Ecological constraints and trait conservatism drive functional and phylogenetic structure of amphibians larvae assemblages in the Atlantic Forest

Investigate how ecological and/or evolutionary factors could affect the structure of ecological communities is a central demand in ecology. In order to better understand that we assessed phylogenetic and functional structure of 33 tadpole communities in the Atlantic Forest coastal plains of Southeastern Brazil. We tested the assumption that phylogenetic conservatism drive tadpole traits. We identied 32 communities with positive values of phylogenetic structure, with 18 of those being signicantly clustered. Twelve of 33 communities showed aggregated functional structure. Trait diversity was skewed towards the root, indicating phylogenetic trait conservatism and evolutionary factors as important drivers of tadpoles community structure. Six out of 11 environmental variables were selected in the best explanatory model of phylogenetic structure. Water conductivity, external and internal diversity of vegetation structure, canopy cover, and dissolved oxygen were negatively related with phylogenetic clustering, whereas presence of potential sh predators was positively related. Four of those environmental variables and area were also included in the best explanatory model of functional structure. All variables represent factors related to performance, survivorship, and distribution of anuran communities. From the 12 functionally structured communities, 10 were also phylogenetically structured. Thus, environmental factors may be acting as lters, interacting with phylogenetically conserved species traits, and driving linage occurrence in tadpole communities. Our study provides evidence that phylogenetic and functional structure in vertebrates are a result of interacting ecological and evolutionary agents, resulting in structured anuran assemblages.


Introduction
Communities structure are the expression of assembles rules. A crucial demand to ecologists is to disentangle the roles of contemporary and historical factors and their interactions as key determinants of community structure [1][2][3][4] . In the last decades, there has been an increased attention on the importance phylogenetic information to enhance our understanding of community structure 3,4-8. The complex relationship between these intertwined factors makes understanding the processes driving the functional structure of ecological communities a major scienti c challenge [9][10][11] . Therefore, it is interesting to integrate different approaches in order to expand our comprehension of the underlying processes originating and maintaining assemblages, including niche-based and evolutionary processes. Assessing phylogenetic structure of ecological communities gives us insights on evolutionary (e.g. natural selection) and ecological (e.g. competitive exclusion) mechanisms and processes, which can determine or in uence community structure 3,8,[12][13][14][15] . Additionally, functional structure of communities can reveal potential ecological mechanisms of community assemblage, as this structure is potentially the interaction of species functional traits and their habitats 4,9,16,17. Phylogenetic structure of ecological communities can be assessed by quantifying phylogenetic distances among species in a given assemblage and comparing the observed distances to null models 3 . If the observed phylogenetic mean distance is higher or lower than expected by null scenarios, one can state that there is a signi cant structure in that assemblage. The Mean Pairwise Distance (MPD) is a standardized metric of mean pairwise phylogenetic distance of taxa in a sample, which quanti es the phylogenetic clustering (positive values) or overdispersion (negative values) of taxa in a particular community 3,18 . This metric and its relationship with ecological variables has been used in numerous studies and is considered as the rst step to uncover ecological (i.e., competitions or environmental ltering) and evolutionary mechanisms (i.e., in situ speciation or character displacement) driving phylogenetic structure of communities 14,[19][20][21][22] It has been suggested that environmental variables can act as ecological lters, for instance, constraining the assemblage of communities, i.e., progressively selecting species better adapted to local conditions from the regional pool, so, if species traits associated to these variables are phylogenetically conserved, it could lead to a phylogenetically clustered structure community 3 . However, interspeci c competition could also be a main process leading to a clustered phylogenetic pattern, for instance, if closely related species have ecological differences that confer competitive advantage to one species in relation to another 18 . Thus, to correctly investigate drivers of assemblage structure, we should not just infer the processes from patterns of phylogenetic structure but evaluate the relationships among species traits and their environment.
