Explaining the anuran beta diversity by pond‐living tadpoles: the role of dispersal limitation and environmental gradients through multiple scales

Determining drivers of beta diversity is a hugely complex task, as it involves processes acting synergistically across multiple scales. We employed a large‐scale standardized protocol to tease apart the environmental and spatial processes driving beta diversity patterns of pond‐living tadpoles across multiple scales.


| INTRODUC TI ON
The metacommunity processes associated with the spatial and temporal variation in species abundance are scale dependent (Chase et al., 2018), so the range of scales that affect species must be explicitly considered (Levin, 1992).This integration essentially includes variations in geographical grain and extent.Increasing grain size results in increases within-grain heterogeneity while decreasing intergrain variation (Barton et al., 2013).In contrast, increasing spatial extent amplifies environmental gradients and increases the distance between sampling units (Barton et al., 2013;Wiens, 1989).Although several studies have been conducted to investigate beta diversity patterns at various scales, most of them have focussed on snapshots at a single scale (e.g.Dalmolin et al., 2019) or multiple scales within the same biogeographical region (e.g.Loewen et al., 2020, but see Gálvez et al., 2023).More importantly, large-scale studies are rarely conducted with a standardized design, which adds extra challenges to distinguishing the effects caused by differences in sampling design in beta diversity from real differences in compositional changes between sites affected by niche or neutral processes.
Several studies have demonstrated that niche-based processes (i.e.species sorting) have greater predictive power at metacommunity scales, whereas neutral processes (ecological drift and stochasticity) are prevalent at large spatial scales (e.g.biogeographic; Gonçalves-Souza et al., 2014;Viana et al., 2016;Heino et al., 2017).This prevalence, however, deserves attention because there is a continuous variation in environmental and spatial properties of metacommunities that will change in strength depending on (i) the number of measured predictor variables, (ii) the degree of environmental heterogeneity, (iii) dispersal rates and connectance among patches and (iv) the regional pool size (Leibold & Chase, 2018;Viana & Chase, 2019).On the species perspective, their responses to environmental and spatial processes depend on some characteristics such as the level of specialization, dispersive capacity, rarity or endemism (Pandit et al., 2009;Siqueira et al., 2012).However, some studies have shown that rare and abundant species might respond equally to environmental processes but vary in their responses to spatial processes (Heino & Soininen, 2010;Siqueira et al., 2012).
Dispersal is critical for understanding the context dependence of species responses to different metacommunity processes (Hayes & Anderson, 2018;Vilmi et al., 2017).The geographic range of a species reflects the interaction of limiting environmental circumstances (i.e.resources and conditions) and dispersal-extinction dynamics through time in a metacommunity (Jocqué et al., 2010).In general, it is expected that there is a trade-off between habitat specialization and dispersal limitation of species, allowing us to connect the interplay between spatial and environmental processes while considering geographic grain and extent (Jocqué et al., 2010).On the one hand, reduced dispersion ability should often result in a smaller range size because dispersal ability, niche width and distribution are connected (Brown et al., 1996), in this way narrow-ranged species, that is, those occurring in a single geographic area, tends to support a small range of climatic conditions (Primack, 2006).Conversely, widespread species that have a wider geographical distribution and occur in multiple biomes might support a wide climatic variability (Gaston, 2003;Jocqué et al., 2010;Lester et al., 2007).In metacommunities, generalist species are more influenced by dispersal-based processes, while more specialized species are mainly influenced by environmental gradients (Pandit et al., 2009).However, there is still limited evidence that processes acting at the local scale are also prevalent at larger scales when the spatial grain is kept the same across spatial scales.
Although several studies have assessed the influence of environmental and spatial processes on anuran metacommunities (e.g.Dalmolin et al., 2019;Provete et al., 2014), comparatively few studies investigated their effects at multiple scales (e.g.Melchior et al., 2017) using standardized protocols (e.g.Bitar et al., 2017).
Moreover, standardized protocols comparing multiple metacommunities across different biogeographic regions have been comparatively seldom performed (Gálvez et al., 2023;Siqueira et al., 2020).Studies using standardized sampling protocols were geographically restricted to a specific biome, with no studies on a macroecological scale using primary data.Differences in pond morphology, vegetational structure and predator density have a strong effect on anuran beta diversity at small scales (Dalmolin et al., 2019;Prado & Rossa-Feres, 2014;Provete et al., 2014).Furthermore, although in several organisms the spatial influence on beta diversity increases with increasing spatial extent (Melchior et al., 2017;Soininen, 2016), it has been demonstrated that anurans are spatially structured even at small spatial scales (Dalmolin et al., 2019;Provete et al., 2014).Therefore, recent studies have advocated that research combining multiple processes affecting anurans is urgent to advance the comprehension about one of the most endangered taxa in the world (Melchior et al., 2017;Pelinson et al., 2022).Additionally, amphibians are ectothermic, possess permeable skin and have eggs without shells.Moreover, the Editor: Xuan Liu narrow-ranged and widespread species respond in a consistent manner to environmental and spatial dynamics, showing consistent responses to niche-and dispersal processes.This evidence can offer important guidelines for their conservation even for large-scale management that includes multiple biomes.

