Drivers of fish beta diversity vary by habitat and rainfall period in temporary freshwater habitats of the Atlantic Forest

Introduction

Temporary pools are key components of freshwater landscapes in tropical regions, defined by their cyclical drying and flooding regimes, which create dynamic environments for colonization, extinction and recolonization events (Williams, 2007). Despite their small size and transient nature, they can harbor high biodiversity and play critical roles in dispersal, reproduction and life cycle for several freshwater species (Blaustein and Schwartz, 2001; DiMauro and Hunter, 2002; Espírito-Santo et al., 2013). However, temporary pools are increasingly threatened by a range of anthropogenic pressures, harboring a high number of small-sized threatened fishes (Castro and Polaz, 2020). For example, habitat loss driven by road construction, urbanization and land-use change has pushed several pool-dwelling species toward extinction (Volcan and Lanés, 2018). The disappearance of natural temporary pools and the degradation of floodplains have also led some fishes to occupy roadside ditches, built along dirt roads for water drainage (Costa et al., 2024, 2025). As a result, fish communities from temporary freshwater habitats that are naturally short-lived and structured over brief hydroperiods are now also being reshaped in artificial environmental settings that may facilitate the introduction of non-native species, together with increased exposure to pollutants and warming (Hohausová et al., 2010; Costa et al., 2025).

Despite their ecological importance, studies on fish metacommunity in temporary aquatic habitats remain scarce in the Neotropical region, which harbors the richest freshwater fish fauna on the planet. Most community ecology studies have focused on seasonal and environmental variation in temporary pools from the Amazon and Pantanal biomes (Pazin et al., 2006; Espírito-Santo et al., 2013; Tondato et al., 2013; Espírito-Santo and Zuanon, 2017; Couto et al., 2018), with few assessments from the Atlantic Forest. Notably, the only study investigating a fish metacommunity in roadside ditches in the Neotropics was recently conducted in this biome (Costa et al., 2025).

Metacommunity theory provides a powerful framework for understanding how communities assemble across spatial and temporal scales (Leibold and Chase, 2017; Leibold et al., 2004). According to this theory, community assembly in metacommunities emerges from the interaction among key ecological processes, including dispersal, environmental filtering, biotic interactions and stochasticity, whose relative importance varies across spatial scales, temporal contexts and even across sites and species (Leibold et al., 2022). Different combinations of these processes give rise to distinct but non-mutually exclusive metacommunity paradigms, such as colonization-extinction dynamics, species sorting along environmental gradients, and mass effects under high connectivity. In this context, beta diversity provides a useful framework to investigate metacommunity structure by capturing spatial and temporal variation in species composition. Its partition into turnover (species replacement) and nestedness (species loss or gain) can offer insights into the relative roles of dispersal, connectivity and local constraints shaping assemblages (Baselga, 2010; Declerck et al., 2011; Legendre and De Cáceres, 2013; Legendre, 2014). These components are also relevant for conservation, as high turnover implies spatial complementarity among sites, whereas nestedness may indicate prioritization of species-rich locations (Socolar et al., 2016).

Temporary freshwater habitats of the Atlantic Forest are not only valuable systems for exploring and understanding the components and drivers of beta diversity but also serve as replicable models for conserving metacommunities. In these fragmented environments, species richness is shaped by factors such as pool volume and proximity to permanent water bodies (Pazin et al., 2006), parallel to the principles of island biogeography theory proposed by MacArthur and Wilson (1967). As landscapes become increasingly fragmented, gaining insight into beta diversity patterns in such temporary and extreme aquatic habitats becomes critical, as may represent the new reality for other climate-vulnerable ecosystems in the future (Comte and Olden, 2017).

In this study we investigated how fish community composition and temporal variation in spatial beta diversity components (turnover and nestedness) act across two habitat types and hydrological periods in temporary freshwater habitats of the Atlantic Forest. Our focus was on temporary pools and roadside ditches fish communities in a blackwater sub-basin located in the coastal plains of southeastern Brazil. The region harbors a high endemism resulting from its geographic isolation, bounded by the Atlantic Ocean and coastal mountain ranges that restrict species dispersal between basins (Menezes et al., 2007; Giongo et al., 2023). We also evaluated how local environmental conditions and spatial isolation influence patterns of species turnover and nestedness.

