Present address: Department of Ecology & Evolutionary Biology, University of Arizona, BioSciences West room 310, 1041 E. Lowell St., Tucson, AZ 85721, USA.
Canopy connectivity and the availability of diverse nesting resources affect species coexistence in arboreal ants
Article first published online: 30 NOV 2010
© 2010 The Authors. Journal compilation © 2010 British Ecological Society
Journal of Animal Ecology
Volume 80, Issue 2, pages 352–360, March 2011
How to Cite
Powell, S., Costa, A. N., Lopes, C. T. and Vasconcelos, H. L. (2011), Canopy connectivity and the availability of diverse nesting resources affect species coexistence in arboreal ants. Journal of Animal Ecology, 80: 352–360. doi: 10.1111/j.1365-2656.2010.01779.x
- Issue published online: 5 FEB 2011
- Article first published online: 30 NOV 2010
- Received 30 August 2010; accepted 23 October 2010Handling Editor: Peter Hamback
- arboreal ants;
- competitive exclusion;
- niche differentiation
1. Arboreal ants are both diverse and ecologically dominant in the tropics. Such ecologically important groups are likely to be particularly useful in ongoing empirical efforts to understand the processes that regulate species diversity and coexistence.
2. Our study addresses how access to tree-based resources and the diversity of pre-existing nesting cavities affect species diversity and coexistence in tropical arboreal ant assemblages. We focus on assemblage-level responses to these variables at local scales. We first surveyed arboreal ant diversity across three naturally occurring levels of canopy connectivity and a gradient of tree size. We then conducted whole-tree experimental manipulations of canopy connectivity and the diversity of cavity entrance sizes. All work was conducted in the Brazilian savanna or ‘cerrado’.
3. Our survey suggested that species richness was equivalent among levels of connectivity. However, there was a consistent trend of lower species density with low canopy connectivity. This was confirmed at the scale of individual trees, with low-connectivity trees having significantly fewer species across all tree sizes. Our experiment demonstrated directly that low canopy connectivity results in significantly fewer species coexisting per tree.
4. A diverse array of cavity entrance sizes did not significantly increase overall species per tree. Nevertheless, cavity diversity did significantly increase the species using new cavities on each tree, the species per tree unique to new cavities, total species using new cavities, and total cavity use. The populations of occupied cavities were consistent with newly founded colonies and new nests of established colonies from other trees. Cavity diversity thus appears to greatly affect new colony founding and colony growth.
5. These results contribute strong evidence that greater resource access and greater cavity diversity have positive effects on species coexistence in local arboreal ant assemblages. More generally, these positive effects are broadly consistent with niche differentiation promoting local species coexistence in diverse arboreal ant assemblages. The contributions of this study to the understanding of the processes of species coexistence are discussed, along with the potential of the focal system for future work on this issue.
Understanding the processes of species diversity and species coexistence remain as central, interrelated goals of ecology (Agrawal et al. 2007). Both historical processes and local interactions can influence the species that coexist at a given location, but more work is needed to understand their relative importance at different spatial scales (Agrawal et al. 2007; Ricklefs 2008). Empirical work that focuses on species traits is seen as a priority for tackling these issues, especially because theory development has outpaced empirical advancements (Agrawal et al. 2007). In these new empirical endeavours, particularly diverse, abundant, and ecologically important taxa are likely to be of great value, as they have been in classic empirical work on ecological communities (Simberloff 2004).
Ants are one of the most diverse, abundant and ecologically important animal groups in most terrestrial systems (overview in Hölldobler & Wilson 1990; Lach, Parr & Abbott 2010). Unsurprisingly, considerable research attention has been given to understanding species diversity and coexistence in ants, at global to local scales (reviewed in Dunn et al. 2010; Fisher 2010; Parr & Gibb 2010). At local scales, competition and other interactions are widely viewed as key in structuring ground-nesting and arboreal ant assemblages (reviewed in Hölldobler & Wilson 1990; Parr & Gibb 2010). Nevertheless, demonstrating the importance of particular interactions for local species coexistence remains challenging (e.g. Sanders et al. 2007; Blüthgen & Feldhaar 2010; Parr & Gibb 2010). Moreover, how the interplay between environment and species traits might influence these interactions is poorly understood (Blüthgen & Feldhaar 2010). The focus here is on understanding these issues in the diverse and ecologically dominant arboreal ant assemblages of the tropics.
