Community Structure of Leaf-Litter Ants in a Neotropical Dry Forest : A Biogeographic Approach to Explain Betadiversity

1 Laboratório de Ecologia de Hymenoptera (HECOLAB), Universidade Federal da Grande Dourados, Rodovia Dourados, Itahum, km 12, Cidade Universitária, 79.804-970 Dourados, MS, Brazil 2 Programa de Pós-Graduação em Entomologia e Conservação da Biodiversidade, Universidade Federal da Grande Dourados (UFGD), Cidade Universitária, 79.804-970 Dourados, MS, Brazil 3 Laboratório de Mirmecologia, Centro de Pesquisas do Cacau, Convênio UESC/CEPLAC, 45.651-630 Ilhéus, BA, Brazil


Introduction
The highly diverse ant fauna of the leaf litter in tropical forests has been the focus of many studies investigating the structure of ecological communities, particularly in the last twenty years [1][2][3].Approximately 60% of the entire ant species that are currently known live in leaf litter, where the ant fauna is especially diverse, taxonomically, morphologically, and ecologically [4][5][6].Studies on the biogeography and diversity of Formicidae, as well as the processes affecting their maintenance, can be of great interest for planning effective conservation of the biota at a regional scale.Such studies can also contribute to producing new ecological and taxonomic data, particularly in areas where no previous records exist for the group [7,8].
The Serra da Bodoquena, within the Chacoan subregion, borders the provinces Chaco, Cerrado, Pantanal and Parana Forest [9,10] and is a place with no previous ant records.Prado and Gibbs [11] pointed out that seasonal deciduous forests are remnants of a broader continuous distribution that was present in the past, ranging from north-eastern Brazil to Argentina in the Pleistocene dry period.This currently fragmented structure is the result of the dry, cold climate that caused the retraction of wet forests to riversides and the spread of seasonal forests [12].Deciduous forests comprise discontinuous patches along fertile valleys and basaltic and calcareous rocks in a matrix of Cerrado on the Brazilian Central Plateau.This matrix, intersected by riparian forests, acts as a connection among dry forests in north-eastern Brazil, east of Minas Gerais and São Paulo States, and forest remnants in Pantanal.The vegetation has some floristic similarities to the Amazon and the Paraguayan Chaco [13,14].

Psyche
We investigated whether the leaf-litter ant assemblages at Serra da Bodoquena could be explained by current factors, such as ant dominance and competition, or if the community structure was influenced by its geography, which differs between the northern and southern portions of Park.
Some hypotheses regarding the array of situations found in the region could be tested, assuming that the vegetation properties are also valid for the ant assemblage.Despite the biogeographic relationships of the vegetation, current and evolutionary effects of environmental formations may be reflected in the structure of the ant community.Interspecific competition is usually associated with significant divergence and with the principle of limiting similarity [15].Although niche differentiation is undoubtedly an important concept, it seems insufficient to wholly determine the high levels of local diversity commonly observed in warm climates [5,6,16,17].
Significant aggregations of assemblages have been associated with the presence of environmental filters [18].The coexistence of species would be more frequent than expected if randomly organized, because of environmental conditions that act as environment filters, allowing only a narrow spectrum of species to survive.We discuss the possibility that the structure of the leaf-litter ant community in Serra da Bodoquena could be influenced by neighbouring landscapes, as it is situated at the intersection between the Pantanal, Chaco, Cerrado, Brazilian Atlantic Forest, and Amazon Forest biomes.Alternatively, the fauna could be completely different and specific to this Seasonal Deciduous Forest.
The following issues were based on three sets of arguments; namely, (i) if the similarity between the sampling sites is high, the ant fauna of the north and south portions of the dry forest could be derived from the same historic processes and by the same selective ecological pressures and could be driven by a single colonisation process (this argument assumes that all species have an equal probability of colonisation in all sites); (ii) if the north and south portions of the forest have a distinct fauna, this suggests that the geographic basis is important to the formation of the ant assemblages once the different portions attained a distinct physiographic structure.(Therefore, the question is whether the faunistic similarity of ant communities between the northern and southern portions of Serra da Bodoquena is low, blocks are likely to be formed through different colonisation processes;) or (iii) if the samples are dissimilar among sites, a series of distinct ecological, spatial, and temporal situations may have contributed to the formation of leaf-litter ant assemblage in the region, and the surrounding environments influence the faunistic colonisation.
The goal in the present study was thus to identify associating parameters between the community structure of leaf-litter ants and the phytophysionomic matrix within the two distinct land portions in Serra da Bodoquena National Park.
In central-western Brazil, the expansion of agriculture and intensive cattle farming has led to a dramatic loss of forests.Thus, it is likely, this insect diversity has already been affected before it is has been thoroughly evaluated.Therefore, the assessment of the structure of the diversity of leaf-litter ants is the first step in understanding these patterns in this region of great biogeographic importance.

