Are there co-occurrence patterns that structure snake communities in Central Brazil ?

The main factors that structure Neotropical animal communities have been the subject of discussion in ecology communities. We used a set of null models to investigate the existence of structure in snake communities from the Cerrado in Central Brazil in relation to the co-occurrence of species and guilds concerning specific resources. We used fragments (conservation units) inside the Distrito Federal and neighbor municipalities. In spite of recent human colonization in the region from the end of the 1950s, intense habitat modification and fragmentation has taken place. Sixty three snake species are present in the Distrito Federal. Co-occurrence analysis of species and guilds associated to snake diets and habitats suggested a lack of organization. The homogeneity of habitats in Central Brazil and the minor importance of ecological effects can lead to random arrangement.


Introduction
Understanding processes responsible for the community structure is a central problem in community ecology.The topic has been widely debated by many ecologists studying different taxa and biomes (Cody, 1974;Pianka, 1973;Ricklefs and Schluter, 1993).The comparative observation of ecological interactions between species provides a wide range of evidence concerning the importance of ecology in community structure (Cody and Diamond, 1975;Losos, 1983;Pianka, 1986;Schoener, 1974).However, few studies have investigated the effect of isolation on the reptiles' community structure (Case, 1983;Gainsbury and Colli, 2003;Murphy, 1983).
Human population growth and the consequent implementation and maintenance of a productive infrastructure with the construction of a structured urban area and, mainly, a large rural area of planted pastures and crops, have been the basic factors of fragmentation of natural environments (Sutherland, 2001).Isolation promotes local colonization and extinctions that influence species interactions.The longer they have been in isolation, the more susceptible the species are to extinctions (Case and Cody, 1987;Foufopoulos and Ives, 1999).Studies on the effect of the habitat fragmentation on communities of lizards and snakes have registered a positive association between species diversity and patch size (Cosson et al., 1999;Kjoss and Litvaitis, 2001).
The development of null models that yield null communities generated by randomization of the original data is a powerful tool for investigating community structure and this analysis has been applied to the study of various animal groups (Caswell, 1976;Colwell and Winkler, 1984;Connor and Simberloff, 1979;Gainsbury and Colli, 2003;Gotelli, 2000Gotelli, , 2001;;Jackson et al., 1992;Pianka, 1980).
Herein we use null models to investigate the existence of structure in snake assemblages of fragments in the Distrito Federal region of Central Brazil.This region has been intensively fragmented since the end of the 50's due to building Brasília, the new capital, and cities that have grown around it (UNESCO, 2000).