Early research on arboreal ants suggested that the structure of local assemblages is determined primarily by the cumulative outcomes of pairwise competitive interactions (e.g. Room 1971, 1975; Majer 1972; Taylor 1977; Jackson 1984; Majer, Delabie & Smith 1994). However, this idea is now widely questioned (Floren & Linsenmair 2000; Ribas & Schoereder 2002; Blüthgen & Stork 2007; Sanders et al. 2007), and recent work has adopted a new focus on species interactions with resources. Some researchers have addressed the role of the distribution and supply stability of resources in mediating competition and its structuring influences (e.g. Davidson 1997; Hossaert-McKey et al. 2001; Blüthgen, Stork & Fiedler 2004; Blüthgen & Stork 2007). Others have gone further by suggesting that differences among trees (e.g. resource types and abundance) may be more important for structuring arboreal ant assemblages than pairwise species interactions (Sanders et al. 2007). Yet to date, we know little about how the interactions between resource access and species differences in resource use influence coexistence in arboreal ants (Blüthgen & Feldhaar 2010). Studies that address these interactions promise valuable new contributions towards our understanding of the forces that structure local assemblages.
Longstanding ecological theory suggests that access to resources and species differences in resource use play a central role in species coexistence (Hutchinson 1959; Macarthur & Levins 1967; Connell 1978). In general terms, when species are resource generalists, stable access to shared resources is likely to result in interspecific competitive exclusion, and thus a reduction in the number of coexisting species. Conversely, evolved interspecific differences in resource use, or ‘niche differentiation’, should reduce competition over accessible resources and allow more species to coexist (reviewed in Chesson 2000; Chase & Leibold 2003). Based on existing knowledge, it is unclear whether either of these alternative processes has a significant and detectable effect on diverse arboreal ant assemblages. On the one hand, competitively dominant and generalist species have featured prominently in explanations for the structure of arboreal ant assemblages (e.g. Room 1971, 1975; Majer 1972; Taylor 1977; Jackson 1984; Majer, Delabie & Smith 1994). Moreover, competitive exclusion does structure some simple arboreal ant assemblages (e.g. Longino 1991; Palmer et al. 2000). On the other hand, strong species differences in nesting cavity use, a key resource for arboreal ants, are important for survival in at least one diverse arboreal ant taxon (Powell 2008, 2009).
Here, we address how the interactions between resource access and resource use influence coexistence in arboreal ants. We focus on canopy connectivity as a measure of resource access and the diversity of nesting cavities, defined in terms of cavity entrance size, for resource use. Canopy connections (overlapping branches, vines, etc.) provide the primary pathways by which arboreal ants access tree-based resources, and connectivity can differ dramatically within and among systems. On the trees, most ants nest in pre-existing cavities made initially by wood-boring beetles, with the ants entering the cavities via the exit hole left by the original occupant. These cavities are typically limited resources, and the colonies of many ant species require multiple cavities to grow and reproduce (Carroll 1979; de Andrade & Baroni Urbani 1999; Philpott & Foster 2005; Debout et al. 2007; Powell 2008, 2009).
If cavity generalists are common, theory predicts that greater canopy connectivity (i.e. resource access) should increase competitive exclusion from available cavity resources, reducing the number of coexisting species. In contrast, if differentiation in cavity use is common, greater canopy connectivity should allow more species to access and coexist on available cavity resources. In addition, greater cavity diversity should further promote coexistence. We test these predictions with both observational and experimental studies conducted in the Brazilian savanna or ‘cerrado’. First, we characterize arboreal ant diversity across canopy connectivity and tree size. Variation in cerrado canopy connectivity represents a natural experiment in the effects of resource access on ant diversity. Looking across tree size, a reasonable proxy for resource availability and diversity, addresses whether any effects of resource access on coexistence vary with the resource base. Second, we confront the patterns suggested by the survey with an experiment manipulating canopy connectivity and cavity-entrance diversity at the level of whole trees. The canopy manipulation mimicked natural connectivity levels from the observational study, while cavity-entrance diversity was chosen as one particularly important property of cavity nesting resources (Powell 2008, 2009) that is likely to vary across trees.
Materials and methods
Study Site, Canopy Connectivity and Focal Tree Species
All work was conducted at the Estação Ecológica do Panga (−19·18235°, −48·39359°), a 404-ha reserve 30 km south of Uberlândia, Minas Gerais, Brazil. This site is dominated by cerrado; a unique savanna-like vegetation that covers approximately 2 000 000 km2 of central South America (Oliveira-Filho & Ratter 2002). Cerrado has a grass and shrub-dominated ground layer and a distinct tree-dominated upper woody layer or ‘canopy’, 3–8 m in height. This general composition ranges from areas with scattered trees to areas of dense woodland, often within relatively short distances, providing natural differences in canopy connectivity within the system.