Materials and Methods
This study was carried out in a seasonal deciduous forest area in Serra da Bodoquena National Park (core coordinates: 21 • 07 16 S 56 • 28 55 W).This is the only fully protected Federal Reserve of Mato Grosso do Sul, Brazil.It harbours significant portions of seasonal deciduous and semideciduous forests, transitional areas between Cerrado and Brazilian Atlantic Forest, Cerrado and Tropical Seasonal Deciduous Forest, marshes, rocky fields, and anthropic lands with cattle farms.
The western region of the Bodoquena mountain range is formed by a mosaic of vegetational types; lowlands, including savannas steppe, arborous and gramineous Chaco, plus xeromorphic and mesoxeromorphic forests.To the east, there are many cattle farms within what used to be Cerrado vegetation, to the south, there are soybean farms and islands of semi-deciduous forest, and to the north, there are the Pantanal plains.The island of preserved dry forest areas in this region is the largest of those in the centre of South America.
The Serra da Bodoquena National Park has an area of 77,200 ha, made up of a steep plateau in the west, and comprising two distinct land portions that together cover a 300 × 50 km area.The area is preserved because it is a watershed that supplies the drainage basins of the Western region of Brazil [19].The region divides important water catchments.Salobra River, in the Northern land portion, fuels the Miranda River on Pantanal plains, and Perdido River, in the Southern land portion, fuels the Apa River.Both rivers are tributaries of the Paraguay River, although their respective waters only mix after a thousand kilometres (Figure 1).
The locality is sustained by calcareous rocks of the Corumbá group-Neoproterozoic III.It is characterised by a high rocky massif, with altitudes varying between 200 m and 770 m asl.Exposed limestone from the Tamengo formation predominates in this karstic region, where rivers are found within canyons [20,21].
The annual average temperatures of the area vary between 22 • C and 26 • C. The minimum temperature can be as low as 0 • C. The relative humidity is low and rarely reaches 80%, and rainfall varies between 1300 mm and 1700 mm a year.The hot and rainy season occurs between October and April, and the cold and dry season from May to September [22].
The survey was carried out from September 2005 to February 2008, with samples taken in the dry and wet seasons, at 10 selected sites, in eight collecting expeditions (in two expedictions has two sites) along the Bodoquena ridge (Table 1), covering the microbasin of Salobra River in the Northern land portion, including the Kadiwéu Indian Reserve, and the microbasin of Perdido River, in the Southern land portion (Figure 2).
The leaf-litter sampling ant was carried out according to the ALL protocol [2], with a few adaptations due to   the habitat being composed of limestone rock floors which made it impossible to use pitfall traps in most areas.A total of 262 leaf-litter samples of 1 m 2 were extracted using mini-Winkler sacks [23].A high diversity of microhabitats with a stratified structure was observed in the study areas.Inside the forest, there are calcareous floors and rocks with little-litter accumulation.The sampled points were chosen randomly along each transect of 40 m × 500 m but set at minimum intervals of 20 m.In each transect a minimum of 25 samples were taken.We searched for microhabitats in dry forest with sufficient leaf-litter accumulation so as to obtain approximate 2 kg.The sample exposure time of the material inside the extractor was 24 hours.The ant identifications follow Bolton [24,25], Fernández [26], Baroni-Urbani and De Andrade [27], and LaPolla et al. [28].Voucher specimens were deposited at the Museu de Biodiversidade da Universidade Federal da Grande Dourados (MuBio-UFGD, Mato Grosso do Sul, Brazil) under the Psyche The data were considered for sites independently and grouped by land portions, South (presumably under Atlantic Forest influence-Paraná subregion), and North (presumably under Pantanal influence-Amazonian subregion).
The data analysis was based on the species occurrence in samples (frequency), as accepted for quantifying social insects [29].To estimate species richness, the Chao 2 and Jackknife 2nd-order estimators were calculated using-EstimateS 7.5 [30], which are widely used in ant diversity studies [31][32][33].
Rarefaction curves showing the expected species richness versus species occurrence were used to assess the sampling efficiency for each sample area [34].From the observed species richness per site, we estimated the number of species remaining to be sampled using the second-order Jackknife estimator (incidence based).The expected number of species was plotted against number of species records on the x-axis (individual-based accumulation curve).This plot provides a measure of species diversity which is robust to sample size effects.
In order to verify if there are differences in betadiversity increasing between northern and southern land portions, the two data sets were compared following a north-south axis and were plotted by increasing ant diversity against the distance between successive sample series in the eight localities.
To analyse site similarity, we used a principal coordinate analysis (PCO) using the Bray-Curtis dissimilarity index [35,36].The similarity among the ant assemblages at the different seasons and altitude was assessed using a cluster analysis (Jaccard coefficient of similarity).The resulting similarity matrix was analysed through a sequential, agglomerative, hierarchical, and nested clustering algorithm, described by Sneath and Sokal [37].The option used was the Unweighted Pair-Group Method, arithmetic average (UPGMA).This analysis was conducted using the MVSP 3.1 software [38].
The diversity and similarity analyses were run using EstimateS 7.5 [30].Similarity and distance matrices (Euclidian distance) were compared using a Mantel's test [39].The data set was analysed using R software [40], using the Vegan package [41].The graphic design was constructed with Statistica for Windows 6 [42].The Morisita-Horn index was used too to evaluate the similarity among the localities, pairwise, because this index is not affected by the number of samples or the species richness, except for very small sampling niches [43,44].
We used EcoSim (version 7.72) to compute random matrices of species co-occurrences [45] to determine whether the mean and variance C-score among samples is larger or smaller than expected by chance.Co-occurrences based on averages that were calculated across all possible pairs of species were randomised (5,000 repetitions) within the constraint of fixed marginal totals, which is an appropriate null model for detecting patterns caused by species interactions [46].

