Ant species distribution along a topographic gradient in a " terra ‐ firme " forest reserve in Central Amazonia

In Amazonia, topographical variations in soil and forest structure within “terra-firme” ecosystems are important factors correlated with terrestrial invertebrates’ distribution. The objective of this work was to assess the effects of soil clay content and slope on ant species distribution over a 25 km2 grid covering the natural topographic continuum. Using three complementary sampling methods (sardine baits, pitfall traps and litter samples extracted in Winkler sacks), 300 subsamples of each method were taken in 30 plots distributed over a wet tropical forest in the Ducke Reserve (Manaus, AM, Brazil). An amount of 26,814 individuals from 11 subfamilies, 54 genera, 85 species and 152 morphospecies was recorded (Pheidole represented 37% of all morphospecies). The genus Eurhopalothrix was registered for the first time for the reserve. Species number was not correlated with slope or clay content, except for the species sampled from litter. However, the Principal Coordinate Analysis  indicated  that  the main  pattern  of  species  composition  from pitfall  and  litter  samples was related to clay content. Almost half of the species were found only in valleys or only on plateaus, which suggests that most of them are habitat specialists. In Central Amazonia, soil texture is usually correlated with vegetation structure and moisture content,  creating different microhabitats, which probably account  for  the observed differences in ant community structure.


Introduction
Approximately one third of the animal biomass in Amazonian "terra-firme" forests is composed by ants and termites (Fittkau & Klinge, 1973), and one hectare could contain more than eight million ant individuals and one million termite individuals (Hölldobler & Wilson, 1990).Besides the high abundance, ants are a diverse group, with more than 12,500 described species (Antbase.org, 2009).Ants have also been widely used as bioindicators in land management (Andersen et al., 2002), because of their roles as seed dispersers (Hughes Pesq. agropec. bras., Brasília, v.44, n.8, p.852-860, ago. 2009 & Westoby, 1990), herbivory controllers (Dyer & Letourneau, 1999), and ecosystems engineers (Folgarait, 1998).The use of ants or other organisms as bioindicators is more useful when supported by relationships with environment changes at spatial scales relevant to the management of biological reserves.For example, studies using ants or other terrestrial invertebrates at the Ducke Reserve, a 10,000-ha "terra-firme" forest reserve close to Manaus which has been intensively studied in the last 40 years, were normally targeted on small-scale distributions (Beck, 1968;Penny & Arias, 1982), and basic management questions about how these animals respond to environmental gradients remain unanswered.
In Central Amazonia, the topographic gradient is correlated with soil characteristics, creating a gradient with sandy soils in the valleys and clay soils on the plateaus (Chauvel et al., 1987).Soil traits, such as soil clay content, are correlated with moisture content, with poorly drained areas concentrated on valleys and relatively dry areas concentrated on plateaus (Ranzani, 1980).Humidity can influence the composition and activity of terrestrial ant species even in tropical ecosystems, with an increase in ant activity in more humid seasons and habitats (Levings, 1983;Kaspari & Weiser, 2000).Patterns of litter deposition and decomposition are also associated with topography.Litter production is lower in valleys than in plateaus.Therefore, the amount of litter on the ground is usually lower in valleys and slope areas (Luizão et al., 2004).Litter is an important resource for nesting sites and food, and, in conjunction with moisture availability, it can generate microhabitat variability, which may affect demographic patterns in ground-dwelling ant assemblages (Vasconcelos et al., 2003).
Most previous studies normally targeted small scale or used coarse differences or habitat type (i.e.forests compartments -valleys and plateaus) assuming that microhabitats are relatively homogeneous and discrete.The objective of this work was to assess the effects of soil clay content and slope on ant species distribution along a topographic gradient in a mesospatial scale (25 km 2 ) at the Ducke Reserve, Manaus, AM, Brazil.