Study area
The studied area is an arbitrary circle of 100 km radius around the center of Brasília (15° 47' S and 47° 52' W), including all the territory of Distrito Federal and part of neighboring municipalities in adjacent states of Goiás (Luziânia and Alexânia municipalities) and Minas Gerais (Unaí municipality) covering 31,400 km 2 (Figure 1).
The region is located in the core of the Cerrado morphoclimatic domain, which covers about 2 million km 2 of the Brazilian territory, mostly on the Central Brazilian plateau (Ab'Saber, 1977).The Distrito Federal region harbors headwaters of important tributaries of the three major Brazilian river basins: the Amazon basin to the north (Tocantins River), the São Francisco basin to the east-northeast, and the Paraná-Plata system to the southsouthwest.The climate in the region is type Cwbl follow-ing Köppen's classification (RADAM, 1982).Average annual rainfall is 1,600 mm, concentrated from October to April with a dry season from April to October.
Inside the Distrito Federal there are conservation areas of varying status.Two Conservation Units of integral protection (Brasília's National Park and Águas Emendadas Ecological Station) and the Environmentally Protected Area of Gama and Cabeça de Veado river basins comprise the Cerrados' Biosphere Reserve, created by UNESCO in 1993.
We used species richness for each area in the Distrito Federal region to test for non-random patterns of species co-occurrence, using EcoSim's Co-occurrence Module (Gotelli and Entsminger, 2001).The data was organized in a matrix of presence (1) and absence (0), in which each species represents a row and each site a column.We used the following options in EcoSim: C-score index (Stone and Roberts, 1990) as a quantitative co-occurrence index, fixed sum row and column constraints, "Sequential Swap" algorithm for randomizing matrices, and 10,000 simulations.Gotelli (2000) shows how these parameters act.The C-score is the number of checkboard units for all unique pairs of species and in a structured community should be significantly larger than expected by chance.Using fixed sum row and column constraints produces null matrices with the same number of site occurrences per species (row totals) and the same number of species per site (column total) as observed in the original data set.The sequential swap algorithm reshuffles the original matrix by repeatedly swapping sub-matrices that preserve row and column totals and is not prone to Type I or Type II errors (Gainsbury and Colli, 2003;Gotelli and Entsminger, 2001).
We used diet and habitat information to investigate the presence of non-random patterns of guild co-occurrence.We consider habitat as an organism's general position in the environment.For this, we used the Guild Structure Module in EcoSim (Gotelli and Entsminger, 2001).This module tests a priori hypotheses concerning guilds according to biologically realistic criteria such as taxonom- ic groupings (all species within a genus), resource-based groupings (all species that use a particular food resource), and functional groupings (all species with similar morphology that exploit a shared resource) (Simberloff and Dayan, 1991).All species in an assemblage are assigned to a guild and each species can only be assigned to one guild.We used Cadle and Greene's (1993) system to classify snakes in two guilds (diet and habitat).Nine guilds were identified based on prey category: arthropods, earthworms ('goo-eaters'), amphibians, lizards, fish and amphibians (water vertebrates), snakes and amphisbaenas (elongate body vertebrates), mammals and lizards, mammals and birds (endothermic vertebrates), and generalists (three or more prey items).Six guilds were identified based on habitat: fossorial, cryptozoic, aquatic, terrestrial, arboreal, and terrestrial plus semi-arboreal.We used the same options of the co-occurrence analysis on EcoSim: C-score index, fixed sum row and column constraints, and "Sequential Swap" algorithm for randomizing matrices.In these analyses, EcoSim measures the significance patterns of the co-occurrence indices among the different guilds.It tests whether the mean co-occurrence index among guilds is larger or smaller than expected by chance.It also tests the co-occurrence index's variance among guilds.Unusually large variances imply that guilds are significantly different in levels of co-occurrence: some guilds have species with high levels of co-occurrence, while others have species with low levels of co-occurrence.An unusually small variance means that guilds are similar in levels of co-occurrence.A random result for the variance means that the level of co-occurrence among guilds is as expected if species were randomly assigned to different guilds (Gotelli and Entsminger, 2001).

Results
A total of 63 snake species were found in our study site in Central Brazil (Table 1).Among these, Corallus hortulanus and Imantodes cenchoa were only collected in neighboring cities outside the Distrito Federal.Colubridae was the most diverse family with 51 species distributed into three subfamilies: Colubrinae (10 spp.), Dipsadinae (05) and Xenodontinae (36).Anomalepididae and Leptotyphlopidae presented one species each, and the family of coral-snakes Elapidae has two species in Central Brazil.Both, Boidae and Viperidae presented four species.However, this number is likely to increase for Viperidae, because Bothrops neuwiedi was recently recognized as a complex of different species (Silva, 2004) and this classification scheme was not applied here.Hence, we used the Bothrops neuwiedi Complex for our analysis.
There was a high diversity of natural history characteristics in this snake community.Fourteen prey categories were identified, varying from invertebrates, arthropods, and mollusks to medium size vertebrates (Table 1).Thirty species have specialized diets, eating only one prey category, while six species are full generalists, eat-ing five or more categories.Twenty seven species are strictly terrestrial, six are strictly fossorial, and five are strictly cryptozoic.Sixteen species are found in arboreal and semi-arboreal habitats, while six species use aquatic habitats frequently (Table 1).
Out of 63 species present in the entire region, the protected areas of Distrito Federal presented more richness with 45 species in Brasília's National Park and Environmentally Protected Area of Gama/Cabeça-de-Veado, and 40 species in Águas Emendadas Ecological Station.Other localities varied between 35 species in the Unaí municipality to 23 in the Núcleo Rural, the most disturbed area (Table 2).
The co-occurrence analysis of species richness revealed that Liotyphlops ternetzii, Boa constrictor and Bothrops moojeni formed no checkboard units in the presence-absence matrix.The largest numbers of checkboard units was (15) between Micrurus lemniscatus and Oxyrhopus rhombifer, and (12) observed between Drymarchon corais and Atractus pantostictus, and between Atractus pantostictus and Boiruna maculata.The observed C-score index was 1.15, and was not significant (1.17; P = 0.89, Figure 2).These results are consistent with the hypothesis that local coexistence of snake species in Central Brazil is not structured by deterministic processes.
The analysis of nonrandom patterns of guilds cooccurrence showed similar results.For guilds based on prey categories, the C-score index for each guild varied from 0 for mammal and bird prey guilds to 2.4 for lizardonly prey guild.The index measured between all guilds (1.18) did not differ from the null assemblages mean index (1.15;P = 0.40, Figure 3), and the observed variance of the co-occurrence index among guilds (0.59) did not differ from the simulated variance by chance (0.96; P = 0.69, Figure 3).For guilds based on snake habitat, C-score index for each guild varied from 0 for terrestrial and semi-arboreal snake guild to 1.9 for the aquatic snake guild.The index measured between all guilds (1.19) did not differ from the null assemblage mean index (1.15;P = 0.41, Figure 4), and the observed variance of the co-occurrence index among guilds (0.42) did not differ from the simulated variance by chance (0.57; P = 0.65, Figure 4).

