Spatial and environmental factors shape multiple facets of macrophyte β-diversity in tropical freshwaters, revealing dispersal and adaptive patterns

Abstract

Dispersal processes, reproductive traits and adaptive dynamics shape aquatic plant (macrophyte) metacommunities. In this study, we explore how environmental and spatial factors influence differences in plant communities across sites, a concept known as beta (β)-diversity, which we assess from the following three perspectives: species identity (taxonomic), functional traits (functional) and evolutionary history (phylogenetic). We surveyed tropical lentic (39) and lotic (85) ecosystems, totalling 124 waterbodies. We analysed β-diversity components (replacement and richness difference) using dispersal traits (vegetative propagation, seed dispersal and life history) and the species’ phylogeny. In lentic ecosystems, replacement and richness differences contributed equally to functional and phylogenetic β-diversity. Richness differences dominated functional β-diversity in lotic ecosystems, while phylogenetic β-diversity remained balanced. Taxonomic β-diversity was largely driven by replacement in both ecosystems. Spatial factors structured macrophyte metacommunities in lentic ecosystems, while spatial and environmental factors influenced lotic ecosystems. These findings reveal how connectivity, isolation and dispersal limitations shape macrophyte communities in contrasting environments. Incorporating phylogenetic β-diversity deepened the understanding of evolutionary relationships in community assembly. This study emphasizes the importance of incorporating the multidimensionality of diversity to uncover key mechanisms driving tropical aquatic ecosystems and inform conservation strategies addressing environmental degradation and the ecological services aquatic plants provide.

1. Introduction

Despite many years since community ecology emerged, understanding the mechanisms influencing species distribution and biodiversity patterns at different spatial scales remains unresolved [1]. The metacommunity concept defines local communities as connected by the dispersal of species and their traits in a way that both local interactions and regional processes shape local community assembly, and simultaneously integrate environmental filters, and spatial and stochastic factors [1,2]. Several metrics of biological diversity can be used to understand these processes, including components of beta (β) diversity, which is the variation in composition (taxonomic, functional or phylogenetic) among different sites or units from a geographic area (e.g. biomes, ecoregions and climatic zone), that guide us in understanding the mechanisms for species composition differentiation across different spatial or temporal scales [3,4]. In freshwater ecosystems, metacommunities can be structured by passive dispersion on the waterbodies or active dispersal through species’ reproductive traits [5].

Aquatic ecosystems present a model system to understand different organizational patterns in metacommunities [6]. This happens because they can disperse to places with a structure completely different from the origin (for example, from marine to freshwater and lotic to lentic) with diversified degrees of environmental heterogeneity and spatial connectivity corresponding to the inherent aptitudes of each species [3,6,7]. For example, lotic ecosystems such as rivers and streams are connected by their drainage systems, while lentic sites such as lakes and ponds may be only seasonally connected or completely isolated [7,8]. This directly affects species' dispersal rates among sites, but that also depends on species’ dispersal ability and environmental tolerance [5,7]. Thus, taxonomic groups with well-known dispersal strategies indicate more accurately the metacommunity processes in a mechanistic approach, where dispersal may be dominant concerning habitat filtering (or even the opposite pattern) in communities [2,8].

Among these organisms, the macrophytes comprise one part of the primary producers of freshwater ecosystems, together with the algae community. In addition to primary productivity and being part of the carbon cycle, other functions employed by macrophytes include nutrient (e.g. phosphorus and nitrogen) cycling, water oxygenation and filtering, and matter siltation, as well as playing a role in trophic webs as food and shelter for other organisms [9,10]. Furthermore, macrophytes are important in providing ecosystem services to humans, by indicating and helping maintain water quality and preservation of freshwater that are important to water supplies throughout the world, as well as many regulator and cultural services.

Given the crucial ecological roles and ecosystem services macrophytes provide, understanding their functional structure, particularly their dispersal traits, is essential. These traits influence their ability to colonize new habitats [11,12] and provide insights into the β-diversity patterns observed in freshwater ecosystems [13], as this community has unique characteristics to survive in ecosystems prone to disturbance [14]. Macrophyte communities comprise some freshwater green algae, ferns and lycophytes, but are dominated by angiosperms [15]. Most species had ancestors adapted to terrestrial ecosystems, producing supportive tissue and developing pollen for water-independent reproduction [14]. These adaptations allowed them to colonize and dominate the terrestrial surface, but re-adaptation was needed to inhabit aquatic systems [14]. This led to changes in their reproductive and dispersal strategies [14], as sessile organisms likely depend on passive mechanisms or other agents for propagation [6,16], depending mainly on the spatial extent between local communities and their connectivity with other habitats.

Reproduction via pollen, successful on land, can be challenging underwater. Many species evolved strategies such as flowers covered by air bubbles or keeping reproductive parts above the water column to ensure water-independent pollination and seed dispersal. Another key strategy is vegetative propagation. Aquatic plants rely more on asexual reproduction, exhibiting a wider range of vegetative propagules than terrestrial plants [17]. This may have evolved to reduce water stress (especially, for example, in lotic ecosystems, where water flow is high), making plants more prone to breakage and easier detachment of parts (e.g. fragments, stolons and rhizomes) [11,17]. Life-history strategies, such as annual or perennial cycles, influence macrophyte dispersal and habitat colonization [11,18]. Annual plants reproduce quickly and disperse seeds over a short time, allowing them to exploit temporary and disturbed habitats, while perennials, with extended vegetative growth, enhance their competitive abilities and resilience, thriving in stable, high-nutrient environments [19]. These traits affect macrophytes’ ability to colonize diverse environments (e.g. lotic and lentic) and at different spatial and environmental scales [18,20].

