Freshwater systems are threatened by many anthropogenic effects including habitat fragmentation, land cover change, and disruption of the natural flow regime (Vörösmarty et al., 2010; Heino, 2013). Human effects on taxonomic and functional composition of freshwater communities are well documented (Poff, 1997; Allan, 2004) to the point that they are foundational to the paradigm of contemporary environmental assessment (Karr, 1981; Bailey et al., 2004). Landscape level studies have assessed the dependence of measured shifts in fish communities upon current and past land use changes, with some evidence of generalization in thresholds for urbanization and agriculture effects on compositional and richness changes (Chen and Olden, 2020). While land use variables provide the first evidence that urbanization and agriculture change fish communities, pinpointing the specific variables associated with these land uses can give additional insight into what environmental monitors should be measured when completing bioassessments (Waite et al., 2021; Iacarella, 2022). For example, Stoczynski et al. (2021) found that dam density affected how species replaced one another across the landscape, Ortega et al. (2021) identified impervious surfaces as a proxy for urbanization, and Waite et al. (2021) highlighted how contaminants in water and sediment were just one of multiple urbanization and agricultural stressors. Thus, disentangling anthropogenic effects from ‘natural’ environmental variables in metacommunity analyses may increase our ability to explain factors structuring beta diversity within metacommunities (Duarte et al., 2018; Gianuca et al., 2018). However, few studies have examined stream fish metacommunities from the lens of anthropogenic effects (Erős et al., 2012; Borges et al., 2020; Stoczynski et al., 2021).

Biodiversity is represented by three dimensions: taxonomic, phylogenetic, and functional diversity; patterns of beta diversity can be unique for each dimension. Taxonomic diversity indexes counts and proportions of taxa (typically species or genera). Alternatively, functional diversity represents the range of traits present in a location, regardless of species (Tilman, 2001), while phylogenetic diversity refers to the degree of variation in evolutionary history in a community (Webb et al., 2002). Most analyses of beta diversity have focused solely on taxonomic diversity and have not considered that co-occurring species may be from different phylogenetic positions and/or have different functional roles within the community (Jarzyna and Jetz, 2016). All species have traits which correspond to their roles and habitat preferences within a community. Functional beta diversity describes differences in these traits between communities (Wootton, 2012; Leibold and Chase, 2017). Phylogenetic beta diversity allows us to investigate evolutionary differences among communities (Weinstein et al., 2014; Nakamura et al., 2017, 2021; Li F. et al., 2021; Li Z. et al., 2021). Thus, metacommunity approaches can also use functional and phylogenetic beta diversity to complement taxonomic beta diversity to provide additional insights and a more complete diversity picture (Anderson et al., 2011; Leibold and Chase, 2017; Cai et al., 2019; Jia et al., 2020).

Riverine fishes provide a unique study system for testing metacommunity concepts. The dendritic structure of river networks restricts movement of fishes to the stream network (Brown and Swan, 2010; Tonkin et al., 2018). This framework creates a dual habitat-connectivity gradient in which (a) some headwater streams contain more isolated communities and thus are more prone to the effects of species sorting and stochastic effects, and (b) other headwaters with similar habitats are well connected to mainstem rivers resulting in a greater influence from dispersal processes (Brown and Swan, 2010; Hitt and Angermeier, 2011; Heino et al., 2015a; Henriques-Silva et al., 2019). Moreover, watersheds are arranged hierarchically; small watersheds are nested within larger watersheds separated by discrete spatial boundaries caused by mountainous and marine areas, creating discrete spatial units for delineating metacommunities as statistical units in large-scale analyses (Jelinski and Wu, 1996; Patrick and Yuan, 2019). While watershed boundaries create distinct dispersal barriers for fishes that naturally delineate metacommunities, data availability or constraints often lead researchers to delineate metacommunities based on other features such as political boundaries or natural geomorphic breaks across the landscape such as ecoregions (Cai et al., 2019; Henriques-Silva et al., 2019; Lech and Willig, 2020). However, few studies have investigated how the choice of spatial delineation may affect the outcome and interpretation of results. Conducting separate analyses on the same dataset with different delineation scenarios representative of existing metacommunity studies will provide important methodological context for conducting metacommunity analyses.

