R E S E A R CH P A P E R Diatom β‐diversity in streams increases with spatial scale and decreases with nutrient enrichment across regional to sub‐continental scales Thibault Leboucher1* | William R. Budnick2* | Sophia I. Passy2 | Sébastien Boutry1 | Aurélien Jamoneau1 | Janne Soininen3 | Wim Vyverman4 | Juliette Tison-Rosebery1 1Aquatic Ecosystems and Global Changes Research Unit, IRSTEA, Cestas, France 2Department of Biology, University of Texas at Arlington, Arlington, Texas, USA 3Department of Geosciences and Geography, University of Helsinki, Helsinki, Finland 4Biology Department, Protistology and Aquatic Ecology, Ghent University, Gent, Belgium Correspondence Thibault Leboucher, Aquatic Ecosystems and Global Changes Research Unit, IRSTEA, Cestas, France. Email: thibault.leboucher@irstea.fr Funding information National Science Foundation, Grant/Award Number: DEB-1745348; AFB (French Agency for Biodiversity); French National Research Agency (ANR) Editor: Simone Fattorini Abstract Aim: To quantify the relative contributions of local community assembly processes versus γ‐diversity to β‐diversity, and to assess how spatial scale and anthropogenic disturbance (i.e. nutrient enrichment) interact to dictate which driver dominates. Location: France and the United States. Time period: 1993–2011. Major taxa studied: Freshwater stream diatoms. Methods: β‐diversity along a nutrient enrichment gradient was examined across multiple spatial scales. β‐diversity was estimated using multi‐site Sørensen dissimilar- ity. We assessed the relative importance of specialists versus generalists using Fried- ley coefficient, and the contribution of local community assembly versus γ‐diversity to β‐diversity across spatial scales, with a null model. Finally, we estimated the response of β‐diversity to environmental and spatial factors by testing the correla- tions between community, environmental and geographical distance matrices with partial Mantel tests. Results: β‐diversity generally increased with spatial scale but the rate of increase depended on nutrient enrichment level. β‐diversity decreased significantly with increasing nutrient enrichment level due to the loss of specialist species. Local assembly was an important driver of β‐diversity especially under low nutrient enrich- ment. Significant partial Mantel correlations were observed between diatom β‐diver- sity and pure environmental distances under these conditions, highlighting the role of species sorting in local assembly processes. Conversely, in heavily enriched sites, only spatial distances were significantly correlated with β‐diversity, which indicated a substantial role of dispersal processes. Main conclusions: Nutrient concentration mediated the expected increase in β‐ diversity with spatial scales. Across spatial scales, β‐diversity was more influenced by local assembly processes rather than by γ‐diversity. Nutrient enrichment was associ- ated with an overall decline in diatom β‐diversity and a shift in assembly processes from species sorting to dispersal, notably due to the elimination of some specialists and their subsequent replacement by generalists. *Co-first authors. Received: 30 April 2018 | Revised: 26 November 2018 | Accepted: 11 December 2018 DOI: 10.1111/jbi.13517 Journal of Biogeography. 2019;1–11. wileyonlinelibrary.com/journal/jbi © 2019 John Wiley & Sons Ltd | 1 K E YWORD S β-diversity, γ-diversity, community assembly, diatom communities, dispersal processes, landscape windows approach, multi-scale, null model, species sorting 1 | INTRODUCTION One of the most interesting facets in the study of biodiversity is the consistent shift in species membership as one samples biological communities across space and time. The pattern of variation in local community composition (α‐diversity), given a regional species pool (γ‐diversity), was dubbed β‐diversity by Whittaker (1960, 1972) more than 50 years ago, and was originally defined in his work as ‘the extent of change in community composition in relation to a com- plex‐gradient of environment, or a pattern of environments’. The concept of β‐diversity has since morphed from a simple scalar (i.e. β = γ / α), into numerous mathematical definitions, which now include similarity/distance and variance‐based measures (Tuomisto, 2010a,b; Tuomisto & Ruokolainen, 2006). The development of new β‐diversity indices has thus given researchers powerful tools to study different aspects of Whittaker's concept of community variation (Socolar, Gil- roy, Kunin, & Edwards, 2016). Consequently, β‐diversity has a recent surge of interest because differences in many widely studied macroecological patterns (e.g. species–area relationships and species abundance distributions) are associated with variation in β‐diversity among different metacommunities and regions. β‐diversity is generally caused by an interplay of biogeographic, environmental and spatial gradients (Jamoneau, Passy, Soininen, Leboucher, & Tison‐Rosebery, 2018; Tonkin, Sundermann, Jähnig, & Haase, 2015; Viana et al., 2016). Species sorting (i.e. environmental filtering coupled with biological interactions), and dispersal processes are particularly important drivers of β‐diversity (Chase & Leibold, 2003; Soininen, 2007), however, their relative influence are scale‐ dependent (local vs. regional). Scale effects on measures of β‐diver- sity, and their drivers, have been of particular interest because research has shown that the average distance between sampling units, their spatial configuration, and the spatial extent of the sam- pling region all confound comparisons of β‐diversity between study areas (Conroy & Noon, 1996; Mac Nally, Fleishman, Bulluck, & Betrus, 2004; Noss, 1983; Wilson & Shmida, 1984). Quantitative multi‐scales approaches are thus recommended to better understand how local and regional processes affect β‐diversity (Huston, 1999; Mac Nally et al., 2004), and to also identify the appropriate spatial scales for studying mechanisms that structure communities (Mac Nally et al., 2004; Underwood & Chapman, 1996; Willis & Whittaker, 2002). However, few studies have actually implemented quantitative scale‐explicit frameworks to explore such relationships. β‐diversity tends to vary with spatial scale through different deter- ministic and stochastic processes. Species sorting, in particular, shapes community composition along environmental gradients (Lack, 1976; Leibold et al., 2004) and thus along spatial gradients when environmental