Additionally, these traits can be associated with phylogeny when the ecological differences between organisms are proportional to the divergence time from a common ancestor [21,22]. This pattern reflects a strong phylogenetic signal, where closely related species tend to retain similar traits due to shared evolutionary history, a phenomenon known as trait conservatism [21,23]. Such traits are shaped by long-term evolutionary constraints and ecological pressures that favour their persistence across lineages [24]. However, not all traits exhibit strong phylogenetic signals; some may evolve rapidly in response to environmental changes, resulting in weak correlations with phylogeny [22,25]. Understanding the degree of phylogenetic conservatism in traits can provide insights into evolutionary processes, ecological strategies and the adaptation of species to their environments [21,26].

Beta (β) diversity, defined as the dissimilarity in species composition among sites, is a key measure of biodiversity [27]. It can be partitioned into the following two components: (i) species replacement (β_repl), when one species replaces another without changing overall richness and (ii) richness (β_rich), or the difference in richness due to gain or loss of species [27,28]. Beyond taxonomic diversity, β-diversity also encompasses functional and phylogenetic dimensions, offering a more holistic view of biodiversity. While taxonomic β-diversity (TβD) focuses on species composition differences, functional β-diversity (FβD) evaluates the variation in trait space among communities [29], reflecting ecological strategies and niche differentiation [29,30]. By contrast, phylogenetic β-diversity (PβD) reveals evolutionary relationships among lineages. The replacement component reflects lineage turnover along spatial or environmental gradients, while the richness component captures lineage loss across sites, indicating the extent to which low-diversity assemblages are subsets of high-diversity ones [29,31,32]. Together, these facets provide a comprehensive understanding of community structure and the processes driving biodiversity patterns.

Overall, in macrophyte communities, species replacement contributes more to total β-diversity than richness and is mostly influenced by environmental conditions rather than spatial factors [33–36]. Moreover, environmental variation seems to be the factor driving FβD [35–37], but some contrasting patterns have been found. In lotic ecosystems, replacement is the main component of FβD [36,38], and connected lakes and wetlands have functional richness differences as their main component [35,39]. This may vary depending on the traits examined, such as life form, photosynthetic pathway, leaf size and area, reproduction time, seed length and longevity, and dispersal syndrome, as these characteristics can affect and contribute to dispersal in different ways (or not contribute at all). Furthermore, PβD seems to be jointly affected by environmental and spatial drivers [35,39], with a predominance of the richness component [39] or equal contribution of components [35]. Thus, examining the reproductive traits of macrophytes across different sites, as well as the loss or gain of lineages, would reveal how local environmental and spatial factors influence macrophyte dispersal ability along with evolutionary implications over space and their interactions with the environment.

Considering this, the study aims to determine how environmental (local) and spatial (regional) factors drive TβD, FβD and PβD in lotic and lentic ecosystems. Based on global patterns for lake and stream/river macrophytes [13,33,35,36,39,40], we hypothesize that the components of β-diversity vary by ecosystem type and biodiversity facet. Specifically, (i) for FβD and PβD, we expect richness differences to contribute more to lentic ecosystems. Furthermore, we also expect that space will primarily drive the organization of these communities in lentic ecosystems due to dispersal limitation by lower connectivity between sites. In contrast, in lotic ecosystems, the environment and space will have similar relative contributions to FβD and PβD because of higher spatial connectivity (overland and fluvial) for the dispersal and varying environmental tolerance of these organisms. Streams are typically more dynamic and complex in terms of hydrology and spatial connectivity, while environmental variation in lakes and ponds may be more suitable for macrophyte colonization, but presents greater spatial connectivity limitations.

For TβD, (ii) replacement is expected to contribute more regardless of ecosystem type, driven by environmental variation. We emphasize that spatial effects are scale dependent: our sites span three drainage basins, and we accounted for this by including spatial eigenvectors based on overland distances (e.g. linear, for lentic and lotic ecosystems) and fluvial distances (for lotic ecosystems). Finally, (iii) functional traits related to dispersal strategies exhibit a strong phylogenetic signal, indicating evolutionary conservatism. This pattern would suggest that closely related species share similar dispersal traits. Conversely, a weak phylogenetic signal would suggest trait lability and a decoupling between functional and phylogenetic responses. To explore this, we tested the phylogenetic signal of each trait individually and together.