Materials and Methods
The fieldwork was conducted between June and August 2006 at Reserva Florestal Adolpho Ducke (hereafter Ducke Reserve), located at Km 26 on the Manaus-Itacoatiara highway (3 o 00'S, 59 o 55'W), close to Manaus, AM, Brazil (Figure 1).Most of the reserve ecosystems are "terra firme" (upland) forest, covering topographic and edaphic variation typical for many areas in Central Amazonia (Ribeiro et al., 1999).The reserve terrain is rugged with an altitudinal gradient (30-140 m a.s.l.), covering 10,000 hectares.The nutrient-poor soils located on the plateaus are classified as Latossolo Amarelo (Xanthic Hapludox), which grade to sandy soils -Espodossolo Cárbico (Typic Epiaquods) (Santos, 2006); Espodossolo Ferrocárbico (Typic Endoaquods) in the valleys (Chauvel et al., 1987).The vegetation is "terra-firme" non-flooded rain forest with closed canopy.The understory is relatively open and characterized by the abundance of sessile palms, such as Astrocaryum spp.and Attalea spp.The canopy height is of 30-37 m with some emergent trees reaching up to 40-45 m.The mean temperature is 26ºC and the mean annual rainfall between 1965 and 1980 was of 1,362±400 mm, with a short dry season between July and September (Marques-Filho et al., 1981).
Ducke Reserve has 144 km of trails (a grid formed by 8-km long trails) systematically distributed along the forest and covering the topographical gradient.A regular system of 72 permanent plots was distributed over 64 km 2 of trails, and each plot was at least 1 km distant from any other.Plots were 250-m long and positioned to follow altitudinal contour lines, and thus minimize altitudinal and soil variation within each plot (Magnusson et al., 2005).All plots were at least 1 km distant from the edge of the reserve (Figure 1).
Data were collected in 30 of those permanent plots, covering an area of 25 km 2 (Figure 1).Ants were collected using pitfall traps, sardine baits, and litter samples extracted in Winkler sacks.These methods tend to complement each other and thus their combined use can better characterize diverse ant assemblages (Olson, 1991;Bestelmeyer et al., 2000).In each plot, at 25-m intervals following the center line, litter was sampled in 1-m 2 delimited areas.In the same places, one pitfall trap and half a spoon of canned sardine (used as bait) were placed, totalizing 10 subsamples for each method per plot (total of 900 subsamples).
The ants were extracted from the 1-m 2 sifted litter in Winkler sacks, through a sieve of 1-cm 2 mesh size.The sieved litter was placed into a mesh bag suspended inside a cotton bag for 48 hours.Before the set was suspended, the litter material was mixed to improve chances of the ants falling into the collecting pot.The ants and other invertebrates migrate from the suspended litter sample as a behavioral response to disturbance of their habitat (response to drying), and fall into the pot partially filled with alcohol at the bottom of the bag (Bestelmeyer et al., 2000).The pitfall traps (6.5 cm diameter; 8 cm depth; 200 mL volume), partially filled with water and detergent, were installed and left for 48 hours.The sardine baits were placed on a piece of paper (10x10 cm) on the forest floor and removed after 40 minutes.The invertebrates were preserved in a 75% alcohol solution.The ant species were identified using taxonomic keys and specific papers.Voucher specimens were deposited in the invertebrates collection of Instituto Nacional de Pesquisas da Amazônia (INPA).
Soil characteristics (clay content) were measured in a previous study by collecting six samples to a depth of 5 cm in each plot.Samples were combined for each plot and analyzed at the soil laboratory of the Departamento de Agronomia at INPA.Methodology for textural analyses is described in Claessen (1997).Slope measures were taken every 50 m perpendicularly to the main axis of the plot with clinometers, and summarized as plot average.
In order to avoid problems of site identification, caused during the transportation of the samples to the laboratory, 50 subsamples of litter sorting (Winkler sacks) of five plots at trail L-7 (Figure 1) were excluded from the analyses.All subsequent analyses were based on the results of 30 plots sampled by sardine baits and pitfall traps and of 25 plots sampled with Winkler method.
We used multiple regression analyses to explore the factor or factors accounting for the observed variability in ant species richness per plot.The predictor variables included in the model were the soil clay content and slope.Principal coordinate analysis (PCoA) was used to ordinate plots by their similarity in ant species composition.The Bray-Curtis distance measurement on site-standardized frequency (percentage of occurrence of each species in each site) was used.The soil clay content and slope were used as surrogates for the topographic gradient.An inferential test on the effects of topography on species composition (expressed as ordination scores of the first two PCoA axes) was made by multivariate analyses of covariance (MANCOVA).The predictor variables included in the model were the soil clay content and slope.All analyses were performed using the R software (R Development Core Team, 2008).