heterogeneity increases with scale. Where environ- mental gradients persist at geological time scales (i.e. under natural conditions), local specialization leads to high regional‐level community dissimilarity (Gutiérrez‐Cánovas, Millán, Velasco, Vaughan, & Ormerod, 2013), and thus β‐diversity tends to increase with spatial scale. Disper- sal processes may also influence community dissimilarity by composi- tional homogenization (lower β‐diversity) or differentiation (higher β‐ diversity) as a function of the rate of dispersal (Bottin, Soininen, Alard, & Rosebery, 2016; Dong et al., 2016; Jamoneau et al., 2018; Soininen, 2007). However, it remains poorly understood how species sorting and dispersal processes interact to influence community variation across scales, given that their relative contributions are clearly depen- dent on the spatial context in which they are measured. A further complication in the study of β‐diversity drivers is that β‐diversity itself is constrained by variation in α‐diversity and γ‐diver- sity along the same environmental and spatial gradients (Kraft et al., 2011). Consequently, to assess the impact of local assembly pro- cesses (i.e. α‐diversity pattern) on β‐diversity, one must account for the influence of the regional species pool (i.e. γ‐diversity). Null mod- els have proven to be useful in this regard, and have seen wide- spread application in the analysis of β‐diversity pattern and its drivers (e.g. Chase, Kraft, Smith, Vellend, & Inouye, 2011; Kraft et al., 2011; Myers et al., 2013), but have not been widely adopted into quantitative multi‐scale frameworks. Scale dependence of β‐diversity drivers (i.e. species sorting, local assembly processes, dispersal and the regional species pool) are fur- ther influenced by anthropogenic disturbance, causing extensive impairment of terrestrial and freshwater ecosystems (Vörösmarty et al., 2010). Anthropogenic disturbances have been associated with taxonomic homogenization (loss of β‐diversity), but it remains unclear which are the key mechanisms and processes contributing to such homogenization. One general pattern often reported is the elimina- tion of sensitive species (McKinney & Lockwood, 1999; Olden & Poff, 2003, 2004; Socolar et al., 2016) and subsequent replacement by a few generalist species, usually sharing similar traits (Gutiérrez‐ Cánovas et al., 2013). Under natural conditions, β‐diversity increases with spatial scale owing to increased environmental heterogeneity and habitat specialization (Chase, 2014). However, when anthro- pogenic disturbances eliminate sensitive species, the species pool may contract and become dominated by generalist species with traits for tolerance. Nevertheless, it remains unclear whether anthro- pogenic disturbance, by impoverishing the species pool of specialists and increasing the prevalence of generalists, would attenuate the increase in β‐diversity with spatial scale. The replacement of specialists by generalists is likely the result of differential interactions of anthropogenic disturbance with the 2 | LEBOUCHER ET AL. underlying deterministic and stochastic processes operating on β‐ diversity. If anthropogenic disturbances eliminate specialists and pro- mote generalists with broader niches, one consequence is that the effect of species sorting might weaken and species composition would be mostly affected by stochastic processes, including disper- sal. However, deterministic (e.g. species sorting) and stochastic pro- cesses (e.g. dispersal) are scale dependent, and it remains unknown whether anthropogenic disturbance may shift the dominant pro- cesses with changing spatial scale. Our goal was to assess the relative contributions of local com- munity assembly processes versus γ‐diversity, and species sorting versus stochastic processes, in driving β‐diversity, and whether spa- tial scale and anthropogenic disturbance interact to determine which processes dominate. Diatoms represent an appropriate biological model for such research since their communities exhibit biogeo- graphical patterns (Soininen, Jamoneau, Rosebery, & Passy, 2016; Vanormelingen, Verleyen, & Vyverman, 2008; Vyverman et al., 2007), subject to both local and regional processes. We first explored how stream diatom β‐diversity responded to different levels of agricultural nutrient enrichment from local to sub‐continental scales. We then quantified how β‐diversity was driven by local pro- cesses versus γ‐diversity and by spatial versus environmental effects. We tested the following three hypotheses. 1. β-diversity in diatom communities should generally increase with scale due to greater environmental heterogeneity but such increase would be weak in nutrient-impacted sites if sensitive species are lost and the species pool becomes impoverished. 2. In conditions of impoverished species pool, the influence of γ- diversity on β-diversity would increase at the expense of local assembly processes. 3. If anthropogenic disturbances eliminate specialists and promote generalists, species sorting would have a weaker effect on com- munities under nutrient-impacted conditions and species compo- sition would be mainly constrained by spatial processes, including dispersal. 2 | MATERIALS AND METHODS 2.1 | Diatoms and environmental datasets Diatom datasets from France and the United States included a total of 3,391 spatially distinct localities (Table 1). Samples were collected primarily between May and October in order to reduce seasonal variability in species assemblages. All community and physico‐chemi- cal data were gathered in 2011 by the Water Agencies for the French sites and between 1993 and 2011 by the US Geological Sur- vey's National Water‐Quality Assessment (NAWQA) Program for the US sites. In both countries, coordinates for all samples were calcu- lated with WGS84 datum and projected using a Robinson projection (EPSG: 54030). Environmental variables included water pH, conduc- tivity (at 20°C, mS/cm), dissolved oxygen saturation (%), concentra- tions of total phosphorous (mg/L of 15P), orthophosphate (mg/L of 15P), ammonia (mg/L of 7N), nitrate (mg/L of 7N) and nitrite (mg/L of 7N). Environmental data were not recorded at the time of commu- nity sampling; therefore, we used the median values of all environ- mental variable measurements obtained during the 60 days before and the 15 days after the diatom sampling date. We also included altitude (m, 1 arc‐second) and river slope (%, 1 arc‐second), drawn from the Shuttle Radar Topography Mission (Farr et al., 2007) and climatic variables, that is annual precipitations (mm) and annual air temperature (°C), drawn from the WorldClim database (Hijmans, Cameron, Parra, Jones, & Jarvis, 2005). Climate data covered the years 1950–2000 and the average values were included in the analyses. In France, diatoms were sampled with a standardized protocol, in line with the European standards (EN 13946, European Commission). In the United States, diatoms were sampled following standardized NAWQA protocols. In both countries, cells were identified at 1,000× magnification by examining permanent slides of cleaned diatom frus- tules (400 valves per slide in France and 600 valves per slide in the United States). The French and the US diatom datasets were taxo- nomically homogenized at the species level using Omnidia 5.3 and different databases (http://www.algaebase.org/, https://westerndia toms.colorado.edu/). 