Amphibians occurrences are affected by environmental variables such as canopy cover, that could indirectly affect the performance and occurrence of anurans in ponds [23][24][25] . The Vegetation heterogeneity of ponds is another important determinant of anuran reproduction or promotes micro-spatial heterogeneity, increasing the availability of sites for foraging and for seeking refuge from predators 26-29 . And nally, physico-chemical parameters of water, including pH and dissolved oxygen are known to directly and indirectly in uence several aspects of anuran biology, including some species amphibians tadpole development and survival, consequently affecting occurrence of species in ponds [30][31][32][33][34][35][36] .
This well-established relationship between environmental variables and amphibians is commonly studied via ecomorphological traits or attributes describing behavioral and/or morphological characters 37,38 .
These characters are believed to result from ecological and evolutionary processes acting on species 39,40 . An interesting and innovative approach to evaluate the association between functional traits and environment is to investigate the functional structure of communities (e.g. [41][42][43]. In this context, both traits and their interactions provide information about how communities respond to environmental conditions 44 . This can be achieved by measuring the mean pairwise functional distance between communities (hereafter namely MPD-functional), derived directly from functional diversity (FD, 37,38 ).
Similar to MPD, MPD-functional positive values indicate a more functionally diverse community, expected to show a high degree of niche partitioning among species. On the other hand, negative values of MPDfunctional indicate niche overlap, with species expected to be more functionally similar to each other.
However, in order to improve our understanding and correctly interpret phylogenetic and functional structure of communities, the relationships of phylogenetic relatedness and ecological traits of species, or Phylogenetic Niche Conservatism (PNC) should also be veri ed 40 . This can be accomplished by testing the tendency for closely related species to be more phenotypically similar each other when compared to species drawn randomly from species pool, and by quantifying their phylogenetic signal 40,45-47 . Although there has been an impressive increase in studies of phylogenetic ecology and community structure in the last decades, this approach has only been applied to a few taxonomic groups 3,7,18 . Also, methods that combine both functional traits and phylogenetic data can be even more powerful to unravel assembly rules of communities 48-50 . Anurans are one of the most diversi ed groups of vertebrates in the Neotropical region 51 , particularly in Brazil (N = 1080 species; 52 ), with several recent phylogenetic hypotheses available [53][54][55][56][57][58] . Anurans are an interesting model to investigate the mechanisms related to the phylogenetic structure of communities. Manly due to their complex life cycle, permeable skin, limited dispersal, anurans are highly sensitive to ecological and evolutionary processes, such as environmental control and speciation 28, 29, 59-61 .
Speci cally, tadpoles may occur in a great range of aquatic habitats, from depression in trunks of tree (e.g., tolimns) to lakes or streams, among others. Thus, tadpoles are exposed to many types of biotic and abiotic factors 62 . Also, the occurrence of aquatic habitats by tadpoles also could be determined by historical constraints imposed by colonization and phylogeny of species 63,64 . Therefore, Pond communities provide a good opportunity to study assembly rules, because they have easily recognizable limits, similarly to islands, where assembly rules have historically been studied 65 .
Our aims in this study are threefold. We rst evaluate the phylogenetic and functional structure of anuran tadpoles assemblages in Atlantic Forest coastal plains of southeastern Brazil. We then test the assumption of phylogenetic conservatism to improve our understanding to better understand the phylogenetic and functional structure of tadpoles. And nally, we assess the relative importance of environmental predictors to phylogenetic and functional structures. We expect that: (i) anuran tadpole communities show a clustered phylogenetic and functional structure mainly because (ii) tadpoles exhibit phylogenetically conserved traits; and (iii) the phylogenetic and functional structure of assemblages are strongly related to environmental predictors such as vegetation structure and physico-chemistry parameter of ponds, which in turn affect anuran biology and ecology.