2. Material and methods

(a). Study area and sampling design

We sampled 124 freshwater ecosystems along three major river basins located in the eastern Amazon (State of Pará, Brazil): the Acará, Capim and Gurupi river basins (figure 1). The Acará and Capim rivers are classified as sixth-order rivers [41,42], while the Gurupi is a fifth-order river [43]. According to the Köppen climate classification, the region encompasses three climatic types: equatorial (Af), tropical monsoon (Am), and tropical wet and dry (Aw). Most of the Acará basin includes areas in the Af (identified by high temperatures, high rainfall throughout the year and no true dry season—but there are wetter and drier seasons) and Am climate (with temperatures above 18°C and the presence of a dry season) [41]. The Capim basin includes areas with Am climate, and the Gurupi basin spans Am to Aw zones (with lower rainfall during the dry season and precipitation below 60 mm). The natural vegetation throughout the three basins is tropical rainforest, within the Amazon biome [44].

Figure 1.

Sampling sites of the study performed at the eastern Amazon, Brazil. The panels represent different portions of the study area: panel A shows the northernmost sites, sampled in 2022; panel B shows sites located in the centre-west portion of the study area, sampled in 2023; and panel C shows the southeastern portion, with sites sampled in 2017 and 2022.

All basins are within the ‘arc of deforestation’, a belt-shaped region of roughly 500 000 km² that stretches from northwestern Maranhão through eastern and southern Pará to Acre, crossing Mato Grosso and Rondônia [45]. The natural vegetation of this area was removed due to agricultural and road expansion between 1970 and 1990, and it still suffers from extensive land-use change [45,46]. Each river basin has a distinct pattern of land-use dynamics and dominant human activities.

The Acará River Basin has an area of 13 536.32 km², featuring an elongated shape with lower defluvium concentration and limited exposure to large-scale flooding [41]. The predominant land-use types include agriculture (e.g. oil palm) and livestock, and mining and urban expansion, with kaolin and bauxite industries being notable. The Capim River Basin has an extensive area (37 599.5 km²), with an elongated rectangular shape and a drainage network influenced by regional slope and tectonics. River width varies from approximately 200 m in the upper reaches to up to 3000 m near its confluence with the Guamá River [42]. Land use here is marked by a mosaic of agriculture (e.g. sorghum, Eucalyptus), areas for livestock farming, secondary forests and bauxite mining activities. The Gurupi River Basin (35 058 km²) flows into the Atlantic Ocean. It is characterized by low altitudes (90% below 300 m), a trellis/angular drainage pattern and higher drainage density in its western portion [43]. Major land-use activities include agriculture (soybean and other annual crops) and livestock farming.

Data were collected between July and October, which represents the drier season in 2017, 2022 and 2023. Regarding the sites, 85 were lotic (streams) and 39 were lentic (lakes and ponds) ecosystems. To collect environmental and biological data, we designated a 150 m stretch on each water body to standardize the sampling effort. This length was based on the protocol by Peck et al. [47], which recommends a 150 m stretch to capture habitat heterogeneity in stream ecosystems. We applied the same transect length to lentic ecosystems to ensure methodological consistency and enable comparisons across different waterbody types. All sampling locations were at least 1 km apart from each other to ensure spatial independence. The maximum fluvial distance between sampling sites was 1908 km (for lotic sites), while the greatest linear distance was 277.5 km for lentic sites and 268.5 km for lotic sites.

(b). Environmental and spatial variables

Local environmental variables comprised water parameters such as temperature (°C), conductivity (µS cm⁻¹) and dissolved oxygen (mg l⁻¹), which were measured using a multiparameter probe (AKSO model AK47), as well as canopy cover, which was measured at the margins of the sites using a convex densiometer, which was later converted to a percentage. These variables were selected based on the literature indicating they are the most relevant environmental predictors for the biological group studied (i.e. [4,9,18,35,36]), especially regarding the relation between these variables and macrophyte dispersal and habitat colonization. Additionally, the selected variables were not strongly correlated, to minimize redundancy and maximize explanatory power in the following analyses. Water parameters were measured three times at each site, every 75 m, while canopy cover was measured every 10 m (totalling 15 subsamples).

For the calculation of spatial variables, we built connectivity matrices (neighbourhood graphs) from the geographical coordinates of the locations sampled in the studied sites (package R spdep; [48]). The connectivity matrices were based on two classes of distances: the first, based on overland distances (linear distances), using the Minimum Spanning Tree as a connection between sites [49] to be used for both types of ecosystems. The second, based on fluvial distances (calculated by riverdist R package; [50]), which considers the distances between sites within the drainage network of the basins used, completely connected through the larger rivers [51]. Regarding the connectivity matrices, overland spatial variables (Moran’s eigenvector maps (MEMs)) were extracted from linear connectivity and fluvial spatial variables (asymmetric eigenvector maps (AEMs)) from fluvial connectivity from drainage network distances (adespatial R package) [52]. Considering the lack of river connectivity in lentic ecosystems, AEMs were used as spatial variables in the analyses for lotic ecosystems. These spatial variables, the MEMs and AEMs, are the predictor variables for the variation of the community about space and represent the spatial relationships between the sampling sites at different scales, with lower-numbered MEMs and AEMs representing broader spatial structures, while higher-numbered ones represent finer spatial scales [53,54]. As a spatial autocorrelation matrix, only significant and positive spatial eigenvectors were used [53].