Results
More than 20,000 ants were captured in the seasonal deciduous forest.We recorded 170 species from 45 genera and 11 subfamilies out of the 15 subfamilies of Formicidae known to occur in the Neotropical Region (Table 2).
Rarefaction curves (Figure 3) show the sampling effort in each sample site, for each land portion.Evidence of asymptotes indicates that most leaf litter ant species were sampled with the number of samples used.
The comparison of northern and southern data sets did not reveal any difference in the betadiversity of the ant communities.The two distributions follow more or less the same pattern of species substitution in relation to increasing the distance of sample sites (Figure 4).
The PCO analysis (Figure 5) shows a consistency of groups between northern and southern land portions.The analysis grouped areas of southern (I, II, and V) and also grouped the sites of northern (VI, VIII, and IX); the site X (southern) is close with this group.The samples most similar were taken in the Salobra River (III and IV) in the same season (dry).

Sites
The similarity was compared between areas in relation to season and altitude.The samples made during wet season were richer in species, and samples performed in same seasons (VI, IX, and X) appeared grouped.There was no consistent pattern between the assemblages according the altitude (r = 0.214; P = 0.55) (Figure 7).
The co-occurrence analysis indicated that species cooccurred less often than expected by chance in only two of the localities sampled P (observed ≥ expected) = 0.003, for the Balneário Perdido river, and P (observed ≥ expected) = 0.042 for the Santa Maria Perdido River, both in the southern micro-basin, suggesting that species co-occurrences are random (Figure 8).

