Seasonal variation in the composition of ground-dwelling anuran (Amphibia) assemblages in southern Brazil

Climate seasonality may differently influence habitat quality and heterogeneity depending on habitat type. We examined whether the taxonomic, functional and phylogenetic composition of ground-dwelling anuran assemblages from grassland and forest habitats vary seasonally. We tested the hypothesis that the forest anuran assemblage varies less seasonally than the grassland assemblage. We monitored anurans using pitfall trap arrays in two areas, sampled across four seasons over a two-year period. For the functional composition, we acquired information on species morphology, reproduction, and habitat use to represent the anuran niche. For the phylogenetic composition, we used the most comprehensive anuran phylogeny. We used Principal Coordinate Analysis and Analysis of Variance to evaluate seasonal variations in assemblage composition along the study period. Our data revealed significant seasonal variation in the taxonomic and phylogenetic composition of anuran assemblages. Variation in taxonomic composition was higher in the grassland than in the forest assemblage, while variation in phylogenetic composition was higher in the spring-summer than in the autumn-winter seasons. We did not identify seasonal variation in functional composition. Seasonal variations in taxonomic and phylogenetic composition, but not in functional composition, indicate that the species with a fluctuating seasonal abundance have similar life-history traits, but belong to different lineages.


Introduction
The study of anurans is particularly intriguing because their sensitivity to water loss through their skin entails a physiological challenge for living under variant and extreme conditions.Their evolution has resulted in a diversity of strategies for acquiring resources, reproducing, avoiding predation and competing in variant and extreme environments (Duellman and Trueb, 1994;Haddad and Prado, 2005;Silva et al., 2012).Although biotic processes (e.g., competition and predation) were previously considered to be the main factors influencing the diversity of anuran assemblages, climate seasonality has recently been recognized as a strong contributing factor (Both et al., 2008;Rievers et al., 2014).It has been suggested that seasonality in temperature, photoperiod and rainfall might control the peaks of anuran reproduction and richness in subtropical and temperate regions, whereas abundant rainfall might sustain continuous anuran reproduction in tropical regions (Prado et al., 2005;Both et al., 2008;Canavero et al., 2008).Further, climate seasonality can differentially influence the vegetation heterogeneity and resource quality of an area, as the effects of climate seasonality depend on the habitat type (Maragno et al., 2013;Gonçalves et al., 2015).Here, we evaluated the seasonal variations in the composition of anuran assemblages from forest and grassland habitats, which are two habitats differently influenced by climate seasonality.
Plant architecture, phenology and tolerance to climate seasonality influences the vertical and horizontal stratification of habitats, which in turn alters the niche availability within forest and grassland habitats, offering differential opportunities for anuran reproduction, habitat selection, predator avoidance and food intake (Urbina-Cardona et al., 2006;Both et al., 2008;Oliveira et al., 2013;Gonçalves et al., 2015).The main seasonal changes in the forest structure are due to increases in tree flowering, fruiting and litter production, but not in canopy cover (Liebsch and Mikich, 2009;Longhi et al., 2011).In grasslands, the peak of plant reproduction, biomass production by both grasses and forbs and the consumption of dry biomass by soil microorganisms occur during the warm seasons (spring and summer) (Soares et al., 2005;Overbeck et al., 2006).In contrast, a low production of biomass and the accumulation of unconsumed biomass occur during cold seasons (autumn and winter), which results in large seasonal differences in the vertical and horizontal structure of grasslands (Soares et al., 2005;Overbeck et al., 2006).Since different ecological factors drive the heterogeneity and intrinsic dynamics of grassland and forest habitats, the structure of anuran assemblages should also differ, with more pronounced dynamics in more seasonal habitats (Inger and Colwell, 1977).
Structural particularities of habitats, as well as their intrinsic dynamics, should strongly influence not only species but also the functional and phylogenetic composition of anuran assemblages (Inger and Colwell, 1977;Urbina-Cardona et al., 2006;Maragno et al., 2013).Differences in richness have been observed when comparing anuran assemblages in stable and heterogeneous habitats to assemblages in unstable and homogeneous habitats (Inger and Colwell, 1977;Xavier and Napoli, 2011;Oliveira et al., 2013;Jimenez-Robles et al., 2017).Furthermore, habitats that are less altered by climate might have more stable anuran assemblages than habitats experiencing large seasonal changes in structure (Inger and Colwell, 1977).However, how the taxonomic, functional and phylogenetic dimensions of anuran diversity varies over time and space is virtually unexplored (e.g., Silva et al., 2012;Trimble and van Aarde, 2014;Jimenez-Robles et al., 2017).The lack of life-history traits that facilitate the persistence during climatic extremes (e.g., cold temperatures and drought) can promote the disappearance of a species from a given habitat (Silva et al., 2012).If the anuran species fluctuating in abundance over time have lineage-specific life-history traits, we could expect large seasonal variation in the taxonomic, functional and phylogenetic composition of the anuran assemblages.On the other hand, if the species fluctuating in abundance over time have similar life-history traits but are from different lineages (i.e., evolutionarily convergent), we could expect large seasonal variations in both the taxonomic and phylogenetic composition, but not in the functional composition of the anuran assemblages.Finally, if the anuran species are highly persistent in the habitat over time (i.e., low fluctuations in abundance), we could expect a lack of seasonal variations in the taxonomic, functional and phylogenetic composition of anuran assemblages.Thus, seasonal changes in habitat structure due to climate seasonality may result in anuran assemblages composed of distinct species with different physiological tolerances, reproductive modes and habitat preferences.
We aimed to examine whether the taxonomic, functional and phylogenetic composition of ground-dwelling anuran assemblages from grassland and forest habitats composição funcional, indicam que as espécies de anuros que variam em abundância ao longo das estações do ano possuem atributos de história de vida similares, mas pertencem a linhagens diferentes.
varied over seasons.We tested the hypothesis that the forest anuran assemblage would exhibit less seasonal variation in taxonomic, functional and phylogenetic composition than the grassland assemblage.Such result is expected since forests experience fewer seasonal changes in vegetation height, vertical stratification and productivity than grasslands do (Inger and Colwell, 1977).In contrast, we expected large seasonal variations in the species, functional and phylogenetic composition of the grassland anuran assemblage because the reproduction and biomass production of grassland plants strongly differs between warm and cold seasons (Soares et al., 2005;Overbeck et al., 2006).Thus, the grassland anuran assemblage of warm seasons should differ in its composition of species, lineages and life-history traits when compared to the grassland assemblage of cold seasons.