(c). Biological sampling—species composition, plant functional traits and phylogenetic tree

Macrophyte presence–absence data were collected by noting all species present inside the 150 m stretch. The plants were identified in the field, and the non-identified material was collected and later identified using specialized literature [55–57], consulting specialists and comparing our samples with herbarium material. The collected specimens were deposited in the João Murça Pires (MG) Herbarium, located in the Museu Paraense Emílio Goeldi (State of Pará, Brazil).

For the functional traits, we selected characteristics that reflected the dispersal ability of the species, especially associated with their reproductive and life-cycle traits [58]. They were: (i) dispersal mode/syndrome: the main way seeds or gametes are dispersed, associated with consequences for the distances the species can cover, the routes they can travel and the places it can end up (autochory, anemochory, hydrochory and zoochory) [58], (ii) life history: the species perenniality, related to population persistence, growth time strategy and resource use efficiency (annual, perennial or annual and perennial) [58], and (iii) vegetative propagation: ways the species disperse propagules, also associated with dispersal ability (cuttings, fragmentation, rhizome, sprout, stolon or none) [35]. The information was gathered using specialized literature (e.g. [55–57]), virtual database information (e.g. Flora do Brasil website—https://floradobrasil.jbrj.gov.br/consulta/), as well as data requested from the TRY Plant Trait database (https://www.try-db.org/TryWeb/Home.php) [59].

The phylogenetic tree of the community found in the study was built by classifying taxa into family, genus and species following Angiosperm Phylogeny Group (APG) IV [60]. We used the megatree ‘GBOTB.extended’ present in the V.Phylomaker package [61] from the R program, to build the tree containing the species found in this study. This megatree is a compilation of the phylogenetic trees of seed plants (GBOTB; [62]) and the trees for ferns and fern allies, which has, in total, 74 533 species of vascular plants (for more information, see [61]). The tree built for this study used the ‘scenario 3’ present in the package, where a new genus is linked to the midpoint of the family branch, using the same approach adopted by Phylomatic and BLADJ [61,63]. For a better visualization, we plotted the phylogenetic tree using the FigTree v.1.4.4 software.

(d). Taxonomic, functional and phylogenetic β-diversity

The β-diversity dissimilarities used in this study follow the partitioning method independently developed by Podani & Schmera [64] and Carvalho et al. [65] for the taxonomic diversity of species, which was later expanded to the functional and phylogenetic facets by Cardoso et al. [29].

The total β-diversity of each diversity facet (taxonomic, functional and phylogenetic) was partitioned into the replacement and richness difference components, where β_total = β_repl + β_rich. The first component, replacement, corresponds to the differentiation of communities from different species, traits or lineages between pairs of communities using the smallest number of different species, traits or lineages from one community [27,29]. The richness difference is calculated from taxonomic, functional or phylogenetic composition variation using the different species/traits/lineages surplus from the other site [27,29]. These two components complement each other, and together they were the total β-diversity of the set of communities studied. Finally, the similarity (S) of the species, functional or phylogenetic composition, corresponds to the overlap between species, traits or lineages common to pairs of communities, calculated simply as (S = 1 beta diversity) or (S = 1 − (replacement + richness difference)).

(e). Data analysis

To test our hypotheses, we performed a partial redundancy analysis (pRDA; [54]) to assess the importance of environmental and spatial predictors on TβD, FβD and PβD and its components. The predictor variables for the spatial dataset (only MEMs for lentic ecosystems and both MEMs and AEMs for lotic ecosystems) were initially selected using the forward selection method [51] and the incidence matrix as response variables, while all environmental variables were used in the model. We partitioned total TβD, FβD and PβD (Tβsor, Fβsor and Pβsor) and its components (βsim and βnes, for each diversity aspect) using the same sets of selected predictor variables. Then, each β-diversity matrix (βsor, βsim and βnes) was used as a response variable to calculate the partition of explained variance for each set of selected predictor variables, considering all sampling sites (preserved and altered streams). The analysis decomposes the variation of the taxonomic, functional and phylogenetic dissimilarity matrices (βsor, βsim and βnes) into four fractions from the environmental and spatial matrices: (i) ‘pure’ environmental variation, (ii) shared environmental and spatial variation, (iii) ‘pure’ spatial variation and (iv) unexplained variation [53]. To assess the significance of the pure environment and pure space partitions, a Monte Carlo test with 9999 permutations was used on the results of a distance-based redundancy analysis (dbRDA; [66]) using a predictor matrix and a conditioning matrix for each dissimilarity matrix.

In order to investigate whether the dispersal traits measured in this study were phylogenetically conserved (if they had a phylogenetic signal), we performed an EM–Mantel test, which combines the standard Mantel test with trait simulations under a specified Evolutionary Model (EM) [67]. This analysis applies a Mantel test followed by a test based on the Brownian motion evolutionary model and supports the use of categorical, binary and continuous data in the same matrix, testing the result against one obtained from a null evolutionary model (for more information, see Debastiani & Duarte [67]). The Brownian evolutionary model assumes that the traits have sustained differentiation that is proportional to the evolutionary time among families [68]. We ran one model including a matrix with all traits, as well as each trait separately, with 999 permutations [67].

