Snake assemblages of Marajó Island , Pará state , Brazil

We describe the diversity, natural history and structure of snake assemblages from Marajó Island, state of Pará, Brazil, after analyzing 439 specimens deposited in herpetological collections. We tested the hypothesis that snake assemblages from forest and open areas of Marajó Island are distinct with regard to their structure, composition and functional groups. To compare the snake composition of the forest and open areas of Marajó with other comparable assemblages in Brazil, Principal Coordinate Analysis and Clustering tests were performed. A total of 61 species of snakes was recorded for Marajó, with ten species cited for the first time for the study area (Atractus natans Hoogmoed & Prudente, 2003, A. schach (Boie, 1827), Dendrophidion dendrophis (Schlegel, 1837), Helicops hagmanni Roux, 1910, Hydrops martii (Wagler in Spix, 1824), Lygophis meridionalis (Schenkel, 1901), Erythrolamprus typhlus (Linnaeus, 1758), Philodryas argentea (Daudin, 1803), Siphlophis cervinus (Laurenti, 1768), and Thamnodynastes sp.). The composition and structure of snake assemblages between forested and open were different, with five functional groups of snakes in forest areas, and three groups in open areas, based on habit and habitat. In all, 19 species were exclusive to forest areas, 10 were exclusive to open areas and 26 species were recorded in both areas. Our results revealed greater richness for forested areas, probably due to greater habitat heterogeneity. The species composition for forested area in Marajó was similar to that found in other Amazonian assemblages, while that for open areas was more similar to the Pantanal region than other open area assemblages. The general structure of the snake assemblage of Marajó was dominated by anurophagous, terrestrial and diurnal species. Terrestrial, arboreal and semi-arboreal snakes showed a seasonal offspring production pattern, while the pattern for aquatic and semi-aquatic species was aseasonal. The structure of assemblages was influenced by both historical and ecological factors.

produced a list of 38 snake species.Subsequently, yuki & SantOS (1996) listed 52 species of snakes for these two islands using specimens deposited in the MPEG.This set of data resulted in a list of snake species for the island but did not include details about their structure and functional groups.
Given that Marajó Island includes both densely forested and open areas, we test the hypothesis that their snake assemblages are distinct with regard to structure, composition and functional groups.Additionally, we compared the snake composition of forested versus open areas of Marajó with comparable assemblages in other regions of Brazil.We provided more detailed information on natural history, including diet, reproduction and species distribution throughout the island.