Results and Discussion
A total of 26,814 ant specimens belonging to 11 subfamilies, 54 genera, 85 species and 152 morphospecies were sampled.Fifty-eight species of Pheidole (56 of them classified as morphospecies), 14 of Crematogaster, 13 of Trachymyrmex, 11 for both Solenopsis and Camponotus and 8 for Hypoponera genera were identified.Crematogaster tenuicula was the most abundant species collected (10,907 specimens, approximately 40% of the total) and the most widespread, present in all sampled plots (30), followed by Pyramica sp. 1 and Paratrechina sp. 1, sampled in 26 plots, and Blepharidatta brasiliensis, sampled in 25 plots.As in other tropical ant communities, the number of "rare" species was high, with 72 singletons and 28 doubletons, representing 30 and 12% of the total number of species (Table 1).The genus Eurhopalothrix was registered for the first time in the Ducke Reserve.The soil clay content varied from 1.6% near streams to 87.2% on plateaus, and the slope varied from 0.6 to 26.4 degrees.As the slope was not correlated with soil clay content (Pearson r = 0.184, p = 0.329), both variables were used on multiple regressions and MANCOVA analyses to explore the factors accounting for the observed variability in ant species richness and composition.
The ant assemblage richness and composition sampled with canned sardine baits were not correlated with the environmental gradients (Tables 2 and 3).Baits are widely used to investigate the activity of ground and litter-dwelling ants in response to microclimate changes (Hölldobler & Wilson, 1990;Davidson, 1998).However, this is a selective method that attracts mainly omnivorous species that are recruited to rich food sources.This sampling method provides a momentary estimative of species richness, but is normally sensitive due to the daily variation in ant activity (Bestelmeyer et al., 2000), which may add confounding effects to sampling.
There were no significant relationships between the soil clay content or slope and the ant species richness sampled with baits (R 2 = 0.013; F 2,27 = 0.812; p = 0.454) and pitfall traps (R 2 = 0.046; F 2,27 = 1.713; p = 0.199).However, the number of species sampled with Winkler sacks was significantly and negatively related to the slope (Table 2), with the overall regression model accounting for 23% of the observed variance (R 2 = 0.228; F 2,22 = 4.559; p = 0.022).This method samples cryptic and litter-specialist species that generally have relatively small colonies, much of them occupying short-lived nest sites (Byrne, 1994).Litter-dwelling ants move more frequently than ground-dwelling species, and are probably more sensitive to soil and litter changes.However, as moisture is correlated with soil clay content, the negative relationship between number of species sampled by litter sorting and clay content may be a response to the moisture gradient.
The number of species -or, for some authors, species richness -is probably the most simple and used measurement for diversity.However, this metric gives the same weight to rare and abundant species, and it is not surprising that the results using community composition and species richness were not congruent.Although the percentage of variance captured by the two PCoA ordination axes varied between 20 and 35%, a significant relationship between clay content and ant community composition was observed for pitfall and litter sorting samples, and between slope and pitfall traps (Table 3).Previous studies conducted in Central Amazonia showed patterns similar to those described here.At the Ducke Reserve, Fagundes (2003) detected correlation of the Ponerinae ants, particularly of the Hypoponera genus, with the variation on soil texture.The influence of clay percentage (between valleys and plateaus) on ant community composition was also detected in the Biological Dynamics of Forest Fragments Project at Reserve 1501, located about 80 km north of Manaus (Vasconcelos et al., 2003).However, this pattern also seems to occur in other parts of the Amazon forest.For instance, Souza et al. (2007) detected correlation between clay content and the ant community structure in a relatively flat terrain forest located in Estação Ecológica de Caxiuanã, in eastern Amazonia.
Figures 2 and 3 show the relative frequency of species of ten subfamilies sampled along the environmental gradients.
The Myrmicinae subfamily is too numerous to build such graphical representation.As the percentage of clay content changed, the ant assemblage sampled with pitfall traps and litter sorting also changed.Almost half of the total of species sampled (105) were rare (singletons and doubletons) and occurred mostly in only one end of the clay texture gradient.Half of the singletons and doubletons (48%) was concentrated on plots with less than 5% of clay content (valleys).
As previous studies point out (Vasconcelos et al., 2003;Costa et al., 2005), it is not clear if these results originated from a direct effect of soil texture, of water drainage or from a combination of these variables.Variation in soil texture is closely related to the topography in the region of Manaus  Pesq.agropec.bras., Brasília, v.44, n.8, p.852-860, ago.2009 (Chauvel et al., 1987).At the Ducke Reserve, the highest clay percentage is concentrated on the plateaus, and the soils along stream valleys are normally hydromorphic and particularly poorly drained (Ranzani, 1980).The temporary accumulation or lateral percolation of water near small streams (Chauvel et al., 1987) probably limits the establishment and growth of some ground-nesting species (Vasconcelos et al., 2003).Disturbance probably accounts for the distribution pattern detected, with more singletons in valleys than on plateaus.In addition, other environmental variables are correlated or interact with the topography and, therefore, with the soil texture.Tree density and litter production tends to be lower in valleys than in plateaus (Luizão et al., 2004;Castilho et al., 2006), and moisture content is normally higher in valleys than in plateaus (Kaspari & Weiser, 2000).The interaction of those variables may promote a high environmental heterogeneity along the landscape, creating different microhabitats that maintain such diverse ant communities.Field experiments involving the ant community and these environmental variables are necessary to answer the relationship between the causes and effects of those variables on the reserve's ant community structure.