Study site
We conducted the study at two areas in the municipality of Passo Fundo, Rio Grande do Sul State, in southern Brazil.The elevation of the areas varies between 630 and 740 m.The climate is predominantly subtropical (Cfb climate according to Köppen, 1948), which is characterized by a mean annual temperature of 14-16 °C, an annual rainfall between 1,900 and 2,200 mm with a lack of hydric deficits, and the occurrence of frost and eventually snow during the winter (IBGE, 2002;Alvares et al., 2013).The grassland area (Fazenda da Brigada; 28°14'56" S, 52°19'52" W) comprises one of the last large remnants (≈ 600 hectares) of Aristida jubata grasslands in the region.Aristida jubata (Arechav.)Herter, a tussock grass, dominates the herbaceous strata of the grassland site (Figure 1), whereas patches of Araucaria forest are found along streams.Marshes and water reservoirs are the main types of hydric resources found in the grassland habitat.Araucaria forest covers nearly the entire extent of the forest area (≈ 200 hectares of forest; 28°13'56" S, 52°20'15" W), with Araucaria angustifolia (Bertol.)Kuntze (Araucariaceae) dominating the forest canopy.Small streams with rocks are the main types of hydric resource occurring in the forest habitat.The distance between the two sites is 2 km.

Sampling of ground-dwelling anurans
We monitored the occurrence of anurans in each area during different seasons of the year (autumn, winter, spring and summer) for two years (from May 2001 to We standardized the number, disposition and period of pitfall opening between habitats and among seasons to facilitate comparisons.The forest array was located more than 100 m from the forest edge, while the grassland array was at least 700 m from the nearest forest edge.Repeated sampling of the same habitat over time circumvented the lack of a spatial replication of habitats (i.e.several forest and grassland habitats), allowing the temporal variations in anuran assemblage composition over time to be tested.The average temperature during the sampling periods was below 15 °C in winter and autumn, and mostly warm (average near 20 °C) during spring and summer (Figure 2).These data were sampled by the Brazilian Agricultural Research Corporation (EMBRAPA, 2016), approximately 6 km from the sampling sites.