K E Y W O R D S
biogeography, metacommunity, spatial grain and extent, tadpoles, tropical biomes tadpoles of many species inhabit lentic environments (which have discrete boundaries; Wells, 2007).These characteristics make amphibians a highly valuable biological model for testing metacommunity theory.As far as we know, no other research has evaluated the environmental and spatial processes at multiple scales in such a huge geographical scope using primary data obtained with the same sampling protocol.
In this study, we used a dataset about pond-living tadpoles obtained with a country-wide standardized protocol across five different biomes (5000,000 km 2 ) in Brazil.Here, we fixed the sampling unit (a pond) and varied sampling extent.Specifically, we asked the following questions: 1. What role do pond-level features, climate and spatial variables play in pond-living tadpoles' beta diversity across geographical scales?We expect that at the metacommunity scale, where dispersal is not limited, environmental gradients will contribute more to beta diversity, whereas spatial processes will contribute more at the regional and subcontinental scales due to dispersal limitation (Prediction 1).This suggests that when comparing metacommunities from different biomes, species sorting will be greater at the metacommunity scale, even though low-level processes differ between sites and biomes.On the other hand, we anticipate neutral process-driven beta diversity at regional and subcontinental scales because of dispersal constraints, resulting in a historical contingency (high-level processes, sensu Vellend, 2016).
2. Do environmental and spatial processes have different influences on narrow-ranged and widespread species?We expect that spatial processes will have a greater impact on widespread species, while environmental processes will have a greater impact on narrow-ranged species (Prediction 2).We predict that environmental factors will be more critical for narrow-ranged species, as if they are habitat specialists, they select environments according to specific requirements.Moreover, if narrow-ranged species present dispersal ability limitation, this may prevent them from occupying favourable environments but more distant.On the contrary, spatial restrictions should have a greater influence on the widespread species due to the increase in spatial extension and, consequently, limiting dispersion due to the great distances between the biomes.We will be able to determine whether the environmental processes that drive anuran beta diversity are consistent across biomes by answering these two questions.

| Study sites
During the development of SISBiota ' Brazilian-tadpoles' project (2012Brazilian-tadpoles' project ( -2014)), we sampled tadpole communities in 692 water bodies, ranging from −39.02 to −0.12 and −1.78 to −25.26 decimal degrees of longitude and latitude, respectively, in five Brazilian biomes (Olson et al., 2001): tropical moist broadleaf forests (here called Amazon), tropical and subtropical moist broadleaf forests (here called Atlantic rain forest), xeric shrublands (here called Caatinga), savanna (here called Cerrado) and flooded grasslands (here called Pantanal; Figure 1).To test our predictions, we included sites with at least eight sampled lentic water bodies (Mean 19 ± 16 SD).The water bodies were separated by 4.1 km (±4.4), which can be considered a metacommunity because moving individuals can potentially reach all water bodies within a given site (see details below).We collected in three sites per biome, except to Cerrado where seven sites were sampled (Figure 1; see Appendix S1, Table S1 in Supporting Information).The final dataset consists of 358 ponds, including 129 in the Amazon, 33 in the Atlantic Forest, 49 in Caatinga, 72 in the Cerrado and 75 in the Pantanal.
The Atlantic rain forest and Amazon are both humid tropical forests with annual rainfall ranging from 1300 to 2500 mm in the Atlantic rain forest and 1900 to 3000 mm in the Amazon.Both areas have a humid tropical climate, with no marked dry season.
The average annual temperature in the Atlantic rain forest ranges between 12° and 26°C, whereas it fluctuates between 22° and 26°C in the Amazon (Alvares et al., 2013).Caatinga has a semiarid climate, with most of the region receiving between 600 and 1000 mm of rain per year, concentrated between February and May, and the annual mean temperature ranging between 25 and 30°C (Alvares et al., 2013).The Cerrado is known as Brazilian savanna with distinct phytophysiognomies that are linked by soil properties and waterlogging and fire regime (Xavier et al., 2019).
The Cerrado has a tropical climate with a well-marked dry season with annual rainfall varying between 1300 and 2500 mm (Alvares et al., 2013).The temperature ranges from 12° to more than 26°C.
The Pantanal is the world's biggest floodplain, with recurrent and prolonged flooding.The flood began in December and ended in April.The climate is markedly seasonal, with a hot and rainy period from October to April and a dry period from May to September.The annual precipitation ranges from 1300 to 1900 mm/year, while the average annual temperature ranges from 22° to 26°C (Alvares et al., 2013;Marengo et al., 2016).