Our work was conducted in three adjacent bands of cerrado with distinctly different levels of canopy connectivity. Using well-established criteria and nomenclature for describing cerrado (Oliveira-Filho & Ratter 2002), these areas were categorised recently (Cardoso et al. 2009) as cerrado ralo (grassland with low connectivity between scattered, often isolated trees), cerrado típico (30–50% crown cover, giving a medium level of canopy connectivity) and cerrado denso (60–70% crown cover, giving a high level of canopy connectivity). Here, we simply refer to these areas as having low, medium and high connectivity, respectively.
Ants were sampled only on Caryocar brasiliense, a common tree species in the cerrado biome and all of our levels of connectivity. This served as a control for possible confounding differences in ant diversity among tree species (e.g. species differences in resource abundances and types). Tree girth was used to define tree size. Diameter at base (DAB), standardized to 10 cm above soil level, was used instead of the more common diameter at breast height, because C. brasiliense trees often bifurcate below breast height. Girth is the standard dendrometric measurement for estimating the volume of wood for a focal tree species. Given our focus on wood-based nesting resources, it was therefore the most appropriate measure of tree size. Typical of cerrado tree species, C. brasiliense has a broad, domed crown (Leite et al. 2006), and all focal trees were large enough to be part of the 3 to 8m canopy layer.
Survey: Ants Across Canopy Connectivity and Tree Size
Sixty C. brasiliense trees were sampled in the diversity survey, with 20 trees for each of the three levels of connectivity. To avoid spatial clustering of sampled trees by level of connectivity, and thus pseudo-replication, trees were selected from two transects that both crossed the three bands of different connectivity. The transects were approximately 1 km long, with a minimum distance between trees in different transects of 0·5 km and a maximum distance of 1 km. Sampling was balanced across transects and levels of connectivity, with ten trees per level of connectivity per transect. This approach effectively created two spatially separated sampling areas for each level of connectivity, and an overall spatial arrangement that prevented trees with the same level of connectivity all being closer to each other than to trees with different levels of connectivity. Tree size ranged from 8·1 cm DAB to 32·6 cm DAB. Stratified sampling was used to ensure that tree size distributions were approximately equal for each sampling area, and not significantly different overall among levels of connectivity (anova, F2,57 = 0·32 P = 0·7).
Baited arboreal pitfall traps were distributed throughout the crown of each focal tree to sample the ants. Trap number was chosen to achieve approximately one trap for each of the levels of branching on the tree (8·1–14·9 cm DAB = 4 traps; 15·0–20·9 cm DAB = 6 traps; 21·0–26·4 cm DAB = 8 traps; 26·5–32·6 cm DAB = 10 traps). This gave a balanced sampling effort across trees of different sizes, with a high trap density relative to crown size. On each tree, half of the traps had a sugar-rich bait (honey diluted 1 : 7 with water) and half had a nitrogen-rich bait (urine diluted 1 : 1 with water). In preliminary trap tests, this bait pairing trapped individuals from all established resident colonies (i.e. not including newly founded ‘incipient’ colonies), and outperformed commonly used protein baits of sardine and tuna. Each trap consisted of a small plastic pot (6 cm high, 5 cm in diameter) wired to a tree limb such that the mouth was horizontal and touching the tree. Bait liquids were filled to the one-quarter mark and one drop of detergent was added to increase the killing efficacy. Day and night hand-sampling at baits was tested as a second collection method, but it only subsampled species collected in the pitfall traps. Sampling was completed between 2 and 19 April 2007. Traps were set on 20 trees on each of three sampling days. Trees from both transects, all levels of connectivity, and a range of sizes were sampled per sampling day. Traps were set in the morning and collected 48 h later, and the contents were processed in the laboratory.
Experiment: Canopy Connectivity and Cavity Diversity
Forty mid-sized trees (14·1–25·7 cm DAB) were selected in the band of cerrado with natural medium connectivity. A two by two factorial design was used, with canopy connectivity as one factor and the diversity of cavity entrances as the other, providing the following four experimental treatments: (1) high connectivity, diverse cavity entrance sizes; (2) high connectivity, uniform cavity entrances; (3) low connectivity, diverse cavity entrances; (4) low connectivity, uniform cavity entrances. Ten trees were assigned randomly to each of the four treatments. The ants were surveyed on all focal trees (methods above) before the treatments were applied. All focal trees had major connections with the surrounding canopy severed, by cutting the conspicuous connecting structures (e.g. branch tips) on non-focal trees. Trees receiving the high-connectivity treatment were then reconnected with four ropes, each making multiple connections with the surrounding canopy. As a control, trees receiving the low-connectivity treatment had four ropes tied to them that did not connect with the canopy or the ground. All trees then had 21 supplementary bamboo cavities of equal length (100 mm) and diameter (c. 10 mm) attached throughout the crown (see Powell 2009 for image of cavity design and attachment). The trees receiving the treatment of a diverse array of cavity entrances had three cavities of each of seven controlled entrance sizes (1, 2, 4, 8, 16, 32, 64 mm2), while all cavities in the uniform cavity treatment had the largest entrance size (64 mm2). Crucially, the largest entrance was of a size that does not exclude any species based on body size [e.g. a size used naturally by Cephalotes atratus (Powell 2008), one of the largest arboreal species at the reserve]. The cavity treatments thus provided either a single non-exclusive potential niche or a diverse array of potential niches. The experiment ran for 9 months (June 2007–March 2008), with monthly checks on the integrity of cavity and connectivity treatments. The trees were then resurveyed with pitfall traps, and the cavities were harvested with the traps. Trap and cavity contents were processed in the laboratory.