All analyses were performed using the R environment [69], and all the steps described here were performed separately for each ecosystem type. The functional trait dissimilarity matrix was calculated using the ‘dist.ktab’ function in the ade4 package [70], while the phylogenetic and taxonomic dissimilarity matrices were calculated using the ‘beta.multi’ function from the BAT package [71]. Spatial vectors for the geographic coordinates of the sampling sites were obtained using the Moran test (function moran.randtest from the spdep package), and only the significant and positive ones were considered. Then, the vectors used in the redundancy analysis were selected based on the species composition matrix using the forward selection method (function forward.sel from the adespatial package [52]). pRDAs for diversity facet were performed using the ‘varpart’ function, while the significance of each partition was tested using the ‘dbrda’ function and ANOVA tests for each model, all from the vegan package [72].

3. Results

We found 122 macrophyte species, including two lycophytes, 16 ferns and 104 angiosperms, 2 basal angiosperms, 73 monocots and 29 eudicots (electronic supplementary material, figure S1 and table S1). Seventy-seven species were recorded in lentic ecosystems, 12 of which were exclusive to these environments. In lotic ecosystems, 110 species were found, 45 of which were exclusive. Sixty-five species were shared between both ecosystem types.

For the functional traits, the most recorded dispersal mode among species was anemochory (46 species; 37.7%), followed by autochory (43 species; 35.2%), zoochory (22 species; 18.0%) and hydrochory (11 species; 9.0%). Regarding vegetative propagation, rhizome propagation was the most common (69 species; 56.6%), followed by propagation by stolon and cuttings (14 species each; 11.5%), fragmentation (13 species; 10.7%), sprout (7 species; 5.7%) and species without vegetative propagation (5 species; 4.1%). In terms of life history, most species were perennials (92 species; 75.4%), with fewer annuals (17 species; 13.9%) and annual-perennials (13 species; 10.7%).

In lentic ecosystems, autochory was the prevalent dispersal mode (32 species; 41.6%), rhizome propagation was dominant (38 species; 49.4%) and perennials dominated the life-history category (54 species; 70.1%). In lotic ecosystems, anemochory was the dominant dispersal mode (42 species; 38.2%), rhizome was the most common form of vegetative propagation (66 species; 60%) and perennials also prevailed in life history (54 species; 70.1%).

For the analysis to detect phylogenetic signals among the dispersal traits, the Mantel test for lentic ecosystems did not detect any significant correlations between the matrix with all traits (R = 0.095, p = 0.081), as well as between dispersal mode (R = 0.056, p = 0.077), vegetative propagation (R = 0.039, p = 261) and life-history traits (R = 0.103, p = 0.542). For lotic ecosystems, the Mantel test showed no significant correlations between all traits (R = 0.013, p = 0.39), vegetative propagation (R = −0.068, p = 0.899) and life history (R = −0.009, p = 0.542). In contrast, the dispersal mode showed significant correlations (R = 0.084, p = 0.004). Still, the EM–Mantel tests indicated that the phylogenetic signal detected in this trait was not higher than expected by the Brownian motion evolutionary model (EM–Mantel p = 0.738).

In lentic ecosystems, TβD (0.73) was primarily driven by species replacement (Tβ_Repl = 0.42; 58.28%), with Tβ_Rich amounting to 41.72% (0.30) of total β-diversity. FβD (0.48) showed a nearly balanced influence of Fβ_Rich (0.25; 52.92%) and Fβ_Repl (0.23; 47.08%). PβD (0.43) had almost equal contributions from Pβ_Repl (0.22; 50.13%) and Pβ_Rich (0.22; 49.87%; figure 2). In lotic ecosystems, TβD (0.84) was also driven by Tβ_Repl (0.49; 58.36%), with Tβ_Rich contributing 41.64% (0.35). FβD (0.53) was dominated by Fβ_Rich (0.34; 64.20%), with Fβ_Repl accounting for 35.80% (0.19). PβD (0.50) showed a slightly higher contribution from Pβ_Repl (0.27; 53.77%) compared with Pβ_Rich (0.23; 46.23%; Figure 2).

Figure 2.

Percentage of the contribution of each component (replacement and richness difference) to total functional, phylogenetic and taxonomic total β-diversity.

The forward selection method identified broad-scale spatial vectors MEM1, MEM3, MEM4 and MEM8 as predictors for lentic ecosystems, with all environmental variables included in the partitioning analyses. The pRDA results showed that, in lentic ecosystems, Tβ_Total was explained solely by spatial factors (F = 1.985, p = 0.002, adj. R^² = 0.106; figure 3A and electronic supplementary material, table S2), while Tβ_Repl was not explained by any variable (figure 3B). Similarly, Tβ_Rich was explained only by spatial factors (F = 2.726, p = 0.021, adj. R^² = 0.156; figure 3C). For FβD, Fβ_Total was influenced by spatial factors alone (F = 2.146, p = 0.006, adj. R² = 0.120), with no contribution from environmental factors (figure 3D). Fβ_Repl was not explained by either spatial or environmental factors (figure 3E). However, Fβ_Rich was driven exclusively by spatial factors (F = 2.393, p = 0.043, adj. R^² = 0.129; figure 3F). For PβD, Pβ_Total was explained only by spatial factors (F = 2.925, p < 0.001, adj. R^² = 0.179; figure 3G), while Pβ_Repl showed no association with any variable (figure 3H). In contrast, Pβ_Rich was influenced by both spatial (F = 3.094, p = 0.001, adj. R^² = 0.311) and environmental factors (F = 3.202, p < 0.001, adj. R^² = 0.057; figure 3I).

