Amphibians of varzea environments and floating meadows of the oxbow lakes of the Middle Purus River , Amazonas , Brazil

Despite the historical knowledge on amphibians of the Purus River basin, basic information on assemblages is fragmented, with gaps in knowledge existing at various scales, which limits conservation strategies. This study provides information on the composition, richness and abundance of the amphibian fauna in varzea environments and floating meadows of the oxbow lakes of the Middle Purus River between the Boca do Acre and Pauinı́ municipalities, Amazonas, Brazil. We sampled six oxbow lakes using forty-seven 200-meter transects, distributed among the ‘‘floating meadows,’’ ‘‘high varzea’’ and ‘‘low varzea,’’ from April to January 2014. We recorded 59 species, with the family Hylidae being predominant. This amphibian fauna represents approximately 19% of the species known for the Amazon, 28% for Amazonas State and 45% of the species recorded so far in the Purus River, increasing the richness of the basin to 132 species. Eight species were considered rare, and 29 are endemic to the Amazon. This study adds to the knowledge on the amphibian species of the Amazonian lowlands, including the expansion of known distributions, as well as increases the knowledge of several species that are locally rare, endemic and/or that are data deficient regarding distribution and ecology.


Introduction
The complex evolutionary history, intricate topography and climate of South America have produced a rich and highly diverse herpetofauna (Duellman 1979(Duellman , 1999)).There are more than 1,026 amphibian species in Brazil (Segalla et al. 2014), and at least 315 anurans, 16 gymnophionans and five caudates are known in the Brazilian Amazon, representing approximately one-third of the total of amphibians recorded for the country (Avila-Pires et al. 2007, Hoogmoed 2012, Toledo & Batista 2012, Neckel-Oliveira et al. 2013).The state of Amazonas, extending approximately 1,570,745.68km 2 , currently has 212 species of anurans (Toledo & Batista 2012, Hoogmoed 2013), nine caecilians and one caudate (Hoogmoed 2013).Although it represents the third highest richness of anurans among the Brazilian states (Toledo & Batista 2012), this biodiversity can be considered underestimated in number and complexity when considering the political limits, given the existence of a large number of cryptic species (Funk et al. 2011), several sampling gaps due to the concentration of researches in a few areas (Azevedo-Ramos & Galatti 2002) and the scarcity of studies that use appropriate methods for the record of amphibians belonging to different niches (Lynch, 2005).However , the number of recorded species has increased rapidly, with frequent descriptions of new species of anurans (e.g., Nunes et al. 2012, Simo ˜es et al. 2013, Peloso et al. 2014, Rojas et al. 2014) and amphibians of the orders Gymnophiona (Maciel & Hoogmoed 2011) and Caudata (Brcko et al. 2013) for the biome.
There are large, virtually unknown areas in the Amazon that possibly have never been sampled.Localities in the west of the biome, for example, seem to have higher amphibian diversity than other localities (Duellman 1999, Azevedo-Ramos & Galatti 2002).This information has been confirmed by the high number of species recorded in several studies conducted in recent years in the western region of the biome (Franc¸a & Vena ˆncio 2010, Ilha & Dixo 2010, Bernarde et al. 2011, Pantoja & Fraga 2012, Bernarde et al. 2013, Prudente et al. 2013, Waldez et al. 2013).However, few studies focus on the importance of varzea environments (Pantoja & Fraga 2012, Bernarde et al. 2013, Waldez et al. 2013), and no studies have systematically sampled amphibian assemblages in abandoned oxbow lakes and in floating meadows.The available studies on amphibians in floating meadows were conducted in the region of Manaus, in the lower Solimo ˜es River varzea and Janauari Lake (Ho ¨dl 1977, Schiesari et al. 2003), in the Para ´(Suriname), Negro (Bolivia) and Curua ´(Brazil) rivers (Hoogmoed 1993) and, recently, in the Samiria River basin, Peruvian Amazon (Upton et al. 2014).Heyer (1976) conducted the first studies on the amphibian fauna of the Purus River region during two expeditions in the western Amazon.Since then, a number of inventories have been conducted in this river basin, from its highest portion in the Peruvian Amazon (Rodrı ´guez 2003), to the middle portion (Franc¸a & Vena ˆncio 2010), down to its confluence with the Solimo ˜es River (Heyer 1977, Gordo 2003, Waldez et al. 2013), demonstrating a surprisingly high amphibian richness compared to other areas of the Brazilian (Gordo 2003) and Peruvian Amazon (Rodrı ´guez 2003).Such studies indicate the need for priority conservation efforts (Azevedo-Ramos & Galatti 2002, Franc¸a & Vena ˆncio 2010).Despite the existing knowledge on amphibians of the Purus River basin, basic information on the composition and richness and abundance patterns of the assemblages is fragmented, with gaps remaining in the knowledge on amphibians at various scales, thereby limiting conservation efforts.In this study, we present the composition, richness and abundance of the amphibian fauna in varzea environments and floating meadows of the oxbow lakes of the Middle Purus River, Amazonas, Brazil.