Analyses and Statistics
Sample-based expected species accumulation curves were calculated for each level of connectivity using the Mao Tau function. Estimated species richness was calculated using the incidence-based coverage estimator (ICE). Similarity in the observed species composition among levels of connectivity was assessed using the Classic Sørensen incidence-based index. We also identified unique species in each level of connectivity and overall. All diversity-related measures were calculated using EstimateS with default settings (Colwell 2009).
Sample-based expected species accumulation curves for Mao Tau give expected species density (not species richness, see Gotelli & Colwell 2001 for discussion). In contrast, ICE calculations, which use species incidence in sample-based collections, give an estimate of total species richness (Chazdon et al. 1998; Colwell 2009). The contrast of expected species density and estimated species richness was valuable for investigating our data. Alternative individual-based expected species accumulation curves, that do give expected species richness, would have been inappropriate for two reasons. First, the social foraging of ants produces extreme spatial clumping of individual ants, making individual-based measures inappropriate (Longino, Coddington & Colwell 2002). Second, ‘individual colonies’ cannot be used because incidence of a particular species within a sample (a tree, in our case) cannot be equated to the presence of a single colony of that species: multiple colonies of one species can coexist on one tree, and one colony can have nests that span multiple trees. Likewise, incidence-based measures of estimated richness and similarity were used because the social foraging of ants also makes abundance-based richness measures inappropriate (Longino, Coddington & Colwell 2002).
Linear least-squares regression was used to describe the relationship between tree size and the number of coexisting species per tree for each level of connectivity. Bonferroni outlier tests were used to identify statistically significant outliers in these relationships. Differences in the relationship between tree size and number of species among levels of connectivity were addressed using ancova, with Tukey HSD tests used to identify pairwise differences among levels of connectivity. For the experiment, tree size was initially assessed and rejected as a possible covariate using ancova. Consequently, the relationships between number of species per tree and the experimental treatments were analysed using anova. The data sets did not deviate significantly from anova assumptions. anova was also used to address the relationship between overall cavity use among experimental treatments. For the diverse entrance sizes treatment, the relationship between total cavity use and the number of species per entrance size was analysed using Spearman’s rank correlation test. Statistical analyses were run using a standard installation of R v.2·90, plus the car-package for Bonferroni outlier tests.
Specimen Processing and Identification
Trap and cavity contents were identified to genus and morphospecies in the laboratory. Each new morphospecies was added to a reference collection, which was used to verify subsequent determinations. The cavities provided large sequences of individuals, illustrating within-species variation and facilitating further refinement of our reference collection. Finally, the reference collection was compared with the existing ant collection at Universidade Federal de Uberlândia (built from extensive collections at the study site and nearby areas, with determinations made by different specialists). This procedure ensured high precision in the identification of collected taxa. We use ‘species’ as a synonym for our morphospecies throughout, even though final species determinations have not yet been possible for some morphospecies and others have not yet been described. Voucher specimens have been deposited in the zoological collection of the Universidade Federal de Uberlândia. They will contribute to a larger species list for the Estação Ecológica do Panga that will be published elsewhere.
Survey: Ants Across Canopy Connectivity and Tree Size
We collected a total of 48 species from the trees with low connectivity, and 55 species from the trees with both medium and high connectivity (Fig. 1a). The total for all levels of connectivity combined was 73 species from 15 genera and five subfamilies. While the 95% confidence intervals for sample-based expected species accumulation curves overlapped to some extent for all levels of connectivity, there was a clear trend of consistently lower species density for low connectivity (Fig. 1a). However, none of these species accumulation curves reached asymptotes. Concordantly, 15 or more additional species were estimated to be present for each level of connectivity, with very similar and stable estimated species richness among connectivity levels (Fig. 1b; low, mean ICE = 78, SD ± 4; medium, mean ICE = 70, SD ± 2; high, mean ICE = 73, SD ± 3). The number of species found on a single tree (i.e. ‘uniques’ at the level of incidence per tree) was high for all levels of connectivity (low = 44%; medium = 31%; high = 33%). However, 42% or more of the uniques in each connectivity were represented by multiple individuals on the trees where they occurred. Moreover, species sharing among levels of connectivity was high (classic Sørensen similarity index: low-medium = 0·80; low-high = 0·68; medium-high = 0·80), only 21% of all species were found on a single tree across levels of connectivity, and <10% of these were represented by a single specimen.