Figure 3.

Venn diagrams of the results of variation partition illustrating the unique and shared contributions of environmental and spatial variables to taxonomic (A–C), functional (D–F) and phylogenetic (G–I) β-diversity in lentic ecosystems. For each facet, panels represent total β-diversity (A, D, G), replacement (B, E, H) and richness difference (C, F, I). The variation is expressed as adjusted R² values, and bold values indicate significant relationships (*p < 0.05).

For lotic ecosystems, the forward selection method identified seven spatial vectors: five were linear vectors of broader (MEM1, MEM2 and MEM3) and intermediate (MEM7 and MEM13) scales, and two were intermediate-scale fluvial vectors (AEM12 and AEM18) with all environmental variables included in the partitioning analyses. The pRDA results showed that Tβ_Total was explained by both environmental (F = 3.110, p = 0.001, adj. R^² = 0.080) and spatial factors (F = 3.285, p = 0.001, adj. R^² = 0.145; figure 4A). Tβ_Repl was also explained by environmental (F = 3.610, p = 0.001, adj. R^² = 0.094) and spatial factors (F = 3.715, p = 0.001, adj. R^² = 0.152; figure 4B), as well as Tβ_Rich (environmental—F = 2.441, p = 0.023, adj. R^² = 0.061; spatial—F = 2.421, p = 0.007, adj. R^² = 0.101; figure 4C). For FβD, Fβ_Total was influenced by environmental (F = 3.613, p = 0.001, adj. R^² = 0.098) and spatial factors (F = 3.179, p = 0.001, adj. R^² = 0.138; figure 4D). Similarly, Fβ_Repl was explained by environmental (F = 3.886; p = 0.002, adj. R^² = 0.099) and spatial factors (F = 3.994, p = 0.001, adj. R^² = 0.170; figure 4E), and Fβ_Rich, as well, was driven by environmental (F = 2.997, p = 0.013, adj. R^² = 0.084) and spatial factors (F = 2.575, p = 0.012, adj. R^² = 0.111; figure 4F). For PβD, Pβ_Total was explained by environmental (F = 3.022, p = 0.001, adj. R^² = 0.073) and spatial factors (F = 3.861, p = 0.001, adj. R^² = 0.174; figure 4G). Pβ_Repl was also explained by environmental (F = 4.886, p = 0.001, adj. R^² = 0.122) and spatial (F = 4.493, p = 0.001, adj. R^² = 0.185) factors (figure 4H), while Pβ_Rich was solely explained by environmental factors (F = 3.126, p = 0.003, adj. R^² = 0.146; figure 4I).

Figure 4.

Venn diagrams of the results of variation partition illustrating the unique and shared contributions of environmental and spatial variables to taxonomic (A–C), functional (D–F) and phylogenetic (G–I) β-diversity in lotic ecosystems. For each facet, panels represent total β-diversity (A, D, G), replacement (B, E, H) and richness difference (C, F, I). The variation is expressed as adjusted R² values, and bold values indicate significant relationships (*p < 0.05).

4. Discussion

Our findings reveal a balanced contribution of replacement and richness differences to total β-diversity in lentic ecosystems, with near-equal proportions (approximately 50/50) observed for both functional and phylogenetic dimensions. Interestingly, in lotic ecosystems, richness amounted the most to total FβD (nearly 65/35 proportion), while the balance between components remained equal for phylogenetic diversity (a finding similar to [35] in connected floodplain lakes). Taxonomic diversity showed the same pattern in both types of ecosystems, with a higher contribution of replacement to total β-diversity (similar to the global pattern found by [33,73] in lakes, by [74] in permanent and temporary ponds, as well as [40] in a river basin). Additionally, space was the main factor in structuring macrophyte metacommunities in lentic ecosystems, while both environmental and spatial factors were responsible for structuring lotic ecosystems, corroborating β-diversity patterns already found for aquatic organisms (macrophytes, aquatic invertebrates and fishes) in lentic [3,75] and lotic environments [76–78]. Furthermore, both overland and fluvial spatial structures were important in structuring lotic ecosystems, indicating that these communities disperse through these two routes.

Lotic and lentic ecosystems differ significantly in environmental factors such as water flow dynamics, canopy openness and nutrient input. Lotic systems, like streams and rivers, typically exhibit higher water flow, whereas lentic systems, including lakes and ponds, often have more open canopies and greater nutrient input. These factors influence the distribution of species, traits and lineages by selecting adaptations to those specific environmental conditions [79], such as traits necessary for inhabiting shaded, high-flow habitats (e.g. in streams) or lineages (such as families or genera) that evolved to thrive under these conditions. However, connectivity is a critical factor in understanding the distribution and structuring of aquatic communities [5,16,80]. In this study, lentic ecosystems were isolated from one another, while lotic ecosystems were highly interconnected through the drainage network. This connectivity directly influences the dispersal mechanisms of macrophytes: isolated sites favour species with adaptations for long-distance dispersal (zoochory) indicating that linear distances between water bodies can have a strong influence on the dispersion of these organisms, while connected sites facilitate colonization by species with more passive dispersal strategies (e.g. hydrochory and anemochory) by fluvial routes [16].