Conclusions
1.A distinct pattern of ant species diversity and composition is found along the gradient of soil texture and slope at the Ducke Reserve.
2. The distribution of ant species along the topographic gradient indicates that there is no environmental and biotic homogeneity within the Ducke Reserve.

Figure 1 .
Figure 1.Map of the Ducke Reserve RFAD (A).Distribution of the system of trails (fine dotted lines), the position of the 30 plots (black spots) and the field camps (black squares) (B).Shadings indicate altitudinal gradients from the lowest areas (lightest color, <40 m.a.s.l.) to the highest areas (darkest shading, >120 m.a.s.l.).The reserve limit is shown by the dashed line.

Figure 2 .
Figure 2. Distribution of poneromorph species along clay content gradient and slope in the 30 plots at the Ducke Reserve.

Figure 3 .
Figure 3. Distribution of Formicinae, Ecitoninae, Dolichoderinae, and Pseudomyrmecinae species along clay content gradient and slope in the 30 plots at the Ducke Reserve.

Table 1 .
Frequency (%) of ant taxa sampled with pitfall traps, sardine baits in 30 plots and from litter sorting (Winkler sacks) in 25 plots, at Ducke Reserve.The values obtained for frequency represent the number of observed presences divided by the number of subsamples (n = 850).

Table 2 .
Multiple regression models for the effects of clay content and slope on number of ant species sampled with sardine baits, pitfall traps and litter sorting (Winkler sacks), at the Ducke Reserve.Standardized partial regression coefficients (b) represent the relative degree of influence of a predictor variable in the model.Pesq.agropec.bras., Brasília, v.44, n.8, p.852-860, ago.2009

Table 3 .
Results of multivariate analyses of covariance (MANCOVA) models for the effects of clay content and slope on the first two ordination Principal Coordinate Analysis' axes of ant community composition, sampled with sardine baits, pitfall traps and litter sorting (Winkler sacks), at the Ducke Reserve.