Discussion
Snake species richness in Central Brazil is high compared to other Cerrado localities and even other biomes, such as the Atlantic Forest and Amazon localities.The presence of a high number of species in protected areas of Distrito Federal representing almost all species present in Central Brazil (lacking only Apostolepis flavotorquata, Chironius exoletus, Corallus hortulanus, Drymoluber brazili, and Imantodes cenchoa) has important implications for local snake conservation.All five snakes absent in the protected areas are rare throughout the entire region, with a small number of specimens collected, but are likely to be better represented in these Table 1.Summary of information of the natural history of snakes in the Distrito Federal, Brazil.Abbreviations of diet are: abn = amphisbaenian, amp = amphibian, ann = annelids, arn = aranae, bi = birds, ch = chilopoda, cro = crocodylians, fi = fish, gas = gastropode, ins = insecta, li = lizards, mam = mammals, mi = millipede, sn = snakes.Abbreviations of habits are: Arboreal = Arb, Aquatic = Aqt, Criptozoic = Crp, Fossorial = Fs, Terrestrial = Te, Semi-arboreal = Sarb, Semi-aquatic = Saqt.areas as more inventories and studies are made in the region.The small numbers of species from the Núcleo Rural, however, can be a result of rapid loss of a high number of species in disturbed areas.Co-occurrence analysis indicated no specific patterns of coexistence for species or guilds, showing no structure in the snake community in Central Brazil.The coexistence of species can be limited by negative ecological interactions such as interspecific competition, competition for habitats in the past, species that evolved distinct habitat preferences, and predator-prey relationships (Connor and Simberloff, 1979;Gotelli et al., 1997;    Frequency distribution of checkboard C-scores and variance, obtained from 10,000 simulations producing random snakes guild of habits from Central Brazil.Arrow indicates observed mean; P is the probability that the observed mean is larger than the expected mean.Jackson et al., 1992).The importance of these ecological factors is recognized in communities with sympatric species that had great abundance.In Central Brazil, even more abundant snake species such as Bothrops moojeni, Crotalus durissus, Liophis poecilogyrus, Philodryas nattereri and Philodryas patagoniensis are not present in high densities.This low abundance, plus low metabolism rates, that are characteristics for snakes (Greene, 1997), can minimize these ecological factors.Considering the low importance of the negative ecological interactions, the possibility that all 63 snake species occupy the sites of Central Brazil cerrados is equivalent, resulting in a random arrangement of the species.

Species PNB ESECAE APAGCV Unaí Luziânia Fercal NRural Alexânia Apostolepis albicolaris
Nonrandom resource distribution can also influence species coexistence by causing species composition to vary nonrandomly (Case, 1983;Stone and Roberts, 1990).Hence, species presence can be related to habitat specificity and limited resources would increase species competition (Connor and Simberloff, 1979).All localities in Central Brazil present the same Cerrado habitats and only small local differences can be detected, such as the mesophytic dry forests patches in Fercal and buriti palm marshes in Águas Emendadas Ecological Station.However, the intense habitat loss can lead to modification in the landscape and can influence the relations between the species.

Figure 3 .
Figure 3.Frequency distribution of checkboard C-scores and variance, obtained from 10,000 simulations producing random snakes guild of diet from Central Brazil.Arrow indicates observed mean; P is the probability that the observed mean is larger than the expected mean.

Figure 4 .
Figure 4. Frequency distribution of checkboard C-scores and variance, obtained from 10,000 simulations producing random snakes guild of habits from Central Brazil.Arrow indicates observed mean; P is the probability that the observed mean is larger than the expected mean.

Table 2 .
Species occurrence in eight localities of central Brazil.