Space was the primary factor influencing macrophyte β-diversity in lentic ecosystems, affecting both taxonomic and functional total β-diversity and their richness difference component. This finding shows that dispersal limitation is crucial in shaping these metacommunities [81]. Lentic systems, being more isolated, have community composition variations that are primarily driven by the spatial distance between sites and the efficiency of species dispersal [82,83]. Heino et al. [5] found significant spatial structuring on macrophyte metacommunities, which they linked with the stronger dispersal limitation in lake organisms compared with stream organisms, but emphasized the stronger effect of environmental variation on those communities. In this study, the lentic sites were environmentally similar, exhibiting comparable conditions such as increased canopy openness, higher water temperature and low water flow. This environmental similarity may have amplified the role of spatial factors in structuring the communities and may also explain why the replacement component was unaffected by spatial or environmental factors, likely due to the lack of environmental variation among these sites.

In lentic systems, the absence of replacement processes may be attributed to environmental conditions’ low spatial heterogeneity, which reduces species turnover opportunities driven by niche differentiation [84]. Instead, community structure is shaped by the gain and loss of species and traits, possibly influenced by the spatial structure of land-use changes, such as deforestation [85]. These changes might create a patchy distribution of suitable habitats, influencing dispersal and colonization dynamics across the landscape [86,87]. At closer sites (considering overland distances), increased species and trait richness reflect the successful establishment of species with diverse traits that enable short-distance dispersal and efficient colonization. In contrast, only species with specific trait sets adapted for long-distance dispersal (e.g. zoochory, perennial life history or vegetative propagation via cutting and fragmentation) are capable of colonizing more distant sites [16]. Moreover, the richness component of PβD was also influenced by environmental factors, suggesting that the gain of lineages was partially driven by environmental conditions (an adaptation to local environmental conditions) [84]. However, spatial factors remained the primary structuring force, emphasizing the importance of dispersal dynamics in lentic systems with low environmental variability.

In contrast, lotic macrophyte metacommunities were structured by both spatial and environmental factors, with a predominance of spatial effects. This suggests that dispersal processes and the spatial configuration of river networks are the primary drivers of community assembly (mass effect) [83]. Total β-diversity and the replacement components of taxonomic, functional and phylogenetic diversity were influenced by both environmental and spatial factors. These findings demonstrate that mass effects in lotic communities may occur through species, trait and lineage turnover driven by environmental variation (operating at a local scale) and high dispersal events (through spatial distance) [88], indicating large dispersal events and good suitability of species within an environmental gradient ranging from suboptimal to excellent habitats, without altering species, trait or lineage richness. In this case, species capable of dispersing to highly connected sites (by land and rivers) can establish themselves in smaller populations even outside their niche range [8,83,88]. Additionally, many of the streams sampled in our study were significantly degraded by land-use changes such as deforestation, agriculture and husbandry. These disturbances altered stream characteristics, creating conditions more akin to lakes [85], including increased canopy openness, elevated temperatures, higher nutrient input and reduced water flow. Consequently, lineages typically associated with lentic environments, such as Nymphaeales, Alismatales, and genera like Eleocharis and Utricularia [55], were able to colonize those habitats, causing a turnover in the communities.

The pattern observed in lotic PβD, which showed richness difference being driven solely by space, also suggests that dispersal limitation and spatially structured historical processes are key drivers of lineage accumulation or loss across sites [83]. This indicates that the presence or absence of certain phylogenetic lineages is more strongly influenced by regional connectivity and colonization dynamics than by current local environmental conditions. Such a pattern could reflect a legacy of past dispersal events, where certain lineages were established in specific regions and persisted due to limited opportunities for dispersal across drainage boundaries. This spatial structuring may be masking the local loss of specialist lineages adapted to pristine lotic environments, potentially misleading assessments of biotic homogenization and ecological degradation [89].

In both ecosystem types, the interplay between spatial and environmental factors offers important insights into the dynamics of macrophyte communities under environmental change [33,90]. These influences are particularly relevant in the context of land-use change and climate change, where shifts in dispersal and colonization dynamics, driven by habitat fragmentation, deforestation and altered hydrological regimes, may favour invasive species and contribute to community homogenization [91]. Modelling these dynamics can provide valuable insights into the resilience of aquatic ecosystems and the potential for bioinvasions to disrupt evolutionary processes. Moreover, understanding how these factors shape phylogenetic and functional diversity can guide conservation strategies aimed at preserving ecosystem functions and evolutionary potential in the face of ongoing environmental change.