Study area
The study area is located in six oxbow lakes situated along the middle Purus River varzea between the municipalities of Boca do Acre and Paunı ´, Amazonas State (08°51'18''S and 67°44'34''W; 08°19'33''S and 67°20'51''W) (Figure 1).The Purus River basin is located in the southwestern Amazon, and the Purus River is one of the major tributaries of the Solimo ˜es-Amazon system, with its headwaters in the hills of the Fitzcarrald Arch, located in lowland forests in the states of Ucayali and Madre de Dios, Peru.The Purus River is one of the longest rivers in South America, running approximately 3,380 km, entering Brazil in the state of Acre, with its mouth in the Solimo ˜es River (Amazonas State) (Rı ´os-Villamizar et al. 2011).The Purus is classified as a white water river sensu Sioli (1991), meanders, and has water rich in Andean sediments (Rı ´os-Villamizar et al. 2011).The region has a monsoon-type tropical rainy climate (Am following Ko ¨ppen; Brazil 1976).Rainfall in the Purus River basin has an annual cycle marked by a rainy season (between November and March) and a dry season (between May and September).April and October are considered transition months (Silva et al. 2008).

Data collection
Data collection took place between April 2013 and January 2014 for a total of 45 sampling days; the sampling period consisted of the transition between the rainy and dry seasons (6 to 20 April), beginning of the dry season (1 to 17 October) and the rainy season (4 to 20 January).We selected six oxbow lakes along approximately 200 km of river.Three lakes are located on the right bank of the Purus (Santana, Bom Lugar and Flor do Ouro) and three are located on the left bank (Cameta ´, Verde and Guama ´).We sampled forty-seven 200meter long transects in the lakes in each campaign, consisting of the ''floating meadows'' environments on the inside of the lakes and in the varzeas.The varzea environments were classified according to Junk et al. (1989) as follows: (1) high varzea -highest land segment of the plain, free from annual floods but occasionally flooded in larger floods (further from the riverbed); and (2) low flooplain -located in the inner portions of the lakes, between the lakes and the river, flooded during intermediate floods (Table 1).However, comparing vegetation types was not the objective of this study.We established a minimum distance of 1,000 meters between transects in the same environment type in each lake.We standardized samplings into nine transects, three in floating meadows, three in high varzea and three in low varzea in the Bom Lugar, Cameta ´, Flor do Ouro, Guama ´and Santana lakes.Standardization was not possible at Lake Verde because the high sediment accumulation prevented movement inside of the lake; therefore, only one transect in the high varzea and one in the low varzea were sampled at Lake Verde (Table 1).