2.2 | Level of nutrient enrichment To classify the sampling stations along a eutrophication gradient, we first performed principal components analysis (PCA) on power‐ transformed nutrient concentrations (nitrate, nitrite, ammonia, orthophosphate and total phosphorous). We used Box‐Cox maxi- mum likelihood estimation (Box & Cox, 1964) to decide the best power‐transformation for approximating normality. We retained only the components with eigenvalues greater than those ran- domly generated by a broken‐stick model (Jackson, 1993; Legen- dre & Legendre, 2012). Following this procedure, we found that only the first principal component axis (PC1) was significant. PC1 explained 55% of the total variance and represented a gradient of nutrient enrichment (Figure 1). We used an ascending hierarchical classification (AHC, Lebart, Morineau, & Piron, 2000) of the PC1 scores to classify sites into classes of nutrient enrichment based on their Euclidean distances. AHC identified three homogeneous classes optimizing the average silhouette width (Rousseeuw, 1987), that is unimpacted, slightly impacted and heavily impacted sites (Table 1). TABLE 1 Stream diatom datasets from France and the United States No. of unimpacted sites No. of slightly impacted sites No. of heavily impacted sites No. of species Years of sampling France 636 1,163 413 901 2011 United States 555 358 165 946 1993–2011 LEBOUCHER ET AL. | 3 2.3 | Generalist and specialist species To achieve a classification of species along a generalist‐specialist gra- dient, we implemented the approach by Fridley, Vandermast, Kup- pinger, Manthey, and Peet (2007), which is based on species co‐ occurrences and allows species characterization without measuring an N‐dimensional niche for each species. This approach derives from a β‐diversity statistic and relies on the assumption that generalists should co‐occur with many species, whereas specialists should co‐ occur with relatively few species (Fridley et al., 2007), independently of the species pool size (Zelený, 2009). The continuous metric (θ‐value or Fridley coefficient) obtained for each species reflects the relative species niche width: low θ‐values denote a confined niche (specialist species), whereas high θ‐values denote a broader niche (generalist species). We measured the mean θ‐value at each site, weighted by the abundance of each species present at that site, to estimate the relative importance of specialists versus generalists both in France and in the United States, following the equation: θsite ¼ ∑ n 1 θi  AbiAbtot where n is the number of species in the site, θi is the θ‐value of spe- cies i, Abi is the abundance of species i in the site and Abtot is the total abundance of all species in the site. 2.4 | Landscape windows approach β‐diversity among sites was examined across multiple spatial scales with square polygons (i.e. landscape windows), using R 3.4.2 (R Core Team, 2017). Spatial scales are defined by the size of the landscape windows and varied between 200 × 200 km and 2,000 × 2,000 km (0.4 × 105 and 40 × 105 square km respectively). Latitudinal transects at 50 km intervals were used to place the centre of each window. Windows were aligned along these transects at 50 km distances between their centres, resulting in 9,006 possible and potentially over- lapping windows at each scale (Figure 2). We selected windows that contained at least 25 unimpacted, 25 slightly impacted and 25 heavily impacted sites distributed over at least 80% of the window's area to ensure equal sampling effort across categories (see Appendix S1). Among 9,006 possible landscape windows, 2,514 in both France and the United States met our selection criteria. There were no landscape windows below 6.4 × 105 square km in the United States because of the low density of sampling localities. For simplicity, each spatial scale was denoted by the width of its corresponding window, for example the 200 × 200 km window (0.4 × 105 square km) was denoted as ‘200 km window’. The creation of landscape windows in R 3.4.2 (R Core Team, 2017) is provided in Appendix S3. In contrast to a simple distance‐decay approach, our landscape windows approach explicitly accounts for community dissimilarities and spatial distribution of sites and thus renders the spatial‐scaling pattern of β‐diversity more infor- mative, robust and comparable among studies. 2.5 | β‐diversity calculations We estimated the β‐diversity of each window using the multi‐site total β‐diversity of Baselga‐family decomposition of the Sørensen dissimilarity coefficient for species presence–absence data (Baselga, 2010) obtained with the beta.div.comp function (R‐package: ‘adespa- tial’, Dray et al., 2018) using the equation: β ¼ ½∑i x2)−p(x1 < x2). Comparing the degree of overlap between the two groups indicates whether a signifi- cant difference is due to an effective ecological process or to sample size alone. Romano, Kromrey, Coraggio, and Skowronek (2006) inter- preted the magnitude of differences between groups as follows: negli- gible for |d| < 0.147, small for 0.147 < |d| < 0.33, medium for 0.33 < |d| < 0.474 and large for higher |d| values. In this study, we considered |d| > 0.33 as a threshold for significance (see Appendix S2). All data analyses were performed with R 3.4.2 (R Core F IGURE 2 Schematic representation of the landscape windows approach. (a) The centres of the windows were aligned every 50 km along latitudinal transects in the study area. (b) We selected windows that contained at least 25 sites for each nutrient level. (c) For each nutrient level, sites must be distributed over at least 80% of the window's area LEBOUCHER ET AL. | 5 Team, 2017), using the ‘vegan’ and ‘adespatial’ packages (Dray et al., 2018; Oksanen et al., 2018). 