MATERIAL AND METHODS
Marajó Island, located between the Brazilian states of Amapá and Pará (0°59'39"S, 49°44'25"W), has an area of approximately 49,000 km 2 (gOulding et al. 2003), being delimited to the south by the Pará River, to the north by the Atlantic Ocean, to the east by the Marajó Bay and to the west by the mouth of the Amazon River.Three distinct ecological regions can be found on the island: dense forests predominate in the western half; successional formations or alluvial fields cover almost its entire eastern half; and Cerrado (natural grasslands) (BritO & ilkiu-BOrgeS 2013) predominates in small areas of the central-north and south eastern regions of the island (radam 1974) (Fig. 1).Our main objective was to test if there are differences between the snakes assemblages of forested and open areas of the island.To make the comparison more robust we considered both the successional formations and the Cerrado together as comprising open habitat.
There is a distinct relationship between the vegetational distribution on the island and the mean annual temperature, with an annual mean in the eastern region (27.3°C)higher than in the western region (25.7°C)(lima et al. 2005).The mean annual rainfall becomes higher from the eastern region (<2500 mm to 3000 mm) towards the western region (3000 mm to > 4000 mm).In the western half of the island, according Köppen's climate classification, the climate is Tropical Humid (Af) and in the eastern half it is Tropical Monsoon (Am) (alVareS et al. 2014).
Data was obtained from the analysis of 439 specimens from different regions of Marajó Island, collected between 1958 and2009, and 1993).For the resource-use matrix, we considered only the primary records of substrate used and daily activity of each species.Substrate use was considered according to the categories proposed by StrüSSmann (2000); aquatic or semi-aquatic snakes were included in the substrate "water", while the fossorial, cryptozoic and terrestrial snakes were included in the substrate "soil" and arboreal and semi-arboreal snakes in the substrate "vegetation".Three categories of activities were also defined: a) diurnal; b) nocturnal; c) and diurnal/nocturnal for snakes found active both during the day and night.
Cluster analysis (manly 1994) was performed by means of the Sorensen similarity coefficient, which gives greater weight to presence than to absence (kreBS 1989), due to the lack of information for some species.In the cluster analysis we used the Unweighted Pair-Group Method of Arithmetic Averages (UPGMA) or group mean (Sneath & SOkal 1973, kreBS 1989), based on the Jaccard Similarity Index (Cj).For all specimens analyzed, a longitudinal incision was made in the ventral region for observations on diet and reproduction.The presence of food content in the stomach and intestine, and prey ingestion direction was verified.The stomach contents were preserved in vials of 70% ethanol and received the same voucher number as the examined specimen.The number of prey items in each stomach was computed, and collected items were referred to specialists for identification.All prey specimens were deposited in the herpetological collection of MPEG.For reproductive characteristics, direct observations on the presence of eggs or embryos in females were carried out.Males were considered mature when the deferent ducts were opaque and convoluted, indicating the presence of sperm (SantOS-COSta et al. 2015).
The period of offspring production (number of new offspring per individual breeder) was estimated from a circular statistical analysis (zar 2010), using ORIANA 4:02 (kOVaCh 2009).Months were converted into angles (30° intervals) and the number of eggs, embryos and offspring observed each month was considered as their frequency in each angle (month) (PradO et al. 2005, BOth et al. 2008, Prudente et al. 2014).From this method, the following values were estimated: a) mean vector (μ); b) length of the mean vector (r); c) circular standard deviation (SD) related to μ; and d) Rayleigh's Uniformity Test.Rayleigh's Uniformity Test (zar 2010) calculates the probability of the data being evenly distributed (p > 0.05) (Null Hypothesis).A significant result of Rayleigh's Uniformity Test (p < 0.05) indicates that data are not distributed evenly.
For forested areas of Marajo Island (dense forests) five functional groups were found: a group formed exclusively by aquatic species (Group A); two formed exclusively by terrestrial species (Groups B and E); and two groups formed by arboreal and semi-arboreal species (Groups C and D) (Fig. 2).
Three functional groups were found for open areas of island (successional formations and Cerrado), as follows: a group formed uniquely by aquatic snakes (Group A); a group of predominantly terrestrial snakes, but containing one arboreal species, Leptophis ahaetulla (Linnaeus, 1758), and a semi-arboreal species, Leptodeira annulata (Linnaeus, 1758) (Group B); and  E. deschauenseei (Group C) (Fig. 3).The aquatic group occurring in both forested and open areas is formed by nocturnal species that use the aquatic substrate for foraging a diet consisting mostly of fish.In general, its representativeness is relevant because it includes 15% of the species from Marajó -Micrurus surinamensis (Cuvier, 1817), H. martii, two species of Eunectes and five species of Helicops.The terrestrial groups present in forested areas are better defined than those from open areas (in which these species were grouped with arboreal ones), being formed by species that feed mainly on anurans, lizards and small mammals.
Comparing the composition of snakes from the two environments on the island with assemblages from ten other localities, we observed that the first two axes of the Principal Coordinate Analysis jointly explained 46.31% of the variation (axis 1: autovalue 1.24 and 32.69% of the variation; axis 2: autovalue 0.52 and 13.62 of the variation).The analysis ordered the assemblages into two groups: Group 1 with open area assemblages (Exú, Castelo do Piauí, and Sete Cidades); and Group 2 formed by assemblages from predominantly forested and forest-cerrado transition areas (Poconé and open areas of Querência; forested of Espigão do Guiana; Caxiuanã and Manaus) (Figs. 4, 5).
A total of 369 digestive tracts from 53 species were analyzed and 119 contained some type of prey item, enabling the collection of information on the diet of 28 species.Of these, the most frequent prey item was "anuran" (56%), followed by "mammal" (16%), "lizard" (12%), "fish" (8%), "arthropod" (4%), "earthworm" (3%) and "snake" (1%).Among the items that were found in adequate condition to enable the inference about ingestion direction, there was a prevalence of anteroposterior digestion (n = 36, 71%) when compared to posteroanterior (n = 15, 29%) (Table 2).Eggs, embryos and offspring of terrestrial, arboreal semi-arboreal species were recorded mainly in the rainy season, indicating a seasonal pattern for offspring production.Eggs, embryos and offspring of aquatic and semi-aquatic species were recorded in practically all months of the year, indicating an aseasonal pattern for offspring production Table The higher richness snakes observed in forested areas of Marajó appears related to the greater environmental heterogeneity.Forests have greater structural diversity of habitats, enabling snakes to occupy vertical substrates while also finding more areas of shelter and higher richness of potential prey (Pianka 1966, Vitt 1987).
The similarity between the assemblage from forested area of Marajó and other Amazonian assemblages is the result of the high number of common species between the areas (53 species).Specifically, the high similarity of Marajó with the Caxiuanã region is probably related to the source of colonization (see maCarthur & wilSOn 1967), considering its recent formation and its geographical proximity to the mainland.The environmental structure of open areas of Marajó provides favorable conditions for the presence of species adapted to similar environments in other regions, such as the Caatinga of the northeast of Brazil and the Cerrado of central and northeastern Brazil.
Morphological and behavioral characteristics, as well as substrate use, foraging activity and type of prey often share a common evolutionary origin (Cadle & greene 1993, alenCar et al. 2013), and are fundamental in structuring snake communities (França et al. 2008).Thus, snakes from the same phylogenetic lineage tend to group in dendrograms of resource use (see mar-