Reduced hydrological connectivity during the dry period strongly limits dispersal among temporary freshwater habitats, restricting movement largely to fish species capable of overland displacement. Under such low-connectivity conditions, local habitat availability, represented by water volume, is expected to play a central role in structuring communities by influencing persistence and species accumulation. At the same time, physicochemical variables are unlikely to act as strong filters at the present ecological scale, as fish assemblages in these systems are already selectively filtered for tolerance to extreme and fluctuating conditions, a pattern previously reported for temporary aquatic habitats in Central Amazonia (Pazin et al., 2006). Based on these premises, we hypothesized that, during the dry period, beta diversity components would be primarily correlated with water volume in both pools and ditches. In contrast, during the wet period, enhanced connectivity among habitats promotes dispersal and homogenizes environmental conditions across sites, as observed in highly connected floodplain systems such as the Okavango Delta (Bokhutlo et al., 2021). Based on this expectation, we predicted that, during the wet period, none of the measured predictors would be correlated with beta diversity components, especially in pools, since they are highly connected by the flooded forest.

Material and methods

Study area and sampling

Sampling was conducted in the Preto River sub-basin, part of the Itanhaém River basin, within the coastal Atlantic Forest of São Paulo State, Brazil (Figure 1A). This coastal plain region is characterized by sandy-clay substrates, restinga vegetation, and blackwater streams with high dissolved organic carbon (rich in humic substances), low pH and low dissolved oxygen (Camargo and Cancian, 2016; Supplementary Table S1). From January to December 2024, 36 temporary pools (Figure 1B) and 36 roadside ditches (Figure 1C) were sampled, with three distinct sites of each habitat type surveyed monthly. This sampling effort was defined based on environmental permits issued by governmental authorities, which limited the number of sites that could be sampled per month to minimize disturbance to populations of threatened species.

A given site, or another within the same temporarily flooded area, was only resampled in a later month if it had undergone complete drying followed by reflooding, or complete reconnection with adjacent streams. Based on this criterion, a limited number of sites were resampled across the study period, with most locations sampled only once, and a subset of sites resampled between two and six times across different months (on average, each site was sampled 1.75 times, that is, close to two times). This criterion, following Costa et al. (2025), ensured temporal independence among sampling events by requiring complete habitat resetting between surveys, either through full drying or hydrological reconnection with surrounding floodable areas and streams. As a result, fish communities were able to reassemble through natural recolonization, minimizing carryover effects from previous sampling.

Fish were sampled for 15 min at each site by two researchers using hand nets (50 × 40 cm, 1 mm mesh), following Costa et al. (2025), and actively sweeping the entire flooded area. Sampling effort was standardized across sites and was sufficient to cover the whole extent of each habitat, which consisted of shallow and spatially restricted pools and ditches, with maximum lengths ranging from < 1 to ~58 m, widths from < 1 to ~10 m, and depths typically below 30 cm. Given the small size and shallow nature of these habitats, this effort allowed rapid detection of species present at each site. Captured individuals were euthanised with eugenol (1–1.5 ml L⁻¹), fixed in 10% formalin for 48 h, and preserved in 70% ethanol. All procedures were approved by the Ethics Committee for Animal Use at São Paulo State University (UNESP – IB/CLP, protocol no. 15/2023) and authorized under SISBIO license 90241-1. Species were identified and voucher specimens were deposited in the Fish Collection of UNESP in São José do Rio Preto (DZSJRP).