We sampled each transect during the day (1:00-5:00 p.m.) and night (7:00-11:00 p.m.), recording the animals through visual and acoustic searches limited by distance and time (200 meters in one hour), using a modified form of a visual time-constrained search (Crump & Scott Jr 1994, von May et al. 2010).Each transect in the varzea environments was sampled by two people, walking side-by-side, at an approximately constant speed, recording the amphibians on both sides of the trail (Seber 1986, Ro ¨del & Ernst 2004).Whenever necessary, such as during samplings conducted along the lakeshore, special attention was given to the presence of aquatic macrophytes, and the transect was stratified (stratified random sampling) (Seber 1986).
All individuals observed and/or calling (observed and/or not observed) were identified and counted.We account all individuals in calls activity in approximately 20 meters along the transect.The collected individuals were recorded, placed in plastic bags and transported to the laboratory to be photographed and measured.The nomenclature used followed Frost (2015).Voucher specimens were anesthetized and euthanized with 5% xylocaine, fixed in 10% formalin (one part 37% formalin and nine parts water) and preserved in 70% alcohol.Then, specimens were deposited in the herpetological collections of the Federal University of Acre (Universidade Federal do Acre -UFAC), Rio Branco, Acre State, Brazil, and Federal University of Goia ´s (Universidade Federal de Goia ´s -ZUFG), Goia ˆnia, Goia ´s State, Brazil (sampling permit ICMBio #37322-5) (Appendix 1).

Data analysis
Because species richness depends not only on the characteristics of the area but also on the sampling effort, we built a species rarefaction curve (Santos 2006) to analyze amphibian richness according to increased sampling effort.Species richness was estimated by extrapolating the species rarefaction curve, using the mean of the four values generated by the abundance-based richness estimators (ACE, Chao 1, Jack-knife 1 and Bootstrap) (Colwell et al. 2012).In this case, using mean values minimizes variations in the performance of certain estimators, usually related to differences in the sampling, diversity and equitability of assemblages (O'Hara 2005, Waldez et al. 2013).The species accumulation and total estimated richness curves were obtained from 1,000 randomizations using the software EstimateS 9.1.0(Colwell 2013).The dominances were represented by a Dominance Component curve or Whittaker Diagram, obtained by ranking species, starting with the most abundant, along the x-axis and the logarithm abundances on the y-axis.Rare species were those represented by a single individual (singletons), and the same number of species in the upper end of the abundance  (Hammer et al. 2001).
The conservation status of each species was based on the Red List of Threatened Species of the International Union for Conservation of Nature (IUCN 2015).The endemism and association of species with other biomes was based on Toledo & Batista (2012).Species not listed in those sources were evaluated based on their respective descriptions and taxonomic revisions, such as Dendropsophus joannae (Ko ¨hler & Lo ¨tters 2001), Typhlonectes compressicauda (Maciel & Hoogmoed 2011), Elachistocleis muiraquitan (Nunes-de-Almeida & Toledo 2012) and Chiasmocleis royi (Peloso et al. 2014).

Results
We recorded 8,732 amphibian specimens, distributed across three orders, 10 families, 22 genera and 59 species in the oxbow lakes of the varzea of the Middle Purus River.The order Anura was the most rich (57 species), and the families Hylidae (31 species), Leptodactylidae (nine species) and Microhylidae (five species) were the most speciose.Ceratophrydae and Dendrobatidae had only one species each.The orders Caudata and Gymnophiona were represented by only one species, the salamander Bolitoglossa caldwellae and the aquatic caecilian Typhlonectes compressicauda, respectively (Table 2; Figures 4 to 7).
The species rarefaction curve, considering all samples, indicated a rapid increase in species richness with the increase in the number of transects sampled and a trend toward stabilization in the last samplings, with the observed richness representing 89.82% of the expected richness for the region considering the mean of the estimators (62.34 ± 3.03 species).The clear trend toward stabilization of the curves, as shown by the considerable overlap of the confidence intervals of the observed and estimated richness, indicates that the sampling method used was appropriate for determining the local richness (Figure 2).