Biological surveys
The study was conducted in coastal plains of São Paulo state, Southeastern, Brazil, between October We sampled tadpoles and potential sh predators from 33 ponds in the study area (Fig. 1). For this study, each pond was considered as an ecological community. The ponds were selected in order to maximize the diversity of physical and vegetational structure, for example, from ponds in open areas to small puddles in forested areas, but all in coastal areas of the Atlantic Forest (more details abouts sampled ponds, see Environmental data section and Supplementary Table S1). Each pond was sampled three times, from October 2011 to March 2012, at the beginning, middle and end of the reproductive season of the Atlantic forest amphibians. Animals were collected in suitable microhabitats for tadpoles and shes. Specimens were identi ed to species level in laboratory and deposited in the "Coleção Cientí ca de Anfíbios" -Universidade Estadual Paulista "Júlio de Mesquita Filho", São José do Rio Preto, São Paulo, Brazil. The collection and surveys protocol/ experimental protocol were approved and provided by the animal ethical committee Instituto Chico Mendes de Conservação da Biodiversidade (ICMBio) (#315541). All methods were performed in accordance with the relevant guidelines and regulations.
The tadpole abundance in ponds is commonly related to species reproductive modes and strategies 68,69 .
We then used a log transformed abundance matrix, since higher logarithmic bases give less weight to quantities than to presences 70 .
Environmental data -sampling and processing We measured the following local habitat variables: pond area, canopy cover, diversity of internal and external vegetation structure, presence of potential predators ( sh), pH, water temperature, water conductivity, dissolved oxygen and water depth (see Supplementary Table S1 for further details). We transformed continuous variables (area, water depth, water temperature, water conductivity and dissolved oxygen) by Gower standardization as recommended by 71 . Gower's coe cient could be de ned as the mean of squared distances between samples 72 , in this case, environmental continuous variables. Thus, all numerical variables had equal weight in analyses. Variables were not correlated to each other (Pearson correlation, r < 0.60), and were all included in subsequent models. Pearson correlation coe cient are quite arbitrary and tend to depends on the variables analyzed, but as a rule of thumb, we considered an absolute value of r <0.60 as a weak to modest correlation between environmental variables.
Phylogenetic data -sampling and processing In order to calculate MPD and to test phylogenetic signal, we built a pruned tree from our regional species pool, based on a phylogenetic hypothesis from 73 (Fig. 2). Our regional pool was represented by all species recorded in the Serra do Mar coastal forests (N = 232) as registered in 74 .
Trait data -sampling and processing We tested the premise of trait phylogenetic conservatism calculating the mean pairwise functional distance (MPD-f). We measured well-known ecomorphological attributes described by nine morphological characters (see Appendices Fig. A1 for further details) related to habitat use of tadpoles (see 68 ). It was considered tadpoles with stages ranging from 33 to 39 (sensu 75 ) (see 76 ). The ecomorphological attributes were: relative caudal height (RCH = (HCM + HDF + HVF)/BH); body compression (BC = BH/BTL), relative width of caudal musculature (RWCM = HCM/ MCW), relative caudal length (RCL = (BTL-BL)/BL), and relative spiracle size (RSP=SH/BH). The following categorical attributes were also included: body shape (BS), position of oral opening (OR), eye position (EP), number of denticle rows (NDR), spiracle position (SP), and presence or absence of agellum (FP) (for more details about the abbreviations and the nomenclature, see Supplementary Table S1).
Traits were selected based on strong ecological associations and biological features of tadpoles, including habitat use and foraging behaviour, which may in uence ecosystem functions and speci c defence against predators, such as position of the eyes (morphological trait), position on the water column (habitat use) and predation risk (ecological association), but for more details about the relationship between traits and habitats variables, see 62,77,78 . where X obs is the phylogenetic distance between two taxa (the sum of branch lengths among the two taxa) from the phylogenetic pool; mn(X obs ) is the mean value of all possible pairs of n taxa from the phylogeny; mnX(n) is the mean for n taxa randomly distributed on the phylogeny; and sdX(n) are standard deviation expected for n taxa randomly distributed on the phylogeny.