Functional and phylogenetic composition
To evaluate whether functional composition changed with seasonality, we acquired information on species life-history traits from the literature (Table 1).We used information on morphological (snout-vent length), reproductive (reproductive modes) and habitat-use (cryptozoic behavior) traits for each species, since they seemed to best represent the nuances of anuran niches (Silva et al., 2012;Tsianou and Kallimanis, 2016).Snout-vent length (SVL, averaged measures for males and females) relates to the species sensitivity to habitat changes, with larger species being less sensitive to humidity deficits and more vagile, facilitating their ability to find resources and favorable conditions (Trimble and van Aarde, 2014;Tsianou and Kallimanis, 2016).Reproductive modes describe the species choice for specific oviposition sites, which indicates the dependency of the species on water sources (Duellman and Trueb, 1994;Haddad and Prado, 2005;Prado et al., 2005;Hartmann et al., 2010).Finally, cryptozoic behavior describes the habit of living hidden in shelters (holes and leaf litter).Such behavior may be favored over non-cryptozoic behavior in open habitats, since staying protected from solar desiccation and predation may be advantageous, particularly during unfavorable periods.
We used a dated phylogeny to obtain the phylogenetic distance between the anuran species registered over the two years of the study.We pruned the phylogenetic tree of Pyron and Wiens (2011) to obtain the phylogenetic distance between all species found in the forest and grassland habitats.Although the phylogeny from Pyron and Wiens is the most comprehensive phylogeny for the group, we added three species that were not included in the phylogeny (Leptodactylus latrans (Steffen, 1815), Elachistocleis bicolor (Guérin-Meneville, 1838) and Physalaemus henselii (Peters, 1872)).These species were randomly inserted in their respective genera using functions from the 'phytools' package (Revell, 2012) in R (R Core Team, 2017).Scientific nomenclature of Physalaemus aff.gracilis follows Frost (2018).We built a functional dendrogram and a phylogeny to visualize the functional and phylogenetic relationships between species.The functional-dendrogram scale is in Gower distance and the scale of the phylogenetic-tree is given in millions of years (according to Pyron and Wiens, 2011).We built the functional dendrogram using Gower distance (because of the categorical, binary and quantitative trait values) and the UPGMA clustering algorithm.

Data analysis
In order to determine whether the composition of ground-dwelling anuran assemblages from the forest and grassland habitats varied over time, we analyzed matrices of taxonomic, functional and phylogenetic composition where each sampling unit comprised of a month of sampling in a particular habitat.Before the analysis, we used Hellinger's transformation on the matrix of taxonomic composition to decrease the discrepancies in abundance among the sampled species (Legendre and Legendre, 2012) (see Table 1).Seasonal variation in the composition of ground-dwelling anuran (Amphibia) assemblages in southern Brazil We then used the Hellinger-transformed matrix to calculate the functional and phylogenetic compositions.Functional and phylogenetic composition matrices were obtained by fuzzy-weighting species incidences by their functional and phylogenetic relationships (Pillar et al., 2009).Fuzzyweighted functional and phylogenetic composition matrices contain the trait -and phylogeny -weighted abundances of anuran species, defined according to the functional and phylogenetic relationships among species occurring in a given month (Pillar et al., 2009).
We explored seasonal variation in the taxonomic, functional and phylogenetic composition of anuran assemblages using Principal Coordinate Analysis (PCoA) based on the square-rooted dissimilarity between pairs of sampled months; we used the Lingoes correction to avoid the negative eigenvalues (Legendre and Legendre, 2012; Legendre, 2014).We used the Bray-Curtis dissimilarity to express the variation in taxonomic composition, and the Euclidean distance to express the variation in functional and phylogenetic composition (Legendre and Legendre, 2012).We excluded the grassland sampling from the winter of 2002, because the lack of captures biased the variance explained by the first ordination axis.Ordination axes are the best way of interpreting how much a site varies over time in relation to other sites, because averaged pairwise dissimilarities only account for the compositional variation between two time periods (Legendre, 2014;Anderson et al., 2006).
We tested the hypothesis that the forest anuran assemblage would exhibit less seasonal variation than the grassland assemblage by applying the one-way Analysis of Variance test to the eigenvectors of the first and second PCoA axes, which captured most of the seasonal variation in the composition of anuran assemblages.In the Analysis of Variance, we related the seasonal variation in taxonomic, functional and phylogenetic composition (represented by PCoA axes) with the interaction between season and habitat, in order to determine in which season a given habitat would show the largest variation in composition.We summarized the pitfall-trap data to compare composition between the most contrasting climatic conditions found in the region (autumn and winter were considered as 'cold' seasons; spring and summer were considered as 'warm' seasons) because we did not have a sufficient sample size to compare each season.We identified differences in the taxonomic, trait and phylogenetic composition of habitats and seasons using contrast analysis (TukeyHSD Test).We conducted the functional and phylogenetic fuzzy-weighting analyzes in the 'SYNCSA' package (Debastiani and Pillar, 2012), dissimilarity-index analysis in the 'vegan' package (Oksanen et al., 2018), ordination analysis in the 'ape' package (Paradis et al., 2004), and both the Analysis of Variance and TukeyHSD (honestly significant difference) Test in the 'stats' package (R Core Team, 2017).
aff. gracilis and R. icterica were more abundant in the forest habitat, while P. cuvieri and P. henselii were more abundant in the grassland habitat.Species that occurred exclusively in the grassland habitat were L. gracilis, L. latrans, L. plaumanni, and P. henselii.The species Odontophrynus americanus (Duméril and Bibron, 1841) maintained a stable abundance throughout the seasons in both habitats.Exploratory analyses of the functional and phylogenetic relationships between the species showed that P. cuvieri, P. aff.gracilis, P. henselii and L. plaumanni formed a group of small-sized species (Figure 3).Elachistocleis bicolor, Leptodactylus gracilis and O. americanus formed a functional group of cryptozoic species.Rhinella icterica and Leptodactylus latrans formed a group of large-sized species (Figure 3).Different reproductive modes were scattered among these three groups.
The results showed significant seasonal variation in the taxonomic and phylogenetic composition; but we did not find seasonal variation in the functional composition (Table 2).The analysis of the first ordination axis showed that the forest anuran assemblage exhibited less seasonal variation in taxonomic composition than the grassland assemblage did (Figures 4 and 5).The analysis of the second ordination axis showed seasonal variations in taxonomic composition within the grassland assemblage (Figures 4 and 5).In contrast, grassland assemblages of the warm seasons varied more in phylogenetic composition than grassland and forest assemblages of the cold seasons (Figures 4 and 5).