Consistent with the suggestion of lower species density in low connectivity (Fig. 1a), trees with low connectivity had fewer species per tree across the full range of tree sizes (Fig. 2). Two trees with low connectivity were identified as being significant outliers (Bonferroni outlier tests; Tree 1, studentized residual = 3·9, d.f. = 17, P < 0·03; Tree 2, studentized residual = 5·1, d.f. = 16, P < 0·002) and were excluded from all subsequent analyses of the relationship between tree size and species per tree. However, including these two points did not yield qualitatively different findings. Initial analysis indicated that tree size was a significant covariate (ancova, covariate, F1,54 = 26·7, P < 0·0001), but the interaction term was not significant (ancova, interaction term, F2,52 = 1·3, P = 0·3; indicates homogeneity of slopes) and was removed from the final ancova model. In the final analysis, both tree size and connectivity had significant effects on the number of coexisting species per tree (ancova: tree size effect, F1,54 = 26·7, P < 0·0001; connectivity effect, F2,54 = 10·2, P < 0·0002). Pairwise post hoc comparisons revealed that the relationship between tree size and species per tree for low connectivity had a significantly lower intercept than the relationship for both medium (Tukey HSD, P < 0·0009) and high connectivity (Tukey HSD, P < 0·0006). However, the relationships for medium and high connectivity were not significantly different from each other (Tukey HSD, P = 1·0).
Experiment: Canopy Connectivity and Cavity Diversity
Initial analysis of the number of species per tree, conducted before treatments were applied to the four experimental groups of trees, showed that tree size was not a significant covariate (ancova maximal model, interaction term, F3,32 = 1·5, P = 0·2). This was anticipated, given the controlled and limited size range of the trees selected for the experiment. There were also no initial significant differences in species per tree between the four experimental groups of trees (anova, treatment effect, F3,36 = 0·3, P = 0·9). Post-treatments, overall species per tree (trap and cavity contents) was significantly higher for trees with high connectivity (anova, connectivity effect, F1,36 = 6·1, P < 0·02; Fig. 3). However, there was no significant effect of cavity-entrance diversity (anova, diversity effect, F1,36 = 0·02, P = 0·9).
Despite the lack of an effect of cavity-entrance diversity on overall species per tree, cavity-use patterns did differ strongly between treatments. The number of species per tree that used the supplementary cavities was significantly higher for the treatment offering a diversity of entrance sizes (anova, diversity effect, F1,36 = 51·2, P < 0·0001, η2 = 0·5) and explained substantially more of the variance than canopy connectivity (anova, connectivity effect, F1,36 = 4·2, P < 0·05, η2 = 0·04) and the interaction between factors (anova, interaction effect, F1,36 = 5·7, P < 0·03, η2 = 0·06; Fig. 3). The highest mean number of species in the supplementary cavities was recorded for trees with high connectivity and diverse cavity entrances (Fig. 3). Cavity-entrance diversity and its interaction with canopy connectivity also had significant effects on the number of species on each tree that were unique to the supplementary cavities (i.e. overall species minus species from traps; anova, diversity effect, F1,36 = 10·5, P < 0·003, η2 = 0·2; anova, interaction effect, F1,36 = 7·5, P < 0·01, η2 = 0·1; Fig. 3). Canopy connectivity, however, did not have a significant effect on the number of unique species in the cavities (anova, connectivity effect, F1,36 = 0·6, P = 0·5). The highest prevalence of unique species in the supplementary cavities was seen on trees with high connectivity and diverse cavity entrances (8 of 10 trees). For the cavities with unique species, 24% contained either a founding queen with brood or an incipient colony (queen, brood, and first cohort of unusually small ‘nanitic’ workers). The remaining cavities with unique species contained either a small number of workers only, or a small number of workers and brood without a queen, consistent with being new nests of colonies already established on other trees.