3 | RESULTS 3.1 | The effect of scale on β‐diversity values Sørensen dissimilarities ranged between 0.27 (heavily impacted sites) and 0.37 (unimpacted sites) in France and between 0.30 (heavily impacted sites) and 0.38 (unimpacted sites) in the United States. Although the minimum and maximum values were similar between the two countries, β‐diversity patterns across spatial scales were different. β‐diversity increased with spatial scale regardless of enrichment level (Figure 3), however, the rate of increase differed across nutrient enrichment levels in both France and the United States. In France, β‐diversity values in unimpacted sites increased significantly with spatial scale, whereas in slightly and heavily impacted sites, β‐diversity increased to about 400 km. In the United States, β‐diversity increased significantly with scale to about 1,500 km across enrichment levels, but tapered off between 1,500 and 2,000 km. Only the 800 km scale was com- mon to France and the United States, and it is notable that the corresponding β‐diversity values were clearly distinct, especially for unimpacted sites. 3.2 | Impact of nutrient enrichment on β‐diversity and specialist species β‐diversity decreased significantly with nutrient enrichment and even a slight impact had a strong effect on communities (Figure 3). More- over, in France β‐diversity tended to plateau at 400 km in impacted sites (slightly and heavily), but continued to increase in unimpacted ones. In the United States, β‐diversity differences were even more pronounced. For unimpacted sites, β‐diversity increased steeply between 800 and 1,200 km, whereas at the same scales, β‐diversity in impacted sites increased only slightly. As in France, β‐diversity in the United States significantly declined with the level of impact, except at 1,100 and 1,200 km where there was no difference between slightly and heavily impacted sites. The proportion of specialist species declined among heavily impacted sites, given the significant increase in θ‐values with nutri- ent enrichment level in both France and the United States (Figure 4). Although we observed that for the French data, there was a signifi- cant difference between unimpacted sites and slightly impacted sites, such a difference was absent for the US data. 3.3 | Relative contributions of local community assembly processes versus γ‐diversity on β‐deviations The sign and magnitude of β‐deviation indicated that β‐diversity among diatom communities was driven primarily by local community assembly processes rather than by γ‐diversity, regardless of enrich- ment level and spatial scale (Figure 5). Similar to β‐diversity, β‐devia- tion varied with scale. In France, β‐deviation in unimpacted sites increased significantly to 500 km. In slightly impacted sites, no dif- ference in β‐deviation was observed across scales, and in heavily impacted sites, only the first two scales exhibited lower β‐deviation values than the larger scales. In the United States, across enrichment levels, β‐deviation values increased significantly to 1,500 km (im- pacted sites) or 1,600 km (unimpacted sites). Nutrient enrichment had a strong negative effect on β‐deviation, especially at large spatial scales (Figure 5). In France, β‐deviation sig- nificantly decreased from unimpacted to heavily impacted sites. In the United States, β‐deviation in unimpacted sites was always significantly higher than in impacted sites (both slightly and heavily). However, except for the 800 km scale, there were no differences between β‐deviation values in slightly impacted and heavily impacted sites. 3.4 | Characterization of the local community assembly processes Partial Mantel test results diverged according to impact level (Figure 6). However, there were only very few significant differences between spatial scales and we have not considered them further (see Scale (km) β− di ve rs ity 200 400 600 800 1000 1200 1400 1600 1800 2000 0. 28 0. 3 0. 32 0. 34 0. 36 0. 38 France US F IGURE 3 Mean β‐diversity values (symbols) and SD (bands) measured in unimpacted conditions (round symbols), slightly impacted conditions (square symbols) and heavily impacted conditions (triangle symbols) across spatial scales 6 | LEBOUCHER ET AL. Appendix S2). In unimpacted sites, there was a significant correlation between diatom β‐diversity and pure environmental distance. The pure spatial influence was significant in France at 200 and 300 km scales, and in the United States between 800 and 1,500 km. In contrast, in heavily impacted sites, only pure spatial effect was significant at small and intermediate scales (200–500 km) in France, and between 1,000 Fr id le y co ef fic ie nt 0 1 2 3 4 5 6 7 8 9 10 Unimpacted Slightly impacted Heavily impacted France *** *** *** 0 1 2 Unimpacted Slightly impacted Heavily impacted US * F IGURE 4 Boxplot of the mean θ‐ values (Fridley coefficient) measured in unimpacted, slightly impacted and heavily impacted conditions. Differences between nutrient levels were tested with Kruskal– Wallis tests followed up by Nemenyi post hoc tests. Stars represent significant differences between nutrient levels (p < 0.001: ‘***’; and p < 0.05: ‘*’) Scale (km) β− de vi at io n 200 400 600 800 1000 1200 1400 1600 1800 2000 55 65 75 85 95 10 5 11 5 12 5 13 5 France US F IGURE 5 Mean β‐deviation values (symbols) and SD (bands) measured in unimpacted conditions (round symbols), slightly impacted conditions (square symbols) and heavily impacted conditions (triangle symbols) across spatial scales H eavily im pacted conditions S lightly im pacted conditions U nim pacted conditions 200 400 600 800 1000 1200 1400 1600 1800 2000 0 0. 2 0. 4 France US M an te l s ta tis tic (R ) 200 400 600 800 1000 1200 1400 1600 1800 2000 0 0. 2 0. 4 France US Scale (km) 200 400 600 800 1000 1200 1400 1600 1800 2000 0 0. 2 0. 