Twenty-nine species found in this study are endemic to the Amazon, 17 occur in the Amazon and Cerrado (Brazilian savanna) biomes, and 10 have a generalized distribution, occurring in more than two biomes.The distribution or status of three species (Dendropsophus gr.minutus, Elachistocleis sp. and Scinax gr.ruber) were not evaluated due to uncertainties in identification.According to the IUCN Red List of Threatened Species (IUCN 2015), none of the species found are listed in any threat category, two (Allobates gasconi and Dendropsophus joannae) are considered data deficient, and the conservation status of three (Bolitoglossa caldwellae, Chiasmocleis royi and Elachistocleis muiraquitan) has not been evaluated yet (Table 2).
A recent review of the taxonomy and distribution of the salamanders of the genus Bolitoglossa in Brazil enabled the recognition of five species within the genus for the Brazilian Amazon (Brcko et al. 2013).Of these, Bolitoglossa caldwellae was described for the municipalities of Cruzeiro do Sul, Marechal Thaumaturgo and Porto Walter, Acre State.The capture of B. caldwellae individuals in the varzea environments of the Middle Purus River is the first record for Amazonas State and extends its distribution approximately 600 km from the nearest locality (Porto Walter).
The microhylid Chiasmocleis royi, historically mistaken for Chiasmocleis ventrimaculata, was recently described for the state of Acre (Peloso et al. 2014).The previously known distribution for C. royi encompasses the eastern and southern regions of Peru, northern Bolivia and western Brazil in the  The record of Pristimantis lacrimosus and Pristimantis skydmainos in the area studied is relatively important.Pristimantis lacrimosus is a small anuran that occurs in the upper Amazon River, distributed from southern Colombia, through Ecuador, northern Peru to the far west of Brazil in the Serra do Divisor National Park and Alto Jurua Extractive Reserve, Acre State (Duellman & Lehr 2009).It inhabits lowlands of tropical forests and low and humid montane forests, being commonly found in arboreal bromeliads (Duellman & Lehr 2009).Pristimantis lacrimosus was recorded at lake Guama ´, in a varzea environment,   The occurrence of Dendropsophus joannae in Brazil has been suggested since it was first described (Ko ¨hler & Lo ¨tters 2001).However, the species had been misidentified in studies conducted in the state of Acre due to its morphological similarity to Dendropsophus leali, causing it to remain known only for its type locality (Cobija, Bolivia) for more than a decade (Frost 2015).The occurrence of this species in Brazil was recently confirmed by Melo-Sampaio & Souza (2015), which extended its distribution to different places across the state of Acre.In the Middle Purus River, Dendropsophus joannae occurred in four of the six sampled lakes, in varzea environments and always while calling in bushes near the creeks or flooded forests (igapo ´s).These records are the first to Amazonas state and expand the geographical distribution of Dendropsophus joannae by approximately 160 km.
Although this study was conducted strictly in varzea environments and floating meadows, i.e., lowlands, the species richness reported is high compared to most of the studies conducted in the western Amazon (Azevedo-Ramos & Galatti 2002, Souza et al. 2008, Franc¸a & Vena ˆncio 2010, Ilha & Dixo 2010, Pantoja & Fraga 2012, Bernarde et al. 2013, Prudente et al. 2013), corroborating the existence of a diversity gradient where the western plains of the Amazon biome have greater amphibian diversity as a possible effect of the proximity of the highlands of western Amazon and its high patterns of endemism and richness (Heyer 1976, Duellman 1978, 1982, 1988, Azevedo-Ramos & Galatti 2002).This high diversity has also been associated with the occurrence areas of endemism (Silva et al. 2005, Waldez et al. 2013).The Middle Purus River is located in the Inambari endemism area, the second largest in terms of area (1.326.684km 2 ), also considered to be species rich and to have a large number of endemic species with restricted distribution (Silva et al. 2005).