For the total number of cavities used per experimental tree, only cavity diversity had a significant effect (anova, diversity effect, F1,36 = 26·7, P < 0·0001), with the most cavities used on trees with diverse cavity-entrance sizes. The combination of the per-tree differences for cavity use and the number of species in cavities resulted in pronounced differences in total cavity use and total number of ant species between cavity treatments, (i.e. lumping across connectivity treatments, which did not have a significant effect on cavity use). A total of 3% of all the cavities with uniform entrance size were used by seven species from three genera, compared to 26% of cavities with diverse entrance sizes used by 22 ant species from six genera. In addition, not all entrance sizes were used equally within the diverse cavity-entrance treatment (Fig. 4). There was a significant positive correlation between total cavity use and number of species across entrance sizes (Spearman’s rho = 0·98, S = 1·02, P < 0·0001). However, no entrance size was used by more than 50% of the overall total of 22 species from all entrance sizes combined (Fig. 4). Moreover, no entrance size was close to being fully exploited across all trees with the diverse entrance treatment (Fig. 4), and most or all entrance sizes remained available on all trees at the end of the experiment (open cavities remaining: entrance one on all 20 trees; entrances 2, 6, and 7 on 19 trees; entrances 4 and 5 on 17 trees; entrance 3 on 14 trees). A minimum of 11 unused cavity remained on all trees with the uniform entrance treatment. The presence of these unused cavities indicates that the degree to which any species was excluded from using cavities with a particular entrance size was minimal. Consequently, cavity-use patterns can be interpreted reliably as species cavity preferences.
Here, we have addressed how resources access, determined by canopy connectivity, and the diversity of cavity nesting resources affect local species coexistence in arboreal ants. We used both observational and experimental approaches. From our observational data, estimated species richness and species composition was similar across all three levels of canopy connectivity. However, species accumulation curves suggested a consistent trend of lower species density with lower connectivity. This was confirmed at the scale of individual trees, with low-connectivity trees having significantly fewer species across all tree sizes. Our experiment tested and confirmed the relationship between connectivity and species per tree, demonstrating directly that high-connectivity trees have significantly more species than low-connectivity trees. In contrast, the availability of diverse cavity entrances, which may be one key resource that increases with tree size, did not significantly increase overall species per tree in our experiment. Nevertheless, cavity diversity did significantly increase the species using new cavities on each tree, and the species per tree unique to new cavities. Cavity diversity also significantly increased the total species using new cavities and total cavity use across trees. To our knowledge, these results represent the first experimental evidence that greater resource access and cavity diversity have positive effects on species coexistence in arboreal ant assemblages. More generally, these positive effects are consistent with niche differentiation promoting species coexistence at a local scale. These findings, their caveats, and broader implications are discussed in the following paragraphs.
The limited differences in estimated total species richness and composition among levels of connectivity suggest that connectivity does not have strong filtering effects on the regional pool of arboreal species. However, we must acknowledge the caveat that we used a single tree species. This controlled for the potentially confounding influence of resource differences among tree species, but data from more species may yield stronger across-connectivity differences in species richness. Nevertheless, our focal tree species, C. brasiliense, is distributed throughout the Cerrado biome and is often abundant within and across different levels of connectivity (Gribel & Hay 1993; Oliveira 1997; Leite et al. 2006), as it was at our site. Moreover, cerrado with different connectivity can be numerically dominated by the same suite of common tree species (Ratter, Bridgewater & Ribeiro 2006), and cerrado trees typically do not have specific ant mutualists (Oliveira, Freitas & Del-Claro 2002). Broad similarities in ant species richness and composition among adjacent areas that differ in connectivity may therefore be relatively robust to more tree species.
Despite the apparently common species pool among connectivity levels, low connectivity was associated with lower species density, both overall (Fig. 1a) and at the level of individual trees (Fig. 2). Moreover, our experiment tested and confirmed that greater canopy connectivity can drive higher species per tree. Indeed, the difference in mean species per tree between the experimental high- and low-connectivity treatments (1·7 species, lumping non-significant cavity treatments) was equivalent to the difference between natural high and low connectivity in the survey (difference in intercepts of 1·5 species; Fig. 2). An increase in species number with increasing tree density (i.e. canopy connectivity) was shown in a previous cerrado survey (Ribas et al. 2003). However, tree density was treated as a proxy of resource availability, and assumed changes in resource availability were offered as an explanation for more species in denser areas. Resource availability is undoubtedly important for species coexistence in a given area, as supported by our tree size relationship. Nevertheless, our tree-level focus and experimental test provide the new insight that greater canopy connectivity, found in denser areas, drives higher species coexistence irrespective of tree density. These positive relationships between connectivity, tree size, and species per tree are broadly consistent with widespread niche differentiation promoting coexistence in arboreal ants. If overlapping resource use and the related process of competitive exclusion were common, greater connectivity (i.e. resource access) should have reduced species coexistence (reviewed in Chesson 2000).