The MPD is a standardized measure of PD, based on Phylogenetic Diversity (PD), developed by 3 , but it has his roots on Faith's PD 81 . The MPD obs (X obs ) values of each community were compared to permutations of the community matrix (MPD null ) randomized from the regional pool with equal probability 82 . This metric test whether communities are assembled by species which are more (i.e., clustered) or less (i.e., overdispersed) phylogenetically related than expected by chance 3,83 .
To assess the functional structure of tadpoles, we implemented an approach analogous to the MPD, by Mean pairwise Functional Distance (MPD-f) 84 . As in MPD, we calculated the standardized size effect, but using the functional diversity (FD) to generate the MPD-functional and thus its standardized size effect values 82 . The FD was calculated based on sum of branch lengths of the functional dendrogram necessary to connect all the species present in a local community, as follows: FD Q expresses the average functional difference between any two randomly selected individuals; d ij is the difference between the i and j species (d ij = d ji and d ii = 0), and p is the relative abundance vector for each species i (e.g. 79,85 ). In order to generate the functional dendrogram, we rst calculated a distance matrix based on the functional traits. As we had qualitative and quantitative traits, we used a generalization of Gower distance 79 to the treatment of mixed data and then implemented an UPGMA (Unweighted Pair Group Method using arithmetic Averages) hierarchical clustering to create the functional dendrogram 79,86 . The following steps are equivalent to MPD exposed above, comparing the values of MPD-f obs to simulated permutations of the community matrix (MPD-f null ) randomized from the phylogenetic pool with equal probability 83 . Analogously, this metric can also test whether species of a community are more or less functionally similar than expected by chance, generating the MPD-f 83,84,87 . We then used only communities with signi cantly MPD-f values for further analysis.
Spatial dependence between samples is commonly observed in nature and it can cause noise and bias to statistical modeling if not considered 88 . This phenomenon is called spatial autocorrelation, that is, values of variables sampled in neighbor sites are not independent of each other, violating the assumption of independence between samples in statistical models 88 . Moreover, if there is spatial autocorrelation in the models, it will also lead to spatial autocorrelation in residuals 89 . Thus, in order to avoid spatial bias in models, we performed a Moran I test, which describes spatial autocorrelation in the data applied to residuals of a regression between MPD and MPD-f signi cant values and environmental variables. In our case, we did not nd a signi cant spatial autocorrelation in the residuals of the regression model (MPD-phylogenetic= Moran I = -0.758; p = 0.50; MPD-functional = Moran I= 0.937, p = 0.46).
We also selected the best predictive environmental model for each community structure metric using a Stepwise Model Selection based on corrected Akaike information criteria (AIC c ) 90,91 . We employed a Generalized Linear Models approach (GLM), expressed by the adjusted R² statistic (R² adj ; 92 ), to assess the variance of MPD and MPD-f that are potentially explained by each environmental predictor. We also used the Hierarchical Partitioning 93 to evaluate the independent and joint contributions of each environmental predictor 94,95 . Additionality, a Redundancy Analysis (RDA) was conducted in order to explore the relative importance of environmental heterogeneity, functional guilds and phylogenetic clades 79 .

Results
We recorded 25 anuran species (N = 20.762 individuals) belonging to four families (Hylidae, Leptodactylidae, Microhylidae, Bufonidae) and 12 genera (Fig. 2) in all 33 communities (ponds). The Rao's diversity coe cient was 0.135 at coastal plains of Atlantic Forest. Tadpole trait diversity was concentrated in a few nodes of the phylogeny ( Fig. 3; single-node skewness test, observed value = 0.190, P < 0.004; few nodes skewness test, observed value = 0.307, P < 0.001). The tips skewness test revealed that trait diversity is concentrated near the root of the phylogeny (Fig. 3, observed value = 0.420, P < 0.001). As the trait diversity was concentrated in few nodes and close to the root of the phylogeny, we con rm that trait diversity of anuran tadpoles from Atlantic Forest coastal plains are phylogenetically conserved.