Discussion
Results showed that the seasonal variation in the taxonomic composition was more pronounced in the grass-   2. Results of the one-way Analysis of Variance, testing for seasonal variations in the taxonomic, functional and phylogenetic compositions of anuran assemblages from forest and grassland habitats in the municipality of Passo Fundo, Rio Grande do Sul, southern Brazil, according to the interaction between season and habitat type (Factor).

Df
land assemblage than in the forest assemblage, while seasonal variation in the phylogenetic composition was more pronounced in warm (spring-summer) than in cold seasons (autumn-winter).Many studies show evidence for within-habitat variation in anuran assemblage composition across seasons due to the varying environment (e.g., Both et al., 2008;Canavero et al., 2008;Santos et al., 2008;Santos-Pereira et al., 2011;Rievers et al., 2014).However, a few studies attempt to understand whether assemblages from different habitats show distinct degrees of variation in composition across seasons (e.g., Inger and Colwell, 1977;Maragno et al., 2013).
Here, we further support previous findings (Inger and Colwell, 1977;Maragno et al., 2013) that assemblages from structurally less stable habitats suffer high variation in composition across seasons.Seasonality in temperature, rainfall and photoperiod influence the phenology and productivity of plant communities (Soares et al., 2005;Overbeck et al., 2006;Liebsch and Mikich 2009;Longhi et al., 2011), causing changes to habitat heterogeneity and resource availability, which consequently influence the peaks of anuran species reproduction and their patterns of habitat use (Inger and Colwell, 1977;Both et al., 2008;Canavero et al., 2008;Santos et al., 2008;Maragno et al., 2013).As forest habitats are seasonally more stable than grassland habitats in regard to habitat structure, we initially suspected that anuran taxonomic, functional and phylogenetic composition would be more stable throughout the changing seasons in the forest habitat.This pattern was confirmed, indicating that the intrinsic dynamics of habitats might cause seasonal variation in their species and lineage composition.These results do not imply that the forest assemblage remains unchanged over time (e.g., Santos-Pereira et al., 2011), but rather that the forest assemblage changes less than the grassland assemblage.The low seasonal variability in the forest assemblage, which we found to mainly consist of a subset of open-habitat and generalist species (e.g., Physalaemus aff.gracilis and P. cuvieri), could be due to the ability of these species to exploit the less-variant conditions and resources found in forest habitats, allowing them to maintain stable population sizes over time (Haddad et al., 2013, Bolzan et al., 2016).
Seasonal changes in grassland vegetation structure are characterized by the flowering of tussock grasses (e.g.Andropogon lateralis and Aristida jubata), the increase in biomass production and the recruitment/regrowth of woody plants during warm months (Soares et al., 2005;Overbeck et al., 2006).The grassland vegetation structure offers a great diversity of food items, calling sites, and shelters (for protection from predation and high solar exposure) for grassland-specialist anurans (Xavier and Napoli, 2011;Gonçalves et al., 2015).Grasslands are of major importance for Leptodactylidae anurans which include grassland-specialist species (e.g.Physalaemus henselii, Leptodactylus plaumanni) with seasonal reproduction and adaptations to temporary ponds and unstable hydroperiods (e.g., spawning in foam nests) (Duellman and Trueb, 1994;Santos et al., 2008;Kwet et al., 2010;Maneyro and Carreira, 2012).The difference in anuran species dominance in grasslands throughout the seasons matches the reproductive period of the species (Both et al., 2008;Santos et al., 2008), showing that anuran species occupying grasslands are sensitive to seasonal changes in primary productivity and grassland phenology (Soares et al., 2005;Overbeck et al., 2006).Individuals of Physalaemus henselii are more reproductively active than other species during the cold seasons (Maneyro et al., 2008).In turn, individuals of P. cuvieri, L. latrans and Physalaemus aff.gracilis are more reproductively active during warm seasons (Camargo et al., 2005;Kwet et al., 2010;Maneyro and Carreira, 2012), although Physalaemus aff.gracilis tend to be reproductively active throughout the year (Bortolini et al., 2018).Such reproductive patterns of different species explain the seasonal change in composition observed in the grassland anuran assemblage.