Discussion
Our results suggest that the studied sites exhibit different arrangements of ant fauna with a replacement in abundant species across sampling units, which results in high beta diversity.Gotelli and Ellison [47] suggested that speciesenergy relationships, in addition to other factors that are Psyche 11  strongly associated with latitude, elevation, light availability, and vegetation composition, are important at regional spatial scales.Local and regional effects can mask or amplify largerscale latitudinal patterns of species richness.The differences in species diversity among the study sites could be related to Pleistocene events, such as biota interpenetration between two geologically distinct environments, and the phytophysionomic mosaic occurring in the region.According to Johnson and Ward [48], the topography of an ecotone and adjacent ecosystems is the most important factor affecting ant species richness.Here, each area could similarly allow the entrance of species coming from the surrounding matrix, affecting in turn the distribution of species in the core of the study sites.
There are several possible explanations for the inverse relationship between the observed low alpha diversity and high beta diversity, such as the particular characteristics of the forest fragments.The conservation status and potential connections between forest sites affect species persistence and colonisation in each site.These potential connections have a direct influence on the structure delimited by a buffer (considering the establishment of an influence zone for each area), which leads to an increase in species richness in interconnected forest fragments.
The absence of correlation between similarity and distance suggests that the ant species and assemblages are randomly distributed over the region.The low similarity between consecutive sampled sites suggests a strong formation effect and the influence of adjacent areas, which are evidenced by high beta diversity, with different arrangements of the ant fauna and a high turnover in species dominance across samples.
In forest fragments, ant richness depends on the diversity of local microhabitats and other factors acting at a local scale, such as physical and vegetation structure [33,49], relief, humidity, and amount of leaf litter available at the location of the food resources and nesting sites used by ants [50,51].Significant variance in species composition can be explained by notable features that shape leaf litter ant communities, namely, litter biomass, soil stoichiometry, heterogeneous distribution of nutrients, soil moisture, invasive species, ecological disturbance at a small scale, and competition dynamics [52][53][54][55][56].However, it is the ecological and historical biogeography that determines the current composition of the ant assemblages that colonize these micro-habitats (biotic and abiotic filters in the historical evolution of habitats), and also patchiness in space and time which are originated from different sources.
Our results suggest that estimated richness (Chao 2 = 231.7,Jackknife 2 = 250.4) is highly affected by the number of species that were only recorded once ("rare species" = 37).This pattern is in agreement with other studies in the Neotropical Region [5,32,57], which have found a high incidence of rare species in ant communities.
The Kadiwéu indigenous reserve, bordering the Pantanal plain, was the locality with lowest similarity conjunct dataset.The deciduous forest in this area forms an enclave of vegetation, influenced by the transition to Cerrado in this area.Transition zones are located at the boundaries between biogeographic regions and represent areas of biotic overlap, which are promoted by historical and ecological changes that allow the mixture of different biotic elements [9,10].
The pattern observed suggests that the structure of the local community is directly affected by the landscape matrix in each region and that it is in fact an ecotone of the Chaco, Cerrado, Atlantic Forest, Amazonian Forest, and Pantanal.
The co-occurrence analysis of leaf-litter ant species indicates that competitive interactions are not the only factors responsible for organising ant assemblages.There is no reason to reject the null hypothesis that the number of checkerboard pairs in the samples is random.
Our results corroborate Andersen [16], agreeing that species coexistence is determined to a significant extent by processes operating during the colonization phase, rather than just by interactions between established colonies, and that competitive outcomes are highly conditioned by

Figure 3 :Figure 4 :
Figure 3: Comparison of incidence-based ant species rarefactions curves showing expected species richness (es) versus sampling effort in ten sampling sites (samples) in the Bodoquena mountain range.

2 Figure 5 :Figure 6 :
Figure 5: PCO analysis of the sites located in the northern and southern portions of Serra da Bodoquena using the Bray-Curtis dissimilarity index (eigenvalues: 0001).

Figure 7 :
Figure 7: Dendrogram of the similarities (Jaccard coefficient, UPGMA cluster analysis) among 10 sampling sites in the Bodoquena mountain range to leaf-litter ant's fauna.Wet and dry mean seasons, and numbers mean altitude (als).

Figure 8 :
Figure 8: Co-occurrence analysis comparing 262 Winkler's samples of leaf-litter ants in Serra da Bodoquena.(Full arrow represents the expected values, and empty arrow represents the observed values for C-score measures.)

Table 1 :
Sampling localities, number of samples per site, altitude above sea level, sampling season, and geographic coordinates.

Table 2 :
Records of occurrences of the 170 species collected at 10 sampling sites in Serra da Bodoquena National Park, Mato Grosso do Sul, Brazil.

Table 4 :
Similarity analysis, richness, species shared and distance between sample sites in a pairwise comparison of the leaf-litter ant assemblages from Serra da Bodoquena, Brazil.