The prevalence of the families Hylidae and Leptodactylidae is a pattern known for Neotropical environments (Duellman 1999, Segalla et al. 2012), including the Brazilian Amazon (Azevedo-Ramos & Galatti 2002, Neckel-Oliveira 2013, Toledo & Batista 2012).Hylids predominate in the studied environments, becoming more dominant in floating meadows (81% of the species), where species of other families were only found occasionally.Apparently, there is no fauna adapted to this floating vegetation, which is mainly invaded by species of the family Hylidae, also present in other open herbaceous vegetation of humid environments, or by aquatic amphibians (Hoogmoed 1993).Although the objective of this study was not to test hypotheses that support the proposal of Hoogmoed (1993), the phylogenetic and ecomorphological similarities of the species associated with the floating meadows indicate a possible influence of the flood pulse, which selects species with ecomorphological characteristics adapted to seasonal flooding (Junk et al. 1989).Hylid species common in other areas, such as Dendropsophus triangulum and Hypsiboas punctatus, have been found in high densities in floating meadows environments, being rare in adjoining terrestrial habitats, which highlights the importance of floating meadows for many species of the family Hylidae (Upton et al. 2014).Furthermore, the occasional record of Dendropsophus triangulum, Hypsiboas punctatus, Leptodactylus leptodactyloides and Rhinella marina individuals on floating meadows moving between the lake and river during the drawdown period supports the suggestion that several species use floating meadows banks as dispersal vectors over long distances in Amazon rivers (Schiesari et al. 2003).
Richness, abundance and, consequently, species composition in assemblages are limited by environmental conditions (e.g., the humidity, temperature, rainfall, nutrient availability, and physical structure of the habitat) and by biotic processes (e.g., the predation, competition, dispersal, disturbances, and diseases) (Ricklefs 1987, Parris 2004).In this study, we did not evaluate changes in species composition among the lakes, but the data show possible influence of these biotic and abiotic factors on amphibian communities.However, differences in richness and the presence of species exclusive to some lakes indicate the influence of diversity structuring mechanisms on the amphibian assemblages along the area studied in the Purus River.For amphibians associated with floating meadows, these variations may be associated with the high beta diversity of macrophytes known among the lakes (Mormul et al. 2013), since there are specific association between plant species and amphibian species (Ro ¨dl 1977, Upton et al. 2014).
With the exception of the aquatic caecilian Typhlonectes compressicauda, all species considered rare were recorded exclusively in varzea environments.This rarity may be attributed to the association of these species to terra firme forests that are not influenced by seasonal flood pulses (Junk et al. 1989, von May et al. 2010, Bernarde et al. 2013, Waldez et al. 2013).The absence of these species in floating meadows can be attributed to their morphological and behavioral characteristics adapted to the vertical stratum, which limit or prevent the use of floating meadows as a resource.Furthermore, the characteristics of the floating meadows systems, since it is a horizontal design environment, directly exposed (no vegetation cover), and unstable (movement dependent of wind direction), can act as environment filters to prevent the occurrence of the species dependent on the terrestrial ecosystem.In contrast, all species considered abundant in the lakes studied have a wide distribution in the Amazon (AmphibiaWeb 2015, Frost 2015), four (Adenomera hylaedactyla, Hypsiboas fasciatus, Leptodactylus leptodactyloides and Scarthyla goinorum) are more abundant in varzea environments, and four (Dendropsophus triangulum, Hypsiboas punctatus, Sphaenorhynchus dorisae and Sphaenorhynchus lacteus) are associated with floating meadows.These associations of the amphibians with floating meadows and varzea environments were also found in other studies (Ho ¨dl 1977, Hoogmoed 1993, Upton et al. 2011, Bernarde et al. 2013, Catenazze et al. 2013, Waldez et al. 2013).
Species abundance distributions may reflect the processes that determine the biological diversity of an assemblage (Magurran 2011), due to the assumption that the abundance of a species reflects its success in competing for limited resources.The little sloped curve produced for the assemblages sampled in the lakes of the Middle Purus River indicates that few species have extreme abundances and that most have intermediate abundances, generating a lognormal distribution model.Lognormal curves result from different population growth rates of many coexisting species (MacArthur 1960).This type of distribution is found in assemblages controlled by ecological factors, with equitable sharing of available resources, as occurs for most assemblages of tropical forests (Ugland & Gray 1982, Ferreira & Petrere-Jr 2008, Magurran 2011).