| Sampling protocol
We employed standardized protocols for sampling of the tadpoles' communities and for morphology characterization of the ponds.
Tadpoles were sampled using a wire mesh dipnet (3 mm 2 ) with 30 cm diameter.The sites and ponds were chosen to showcase a wide range of aquatic environments according to availability and logistical options.There were sampled ponds smaller than 5000 m 2 with distinctive features (e.g.aquatic vegetation, depths and hydroperiods) and those that are not connected to other ponds.With a one-hour sampling window, the largest number of microhabitats was sampled in large water bodies at the margin and inside ponds.Tadpoles were euthanized by immersion in a 10% lidocaine and conserved in a preservative solution (1:1 alcohol (70%) and formalin (15%)).
This procedure was needed considering the distance between the sampled pond and the laboratory of researchers, since tadpoles are very sensitive, and it is difficult to transport them safely through large distances.Moreover, as aquatic organisms constituted only by soft tissues, they suffer a fast decomposition process after dying.
The tadpoles were identified by comparing them to species from biological collections, using taxonomic keys (e.g.Rossa-Feres & Nomura, 2006) or expert knowledge.When individual identification at species level was not feasible, the species was compared with all other species collected in the entire database to ensure that it was not the same species already collected elsewhere.The tadpoles are deposited in biological collections (see Appendix S1).

| Spatial grain and extent
In this study to define the spatial scale, we used ponds as grains, which naturally vary in size.However, we consistently employed the same sampling grain across all three spatial scales analysed, thereby varying the scales solely based on geographic extent.The spatial scale was defined as a grain, in this case pond and varying extent, and did not include multiple ponds as the same sampling unit.We therefore assumed that ponds, despite having different areas, represented the same grain because we did not vary the sampled area by sampling them over the course of an hour.We included pond size as a predictor in the analysis to see whether the variation in grain size at different spatial extents could explain the variation in the structure of metacommunities.The spatial extent, in turn, varies hierarchically from the metacommunity scale to regional and subcontinental.The metacommunity scale (within biome) includes at least eight lentic water bodies and the mean distance among them was 4.1 km (±4.36).The regional scale (biome) includes all ponds sampled within a given biome, and the average distance among these ponds was 127.9 km (±247.6).Lastly, the subcontinental scale (biogeographic) considers all sampled ponds in the five different biomes, which have a mean distance of 1889.5 km (±967.36).

| Defining narrow-ranged and widespread species
We grouped species into 'narrow-ranged' or 'widespread' according to their spatial distribution across the Brazilian biomes.All species that occurred in more than one biome was considered as widespread, whereas those species occurring in a single biome were considered as narrow-ranged (see similar definition in Leibold et al., 2022).Due to the large distances involved, our field samplings did not cover a large area in each biome.In this way, species considered narrowranged based on primary data may occur in other biomes based on secondary data.Thus, we checked the categorization based on primary data comparing to that one based on secondary databases (AmphibiaWeb, 2021;Frost, 2021), and the results were virtually the same.Species that were not identified to the species level were excluded from these analyses.Based on a primary data source, 49 and 57 of the 131 sampled species were classified as narrow-ranged and widespread species, respectively, while 25 were not examined because they were not identified to the species level.