In this study, we used an extensive stream fish abundance dataset covering the state of South Carolina, USA, to quantify variation in the three dimensions of beta diversity attributable to spatial, natural environment, and anthropogenic effects. Our first objective was to quantify variation in stream fish taxonomic, functional, and phylogenetic beta diversity. We expected to see more explained variation in functional and phylogenetic beta diversity because of the association between species traits and common resource utilization. Additionally, relatedness between those traits may be masked by taxonomic diversity in functionally redundant communities (Leibold and Chase, 2017; Gianuca et al., 2018; Cai et al., 2019). Our second objective was to measure the relative variation explained by spatial, natural environmental, and anthropogenic variables. We expected that the use of anthropogenic variables more specific than land use would explain appreciable amounts of variation in all three beta diversity dimensions because of the extensive alterations to stream systems in South Carolina in recent centuries (Allan, 2004; Walters et al., 2005; Wenger et al., 2008; Marion et al., 2015; Langerhans and Kern, 2020). We conducted analyses for these objectives separately for three different spatial delineations: one based on the entire state, one based on a natural geomorphic ecoregion break, and one based on discrete watershed boundaries. In doing so, we highlight the importance of a priori methodological decisions when conducting beta diversity analyses.

Variation partitioning explained 25–81% of variation in stream fish beta diversity, depending on the metacommunity delineation and dimension of beta diversity (Figure 3 and Table 4). All db-RDA analyses resulted in significant explained variance by the explanatory variables (all p-values <0.001; Table 5). We observed a strong phylogenetic signal in our fish traits for each metacommunity (Supplementary Table S5). We found that overall explained variation in stream fish beta diversity, the relative contribution of the three suites of variables to beta diversity, and the contribution of individual variables, all differed depending on how metacommunities were delineated—whether the entire state, above and below the fall line, or based on watershed boundaries. Breaking up our study region into the upstate and lowland metacommunities resulted in the least explained variance; however, the explained variance for anthropogenic portions alone was much higher than when metacommunities were distinguished based on major watersheds or the whole state (Figure 3 and Table 4).

We investigated how spatial, natural environmental, and anthropogenic variables influence taxonomic, functional, and phylogenetic beta diversity for stream fish communities across South Carolina, USA. Our measured variation explaining beta diversity ranged from 25–81%, which is high relative to other studies investigating similar questions (Gianuca et al., 2018; Burdon et al., 2019; Benone et al., 2020). Many freshwater studies investigating drivers of beta diversity using variation partitioning have explained around 30% of the variation for a variety of taxa including stream macroinvertebrates, ostracods, fish, and macrophytes (Alahuhta and Heino, 2013; de Campos et al., 2018; Huang et al., 2019; Krynak et al., 2019). Accounting for biogeographical features of metacommunities while investigating taxonomic and functional beta diversity of Amazonian stream fishes resulted in explained community variance above 60% (Benone et al., 2020). Studies that added anthropogenic effects also increased the explained variance, similar to our results, where researchers described over 60% of the variance in freshwater macroinvertebrate community structure in Switzerland (Burdon et al., 2019). Variables including road crossings, impervious surfaces, fertilizer application, and forest loss were selected over generic land use variables. Our study, in conjunction with previous studies, indicates that the unexplained variance of past studies was due partly to information that the inclusion of anthropogenic effects provides. Our research highlights the importance of including anthropogenic variables in beta diversity studies and how these variables can provide better context for understanding human effects on stream fish beta diversity.

Quantifying beta diversity in functional and phylogenetic dimensions more fully captures the factors driving beta diversity by introducing traits and relatedness among taxa that is not captured by using taxonomy alone. We explained more variation using taxonomic beta diversity relative to the other diversity dimensions for all but one metacommunity. Large regional species pools provided a greater difference in species composition among sites to account for the increased explained variance of the taxonomic beta diversity compared to differences in traits or evolutionary relatedness, indicating higher redundancy of traits and evolutionary history for stream fish compared to other taxa. Substantial differences in explained variation between the diversity dimensions has been observed among macroinvertebrates (Gianuca et al., 2018; Hill et al., 2019; Bispo et al., 2021; Li Z. et al., 2021) and macrophytes (Garcia-Giron et al., 2019). Gianuca et al. (2018) observed up to three times more explained variation in functional and phylogenetic beta diversity compared to taxonomic diversity of zooplankton metacommunities over an urbanization gradient. Therefore, the inclusion of more than one diversity dimension can not only provide more explained variation but can offer differential insights on the factors structuring beta diversity. For example, functional diversity can provide insights into how environmental stressors effect community structure in stream fish (Jia et al., 2020; Zhang et al., 2020). The only other study to our knowledge which also investigated all three diversity dimensions in stream fish observed that instream habitat variables had a weak relationship with taxonomic beta diversity, while phylogenetic beta diversity values failed to show deterministic structuring processes (Roa-Fuentes et al., 2019). In contrast, our study showed phylogenetic beta diversity was best explained by varying degrees of spatial, natural environmental, or anthropogenic factors based on the spatial delineations within our study. Thus, the amount of explained variation in beta diversity is context dependent because even similar taxa can show different outcomes. Similar studies conducted on other regions of the world with varying climates, geography, and evolutionary histories would improve our understanding of context dependency.