MPD-phylogenetic values of the tadpole communities ranged from -2.18 to 0.57 (mean = -1.06 ± 0.51). Among the 33 sampled tadpole communities, 18 (approximately 55%) showed a phylogenetic clustered structure, with species more closely related than expected by chance. The 14 remaining communities had non-signi cant negative MPD values.
The MPD-functional values of tadpole communities ranged from -4.38 to 1.45 (mean= -2.07 ± 0.89). Among the 33 tadpole communities (ponds) sampled, 12 (approximately 37%) showed an aggregated functional structure with species more functionally similar than expected by chance. The 21 remaining communities had non-signi cant negative MPD values. Notably, of the 12 communities with aggregated functional structure, 10 also showed a clustered phylogenetic structure.
The Redundancy Analysis (RDA) revealed that variance of the relationship between phylogenetic clades and environment explained by the rst two canonical axes was 87.4% (53.9% for axis 1 and 33.5% for axis 2, respectively, see Fig. 4a). The variance of functional guilds-environment relation was 83.0% (55.2% for axis 1 and 27.8 % for axis 2, respectively, see Fig. 4b).
The overall model with six environmental variables explained 58% (R²adj = 0.58) of total variance of phylogenetic structure. The best environmental tting the MPD-phylogenetic included six variables: presence of potential sh predators, internal and external vegetation diversity, canopy cover, dissolved oxygen and water conductivity (Table 1a). The most important environmental predictors in GLM model were presence of potential sh predators (predator), followed by water conductivity, external vegetation structure diversity (external pond vegetation), canopy cover, internal vegetation structure diversity (internal pond vegetation) and oxygen dissolved (Table 1a, Fig. 5a). Predator presence was positively correlated to MPD values, whereas water conductivity, canopy cover, internal vegetation, external vegetation and oxygen dissolved were negatively associated (Fig. 5a).
On the other hand, the best environmental model to MPD-f selected trough stepwise model selection based on AICc included ve variables: external vegetation diversity, canopy cover, area, dissolved oxygen and presence of potential sh predators. The overall model with these ve variables explained 41% (R²adj= 0.41) of total variance of functional structure. The most signi cant environmental variables in GLM model were external vegetation, followed by canopy cover, dissolved oxygen and predator (Table 1b, Fig. 5b). Finally, external vegetation, canopy, area and dissolved oxygen were negatively associated with the MPD-functional values, whereas predator was positively correlated (Fig. 5b).

Discussion
Our general aim in this study was to investigate the phylogenetic and functional structure of tadpole communities (ponds) from coastal plains in southeastern Brazil and what ecological and evolutionary processes (such as ecological ltering or trait conservatism) potentially affect their structure. Earlier studies on plant communities argue that when taxonomic levels increase, phylogenetic structure shifts from overdispersion to clustering 13 . Recent studies found that phylogenetic clustering tends to increase with spatial extent from local to landscape scales, probably due to higher environmental variation at broad spatial scales 13,95,96 . In this scenario, assemblages would be driven by niche conservatism in more regional scales, while species interactions would determine community structure at local scales 13 . However, according to our predictions, we found a clustered phylogenetic structure in most sampled ponds (18 out of 33 communities). This pattern was also already found in Pantanal at regional and local scales, although the processes underlying remains unclear at local scales 97 . In fact, clustered phylogenetic structure has been considered fundamental if not the dominant pattern in recent studies of phylogenetic structure of vertebrate assemblages (e.g. 5,7,14,98 ). We then argue the possibility of, contrary to plant communities, clustered phylogenetic structure as an important pattern for vertebrates, even at local scales.
Our results also showed that 12 communities were composed by species more functionally similar than expected by chance, and 10 of these communities also displayed a clustered phylogenetic structure.