Because species with small geographic ranges tend to be more threatened than species with wide geographic ranges (Toledo & Batista 2012), the high number of species endemic to the Amazon biome recorded in this study (49.15%) reinforces the need for conservation efforts for the amphibian fauna in the region of the Middle Purus River (Franc¸a & Vena ˆncio 2010).Many species are data deficient due to gaps in knowledge on their ranges and populations sizes and population trends (Morais et al. 2013).The data deficiency can be partially explained, for example, by factors such as data uncertainty during the evaluation process, reduced body size, cryptic behavior, fossorial habits, nocturnal activity and lack of studies on the natural history of species (Butchart & Bird 2010).Two species (Allobates gasconi and Dendropsophus joannae) from the oxbow lakes of the Middle ).Another three species (Bolitoglossa caldwellae, Chiasmocleis royi and Elachistocleis muiraquitan), all described recently, are still not included in the IUCN database.Such situations can become a problem for conservation because the extinction risk of these species becomes unknown (Morais et al. 2013).
The high amphibian richness recorded in this study for the oxbow lakes of the Middle Purus River contributes to the knowledge on species inhabiting lowlands in the biome, including the expansion of spatial distributions, habitat use and population sizes.Although the sampling effort was enough to represent the regional amphibian assemblages, future studies should assess the importance of varzea environments and floating meadows for amphibian conservation as well as the association of the several species directly or indirectly associated to floating meadows and their ecological interactions.

Figure 1 .
Figure 1.Location of the six oxbow lakes studied in the Middle Purus River, Amazonas, Brazil.
) and Rondo ˆnia (Guajara ´-Mirim).We were the first to record this species for the state of Amazonas, where it was recorded in four of the six lakes studied, expanding the distribution of the species by approximately 200 km.Another microhylid, Elachistocleis muiraquitan, previously known only from southeastern Peru, northwestern Bolivia and Brazil only in the municipalities of Rio Branco and Xapurı ´, Acre State (Nunes-de-Almeida et al. 2012, Allen et al. 2014), was also recorded in four of the lakes studied, thus being the first recorded for the state of Amazonas, extending its distribution by approximately 250 km.

Figure 2 .
Figure 2. Rarefaction curve based on sampling effort (number of transects sampled, red line) and mean of the richness estimates (black line) of four different richness estimators (ACE, Chao 1, Jack-knife 1 and Bootstrap) for the amphibian assemblages recorded in the varzea environments and floating meadows of the oxbow lakes of the Middle Purus River, Amazonas, Brazil.Dashed lines represent the 95% confidence intervals.

Table 1 .
(Magurran 2011led in six oxbow lakes of the Middle Purus River, Amazonas, Brazil.distribution were used to define species as common.The other species were classified as having intermediate abundance.The pattern of the species abundance distribution was fitted to the logarithmic, geometric, lognormal and broken-stick models.Model fit was assessed by the chi square adherence test(Magurran 2011) using the software PAST version 2.17c

Table 2 .
Amphibian species recorded in the varzea environments and floating meadows of the oxbow lakes of the Middle Purus River, Amazonas, Brazil.N ¼ Abundance; Bl ¼ Bom Lugar; Ca ¼ Cameta ´; Fo ¼ Flor do Ouro; Gu ¼ Guama ´; St ¼ Santana; Ve ¼ Verde; Fm ¼ Floating Meadows; Vz ¼ Varzea; Gen ¼ occurrence in more than two biomes; Am, Ce ¼ Occurs in the Amazon and Cerrado; End ¼ Endemic to the Amazon; Status ¼ IUCN conservation status; LC ¼ Least Concern; DD ¼ Data deficient; NA ¼ Not available.