In contrast to our findings, competitive exclusion does seem to play a key role in species coexistence in some simpler ant assemblages (Palmer et al. 2000). Small myrmecophytic Acacia drepanolobium trees can contain founding queens of a number of species. But they become dominated by one of four ant species as they grow, and the most dominant species ultimately controls many of the largest trees. A similar decrease in species coexistence with increasing tree size has been reported for the myrmecophytic tree Cecropia obtusifolia (Longino 1991). In our more diverse system, the number of coexisting species increased with tree size in all three levels of canopy connectivity. Moreover, an increase in canopy connectivity, which should allow competitive dominants greater opportunity to exclude other taxa, further increased the number of coexisting species per tree. This suggests an interesting switch in the main structuring processes between systems: competitive exclusion may be the main process in simple ant assemblages of myrmecophytic trees, whereas niche differentiation may promote coexistence in diverse tropical assemblages.
Both our survey and experiment support a consistent, positive relationship between the level of connectivity and species per tree, but how does this happen mechanistically? Our cavity-use patterns suggest that an increase in resident species played a central role. Specifically, the number of species in, and unique to, the cavities was highest when they had diverse entrances and were on high-connectivity trees (Fig. 3). Connectivity thus increases residency. The species unique to new cavities could have been artefacts of our pitfall-trapping method. However, the contents of these cavities were either consistent with being newly founded colonies or new nests of established colonies on other trees. Both nest types have zero or weak forager activity that is probably invisible to pitfall-trap sampling. It is therefore more likely that our traps did effectively sample established resident colonies, and species unique to the cavities were new tree colonizations. Non-resident foragers from other trees could have contributed also to the consistently higher number of species on high-connectivity focal tree. However, their presence would still be consistent with niche differentiation generally promoting coexistence. If competitive exclusion were a dominant process, significant numbers of non-resident foragers would be prevented by resident dominant species (e.g. Adams 1994; Palmer 2004).
While cavity diversity did not have a significant effect on the overall number of coexisting species per tree, it did drive very different cavity-use patterns. First, the number of species per tree that used, and were unique to, supplementary cavities was significantly higher when the cavities had diverse entrance sizes (Fig. 3). Second, there was an eightfold difference in total cavity use between the diverse and uniform treatments (26% vs. 3%), and a threefold difference in total species (22 species vs. 7 species). Crucially, cavities remained available on all trees at the end of the experiment, allowing cavity-use patterns to be interpreted reliably as evolved preferences (i.e. fundamental niches). The differences between treatments therefore indicate that many species use only specific entrance sizes (Fig. 4), not any entrance large enough to allow them to enter. Put in more general terms, our data provide direct assemblage-level evidence that entrance size is a key niche axis along which arboreal ants have differentiated. Why, then, did cavity diversity not significantly increase coexistence? Potentially, a longer period than our 9-month experiment may be required for residency increases to accumulate to a statistically significant level. Future work needs to test this possibility. Species-level evidence of niche differentiation should also be a focus. This will be effort-intensive, but should identify important interspecific niche differences (e.g. Powell 2008) crucial for understanding of how such differences promote coexistence.
A number of previous experimental studies have demonstrated that a range of supplementary nesting resources will be used rapidly by a diversity of arboreal ant species (e.g. Carroll 1979; Philpott & Foster 2005; Armbrecht, Perfecto & Silverman 2006; Davidson, Arias & Mann 2006). These studies thus provided the important insight that cavities are typically a limited resource for arboreal ants, and some suggested that cavity availability may limit diversity in some systems (Philpott & Foster 2005; Armbrecht, Perfecto & Silverman 2006). Our experiment further confirms cavity limitation. More importantly, however, it contributes the new understanding that cavity diversity is a crucial component of cavity limitation: even small differences in entrance size affect the number of species that will use available cavities (Fig. 4). We further show that the availability of diverse cavities, in combination with species differences in cavity use, has important implications for colony founding and growth.
Interestingly, twigs matched in dimensions but derived from a range of tree species have been shown to significantly increase litter-nesting ant diversity, compared to twigs from one tree species (Armbrecht, Perfecto & Vandermeer 2004). This suggests a common importance of cavity diversity in species coexistence in otherwise strongly contrasting realms and ant assemblages. It is not known why cavities from a diversity of tree species attract more species of leaf litter ants (Armbrecht, Perfecto & Vandermeer 2004). However, there is a clear functional explanation for why arboreal ants have differentiated along the niche axis of entrance size, and thus why it affects assemblage structure. In the genus Cephalotes, the relationship between ant size and entrance size can significantly impact nest survival and colony fitness (Powell 2009). The functional relationship between hole and ant size can, therefore, be a strong source of selection for more specific entrance size use and, potentially, related defensive traits. Indeed, this does appear to be the case within Cephalotes (Powell 2008). This evolutionary explanation for differentiated entrance size use will be valuable for designing future experiments on coexistence in this system.