We conducted analyses separately for three spatial delineations that are representative of commonly applied approaches—delineations based on political boundaries, natural geomorphic breaks, and watershed boundaries. Our results indicate that the choice of spatial delineation affects the outcome of and interpretation of variation portioning analyses, resulting in alternative interpretations of overall variance explained in metacommunity structure, as well as relative variable contributions to changes in beta diversity. Analyses based on the whole state delineation (based on political boundaries) emphasized the importance of natural environmental factors and spatial gradients indexing large-scale dispersal. Alternatively, analyses based on the fall line delineation provided different inference in which the importance of anthropogenic and spatial factors far outweighed the importance of environmental factors. This delineation also shed light on the fact that the relative contributions of anthropogenic and spatial dispersal gradients differ appreciably between the two regions, indicating different assembly rules structure stream fish metacommunities in different parts of the state. Finally, analyses based on watershed boundary delineations highlighted the synergistic effects of anthropogenic factors, spatial dispersal gradients, and natural environmental variables for structuring stream fish metacommunities. The latter analyses exposed differences between the three dimensions of stream fish beta diversity that were not evident in analyses based on the other two delineations. Many metacommunity studies are often applied to datasets that were collected without the expressed intent of quantifying beta diversity across the landscape, and are subject to variation according to methodological attributes of the dataset such as sampling density and intensity (Stoczynski et al., 2021). Likewise, our results show that user-defined choices of how metacommunities are defined a priori have important bearing on model results and ecological inference, and should be considered carefully before conducting metacommunity analyses.

Differences in the results of analyses based on different spatial delineations bring to the forefront how past geologic change affects community assembly processes across the state. The fall line represents the farthest inland the coast has been since the Cretaceous period when the sea levels were at their highest. The Appalachian uplands including the majority of the Piedmont ecoregion has been continuously above sea level since that time, which predates the modern ostariophysan fauna (Cavender, 1998). Accordingly, the faunas of river basins above the fall line have had an essentially uninterrupted evolutionary history of isolation and speciation, with geologic events such as stream capture permitting some dispersal across basins (Swift et al., 1986). This uninterrupted evolutionary history is evident in the occurrence of numerous sister species across these basins (e.g., in the families Leuciscidae and Percidae), and is reflected in the fact that factors affecting the three dimensions of beta diversity differed among basins and between the Upstate and Lowlands. In contrast, the lowering of sea level during periods of glaciation including the Pleistocene would have allowed gene flow in coastal plain drainages as rivers merged further out on the Atlantic Slope, with flood events in these low topographic relief areas enhancing interbasin dispersal. Rohde et al. (2009) noted that similarity was higher in the faunas of the Santee and Pee Dee basins, and in the Savannah and ACE basins than between those two basin pairs, indicating these watershed pairs shared more recent connections permitting dispersal. As recently as 20,000 years ago, potential basin connections occurred as sea level retreated to roughly 120 m below current sea level, placing the coast near the edge of the Continental Shelf (Miller et al., 2020). Fluctuations in shoreline location through geologic time resulted in observed differences in taxonomic beta diversity, including evidence that anthropogenic effects are greater in the upland region due to less historic mixing of communities compared to below the fall line on the Atlantic Coastal Plain.