| Pond-and site-level predictor variables
All sampled ponds were characterized using a standardized environmental descriptor of the structural heterogeneity of the habitat and the presence/absence of fish predators.The pond features were assessed using the following variables: (1) hydroperiod (temporary or permanent); (2) depth (maximum depth); (3) percentage of each marginal vegetation type (none, grass, erect herbaceous, shrub, cattail and tree); (4) percentage of canopy coverage over the water body; and (5) presence or absence of predatory fishes.The types of vegetation inside the water body and along the 1 m width of the perimeter of pond were estimated visually and classified into the categories 1%-20%, 21%-40%, 41%-60%, 61%-80% and 81%-100%.The canopy coverage estimate was calculated based on five measurements: north, south, east, west and in the centre of the water body, using a spherical densiometer.Canopy cover was not measured in the Cerrado, so it was not included in the analyses of this biome.
The influence of climate on beta diversity was only evaluated at regional and subcontinental scales since all ponds are subject to the same climatic effect at the metacommunity scale.Previous empirical studies have shown that climate variables, particularly temperature and precipitation, are the primary predictors of anuran species distribution and richness, as they affect anuran their performance, development and survival (Buckley & Jetz, 2008;Ceron et al., 2020;Chejanovski & Wiens, 2014;Wells, 2007).Climatic variables, particularly temperature and precipitation, have a profound impact on anuran reproduction (Dastansara et al., 2017;Wells, 2007).Therefore, we used six temperature and precipitation variables extracted from Worldclim (Fick & Hijmans, 2017): (1) average annual temperature (bio1), as an indicator of the energy of the environment; (2) seasonality in temperature (bio4), related to the thermal tolerance of organisms; (3) difference between the average maximum monthly temperature and the minimum monthly temperature (bio2), as measures of the limit of tolerance to cold and heat; (4) annual precipitation (bio12), a measure of water availability; (5) seasonality in precipitation (bio15), as a measure of seasonal water availability; and (6) precipitation in the driest quarter (bio14), as a measure of water restriction (see Table S2 in Appendix S1).We performed a preanalysis procedure to reduce multicollinearity between climatic and pond-level predictors (see details in Appendix S1).

| Spatial predictor variables
Moran's eigenvector maps (MEM) were used to describe the spatial relationships among sites and to provide a set of spatial predictors (Dray et al., 2006).These predictors are produced by performing an element-wise product of a centred spatial weighting matrix (A) and a connectivity matrix (B), yielding the matrix W (Dray et al., 2006).
There are several possible metrics to generate the matrix B (e.g.Delaunay triangulation, Gabriel graph and relative neighbourhood graph) and A (e.g.neutral, linear or concave-down function; Dray et al., 2006;Bauman et al., 2018), but most studies have been arbitrarily selected among them.However, we applied the approach proposed by Bauman et al. (2018), which enables the selection of the best-performing method for each set of spatial points (see more details in Appendix S1).

| Dependent variables: beta-diversity estimates
Many studies have shown that when comparing various sites, changes in the size of local (α) or regional (γ) species pools or species abundance affect beta-diversity patterns (Myers et al., 2013).This indicates that increasing α diversity along a gradient may change β diversity without impacting ecological differences across sites (e.g.Chase et al., 2011).Even though we standardized the sample in all ponds collecting tadpoles, differences in the number of individuals that vary with pond size, or the regional pool may limit our ability to identify metacommunity assembly processes that affect beta diversity (Chase et al., 2011;Myers et al., 2013).Therefore, we used the Raup-Crick (βRC) method with presence-absence data to distinguish between two different sources of beta-diversity variation: (1) the variation caused by differences in alpha diversity (random sampling effects, which is affected by pond size) or (2) the variation affected by niche-based processes (Chase et al., 2011;Myers et al., 2013).This method compares whether two communities are more dissimilar than expected by chance, as follows: (i) compute the observed alpha diversity for all pair of sites (α1 and α2) and the number of shared species between them; (ii) compute the number of species in the regional pool and the occupancy of each species (i.e. the proportion of sites occupied by each species); (iii) randomize 10,000 times the values of α1 and α2 using a probability value that takes into account the proportional occupancy of each species (Chase et al., 2011).Although the Raup-Crick index is conditioned to the variations in alpha diversity between ponds, its results are similar with classic indices (e.g.Jaccard and Sorensen), with values ranging from 0 (ponds with entirely similar species composition) to 1 (ponds with entirely distinct species composition) in a triangular matrix pond by pond.