Although our study was limited to two sites, our data of anuran abundances collected over two years revealed seasonal variations in the structure of the anuran assemblages.Our main result was the identification of seasonal variation in the taxonomic and phylogenetic composition of ground-dwelling anuran assemblages, but this variation was not found for their functional composition.Such results can be explained by the ability of functionally similar and generalist species, which are less dependent on water availability, to transit between habitats.For example, Physalaemus aff.gracilis and P. cuvieri, occurring in both the forest and grassland habitats, are habitat-generalist species which reproduce in one habitat (where the species are more abundant) and acquire food in the other (Maragno et al., 2013).The movement of generalist species decreases the differences in functional composition over time due to the species' high similarity in size, reproductive mode and habitat use (Maragno et al., 2013).Finally, we observed that the seasonal variation in the phylogenetic composition was more pronounced in warm (spring-summer) than in cold seasons (autumn-winter).The seasonal variation in the phylogenetic structure was caused by the occurrence of Elachistocleis bicolor, the most phylogenetically distinct species (Figure 3).Elachistocleis bicolor occurs mainly in open areas and shows explosive reproduction during the rainy periods of warm seasons (Kwet et al., 2010).
We observed seasonal variations in the taxonomic and phylogenetic composition, but not in the functional composition, of the anuran assemblages.Changes in habitat structure due to climate seasonality influenced the taxonomic and phylogenetic structure of anuran assemblages by promoting the replacement of species with similar functional traits.We perceived that the abundance of species that are adapted to and reproductively active during the cold seasons decreases with the increase of the abundance of species that are adapted to and reproductively active during the warm seasons.Overall, the results indicate that the species of anurans that fluctuate in abundance over the seasons belong to different lineages but have similar life-history traits.The patterns registered in our study provide a benchmark on how communities from native forests and grasslands vary in composition through time.

Figure 1 .
Figure 1.Study areas located in the municipality of Passo Fundo, Rio Grande do Sul, southern Brazil.Dashed lines represent the extent of the grassland and forest habitats, and the empty white circles represent the position of the T-shaped pitfall-trap arrays.

Figure 2 .
Figure 2. Mean monthly temperature and rainfall in the municipality of Passo Fundo, Rio Grande do Sul State, Brazil, during the sampling period.Arrows represent the sampled months.Data were obtained from EMBRAPA (2016).

Figure 3 .
Figure 3. Functional and phylogenetic relationships among the ground-dwelling anuran species registered in forest and grassland habitats in southern Brazil.The functional-dendrogram scale is in Gower distance and the scale of the phylogenetic-tree is given in millions of years.

Figure 4 .
Figure 4. Principal Coordinate Analysis plot, showing seasonal variation in the taxonomic (left) and phylogenetic (right) composition of anuran assemblages in forest and grassland habitats from southern Brazil.

Figure 5 .
Figure 5. Box-plot of Analysis of Variance and contrast analysisshowing significant seasonal variation in the taxonomic and phylogenetic composition of anuran assemblages from southern Brazil over two years.Seasonal variation in taxonomic and phylogenetic composition was represented by ordination axes (axis I-above, axis II-below).Boxes represent the mean, 1 st and 3 rd quartiles of ordination eigenvectors.Boxes followed by different letters indicate significant differences (P ≤ 0.05) in the contrast analysis.

Table 1 .
Life-history traits and the number of individuals of species of ground-dwelling anurans, presented according to habitat type and season in southern Brazil.SVL= snout-vent length.See reference list in the table notes.