At each site, we measured environmental variables including maximum pool length, width, and depth (in meters) using a fiberglass measuring tape. Pool or ditch volume (m³) was estimated assuming a half-ellipsoid shape. Water temperature (°C), pH and dissolved oxygen (mg·L⁻¹) were measured in situ, before fish sampling, using a portable multiprobe (Hanna HI98194); these three variables were selected for their ecological relevance in temporary habitats, where exposure to direct sunlight and the absence of water flow can lead to thermal and oxygen extremes. The distance from each sampling site to the nearest permanent stream was measured as the straight-line distance, expressed in meters. Distances were obtained using high-resolution satellite imagery in Google Maps and, when sites were sufficiently close to streams, were measured in the field using a fiberglass measuring tape during sampling. Sampling months were grouped into: wet (November-March) and the dry (April–October) periods, based on regional rainfall patterns described in Costa et al. (2025). It is important to note that despite there being strong differences in precipitation between these periods (Costa et al., 2025), there is rain in both periods (see monthly water volume per aquatic habitat in Supplementary Figure S1), typical of the humid coastal plains of southeastern Brazil. Summary statistics (range, mean, standard deviation) for each environmental variable per habitat type and period are provided in Supplementary Table S1, together with the number of sites sampled per habitat-period combination.

Data analysis

To assess variation in fish assemblage composition across habitats and hydrological periods, we performed a non-metric multidimensional scaling (NMDS) based on Bray–Curtis dissimilarities calculated from species abundance data (counts). Differences in fish community composition between habitat types (temporary pools and roadside ditches) and hydrological periods (wet and dry) were tested using Permutational Multivariate Analysis of Variance (PERMANOVA). Analyses were conducted using two complementary data structures: (i) abundance-based matrices (“count”), using Bray–Curtis dissimilarities to account for differences in species abundances; and (ii) presence–absence matrices (“pa”), using Jaccard dissimilarities to assess compositional differences independent of abundance. For each data structure, we also tested the homogeneity of multivariate dispersion among groups using PERMDISP, ensuring that any detected differences in community composition were not driven by unequal within-group variability. All tests were performed with 999 permutations.

Beta diversity was further quantified using the Sørensen dissimilarity index, which was partitioned into its turnover and nestedness components using the betapart R package (Baselga and Orme, 2012). All analyses were performed separately for each habitat-period combination using presence–absence data. To explore the drivers of beta diversity components, we calculated pairwise Euclidean distances for selected predictors. Temperature, dissolved oxygen, and pH were used to represent potential environmental filtering; volume was considered a proxy for habitat availability; and distance to the nearest stream was used as a proxy for spatial isolation. These predictor dissimilarity matrices were related to pairwise beta diversity components (turnover and nestedness) using partial Mantel tests, accounting for spatial autocorrelation by controlling for geographic distance among sites, with 999 permutations. All statistical analyses were performed in R.

Results

A total of 1,545 individuals were sampled, with 758 individuals in temporary pools and 787 in roadside ditches, among a total of 20 species. Of these, 18 species occurred in pools and 17 in ditches. In temporary pools, 199 individuals and 12 species were recorded during the wet period, whereas 559 individuals and 16 species were recorded during the dry period. In roadside ditches, 333 individuals and 16 species were sampled during the wet period, and 454 individuals and 12 species during the dry period. The most abundant species in temporary pools were Mimagoniates lateralis (Nichols, 1913) (n = 281), Atlantirivulus peruibensis Ywamoto, Nielsen, and Oliveira, 2025 (n = 157), Hyphessobrycon boulengeri (Eigenmann, 1907) (n = 92) and Phalloceros reisi Lucinda, 2008 (n = 79). In roadside ditches, the dominant species were A. peruibensis (n = 212), H. boulengeri (n = 131), P. reisi (n = 104), Poecilia reticulata Peters, 1859 (n = 77) and Hyphessobrycon bifasciatus Ellis, 1911 (n = 72) (Table 1). Non-native species, P. reticulata was exclusive to roadside ditches, while Nannostomus beckfordi Günther, 1872 occurred in both aquatic habitats, but mainly in ditches. A single specimen of Hoplosternum littorale (Hancock, 1828) also occurred in a roadside ditch; however, its origin as native or non-native to the southeastern Brazil basins is not fully understood yet. In addition, we recorded three endangered species: Leptopanchax itanhaensis (Costa, 2008), classified as Critically Endangered, and more abundant on roadside ditches than in temporary pools (Table 1); Rachoviscus crassiceps Myers, 1926 and M. lateralis, both classified as Vulnerable, the first more abundant in ditches and the second in pools. Fish assemblage composition did not differ significantly between habitats or across sampling periods when considering both count and presence-absence data (PERMANOVA_count: F_{3, 68} = 1.333, p = 0.085; PERMDISP_count: F_{3, 68} = 0.292, p = 0.831; PERMANOVA_pa: F_{3, 68} = 1.51, p = 0.07; PERMDISP_pa: F_{3, 68} = 0.05, p = 0.98; Figure 2).