DISCUSSION
The large number of species recorded for Marajó Island reflects the high richness of snakes in the Amazon region, as demonstrated in other studies (e.g., duellman 1978, dixOn & SOini 1986, martinS & OliVeira 1999, maSChiO et al. 2009).In this study, the presence of ten species that had not been previously recorded for the island indicates that new records can be added even in areas that have been well inventoried (naSCimentO et al. 1991, yuki & SantOS 1996).other hand, in grouping analysis, it was possible to observe phylogenetically distinct species that converged on the same resource, as observed, for example, in Group C for open areas formed by four subfamilies (Boinae, Colubrinae, Crotalinae and Xenodontinae) that forage both in the soil and on vegetation, preying primarily on anurans and lizards.The overlap in the use of resources by snakes from the same community indicates that factors such as prey abundance and predation may aid in avoiding competition between syntopic species (Bernarde & aBe 2006).Thus, the fact that there are groups formed by both phylogenetically close species and distant species may indicate that the community is being structured by both historical and ecological factors.
The record of anurans in the diet of 54% of the species from Marajó Island does not correspond to the pattern observed for central and eastern Amazonia, where a greater number of saurophagous snakes were observed, 42% and 58%, respectively (martinS & OliVeira 1999, Bernarde & aBe 2006).The high frequency of amphibians in the diet was a result of the high frequency of species that feed primarily and exclusively on anurans, such as Chironius, Erythrolamprus and Thamnodynastes (Cadle & greene 1993, dixOn et 1993).
The proportion of aquatic on Marajó is relatively high (20%), and is probably associated to the extensive and prolonged flood periods that occur on the island, mainly from January to June (gOulding et al. 2003).The same pattern of aquatic species (15% of the assemblage) was observed in the Pantanal, a region that also remains flooded for long periods of the year (February to June) (StruSSmann & Sazima 1993).
In aquatic environments, seasonal variation in the availability of food resources is lower when compared to terrestrial environments (Junk 1997).With the rise in water level during the rainy season, there is a natural expansion in the physical space and, consequently, an increase in the availability of habitats for fish prey, which reduce in density due to dispersion.However, with the lowering of the water level during the dry season, the opposite occurs, that is, the physical space decreases as does the habitat availability for prey, increasing their density (see Okada et al. 2003, luz-agOStinhO et al. 2009).This variation in water level probably exerts some influence on aquatic species such as E. murinus, H. leopardinus and H. gigas that use fish as their main food resource (StaraCe 1998, lóPez & giraudO 2003, ÁVila et al. 2006), or on species that forage in aquatic environments such as E. taeniogaster and E. poecilogyrus, which consume tadpoles and adult anurans.
The relative ease of finding food throughout the year in aquatic environments explains the noted aseasonal offspring production for aquatic species of Marajó.The opposite situation is observed with terrestrial, arboreal and semi-arboreal species, where offspring production occurs almost exclusively in the rainy season, when the availability of resources in these substrates increases (martinS & OliVeira 1999).
Marajó Island corresponds to a relatively small portion of the Amazon biome, and yet its unique formation process has resulted in distinctive phytophysiological characteristics, producing environmental specialization and hence a diverse ophidofauna.The results presented here form the basis for future studies of Amazon biodiversity in a region under severe processes of fragmentation and anthropic change.
Figures 6-7.Seasonal variation in frequency of eggs, embryos and offspring of aquatic and semi-aquatic species (6) and terrestrial, arboreal and semi-arboreal species (7) in the Marajó Island, Pará, Brazil.
Figure 1.Map showing the location of the Marajó Island, its main plant formations and the territorial boundaries of its municipalities.Modified from Projeto RADAM (1974).

Figure 3 .
Figure 3. Grouping dendrogram of resource use based on diet and use of foraging substrate data from 61 species recorded in forested areas at Marajó Island, Pará, Brazil.

Table 2 .
Food content (and respective direction) of Island, Pará, Brazil.Legend: n = number of snakes with respective content; A-P = Antero-posterior; P-A = Postero-anterior; Arabic algorism between parenthesis = number of prey in the same individual.

Figure 4 .
Figure 4. Ordination diagram of the Principal Coordinate Analysis resulting from the composition of snake species (presence or absence).Data for Ilha de Marajó was entered in two columns: forested area and open area.Axis 1: autovalue 1.24 and 32.69% of variance; Axis 2: autovalue 0.52 and 13.62% de variance.

Figure 5 .
Figure 5. Grouping dendrogram resulting from the composition of snake species (presence or absence), analyzed by separating the two physionomic units (forested area and open area) in the Marajó Island, Pará, Brazil.

Table 3 .
Results of the circular statistical analysis testing for seasonality in the snakes offspring production at Marajó Island, Pará, Brazil.