| Statistical analyses
We performed the analysis at three spatial scales: (1) metacommunity scale, which compares beta diversity in n ponds within each site.
At this scale, just environmental (pond-level) and spatial variables (MEMs) were included in the model; climate was excluded because it was the same for all ponds at this scale; (2) regional scale compares beta diversity in n ponds across m sites within each biome; in these scale, besides pond and spatial variables, we also considered climate variables as predictors; (3) subcontinental scale, which compares beta diversity in 19 sites from five biomes utilizing the same variables used in the regional scale as predictors.Furthermore, at the subcontinental scale, we evaluated the effects of pond features and climatic and spatial variables on the beta diversity of narrowranged and widespread species separately.The first step is to assess the influence of the predictors (environmental, climatic and spatial variables) on beta diversity using partial distance-based redundancy analysis (dbRDA).To test the effects of the environmental and spatial predictor variables on beta diversity, we utilized a permutationbased procedure to assess their significance at each scale (anova.ccafunction from the vegan package: Oksanen et al., 2019).We then used the variance partitioning method on each scale to determine the relative importance of environmental (pond-level heterogeneity) and climate and spatial variables (MEMs) as drivers of anuran beta diversity.
The variation partitioning method computes the variation explained by individual and shared fractions after obtaining the R 2 values for each step (Peres-Neto et al., 2006).The fractions [a] and  (Clappe et al., 2018).The main cause for both aforementioned issues is the spatial autocorrelation, which increases the estimation of neutral dynamics in the fraction [ab] (i.e.environmental pure + spatially correlated environmental variation).Therefore, we used the method proposed by Clappe et al. (2018) to obtain unbiased estimators, which were detailed in Appendix S1.We used the R version 3.6.3and the packages vegan (Oksanen et al., 2019), ade4 (Dray & Dufour, 2007) and adespatial (Dray et al., 2019) to run all analyses.

| RE SULTS
We collected tadpoles of 130 species belonging to 10 families (Table S3 in Appendix S2), which represents 12.5% of the anuran species known to Brazil.The most species-rich family was Hylidae (76 species), followed by Leptodactylidae (22 species), Microhylidae (nine species), Bufonidae (eight species), Phyllomedusidae (six species), Odontophrynidae (four species), Aromobatidae (two species) and Centrolenidae, Ceratophryidae and Dendrobatidae with just one species each.We registered 2527 individuals from 20 species in the Amazon, 18,933 individuals from 24 species in the Atlantic rain forest, 7727 individuals from 36 species in Caatinga, 15,884 individuals from 47 species in Cerrado and 27,181 individuals of 23 species in the Pantanal.

| Effects of varying geographical extent on the relative importance of pond-level features, climate and spatial variables to anuran beta diversity
At the subcontinental scale, we found that spatially structured environmental variation in pond and climate features explained 23% of the variation in beta diversity, while the pure spatial component explained 16%, and the pure environmental component explained just 2% (p = .001,Figure 2).Both climate and pond-level features had a weak effect on beta diversity at subcontinental scale (Table 1; Table S4 in Appendix S2).However, all variables (except precipitation in the driest quarter and canopy cover) were significant (Table 1), whose joint effects can explain the great influence of the spatially structured environmental variation.At the regional scale, we found that the pure spatial component explained a higher portion of the metacommunity structure (11% ± 5%), followed by the component representing the spatially structured environmental variation (10% ± 4%) and the pure environmental component (4% ± 4%; Figure 2).The pure spatial component explained 16% of the beta diversity of anurans in Amazonia, 14% in Pantanal and 12% in Cerrado (Figures 2 and 3; Table 1; Tables S4 and S5 in Appendix S2).
On the contrary, the spatially structured environmental variation explained a greater proportion of the metacommunity structure in Caatinga (12%) and Atlantic Forest (6%; Figures 2 and 3).
Our findings at the metacommunity scale revealed that the environmental component was more prevalent than the spatial component in determining the beta diversity of anurans (Tables S4 and   S5; Figures 2 and 3).The environmental component was the primary driver of metacommunity structure in 50% of the studied sites (two in Atlantic rain forest, Caatinga, Cerrado and Pantanal and one in the Amazon; Figure 3).In 33% and 17% of metacommunities, respectively, spatially structured environmental variation and spatial components were the greatest beta diversity predictors.The environmental component did not explain the variation in beta diversity in five metacommunities (26%), and the pure spatial component and the spatially structured environmental variation did not explain any of the variation in species composition in 53% (n = 10) and 21% (n = 4) of metacommunities, respectively.The importance of environmental predictors on each metacommunity varied greatly across metacommunities.Depth was crucial in four metacommunities, whereas canopy cover and marginal vegetation were relevant drivers in two metacommunities, and predatory fish was important in a single metacommunity.

| Effects of pond-level features, climate and spatial variables on narrow-ranged and widespread species
We found that the effects of environmental and spatial components on species differed slightly depending on species range size.For narrow-ranged species, environmental and spatial variables explained 44% of the variation in beta diversity (p = .001,Table S6 in Appendix S2).From this explained variation, the pure environment, the spatially structured environment and the pure spatial component accounted for 7%, 47% and 46%, respectively.More specifically, the climatic variables (average annual temperature, difference between the average maximum monthly temperature and the minimum monthly temperature, seasonality in temperature, annual precipitation and seasonality in precipitation) and pond-level variables (depth, PC1 veg and PC2 veg ) were relevant in driving beta diversity of narrow-ranged species, being average annual temperature, seasonality in temperature and annual precipitation the three most important environmental variables (Table 2).