4 France US F IGURE 6 Mean partial Mantel coefficients (symbols) and SD (bands) of pure environmental (square symbols) and pure spatial effect (triangle symbols) on β‐diversity along a spatial gradient. Filled symbols represent significant relationships (Cliff's |d| > 0.33) and open symbols, non‐ significant ones (Cliff's |d| ≤ 0.33) LEBOUCHER ET AL. | 7 and 2,000 km in the United States. Pure environmental effect, on the other hand, was non‐significant in both France and the United States. Results for slightly impacted sites differed between France and the United States. In France, we observed significant correlations between diatom β‐diversity and pure environmental distances between 300 and 700 km. Notably, the pure environmental effect in France was signifi- cantly greater in unimpacted sites than in slightly impacted sites regardless of the scale. A pure spatial influence was significant at the 200 km scale only. In the United States, pure environmental effect was not significant in slightly impacted sites except at the 2,000 km scale, whereas pure spatial influence was significant across all scales. 4 | DISCUSSION This work provided information on diatom β‐diversity patterns and drivers along a nutrient enrichment gradient and across multiple spa- tial scales. First, we confirmed that β‐diversity increases with spatial scale, but nutrient enrichment clearly modulated this increase. Sec- ond, in both continents, local assembly processes had a stronger contribution to β‐diversity than γ‐diversity regardless of scale and nutrient enrichment level. However, nutrient enrichment played a clear role in determining whether species sorting and/or dispersal dominated assembly processes along the spatial gradient. We discuss the implications of these findings below. 4.1 | β‐diversity across scales and nutrient levels We confirmed our first prediction that β‐diversity of diatom commu- nities increases with spatial scale, which is generally consistent with prior literature (Gabriel, Roschewitz, Tscharntke, & Thies, 2006; Mar- tiny, Eisen, Penn, Allison, & Horner‐Devine, 2011). We also revealed that regardless of spatial scale, β‐diversity was dependent on nutri- ent level and that greater enrichment more strongly depressed β‐ diversity. Anthropogenic and natural stress gradients differ in type and variability, which ultimately drive richness and spatial distribu- tion patterns of sensitive and tolerant species across scales (Clavel, Julliard, & Devictor, 2011). Acute and persistent anthropogenic stress, such as eutrophication, may offer insufficient time for adapta- tion by specialists and potentially eliminate them from the species pool (Gutiérrez‐Cánovas et al., 2013; Kaspari, Stevenson, Shik, & Kerekes, 2010). The loss of specialists benefits subsequent coloniza- tion by more physiologically plastic generalists, which are normally poor competitors for resources and niche space compared to special- ists (Bolnick, Svanbäck, Araújo, & Persson, 2007; Olden & Rooney, 2006). Our data thus showed that the strength of spatial depen- dence of β‐diversity was likely regulated by how strongly anthro- pogenic gradients shifted the species pool towards greater prevalence of species with generalist versus specialist traits. Analyses of the Fridley coefficient, θ, indeed indicated that these patterns were attributable to the fact that among impacted streams, relatively fewer new species (particularly those with specialist traits) were added to the pool compared to unimpacted streams. Our results thus support widely reported observations that agriculture‐driven nutrient enrichment is a major driver biotic homogenization (Blüthgen et al., 2016; Ekroos, Heliölä, & Kuussaari, 2010; Karp et al., 2012; Vallejos, Padial, & Vitule, 2016) but further demonstrate that the mechanism of this homogenization is through changes in trait composition. However, these findings run counter to studies that reported a positive correlation of β‐diversity with productivity (e.g. Chase, 2010). Nutrient levels in our study systems were substantially high (i.e. max TN ~50 mg/L and max TP ~2 mg/L versus max TN ~2 mg/L and max TP ~0.2 mg/L in Chase's study), and might have exceeded the physiological thresholds of oligotrophic and mesotrophic species, thus resulting in diminished species pool and low β‐diversity. 4.2 | Roles of local assembly versus γ‐diversity in driving β‐diversity The role of γ‐diversity compared to local assembly in driving β‐diver- sity generally increased with impact, given the lower β‐deviation val- ues at higher nutrients levels in both datasets. Species pools tended to be more impoverished with increased impact, and dominated by gen- eralists, thus smaller β‐deviation at higher impact was likely indicative of increased stochastic sampling of generalists from the regional spe- cies pool (Karp et al., 2012; Kraft et al., 2011). Therefore, under stress- ful conditions, β‐diversity was more strongly constrained by regional processes, affecting the distribution of specialists and the size of the regional species pool. A similar finding was reported for diatoms in an acid‐impacted region, where acid stress increased the importance of the regional species pool for β‐diversity due to the elimination of acid‐ sensitive species (Pound, Lawrence, & Passy, 2018). Furthermore, we generally observed weak scale dependence of β‐deviation in heavily impacted sites. This suggested that the relative roles of γ‐diversity and local assembly processes may be decoupled from spatial scale if anthropogenic disturbance is high. Our results disagree to a certain extent with prior research, reporting that the relative importance of community assembly processes in wood plant communities is strongly scale dependent along elevational gradients (Tello et al., 2015). It is thus possible that the scale dependence of the mechanisms underlying β‐diversity may vary along natural (e.g. elevation) versus anthropogenic gradients (here, eutrophication). 