Table 1. Species recorded in temporary pools (TP) and roadside ditches (RD) of the Preto River sub-basin, in the Atlantic Forest of Southeastern Brazil.

Voucher specimens are deposited in the DZSJRP collection of UNESP.

^*non-native species.

Beta diversity was high in all groups (β_S*rensen* ≥ 0.86), with species turnover as the dominant component (β_turnover ≥ 0.71) and a smaller contribution from nestedness (β_nestedness ≤ 0.18). This pattern was consistent across habitats and hydrological periods (Figure 3). Water volume was significantly correlated with nestedness in both temporary pools and roadside ditches during the dry period (temporary pools: r = 0.44, p = 0.001; roadside ditches: r = 0.21, p = 0.009; Figure 4). Differences in distance to the nearest stream were positively correlated with turnover in temporary pools during the wet period (r = 0.39, p = 0.02; Figure 4). A significant correlation was also found between differences in pH and species turnover in roadside ditches during the wet period (r = 0.33, p = 0.002; Figure 4). No significant correlation was detected between pairwise beta diversity components and differences in temperature or dissolved oxygen across habitats or hydrological periods (Table 2).

Table 2. Partial Mantel test results showing correlations (r) and associated p-values between pairwise beta diversity components (turnover and nestedness) and predictors dissimilarities (temperature, dissolved oxygen, pH, volume and distance to the nearest stream), while controlling for spatial distance among sites, across habitat and sampling period of temporary freshwater habitats from the Preto River sub-basin, in the Atlantic Forest of Southeastern Brazil.

Significant results (p < 0.05) are shown in bold.

Discussion

Our study represents the first effort to investigate beta diversity components of fish in temporary pools and roadside ditches of the Atlantic Forest. It also provides the first report of fish communities in temporary pools from this biome, complementing recent findings on roadside ditches (Costa et al., 2025). We found a similar total number of individuals and species in both habitat types throughout the year, with A. peruibensis emerging as the second most abundant species in pools and the most abundant in ditches. This aligns with the species' highly amphibious lifestyle, including high emersion tolerance and terrestrial locomotion via leaping, which facilitates both escape from predators and colonization of new temporary habitats, as evidenced in other Rivulidae species (Espírito-Santo et al., 2019). However, in the pools, A. peruibensis was outnumbered by M. lateralis, a typical stream-dwelling tetra endemic from the region (Moraes et al., 2021; Selinger et al., 2025), that occurs mainly in large schools in floodplain pools, with up to 90 individuals recorded at a single site.

We found no significant differences in species composition between habitat types or hydrological periods, regardless of their origin (natural or man-made) or seasonal variation. This pattern is consistent with findings from temporary aquatic systems in subtropical Africa, where hydrological regimes act as major drivers of community structure and increased connectivity during wet periods promotes assemblage homogenization (Bokhutlo et al., 2021). In addition, in southeastern Brazil, the dry period includes unregular rainfall events, and not all pools or ditches dry synchronously and completely, as evidenced in Supplementary Figure S1. The presence of P. reticulata, a non-native species introduced for mosquito control and known for its broad environmental tolerance (Silva et al., 2023), together with N. beckfordi, a species native to the Amazon basin that has been widely exported as an ornamental fish and introduced in various regions (Magalhães et al., 2019), with both species occurring mostly in ditches, is of particular concern. Another noteworthy species is H. littorale, recorded by a single individual in a roadside ditch; its native or non-native status in southeastern Brazil remains uncertain, as regional records are recent (Reis, 1997; Ferreira and Petrere, 2009), and the species is already reported as non-native in other Brazilian river basins (Moreira et al., 2023), suggesting a possible recent dispersal to coastal systems.