F I G U R E 2
Results of variation partitioning with partial redundancy analysis showing the percentage of explanation of the pure environmental [E], pure spatial [S] and spatially structured environmental variation [E | S] components on anuran beta diversity across spatial extents: subcontinental (biogeographic), regional (biome) and metacommunity (within biomes).
TA B L E 1 p-Value of environmental and macroclimatic variables predicting anuran beta diversity at a subcontinental and biome scales, resulting from partial distance-based redundancy analysis.Conversely, environmental and spatial variables explained 36% of the beta-diversity variation in widespread species (p = .001,Table S6 in Appendix S2).From this explained variation, the pure environment, the spatially structured environment and the pure spatial component accounted for 7%, 53% and 40%, respectively.
In particular, the climatic variables and pond-level variables were relevant in driving beta diversity of widespread species.Accordingly, depth, marginal vegetation types (PC2 veg ), average annual temperature and difference between the average maximum monthly temperature and the minimum monthly temperature were the most important environmental variables (Table 2).In summary, both narrow-ranged and widespread species were influenced by the same climatic and pond variables; however, seasonality in precipitation and water body depth were only important for narrow-ranged species.

| DISCUSS ION
This study uses a standardized sampling protocol in a subcontinental scale, including five Brazilian biomes to investigate how the geographical extent explains the relative contribution of niche-and Results of the variation partitioning of partial distance-based redundancy analysis of the anuran beta diversity showing the influence of environmental (Env) and spatial variables (Spa) and spatially structured environmental variation (Env|Spa) in each metacommunity (circles) and the median of each variable for each biome (diamonds).TA B L E 2 Partial distance-based redundancy analysis for the narrowranged and widespread species separately.

Narrow-ranged
dispersal-based processes to beta diversity of pond-living anuran tadpoles.We revealed that when the grain is kept the same across the scales of analysis and the geographical extent varies, the relative impact of niche-based processes decreases whereas dispersal-based processes gain relevance for beta diversity from metacommunity to subcontinental scales, supporting Prediction 1.
We verified that spatially structured environmental variation was more or so important than the pure spatial component at regional and subcontinental scales, respectively.However, we did not support Prediction 2 because narrow-ranged and widespread species responded equally to pure environmental variables but differently to the spatially structured environmental variation (that was 1.13 times more important to widespread species) and pure spatial (that was 1.15 times more important to narrow-ranged species) components.
Our findings reinforce that spatial scale matters and it is pivotal in understanding metacommunity dynamics across scales (Gonçalves-Souza et al., 2014;Melchior et al., 2017;Pelinson et al., 2022;Viana & Chase, 2019).Anurans' tadpoles responded strongly to spatial environmental gradients across scales, whereas niche-based processes (such as habitat filtering and biotic interactions) were only relevant at small spatial scales.As a result, dispersal limitation is an important aspect of anuran life history because it drives metacommunity structure.However, at small spatial scales where dispersion is not limited, and surprisingly to regional scales (Pantanal, Caatinga and Atlantic Forest biomes) niche-based processes were important, showing the joint action of dispersal and niche-based processes in the dynamics of metacommunities.