4.3 | Relative influence of species sorting versus dispersal processes With heavy nutrient enrichment, the correlation between community dissimilarity and pure environmental distance virtually disappeared whereas the correlation between community dissimilarity and spatial distance increased. This suggested that nutrient enrichment weakened the influence of species sorting and dispersal processes became a more important driver of community composition. Mass effects and dispersal limitation are the two principal dispersal processes that can generate spatial structure of community dissimilarity, with mass effects treated as a community homogenizer (Mouquet & Loreau, 2003; Shmida & Wilson, 1985) and dispersal limitation, as a 8 | LEBOUCHER ET AL. community differentiator (Heino et al., 2015; Hubbell, 2001; Jamo- neau et al., 2018). However, it is difficult to disentangle these two pro- cesses with Partial Mantel tests because they both can be inferred from correlations between spatial variables and community composi- tion. It is also possible that the spatial effects are due to unmeasured environmental factors, but this may be less of a problem in our study, given that we measured dominant environmental gradients. The results of the partial Mantel tests indicated that species sort- ing was the dominant mechanism of local assembly in unimpacted sites. In the United States, the strength of species sorting increased with spatial scale. Özkan, Svenning, Jeppesen, and Webb (2013), working with bird communities, also documented stronger species sorting compared to dispersal processes across spatial scales. Inter- estingly, local assembly and species sorting increased only weakly with spatial scale in the French data. A likely reason for this differ- ence can be the biogeographical context of the two datasets, with longer latitudinal and longitudinal gradients (i.e. study extent) and greater total richness in the US dataset. In sum, the increase in β‐diversity with spatial scale was appar- ently stronger when species sorting processes dominated but weaker when dispersal processes prevailed. Severity of anthropogenic impact was also important in determining whether species sorting (unimpacted conditions) or dispersal processes (heavily impacted) dominated across spatial scales. However, under slightly enriched conditions, spatial correlations were significant only among US sites regardless of spatial scale whereas environmental correlations were significant for French sites. Thus, the severity of nutrient enrichment, along with biogeographical context, may influence the roles of envi- ronmental and dispersal processes in structuring community dissimi- larity and ultimately how fast β‐diversity increases with scale. ACKNOWLEDGEMENTS Support from the National Science Foundation (grant NSF DEB‐ 1745348 to S.P.) is gratefully acknowledged. We also thank AFB (French Agency for Biodiversity) and all French Water Agencies for data contribution and financial support. This document has also been carried out with financial support from the French National Research Agency (ANR) in the frame of the Investments for the future Pro- gramme, within the Cluster of Excellence COTE (ANR‐10‐LABX‐45). We also thank S. Fattorini, and the three reviewers, for helpful com- ments which improved the clarity of this work. DATA ACCESSIBILITY All environmental and diatom data for the United States are avail- able: https://water.usgs.gov/nawqa/. All environmental and diatom data for France are the property of the French Water Agencies. ORCID Thibault Leboucher http://orcid.org/0000-0003-0353-8896 REFERENCES Baselga, A. (2010). Partitioning the turnover and nestedness components of beta diversity. Global Ecology and Biogeography, 19, 134–143. https://doi.org/10.1111/j.1466-8238.2009.00490.x Bennett, J. R., & Gilbert, B. (2016). Contrasting beta diversity among regions: How do classical and multivariate approaches compare? Glo- bal Ecology and Biogeography, 25, 368–377. https://doi.org/10.1111/ geb.12413 Blüthgen, N., Simons, N. K., Jung, K., Prati, D., Renner, S. C., Boch, S., … Gossner, M. M. (2016). Land use imperils plant and animal commu- nity stability through changes in asynchrony rather than diversity. Nature Communications, 7, 10697. https://doi.org/10.1038/ncomm s10697 Bolnick, D. I., Svanbäck, R., Araújo, M. S., & Persson, L. (2007). Compara- tive support for the niche variation hypothesis that more generalized populations also are more heterogeneous. Proceedings of the National Academy of Sciences (USA), 104, 10075–10079. https://doi.org/10. 1073/pnas.0703743104 Bottin, M., Soininen, J., Alard, D., & Rosebery, J. (2016). Diatom cooccur- rence shows less segregation than predicted from niche modeling. PLoS ONE, 11, e0154581. https://doi.org/10.1371/journal.pone.0154581 Box, G. E. P., & Cox, D. R. (1964). An analysis of transformations. Journal of the Royal Statistical Society Series B (Methodological), 26, 211–252. https://doi.org/10.1111/j.2517-6161.1964.tb00553.x Chase, J. M. (2010). Stochastic community assembly causes higher biodi- versity in more productive environments. Science, 328, 1388–1391. https://doi.org/10.1126/science.1187820 Chase, J. M. (2014). Spatial scale resolves the niche versus neutral theory debate. Journal of Vegetation Science, 25, 319–322. https://doi.org/ 10.1111/jvs.12159 Chase, J. M., Kraft, N. J. B., Smith, K. G., Vellend, M., & Inouye, B. D. (2011). Using null models to disentangle variation in community dis- similarity from variation in α‐diversity. Ecosphere, 2, 1–11. https://doi. org/10.1890/ES10-00117.1 Chase, J. M., & Leibold, M. A. (2003). Ecological niches: Linking classical and contemporary approaches. Chicago, IL: University of Chicago Press. https://doi.org/10.7208/chicago/9780226101811.001.0001 Chase, J. M., & Myers, J. A. (2011). Disentangling the importance of eco- logical niches from stochastic processes across scales. Philosophical Transactions of the Royal Society of London B: Biological Sciences, 366, 2351–2363. https://doi.org/10.1098/rstb.2011.0063 Clavel, J., Julliard, R., & Devictor, V. (2011). Worldwide decline of special- ist species: Toward a global functional homogenization? Frontiers in Ecology and the Environment, 9, 222–228. https://doi.org/10.1890/ 080216 Cliff, N. (1993). Dominance statistics: Ordinal analyses to answer ordinal questions. Psychological Bulletin, 114, 494–509. https://doi.org/10. 1037/0033-2909.114.3.494 Conroy, M. J., & Noon, B. R. (1996). Mapping of species richness for con- servation of biological diversity: Conceptual and methodological issues. Ecological Applications, 6, 763–773. https://doi.org/10.2307/ 2269481 Dong, X., Li, B., He, F., Gu, Y., Sun, M., Zhang, H., … Cai, Q. (2016). Flow directionality, mountain barriers and functional traits determine dia- tom metacommunity structuring of high mountain streams. Scientific Reports, 6, 24711. https://doi.org/10.1038/srep24711 Dray, S., Blanchet, G., Borcard, D., Clappe, S., Guenard, G., Jombart, T., … Wagner, H. H. (2018). adespatial: Multivariate multiscale spatial analysis. R package version 0.1-1. Retrieved from https://CRAN.R- project.org/package=adespatial. Ekroos, J., Heliölä, J., & Kuussaari, M. (2010). Homogenization of lepi- dopteran communities in intensively cultivated agricultural land- scapes. Journal of Applied Ecology, 47, 459–467. https://doi.org/10. 