Our findings demonstrate that fish beta diversity in temporary freshwater habitats of the Atlantic Forest is consistently high and predominantly driven by species turnover, regardless of habitat type or hydrological period. This indicates that species replacement is the primary process structuring fish assemblages across space and time in these systems. Although turnover is often associated with environmental filtering and biotic interactions (Legendre and De Cáceres, 2013), previous studies in temporary aquatic environments (Pazin et al., 2006; Bokhutlo et al., 2021; Costa et al., 2025), together with our results, suggest that instantaneous physicochemical conditions such as temperature and dissolved oxygen are weak predictors of compositional change when measured at the site scale. While we acknowledge that diel variation in these variables was not explicitly quantified, the lack of consistent relationships indicates that short-term limnological fluctuations are unlikely to be the primary drivers of beta diversity patterns observed here, as expected by species sorting operating at evolutionary or historical scales (Jovanovska et al., 2022). Instead, habitat availability and spatial isolation, represented by water volume and distance to the nearest stream, emerged as more relevant drivers.

This pattern is consistent with our hypothesis that environmental conditions would have limited influence on beta diversity components, with the sole exception of roadside ditches during the wet period, where differences in pH were positively correlated with turnover. Rather than reflecting species sorting through environmental filtering, this relationship seems to indicate seasonal mass-effect dynamics, in which recurrent rainfall pulses and temporarily increased hydrological connectivity during the wet period allow species with good dispersal capacities to occupy habitats with less acidic pH levels that are otherwise suboptimal and atypical for the basin. These conditions do not seem to exclude acid-tolerant taxa, what would be expected in an environmental filter, as species such as L. itanhaensis, M. lateralis and R. crassiceps, typically found in pools (Costa et al., 2024) and streams (Ferreira et al., 2014; Giongo et al., 2023) with highly acidic waters, were recorded with considerable abundance (Table 1) in roadside ditches, typically with higher pH values due to higher distance to blackwater streams and more rainfall water input (Costa et al., 2025). In addition, for temporary pools during the wet period, turnover increased with differences in distance to the nearest stream, suggesting spatially structured colonization from distinct source habitats. This pattern likely reflects a differentiation between pool-dwelling taxa occurring farther from streams and stream-associated taxa occurring closer to streams, as flooded areas may be supplied by different sources. Thus, stream-dwelling species may enter pools via flood pulses, whereas pool-adapted taxa likely persist along marginal flooded habitats and avoid being transported to streams as water recedes, consistent with colonization by highly tolerant and dispersal-capable taxa in more isolated pools. However, it is important to emphasize that this wet-period pattern was detected in the year sampled and may vary with inter-annual climatic variation, especially precipitation intensity and flood extent; long-term monitoring is needed to test the consistency of these relationships in metacommunity structuring, and if its consistent between sites and species with distinct dispersal capacity traits (Leibold et al., 2022).

During the dry period, competitive exclusion may play a more immediate role than environmental conditions or colonization source distance. This pattern was evident in both temporary pools and roadside ditches, where differences in water volume were significantly correlated with nestedness, supporting our hypothesis that habitat availability modulates species accumulation under low connectivity in both environments. Under such conditions, dispersal is likely restricted to a limited subset of species with functional traits that confer tolerance to emersion or hypoxia/anoxia. In our system, this potential is most evident in A. peruibensis, a highly amphibious and abundant species able of short-distance terrestrial locomotion, as well as in C. callichthys and H. littorale, which can move across moist substrates and rely on facultative air-breathing. Other species, such as L. itanhaensis, S. marmoratus and P. reisi, may tolerate brief desiccation or air exposure, but are unlikely to sustain active overland dispersal. Therefore, any contribution of terrestrial displacement to metacommunity dynamics during dry periods should be viewed as limited to a few functionally specialized taxa.

In this context, smaller sites tend to harbor subsets of species present in larger ones, consistent with nestedness driven by volume-mediated constraints on coexistence. While roadside ditches generally retain water for longer periods than temporary pools (Costa et al., 2025), variation in ditch volume during the dry season still appears sufficient to structure assemblages, likely reflecting differences in microhabitat availability and carrying capacity, which can constrain the number of species that co-occur locally. This can be expected since increased habitat availability is often associated with greater opportunities for niche differentiation and species accumulation, promoting coexistence at the site level, as suggested by classical niche theory (Taylor, 1997; Spencer et al., 1999). However, inferring competitive interactions among co-occurring species would require approaches such as joint species distribution modeling to investigate the metacommunity internal structure, that is, how these factors operate at site and species level (Leibold et al., 2022). Meanwhile, smaller temporary freshwater habitats likely support only highly amphibious or tolerant species, such as A. peruibensis (Costa et al., 2025), with a selective loss of species. Thus, even in these more persistent artificial habitats, reduced water availability during dry periods may intensify biotic interactions and promote nested patterns in community composition.