| Effects of varying geographical extent on the relative importance of pond-level features, climate and spatial variables on tadpoles' beta diversity
As expected, we found an effect of spatial extent on the prevalence of niche-based and dispersal processes.The increasing of the spatial extent captures greater environmental variability by incorporating different biogeographical areas increasing habitat heterogeneity (Leibold & Chase, 2018;Nekola & White, 1999;Wiens, 1989), resulting in a stronger influence of niche-based processes.However, some review studies have indicated that increasing geographical extent implies a reduced environmental effect, as well as a larger spatial effect due to dispersal limitation (Cottenie, 2005;Soininen, 2014Soininen, , 2016)).As a result, there is an increase in the importance of biogeographic processes, such as historical contingency and speciation-extinction events (Barton et al., 2013;Hortal et al., 2010).Indeed, long distances among ponds on the subcontinental and regional scales limit species' ability to colonize new habitats due to geographic barriers (e.g.isolation by mountain ranges and rivers) resulting in greater beta diversity at large scales (Barton et al., 2013;Godinho & da Silva, 2018).Previous studies reveal that spatial processes outperform the environmental ones in shaping beta diversity of amphibians (Knauth et al., 2019;Provete et al., 2014;Silva et al., 2014), indicating the influence of dispersal limitation on these organisms, even when accounting for different spatial extent and biomes.However, differing from those previous findings, the spatially structured environment variation was the most important driver of anuran beta diversity at subcontinental scale and so important as the spatial component at regional scale.
This result evidences a great influence of environmental variation across space, indicating that the species dispersal limitation is due mostly to environmental features than to simply the distance among habitats.
The nonrandom distribution of species along spatially structured environmental variation supports a joint action of climaterelated dispersal limitation at both biome and subcontinental scales (Benício et al., 2021;Jocqué et al., 2010).This predominance of the spatially structured environmental variation was also found in most metacommunities at the regional scale.Indeed, the relationship between environmental factors and spatial distance for animals and plants within and across biomes has been documented because many environmental variables are spatially structured (Legendre, 1993;Legendre et al., 2009).Thus, it seems that the interaction between species sorting with efficient dispersal (i.e. species tracking preferred environmental conditions) typical at the metacommunity scale might be generalized to understand patterns at the biogeographic scale (Declerck et al., 2011;Gonçalves-Souza et al., 2014;Van der Gucht et al., 2007).Furthermore, this strong effect of spatially structured environmental variation on beta diversity may be explained by the trade-off in the natural history of anurans.On the one hand, skin respiration and desiccation risk make them dependent on climatic and pond features (e.g.Wells, 2007, da Silva et al., 2011, da Silva et al., 2012).On the other hand, these characteristics also limit their dispersion ability (e.g.habitat split; Becker et al., 2007).In addition, by sampling tadpoles, we are partially evaluating reproductive success, which theoretically indicates success in habitat selection; therefore, habitat selection would have already passed through very strict environmental filters, related to reproduction habitat and survival of spawns and tadpoles.Reinforcing this interpretation, even if we downscale to the regional scale, three from five biomes (Caatinga, Cerrado and Atlantic Forest) had a relevant contribution of the spatially correlated environmental variation.
On the metacommunity scale, niche-based processes were important in all biomes (except for metacommunities in the Amazon).The importance of pond features on the metacommunity scale has been observed in several studies in different biomes and ecoregions (e.g.Dalmolin et al., 2019;Gálvez et al., 2023;Knauth et al., 2019;Provete et al., 2014;Ramalho et al., 2021).Pond depth, hydroperiod, canopy cover and types of marginal vegetation determine species richness and composition (Silva et al., 2014;Provete et al., 2014;Queiroz et al., 2015).However, the importance of each pond-level feature that was important in structuring metacommunities varied between metacommunities within the same biome and between metacommunities from different biomes.For example, in the metacommunities of Barranco Alto and Porto Murtinho, both in the Pantanal, beta diversity is driven by different pond-level features, such as depth and marginal vegetation, respectively.In this way, idiosyncratic variations in habitat selection for breeding by adult anuran species result in variation in the relationship between pond-level features and species composition among ponds on a small scale.We also confirmed that the widespread pattern of niche-based processes is more relevant from large to small scales in many organisms (Cottenie, 2005 Herein, even though we found a significant effect of pondlevel features at the metacommunity scale, the strong influence of environmental variation along the space at subcontinental and regional scales reinforces the role of niche differentiation in determining anuran patterns of beta diversity in wide spatial scales.However, the influence of the sampling design cannot be disregarded.Because we collected pond-living tadpoles regardless of the spatial extent, this might reinforce the relevance of biogeographical signatures of amphibian communities that correlate with species tracking preferred environmental gradients over large temporal and spatial scales.Furthermore, because our models only evaluate linear correlations between spatial and environmental variables, our findings could not be generalized to account for potential nonlinear patterns between these predictors and beta diversity.