1111/(ISSN)1365-2664 LEBOUCHER ET AL. | 9 Farr, T. G., Rosen, P. A., Caro, E., Crippen, R., Duren, R., Hensley, S., … Alsdorf, D. (2007). The shuttle radar topography mission. Reviews of Geophysics, 45, RG2004. https://doi.org/10.1029/2005RG000183 Fridley, J. D., Vandermast, D. B., Kuppinger, D. M., Manthey, M., & Peet, R. K. (2007). Co‐occurrence based assessment of habitat generalists and specialists: A new approach for the measurement of niche width. Journal of Ecology, 95, 707–722. https://doi.org/10.1111/j.1365- 2745.2007.01236.x Gabriel, D., Roschewitz, I., Tscharntke, T., & Thies, C. (2006). Beta diver- sity at different spatial scales: Plant communities in organic and con- ventional agriculture. Ecological Applications, 16, 2011–2021. https://doi.org/10.1890/1051-0761(2006)016[2011:BDADSS]2.0. CO;2 Gilbert, B., & Bennett, J. R. (2010). Partitioning variation in ecological communities: Do the numbers add up? Journal of Applied Ecology, 47, 1071–1082. https://doi.org/10.1111/j.1365-2664.2010.01861.x Greenwood, P. E., & Nikulin, M. S. (1996). A guide to chi-squared testing (vol. 280). New York, NY: John Wiley & Sons. Gutiérrez-Cánovas, C., Millán, A., Velasco, J., Vaughan, I. P., & Ormerod, S. J. (2013). Contrasting effects of natural and anthropogenic stres- sors on beta diversity in river organisms. Global Ecology and Biogeog- raphy, 22, 796–805. https://doi.org/10.1111/geb.12060 Heino, J., Melo, A. S., Siqueira, T., Soininen, J., Valanko, S., & Bini, L. M. (2015). Metacommunity organisation, spatial extent and dispersal in aquatic systems: Patterns, processes and prospects. Freshwater Biol- ogy, 60, 845–869. https://doi.org/10.1111/fwb.12533 Hijmans, R. J., Cameron, S. E., Parra, J. L., Jones, P. G., & Jarvis, A. (2005). Very high resolution interpolated climate surfaces for global land areas. International Journal of Climatology, 25, 1965–1978. https://doi.org/10.1002/(ISSN)1097-0088 Hubbell, S. P. (2001). The unified neutral theory of biodiversity and biogeog- raphy (MPB-32). Princeton, NJ: Princeton University Press. Huston, M. A. (1999). Local processes and regional patterns: Appropriate scales for understanding variation in the diversity of plants and ani- mals. Oikos, 86, 393–401. https://doi.org/10.2307/3546645 Jackson, D. A. (1993). Stopping rules in principal components analysis: A comparison of heuristical and statistical approaches. Ecology, 74, 2204–2214. https://doi.org/10.2307/1939574 Jamoneau, A., Passy, S. I., Soininen, J., Leboucher, T., & Tison-Rosebery, J. (2018). Beta diversity of diatom species and ecological guilds: Response to environmental and spatial mechanisms along the stream watercourse. Freshwater Biology, 63, 62–73. https://doi.org/10.1111/ fwb.12980 Karp, D. S., Rominger, A. J., Zook, J., Ranganathan, J., Ehrlich, P. R., & Daily, G. C. (2012). Intensive agriculture erodes β‐diversity at large scales. Ecology Letters, 15, 963–970. https://doi.org/10.1111/j.1461- 0248.2012.01815.x Kaspari, M., Stevenson, B. S., Shik, J., & Kerekes, J. F. (2010). Scaling community structure: How bacteria, fungi, and ant taxocenes differ- entiate along a tropical forest floor. Ecology, 91, 2221–2226. https://doi.org/10.1890/09-2089.1 Kraft, N. J. B., Comita, L. S., Chase, J. M., Sanders, N. J., Swenson, N. G., Crist, T. O., … Myers, J. A. (2011). Disentangling the drivers of β diversity along latitudinal and elevational gradients. Science, 333, 1755–1758. https://doi.org/10.1126/science.1208584 Lack, D. (1976). Island biology; illustrated by the land birds of Jamaica. Oxford, UK: Blackwell Scientific Publications. https://doi.org/10. 2307/2418771 Lebart, L., Morineau, A., & Piron, M. (2000). Statistique exploratoire multi- dimensionnelle. Paris, France: Dunod. Legendre, P., & De Cáceres, M. (2013). Beta diversity as the variance of community data: Dissimilarity coefficients and partitioning. Ecology Letters, 16, 951–963. https://doi.org/10.1111/ele.12141 Legendre, P., & Legendre, L. F. J. (2012). Numerical ecology. Amsterdam: The Netherlands: Elsevier. Leibold, M. A., Holyoak, M., Mouquet, N., Amarasekare, P., Chase, J. M., Hoopes, M. F., … Gonzalez, A. (2004). The metacommunity concept: A framework for multi‐scale community ecology. Ecology Letters, 7, 601–613. https://doi.org/10.1111/j.1461-0248.2004.00608.x Mac Nally, R., Fleishman, E., Bulluck, L. P., & Betrus, C. J. (2004). Com- parative influence of spatial scale on beta diversity within regional assemblages of birds and butterflies. Journal of Biogeography, 31, 917–929. https://doi.org/10.1111/j.1365-2699.2004.01089.x Martiny, J. B. H., Eisen, J. A., Penn, K., Allison, S. D., & Horner-Devine, M. C. (2011). Drivers of bacterial β‐diversity depend on spatial scale. Proceedings of the National Academy of Sciences, 108, 7850–7854. https://doi.org/10.1073/pnas.1016308108 McKinney, M. L., & Lockwood, J. L. (1999). Biotic homogenization: A few winners replacing many losers in the next mass extinction. Trends in Ecology & Evolution, 14, 450–453. https://doi.org/10.1016/S0169- 5347(99)01679-1 Mouquet, N., & Loreau, M. (2003). Community patterns in source‐sink metacommunities. The American Naturalist, 162, 544–557. https://d oi.org/10.1086/378857 Myers, J. A., Chase, J. M., Jiménez, I., Jørgensen, P. M., Araujo-Murakami, A., Paniagua-Zambrana, N., & Seidel, R. (2013). Beta‐diversity in tem- perate and tropical forests reflects dissimilar mechanisms of commu- nity assembly. Ecology Letters, 16, 151–157. https://doi.org/10.1111/ ele.12021 Noss, R. F. (1983). A regional landscape approach to maintain diversity. BioScience, 33, 700–706. https://doi.org/10.2307/1309350 Oksanen, J., Blanchet, G., Friendly, M., Kindt, R., Legendre, P., McGlinn, D., … Wagner, H. (2018). vegan: Community Ecology Package. R package version 2.4-6. Retrieved from https://CRAN.R-project.org/ package=vegan. Olden, J. D., & Poff, N. L. (2003). Toward a mechanistic understanding and prediction of biotic homogenization. The American Naturalist, 162, 442–460. https://doi.org/10.1086/378212 Olden, J. D., & Poff, N. L. (2004). Ecological processes driving biotic homogenization: Testing a mechanistic model using fish faunas. Ecol- ogy, 85, 1867–1875. https://doi.org/10.1890/03-3131 Olden, J. D., & Rooney, T. P. (2006). On defining and quantifying biotic homogenization. Global Ecology and Biogeography, 15, 113–120. https://doi.org/10.1111/j.1466-822X.2006.00214.x Özkan, K., Svenning, J., Jeppesen, E., & Webb, T. (2013). Environmental species sorting dominates forest‐bird community assembly across scales. Journal of Animal Ecology, 82, 266–274. https://doi.org/10. 