assembly of algal communities in river-connected lakes. Most of the algal metacommunities for each lake follow a Clementsian structure, characterized by a continual change in algal composition at the genus level along environmental gradients. EMS produced three regional compartments (upstream, midstream and downstream) by reciprocal averaging score. CCA revealed that three compartments were associated with different variables of water quality. Therefore, algal communities along the river were generally assembled depending on the water quality of the region, even though algal communities were dispersed and the species were shared through hydrological connections.
The EMS and the conventional diversity approach were compared to understand the importance of biogeological features on the algal community in river-connected lakes. The alpha diversity (richness) varied signi cantly depending on individual lakes (Fig. 2a), but the beta diversity and EMS approach could provide clear clustering by biogeographical features (Figs. 2b and 3b). Since beta diversity measures the changes in diversity of species from one site to another 19 , beta diversity should provide similar clustering results to the EMS approach. Nevertheless, it is worth noting that the number of regional partitions was different in the two approaches. Because EMS is based on site-by-species incidence, matrices consider whether the community responds to environmental gradients by measuring the proportional species turnover 1 , thus the EMS approach could provide discriminatory information compared to beta diversity.
Metacommunities in biogeographical regions or individual lakes were either Clementsian or quasi-Clementsian (Table 1 and S1).
Clementsian structures arise when communities are actually changing consistently through groups of species that respond in a similar way to environmental gradients 20 . Synchronous species turnover is a phenomenon that occurs in ecosystems that share a signi cant proportion of species 21 . Clementsian structure is not rare, and they have already been reported for other aquatic organisms 8,22 . Most species found in riverine ecosystems are generally regulated by species dispersal and sorting 12 , so that the downstream sites shared high proportions of species, while upstream sites showed signi cant differences ( Fig. 3b and Fig. S2). The lakes located in midstream (CP, UM, and CC) shared highly similar distributions of algal species but were signi cantly different from downstream (PD). PD, at the con uence of three rivers, is prone to dispersal of other species from other rivers. These partially explain the quasi-Clementsian structure and distinct patterns compared to midstream.
One of the advantages of the EMS approach is identifying the environmental variables that in uence community assembly by correlating reciprocal averaging and environmental variables. CCA, which is based on reciprocal averaging and multiple regression, was used to determine which environmental variables were associated with gradients along which metacommunities were structured 18 . The algal communities across broad geographical gradients showed consistent Clementsian structure. Clementsian structure emphasizes discrete 'community types' along ecological gradients, such that subgroups of species replace other subgroups in space 20 . Such variation also suggests that subgroups of species either respond similarly to environmental variation, or are affected by similar historical effects 23 .
Conductivity, COD, and BOD were found to be the most important variables (Loading > |0.5|) of the entire algal community assembly.
Previous reports also identify conductivity, COD, and BOD as the main drivers of the algal community composition 13 , indicating the importance of these factors as a driver of algal species composition in the rivers. This contradicts, in part, the work of Padisak et al. 24 who found TN and TP to be important drivers of functional groups in the river, while conductivity and COD were not signi cantly correlated with functional groups. However, untangling these communities and analyzing each lake type classi ed by the EMS approach revealed that the algal community could be distinguished by presenting a different relationship with temperature. The compositions of algal communities are remarkably in uenced by temperature in a single lake 14,25 . Since lakes classi ed through the EMS approach had a similar algal composition (Figs. 2 and 3), it is possible to explain that temperature acted as an important variable, unlike the results where the entire algal community is analyzed. Except for temperature, the variables strongly related to algal communities were conductivity, COD, and BOD, which concurs with the entire algal community analysis, but that importance differs depending on the location of the lake. Depending on land use and population density along the river, the types and concentrations of organic matter owing into rivers vary, and the algal communities, which are strongly affected by differences in organic matter, are sensitive to regional differences 26 . This may explain why environmental variables were found to regionally in uence the algal community assembly in river-connected lakes.
The relationship between the main structure of the entire metacommunity and the three lake types classi ed by biogeographical regions (up-, mid-, and downstream) reveals the role of spatially structured factors on species composition. Previous research on lakes has shown that geographical distance strongly in uences the algal community distribution 15 . The results of the current study also provide evidence that more than half of the species are shared regardless of the lake position as species dispersal is a main driver of community assembly in a riverine ecosystem. The uniqueness of the algal communities in each lake supports that the assembly of algal community is affected by species sorting. These results are consistent with previous ndings that algal communities are determined by species dispersal when habitats are shared in aquatic environments 12 . The EMS approach is powerful in detecting compartmentalized structures according to spatial distribution and provides a fruitful interpretation of algal communities at the species and community levels.

Conclusion
River-connected lakes were used to address patterns and the underlying process of metacommunity organization of algal communities in freshwater. The approach based on metacommunity used ecological features, providing a fruitful starting point for more sophisticated analyses of variations in algal community structure. Our ndings strongly suggest that algal metacommunities showed Clementsian structures in broad spatial extents. The EMS approach combined with CCA facilitated the interpretation of the effect of environmental variables on the variation of the algal community assembly, and its effects across biogeographic regions in riverine ecosystem. In addition, the results also provide insight into biogeographical patterns of algal community structure in freshwater by comparing the beta diversity and EMS approach. This nding may also be applicable in aquatic ecosystems when studying local communities across large spatial scales.