| Effects of pond-level features, climate and spatial variables on narrow-ranged and widespread species
Narrow-ranged and widespread species respond to niche and dispersal-based processes almost in the same way.Previous studies comparing narrow-ranged and widespread species, or even rare and common species, have validated this finding (Pandit et al., 2009;Siqueira et al., 2012;Newbold et al., 2018).Narrowranged species are more spatially constrained than widespread species at the subcontinental scale.Despite the differences in spatial range distribution, spatially structured climate variation was 1.1 times more relevant to widespread than to narrow-ranged species.Interestingly, both narrow-ranged and widespread species were under the influence of same set of climatic variables, related to temperature and precipitation, with the exception of the seasonality in precipitation that was important only for narrowranged species.The effect of the pure environmental component on both species groups was the only similar effect.Whereas widespread species are subject to greater climatic variations across their geographic range compared with narrow-ranged species, surprisingly both groups of species seem to respond to the same climatic filters.These findings suggest a promising direction for further research.
First, the differential response of narrow-ranged and widespread species may not be strong as previously expected.Indeed, previous studies have shown that species restricted to a given habitat or that are geographically rare respond to environmental changes in the same way that widespread species (Siqueira et al., 2012).The inclusion of dispersal and biological traits, as reproductive modes, is a further step in determining whether narrow-ranged and widespread species respond differently to environmental and spatial processes (Kneitel, 2018).Second, regardless of whether species are narrow-ranged or widespread, spatial processes (pure or correlated with environmental variables) were critical in determining amphibian community structure.The key component impacting widespread species was the spatially structured environmental variation, implying that these species are tracking large-scale climate variation caused by spatial gradients.At a subcontinental scale, neutral dynamics (dispersal limitation) and species sorting via efficient dispersal (i.e.species tracking preferred environmental features that are spatially structured; Winegardner et al., 2012;Provete et al., 2014) may be the primary drivers of amphibian species distribution.In contrast, the pure spatial process was the most powerful factor influencing the beta diversity of narrow-ranged species, probably reflecting their more specific climatic and habitat requirements.Taken together, these findings highlight that dispersal ability and species range size (or ecological specialization) determine beta-diversity patterns at large scales, such as the latitudinal gradient (Jocqué et al., 2010).
Large-scale standardized observational and experimental protocols are fairly rare in ecology (Romero et al., 2018) and even more rare in tropical ecosystems (Clarke et al., 2017;Culumber et al., 2019).Many studies emphasize that these limitations may affect our ability to predict how global climate change and habitat loss alter the megadiverse tropical forests.In this study, we developed a standardized sampling protocol to collect pond-living tadpoles at a broad spatial scale (five million km 2 ) represented by five Brazilian biomes using the same grain and varying spatial extent, which allows us to tease apart how environmental and spatial processes affect the beta diversity of amphibians.
We recorded 12% of the anurans biodiversity recognized for the country (Segalla et al., 2021) and more than 17% of the species that depend on water bodies for reproduction.In particular, we demonstrated that increasing the geographical extent augments the contribution of spatial processes and spatially structured environmental variation of amphibian beta diversity.The dispersal limitation associated with amphibian species tracking preferred environmental conditions highlights the joint interaction of nichebased and neutral processes in metacommunity dynamics, which

B I OS K E TCH
Our research team includes biologists researching on a diverse array of topics related to biogeographic, community ecology, natural history and conservation biology.We aim to increase understanding of the processes that generate and maintain the biodiversity patterns at different spatial scales, in order to support conservation actions for anurofauna.
[c] represent the variation explained solely by environmental and spatial variables, respectively.The fraction [b] captures the shared variation explained by environmental and spatial variables.The unexplained fraction is represented by the remaining variation[d]    (see details inPeres-Neto et al., 2006 andClappe et al., 2018).Although variation partitioning has been widely used in the ecological research, certain studies have demonstrated (i) its limitations in inferring metacommunity processes (e.g.Gilbert & Bennett, 2010, but   see Diniz-Filho et al., 2012)  and (ii) its large Type I errors when estimating the environmental component 3, 422537/2016-0 and 563075/2010-4; Fundação de Amparo à Pesquisa do Estado de São Paulo, Grant/Award Number: 2010/523217; Universidade Federal da Integração Latino-Americana, Grant/Award Number: edital PRPPG 110/2018, PRPPG205/2021 and PRPPG77/2022 Abbreviations: Bio 1, average annual temperature; Bio 2, difference between the average maximum monthly temperature and the minimum monthly temperature; Bio 4, seasonality in temperature; Bio 12, annual precipitation; Bio 15, seasonality in precipitation; Dep, Depth; PC1, first axis of the PCA of marginal vegetation types; PC2, second axis of the PCA of marginal vegetation types.