1111/j.1365-2656.2012.02019.x Pound, K. L., Lawrence, G. B., & Passy, S. I. (2018). Beta diversity response to stress severity and heterogeneity in sensitive versus tol- erant stream diatoms. Diversity and Distributions, 00, 1–11. https:// doi.org/10.1111/ddi.12865 R Core Team. (2017). R: A language and environment for statistical com- puting. Vienna, Austria: R Foundation for Statistical Computing. Retrieved from https://www.R-project.org/. Romano, J., Kromrey, J. D., Coraggio, J., & Skowronek, J. (2006). Appro- priate statistics for ordinal level data: Should we really be using t‐test and Cohen's d for evaluating group differences on the NSSE and other surveys, Annual Meeting of the Florida Association of Institutional Research, pp. 1–33. Arlington, Virginia, October 14‐17, 2006. Rousseeuw, P. J. (1987). Silhouettes: A graphical aid to the interpretation and validation of cluster analysis. Journal of Computational and Applied Mathematics, 20, 53–65. https://doi.org/10.1016/0377-0427(87) 90125-7 Shmida, A., & Wilson, M. V. (1985). Biological determinants of species diversity. Journal of Biogeography, 12, 1–20. https://doi.org/10.2307/ 2845026 Socolar, J. B., Gilroy, J. J., Kunin, W. E., & Edwards, D. P. (2016). How should beta‐diversity inform biodiversity conservation? Trends in Ecol- ogy & Evolution, 31, 67–80. https://doi.org/10.1016/j.tree.2015.11.005 10 | LEBOUCHER ET AL. Soininen, J. (2007). Environmental and spatial control of freshwater dia- toms—a review. Diatom Research, 22, 473–490. https://doi.org/10. 1080/0269249X.2007.9705724 Soininen, J., Jamoneau, A., Rosebery, J., & Passy, S. I. (2016). Global pat- terns and drivers of species and trait composition in diatoms. Global Ecology and Biogeography, 25, 940–950. https://doi.org/10.1111/geb. 12452 Tecchio, S., Chaalali, A., Raoux, A., Tous Rius, A., Lequesne, J., Girardin, V., … Niquil, N. (2016). Evaluating ecosystem‐level anthropogenic impacts in a stressed transitional environment: The case of the Seine estuary. Ecological Indicators, 61, 833–845. https://doi.org/10.1016/ j.ecolind.2015.10.036 Tello, J. S., Myers, J. A., Macía, M. J., Fuentes, A. F., Cayola, L., Arellano, G., … Jørgensen, P. M. (2015). Elevational gradients in β‐diversity reflect variation in the strength of local community assembly mecha- nisms across spatial scales. PLoS ONE, 10, e0121458. https://doi.org/ 10.1371/journal.pone.0121458 Tonkin, J. D., Sundermann, A., Jähnig, S. C., & Haase, P. (2015). Environ- mental controls on river assemblages at the regional scale: An appli- cation of the elements of metacommunity structure framework. PLoS ONE, 10, e0135450. https://doi.org/10.1371/journal.pone.0135450 Tuomisto, H. (2010a). A diversity of beta diversities: Straightening up a concept gone awry. Part 1. Defining beta diversity as a function of alpha and gamma diversity. Ecography, 33, 2–22. https://doi.org/10. 1111/j.1600-0587.2009.05880.x Tuomisto, H. (2010b). A diversity of beta diversities: Straightening up a concept gone awry. Part 2. Quantifying beta diversity and related phenomena. Ecography, 33, 23–45. https://doi.org/10.1111/j.1600- 0587.2009.06148.x Tuomisto, H., & Ruokolainen, K. (2006). Analyzing or explaining beta diversity? Understanding the targets of different methods of analysis. Ecology, 87, 2697–2708. https://doi.org/10.1890/0012-9658(2006) 87[2697:AOEBDU]2.0.CO;2 Underwood, A. J., & Chapman, M. G. (1996). Scales of spatial patterns of distribution of intertidal invertebrates. Oecologia, 107, 212–224. https://doi.org/10.1007/BF00327905 Vallejos, M. A. V., Padial, A. A., & Vitule, J. R. S. (2016). Human‐induced landscape changes homogenize Atlantic forest bird assemblages through nested species loss. PLoS ONE, 11, e0147058. https://doi. org/10.1371/journal.pone.0147058 Vanormelingen, P., Verleyen, E., & Vyverman, W. (2008). The diversity and distribution of diatoms: From cosmopolitanism to narrow ende- mism. Biodiversity and Conservation, 17, 393–405. https://doi.org/10. 1007/s10531-007-9257-4 Viana, D. S., Figuerola, J., Schwenk, K., Manca, M., Hobæk, A., Mjelde, M., … Santamaría, L. (2016). Assembly mechanisms determining high spe- cies turnover in aquatic communities over regional and continental scales. Ecography, 39, 281–288. https://doi.org/10.1111/ecog.01231 Vörösmarty, C. J., McIntyre, P. B., Gessner, M. O., Dudgeon, D., Pruse- vich, A., Green, P., … Davies, P. M. (2010). Global threats to human water security and river biodiversity. Nature, 467, 555. https://doi. org/10.1038/nature09440 Vyverman, W., Verleyen, E., Sabbe, K., Vanhoutte, K., Sterken, M., Hodg- son, D. A., … Wever, A. D. (2007). Historical processes constrain pat- terns in global diatom diversity. Ecology, 88, 1924–1931. https://doi. org/10.1890/06-1564.1 Whittaker, R. H. (1960). Vegetation of the Siskiyou Mountains, Oregon and California. Ecological Monographs, 30, 279–338. https://doi.org/ 10.2307/1943563 Whittaker, R. H. (1972). Evolution and measurement of species diversity. Taxon, 21, 213–251. https://doi.org/10.2307/1218190 Willis, K. J., & Whittaker, R. J. (2002). Species diversity‐scale matters. Science, 295, 1245–1248. https://doi.org/10.1126/science.1067335 Wilson, M. V., & Shmida, A. (1984). Measuring beta diversity with pres- ence‐absence data. Journal of Ecology, 72, 1055–1064. https://doi. org/10.2307/2259551 Zelený, D. (2009). Co‐occurrence based assessment of species habitat specialization is affected by the size of species pool: Reply to. Journal of Ecology, 97, 10–17. https://doi.org/10.1111/j.1365-2745.2008. 01394.x BIOSKETCHES Thibault Leboucher is a PhD candidate at the French research institute Irstea (Aquatic Ecosystems and Global Changes Research Unit). His research interests relate to large‐scale com- munity ecology, especially freshwater diatom communities. William Budnick is a PhD candidate at the University of Texas at Arlington. His research interests include macroecology and applications thereof for the conservation of aquatic biodiversity, particularly fish and crayfish assemblages. SUPPORTING INFORMATION Additional supporting information may be found online in the Supporting Information section at the end of the article. How to cite this article: Leboucher T, Budnick WR, Passy SI, et al. Diatom β‐diversity in streams increases with spatial scale and decreases with nutrient enrichment across regional to sub‐continental scales. J Biogeogr. 2019;00:1–11. https://doi.org/10.1111/jbi.13517 LEBOUCHER ET AL. | 11