The rise and fall of the coast resulted in different geological properties found above and below the fall line. Providing context to the geographical properties of our study region is important for allowing potential comparison between studies in different regions of the world. The South Carolina lowlands are characterized by a mosaic of highly populated areas and unpopulated more agricultural areas (Homer et al., 2015). Lowland streams are dominated by harsh environmental conditions, frequent exposure to extreme weather, and poorly defined stream channels. These streams are also connected more closely to large mainstem rivers and reservoirs and share more species among watersheds due to historical sea level change (Marion et al., 2015; Miller et al., 2020; Denison et al., 2021b). Moreover, the species inhabiting the lowland streams tend to have more dispersal-associated traits and have populations that respond more to connectivity than local-scale abiotic factors (Midway and Peoples, 2019). Humans also historically connected lowland watersheds during the canal boom of the 1790s–1830s that saw over 3,200 km of cross-watershed canal construction, as well as widespread channelization and woody debris removal (Kapsch, 2010). These disturbances have likely contributed to a landscape legacy of increased connectivity, which we observe in the spatial signals in the lowland metacommunity across each diversity dimension (Robinson et al., 2002; Wenger et al., 2008). When comparing rivers, geomorphic variables including the width and gradient of streams dictated community structure with constrained river valleys having more varied rocky substrate and log jams to provide fish cover and wide river valleys showing increased riparian vegetation, slower water velocities, and less woody debris (Walters et al., 2003; Shields et al., 2021). In contrast to the lowlands, beta diversity in the upstate was affected more by anthropogenic and natural environmental factors. In upland regions of the eastern United States, stream fish community structure is often dominated by more specialized species in less disturbed watersheds, which transition to communities dominated by generalists as watersheds become more disturbed (Scott and Helfman, 2001; Petersen et al., 2021), which would result in high beta diversity between the two ends of this continuum. Within the upstate, increased disturbance associated with higher gradient streams can result in increased sediment transport, flash flooding, channel incision, and nutrient or pollution pulses detected by the anthropogenic signal (Marion et al., 2015). The disparity in the anthropogenic signal between the upstate and lowlands may be due to the sensitivity of upstate fishes as well as the lower connectivity for recolonization after disturbance events. The geomorphic characteristics describing streams above and below the fall line may account for the lack of a large environmental signal observed in the fall line spatial delineation compared to the statewide and watershed spatial delineations (Warren et al., 2000; Smogor and Angermeier, 2001).

While we explained a large portion of variation in stream fish beta diversity, we still had some unexplained variance, as well as anthropogenic variables with poor explanatory power. Unexplained variation could be due to unmeasured variables such as biotic interactions and behavior (D’Ambrosio et al., 2009), as well as natural and anthropogenic variables not included in our study such as local hydrologic variation (Marion et al., 2015). We may have seen a lack of an anthropogenic signal in some watershed metacommunities due to the branching complexity within river networks and the environmental heterogeneity provided by geomorphic variables offering a natural defense against human induced environmental changes (Terui et al., 2021). Unexplained variation could be due to homogenization from fish introductions, resulting in disruptions to ‘natural’ riverine processes structuring species traits and relatedness (Peoples et al., 2020). However, delineating metacommunities by geomorphic variables may allow for the anthropogenic mechanisms driving community structure to become more apparent.

In conclusion, we found that human modification of the land and riverscape is evident and important for structuring stream fish beta diversity. These effects operate both independently and in concert with the natural ecological variables that are commonly included in metacommunity analyses, as well as spatial gradients representing potential for species dispersal and homogenization of stream fish community structure across the landscape. Future studies should consider these anthropogenic effects alongside natural environmental effects and dispersal to obtain a clearer picture of regional beta diversity. We also found that user-defined a priori metacommunity delineations have important consequences for the outcomes and interpretation of metacommunity analyses. Analyses based on different spatial delineations highlighted the importance of synergistic spatial and environmental effects, regional differences in assembly rules, and basin-specific differences in the factors affecting different dimensions of stream fish beta diversity. Researchers should give careful consideration to this important methodological aspect, and seek spatial delineations that are ecologically relevant, given ecology of their focal taxa.

The original contributions presented in the study are included in the article/Supplementary material, further inquiries can be directed to the corresponding author.

Ethical review and approval was not required for the animal study because the data collection requiring IACUC approval was not completed in this study. The authors were given a dataset that on stream fish abundance.

LS: conceptualization, formal analysis, writing—original draft, and visualization. LB: resources and writing—review and editing. MS: writing—review and editing and resources. BP: supervision and writing—original draft. All authors contributed to the article and approved the submitted version.

This work was supported in part by the NIFA/USDA, under project number SC-1700599, and represents technical contribution number 7136 of the Clemson Experiment Station. Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement of the U.S. Government.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. ny product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.