Functional structure related to phylogenetic structure was also found in other vertebrate groups, including birds 5,99,100 , bats 101 and shes 102,103 . These ndings point out to a likely general pattern in vertebrates, where phylogenetic and functional structures are interconnected at some level, as phylogeny could inform about functional structure or could be employed as a possible proxy of each other, even if they could not be considered directly interchangeable. This is because part of the functional structure is not related to phylogenetic structure, which may be the outcome of ecological or stochastic processes. On the other hand, phylogenic structure unrelated to functional structure could be a result of evolutionary processes such as genetic drift 40 .
We found that the two best environmental models of MPD and MPD-f included six and ve environmental variables, respectively, indicating that species composition and abundance changes on clades and functional groups are related to environmental variables. The presence of potential sh predators was the most important variable and the only one positively related to clustered phylogenetic structure (MPD, Fig  5a). On the other hand, this variable was less important but positively related to functional structure (MPD-f, Fig. 5b). The in uence of predation on occurrence and co-occurrence of species has been well documented 104 . Predators can cause prey to modify their behavior, morphology, life history, and physiology in an attempt to reduce risks of predation [105][106][107][108][109] . Fish predators can negatively affect the development and growth of tadpoles, as well as select morphological changes. On the other hand, these effects on prey can be regarded as adaptive responses to the presence of predators 108, 110,111 . Predation could also reduce interspeci c prey competition, for instance, by reducing or eliminate entire populations of strong competitors, leaving an empty ecological niche space to competitively weaker species 109,112 .
Some studies reveal that aquatic predators selected prey characteristics in a non-random fashion 111,113 .
We found an increase of potential predator presence associated to clustered phylogenetic structure. Thus, potential predators seem to be negatively associated with anuran species occurrence or even functional guilds of adhrent and macrophagous species (Fig. 4b). On the other hand, predators are apparently favoring the occurrence of species of Bufonidae (Fig. 4a). It could re ect speci c defenses that many species of this family display, such as toxins present in the cutis and body uids of several Bufonid species 114 .
Water conductivity was a signi cant explanatory predictor for taxonomical and functional beta diversity of anuran tadpoles 115 and was the second most signi cant explanatory variable of phylogenetic structure (Fig. 5a). It is known that water conductivity can be a surrogate to food availability or productivity in a pond, as higher productivity provides more hydrogen ions, and thus a higher conductivity 116 . The higher the productivity, the higher the availability and quality of food resources (periphyton) will be. In turn, high productivity will generate an increase of species performance in aquatic communities, a biological aspect commonly associated to certain species traits, such as those related to caudal and oral morphology of tadpoles 23,24,[117][118][119] . Conductivity could be indicating a positive association of productivity to certain lineages as Microhylids (Fig. 4a) by these mechanisms. However, the same variable shows a negative relationship to Leptodactylids, even though such association remains to be further investigated.
We also found an important in uence of external vegetation diversity to MPD, and this was the most important variable to MPD-f (Fig. 5b). This variable probably determines the availability of suitable sites for vocalization, amplexus, and oviposition for adults, interacting with species traits in ponds. Thus, external vegetation structure diversity of a pond is expected to be associated with distribution and reproduction of anurans, mainly for Hylidae 27,120-122 , but was more strongly associated with Leptodactylids than Hylids (Fig. 4a), which a result that still demands ecological interpretations. Canopy cover was the fourth most important explanatory variable of phylogenetic structure and the second more important of functional structure (Fig. 5a,b). In fact, canopy cover is claimed to indirectly drive species diversity, and the growth and development of tadpoles in aquatic habitats, because open-canopy ponds tend to have higher productivity, and as discussed here, ponds with higher productivity show higher resource availability and quality [23][24][25]119,123 . It is worth noting that canopy cover was already recovered as an important predictor of anuran functional and phylogenetic beta diversity 115 and metacommunity structure 124 . Also, there is evidence that canopy cover may be a signi cantly source of nutrients, acting as key predictor for tadpoles growth and development 125 . In our sampled ponds, this variable was positively associated to MDP, indicating that ponds with more closed canopies show phylogenetic clustering. Canopy cover seems to be segregating forest interior clades, such as Microhylids, from open area clades, as Hylids (Fig. 4a); and it is positively associated with adhrent and microphagous tadpoles, functional groups in uenced by food accessibility, strongly affected by canopy cover (Fig. 4b) 64,65,126 .