Moreover, lotic Fβ_Total, Fβ_Repl and Fβ_Rich were also influenced by environmental and spatial factors, indicating that multiple processes operate simultaneously to shape FβD. The strong spatial effect on functional replacement indicates a trait turnover among sites is structured by dispersal-related processes [29]. This may reflect a shift from traits suited to short-distance dispersal (e.g. autochory, annual life cycles and vegetative propagation via stolons) to traits facilitating long-distance dispersal (e.g. anemochory, zoochory or vegetative propagation via cutting and fragmentation) [16,20,35]. Additionally, the connectivity and flow of streams may act as a mechanism of passive dispersal for aquatic plants, favouring hydrochory regardless of their intrinsic dispersal traits [16]. The balanced influence of environmental and spatial factors on functional richness suggests that changes in environmental conditions drive the addition or reduction of trait richness, but the spatial configuration of sites is still the main factor limiting trait composition. This pattern likely reflects a combination between environmental filtering and dispersal limitation, where only species with traits suited to specific environmental niches can establish and persist, but their traits also need to support long-distance dispersal in order for individuals to reach those environments [5,7,20,88]. This pattern implies that even when environmental filtering occurs, the ability of species with diverse traits to reach and persist in new locations is strongly influenced by spatial processes, reinforcing the joint role of dispersal and environmental constraints in shaping macrophyte FβD in lotic ecosystems.

Finally, the lack of strong phylogenetic signals also reinforces the pattern found in the region. Of all traits evaluated, only the dispersal mode showed a significant, albeit very weak, correlation, indicating a weak phylogenetic signal. This suggests trait lability, a condition where traits are phylogenetically structured but not strongly conserved, allowing closely related species to exhibit functional variability [24], or possibly convergent evolution, where unrelated species independently develop similar traits under comparable environmental pressures [92]. This finding is most likely linked to the polyphyletic origin of aquatic plants, which include algae, ferns and flowering plants, organisms that can be morphologically, physiologically and phenologically distinct [15]. This suggests a decoupling between the observed patterns of FβD and PβD, indicating that differences in trait composition among sites do not reflect differences in lineage composition.

Thus, our study provides valuable insights into macrophyte community patterns, dispersal mechanisms and evolutionary dynamics in contrasting environments, highlighting their responses to ongoing environmental degradation. By focusing on reproductive and life-history traits that were both easy to measure and effective proxies for dispersal abilities, we successfully revealed differing patterns across lentic and lotic ecosystems. This approach allowed us to better capture the effects of spatial processes, particularly in environments where habitat connectivity and isolation play critical roles. Additionally, incorporating PβD into our approach provided a deeper understanding of how evolutionary relationships influence community assembly and species turnover in these ecosystems.

However, comparing lentic and lotic systems presents inherent challenges due to their fundamentally different habitat structures. While these differences complicate direct comparisons, the ecological and evolutionary processes influencing species distribution, trait composition, and lineage turnover remain conspicuous and provide a robust framework for understanding these systems. Despite these challenges, our study demonstrates the utility of combining taxonomic, trait-based and phylogenetic approaches to identify dispersal mechanisms and community assembly patterns, even in complex tropical ecosystems such as the Amazon forest.

For future studies, we recommend incorporating morphological and physiological traits of macrophytes to better understand how multiple trait dimensions interact with spatial and especially environmental processes, since our findings achieved great spatial representation using traits related to dispersion. Traits related to tolerance to water stress (e.g. stomatal conductance and density, leaf cuticle density, aerenchyma density and root-to-shoot ratio), phenotypic plasticity (in leaf morphology, e.g. area and thickness, as well as root length) and functional responses to environmental variation (photosynthetic efficiency, nutrient—phosphorus, nitrogen— and carbon content and ratios) could unveil additional ecological and evolutionary processes underlying observed diversity patterns, especially in the face of climate change and altered hydrological regimes. Additionally, exploring the impacts of the climate crisis and forest fragmentation through ecological and dispersal modelling would provide valuable insights regarding species distribution and possible ecological invasions. However, we also suggest that taxonomic studies should continue to be done, especially in places with little information, such as tropical freshwater ecosystems (because, as our results showed, the taxonomic facet explained most of the variance in both ecosystem types). We also emphasize the importance of integrating multiple aspects of diversity into the study of ecosystem services provided by macrophytes, such as water quality regulation and biodiversity maintenance, to inform effective conservation and management strategies.

Ethics

This work did not require ethical approval from a human subject or animal welfare committee.

Data accessibility

Data from this study as well as custom R code used to perform statistical analyses are available at [93].

Supplementary material is available online [94].

Declaration of AI use

We have not used AI-assisted technologies in creating this article.

Authors’ contributions

A.L.F.: conceptualization, data curation, formal analysis, investigation, methodology, writing—original draft, writing—review and editing; E.J.R.C.: conceptualization, formal analysis, investigation, methodology, writing—original draft, writing—review and editing; R.S.C.: data curation, investigation, methodology, visualization, writing—review and editing; T.S.M.: conceptualization, data curation, funding acquisition, investigation, methodology, project administration, supervision, validation, visualization, writing—review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Conflict of interests

We declare we have no competing interests.

Funding

The research leading to these results received funding from Conselho Nacional de Desenvolvimento Científico e Tecnológico—CNPq (process: INCT-SinBiAm [406767/2022-0]; PPBio-AmOr [441257/2023-2] and PELD-AmOr [445970/2024-3]), Coordenação de Aperfeiçoamento de Pessoal de Nível Superior—Brasil (CAPES)—Finance Code 001, Norsk Hydro, Conservation International Brazil (CI) and Agropalma.