In summary, our study provides novel insights into beta diversity components of fish metacommunities inhabiting temporary freshwater habitats, showing that species turnover consistently predominates across both natural and man-made environments, regardless of hydrological season. This pattern highlights the central role of spatial and niche processes, particularly dispersal limitation, episodic connectivity and habitat availability, in structuring communities in dynamic freshwater systems. While environmental filtering via physicochemical variables appears weak, our results indicate that mass effects and local habitat features, such as habitat availability and spatial isolation relative to permanent water bodies, can modulate species replacement and nestedness, especially during periods of reduced connectivity. Although pure spatial distance among sites is also known to strongly influence community dissimilarity in temporary freshwater systems (Costa et al., 2025), our findings emphasize that habitat-related and connectivity-driven processes operate on top of this underlying spatial structure. Together, these results reinforce the importance of considering temporary freshwater habitats as spatially dynamic mosaics in community assembly, a perspective that is broadly applicable to seasonal aquatic ecosystems worldwide.

Implications for conservation

The consistently high turnover observed across temporary habitats readily suggests that conservation strategies should focus on protecting multiple sites rather than prioritizing only species-rich locations (Socolar et al., 2016). However, applying this principle to temporary pools and roadside ditches is challenging due to their inherently dynamic nature. Pools and water-filled ditches can form, dry, and reappear in different locations, with distant shapes and volumes. This spatial and temporal variability implies that protecting isolated sites is unfeasible. Instead, conservation efforts should target the entire floodable mosaic within a basin or sub-basin, encompassing the naturally floodable forest areas where these habitats form and reconfigure over time. Such a systemic approach is particularly urgent in the Preto River sub-basin, since Brazilian Law 12.651/2012 (Art. 61-A) (the so-called Brazilian Forest Code) designates Permanent Preservation Areas only along narrow strips, 20 to 30 meters on each side for small streams (< 10 m of width), intended to protect water resources and biodiversity. This legal framework, even if strictly enforced, would fail to safeguard many pools that arise beyond these strips in floodable forests, as well as virtually all fish assemblages inhabiting roadside ditches along dirt roads. This gap reveals the inadequacy of current regulations to encompass the ecological reality of temporary freshwater habitats both in pristine and anthropogenically altered landscapes.

A more ambitious conservation strategy is needed, that explicitly recognizes temporary habitats as integral components of coastal freshwater systems. This could involve expanding the width of legally protected floodplain vegetation to include the full inundation zone of basins, effectively buffering both natural and man-made temporary sites. In the case of roadside ditches, management actions should focus on monitoring and preventing the spread of non-native species, while also avoiding road paving or ditch channelization, which would disrupt their role as temporary refugia and recolonization sources during extreme periods. Although this study focuses on a single sub-basin, extending similar assessments to other coastal basins is essential. Given the high turnover observed within and likely between basins, a network of smaller protected areas across multiple basins may be more effective than concentrating protection within a single large area, as it would better capture the spatial complementarity of fish metacommunities in temporary freshwater habitats. Such an approach not only aligns with the ecological patterns observed but also challenges the current scope of the Forest Code, highlighting the need to revise or reinterpret existing legal instruments to ensure the conservation of these dynamic and overlooked environments.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher's note

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

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/ffwsc.2026.1776529/full#supplementary-material

Supplementary material Supplementary material at the source

Supplementary material Supplementary material at the source

Footnotes

Edited by: Davide Taurozzi, Roma Tre University, Italy

Reviewed by: Maria Filomena Magalhães, University of Lisbon, Portugal

Swann Felin, National Research Institute of Science and Technology for Environment and Agriculture (IRSTEA), France

Andre Martins, Seduc Amazonas, Brazil