Internal vegetation structure diversity was the fth explanatory variable selected for MPD (Fig. 5a). This variable has been identi ed as important descriptor of functional, taxonomical and phylogenetic beta diversities 115 . This predictor is vital to anurans, as it provide suitable sites to vocalization, amplexus, oviposition, and even predator protection to adults and tadpoles 27,[117][118][119][120][121][122]127 . Therefore, increased heterogeneity of overall vegetation structure on ponds, such as higher presence of shrubs and trees, may be favoring the presence of certain species, including treefrog members of Hylidae family. Hylids are the most representative family in our study region and they use mainly hanging vegetation as vocalization sites 27,69,128 . This is mostly due to their adhesive disks on ngers, a key trait to anurans, allowing the exploration of three-dimensional habitat 69 . Given that, vegetation structure may affect the occurrence of this anuran clade. Higher vegetation diversity was then positively related to treefrogs, affected by vertical vegetation availability, and negatively related to toads (Bufonidae), that typically reproduce on the ground ( Fig. 5a; 69 ). Therefore, it suggests that external and internal vegetation structure may be affecting the phylogenetic structure of tadpole communities.
Dissolved oxygen in water (DO) was the sixth most important explanatory predictor of MPD, but the fourth more important to MPD-functional ( Fig. 5a and b). This physical parameter of water has a strong in uence on tadpole species-speci c performance and survivorship 116, [129][130][131] . Thus, it could be affecting the presence of species of distinct lineages, such as Bufonids, which does not have functional lungs in early larval stages 130 . In our study, the DO was negative linked to Leptodactylid, Hylids and benthic tadpoles, but positive associated to Microhylid and Nektonic tadpoles (Fig. 4a, b). However, the impacts of long-term hypoxic conditions on growth and development of tadpoles are not yet well understood 23 .
Surface area was the third most important explanatory variable to the functional structure (Fig. 5b). Physical parameters of water bodies are known to in uence the diversity and distribution of tadpoles, but in complex ways. The Island Biogeography theory (IBT) predicts that species richness increases with island size at a certain distance from a source of colonizers. However, stream volume (a variable strong related to surface area) was negatively and weakly associated to tadpoles richness 132 , the inverse of the prediction of IBT, if we consider water bodies as islands within land habitats as a matrix. But the increasing surface-area-to-volume ratio was also positively associated to tness of Bufo americanus. Likewise, in coastal plains ponds, benthic tadpoles, such as Rhinella ornata were positive associated to area (Fig. 4b), but nektonic species, such as Scinax spp. were negative related to this variable.
In summary, our best predictors were mainly negatively related to MPD and MPD-f structure of tadpole communities, indicating that a decrease in values of these variables generally is associated to more clustered assemblages, both in functional and phylogenetic diversity. In other words, ponds with lower environmental complexity or heterogeneity show more closely related species than expected from random assemblages. Therefore, environmental variables are probably constraining community composition, by ltering tadpole lineages or functional guilds, and in association with the species phylogenetic conserved traits, generate an emergent clustered phylogenetic structure. In fact, these factors are acting particularly in some clades, limiting the occurrences of Microhylids and Bufonids, and favoring the occurrence of Hylids, and acting as assembly rules to tadpole communities. Our study provides an important step to further understand anuran community structure and their underlying assembly processes. Anurans assemblages are the outcome of interactions between ecological and evolutionary processes, such as environmental lters and phylogenetic trait conservatism.

Declarations
Author Contributions: TALP, AML and RJS originally formulated the idea, TALP and AML developed methods, TALP and AML conducted eldwork, TALP performed statistical analyses, and TALP, AML and RJS wrote the manuscript. All authors reviewed the manuscript.