Niche differences are still widely seen as crucial for explaining species coexistence and diversity (reviewed in Chase & Leibold 2003), even after challenges by the neutral theory of biodiversity (Hubbell 2001), and new support continues to accumulate (e.g. Levine & HilleRisLambers 2009; Adler, Ellner & Levine 2010). Despite the prevalence of the niche concept in the coexistence literature, research on coexistence in arboreal ants has focused strongly on spatial structuring from the competitive process itself (e.g. Room 1971; Majer 1972; Adams 1994; Floren & Linsenmair 2000). Little attention has been given to how evolved interspecific differences in resource use might interact with the environment to influence coexistence and diversity (Blüthgen & Feldhaar 2010). Our data give the first suggestion that cavity nesting resource may provide well-defined and finely graded axes for niche differentiation within the canopy. Our focus has been on the axis of cavity entrance size, but other cavity properties are also likely to be important. The unusually good experimental tractability that is possible in cerrado, as demonstrated here, is likely to be valuable for advancing understanding of niche differentiation and coexistence in arboreal ants. The tractability of this study system may also hold considerable potential for general experimental tests of the role of niche differentiation in species coexistence.
We thank the members of the Vasconcelos Lab at the Universidade Federal de Uberlândia for their invaluable help in the field. We also thank the members of the Marquis Lab at the University of Missouri, St. Louis, and three anonymous reviews for valuable comments on earlier versions of this manuscript. This work was funded primarily by a research grant from the Committee for Research and Exploration of the National Geographic Society. S. Powell was also funded by an 1851 Research Fellowship from the Royal Commission for the Exhibition of 1851, UK and H. Vasconcelos by the Brazilian Council for Research and Scientific Development (CNPq grant 301255/2007-5).
- 1994) Territory defense by the ant Azteca trigona: maintenance of an arboreal ant mosaic. Oecologia, 97, 202–208. (
- 2010) Coexistence of perennial plants: an embarrassment of niches. Ecology Letters, 13, 1019–1029. , & (
- 2007) Filling key gaps in population and community ecology. Frontiers in Ecology and the Environment, 5, 145–152. , , , , , , , , & (
- 1999) Diversity and adaptation in the ant genus Cephalotes, past and present. Stuttgarter Beiträge zur Naturkunde Serie B (Geologie und Paläontologie), 271, 1–889. & (
- 2006) Limitation of nesting resources for ants in Colombian forests and coffee plantations. Ecological Entomology, 31, 403–410. , & (
- 2004) Enigmatic biodiversity correlations: ant diversity responds to diverse resources. Science, 304, 284–286. , & (
- 2010) Food and shelter: how resources influence ant ecology. Ant Ecology (eds L.Lach, C.L.Parr & K.L.Abbott), pp. 115–136. Oxford University Press, Oxford. & (
- 2007) Ant mosaics in a tropical rainforest in Australia and elsewhere: a critical review. Austral Ecology, 32, 93–104. & (
- 2004) Bottom-up control and co-occurrence in complex communities: honeydew and nectar determine a rainforest ant mosaic. Oikos, 106, 344. , & (
- 2009) Mudanças fitofisionômicas no Cerrado: 18 anos de sucesão ecológica na Estação Ecológica do Panga, Uberlândia - MG. Caminhos de Geografia, 10, 254–268. , , & (
- 1979) A comparative study of two ant faunas: the stem-nesting ant communities of Liberia, West Africa and Costa Rica, Central America. American Naturalist, 113, 551. (
- 2003) Ecological Niches: Linking Classical and Contemporary Approaches. University of Chicago Press, Chicago. & (
- 1998) Statistical methods for estimating species richness of woody regeneration in primary and secondary rain forests of NE Costa Rica. Forest Biodiversity Research, Monitoring and Modeling: Conceptual Background and Old World Case Studies (eds F.Dallmeier & J.A.Comiskey), pp. 285–309, Parthenon Publishing, Paris. , , & (
- 2000) Mechanisms of maintenance of species diversity. Annual Review of Ecology and Systematics, 31, 343–366. (
- 2009). EstimateS: Statistical Estimation of Species Richness and Shared Species from Samples. Version 8.2. User’s Guide and application published at: http://purl.oclc.org/estimates. (
- 1978) Diversity in tropical rain forests and coral reefs. Science, 199, 1302–1310. (
- 1997) The role of resource imbalances in the evolutionary ecology of tropical arboreal ants. Biological Journal of the Linnean Society, 61, 153–181. (
- 2006) An experimental study of bamboo ants in western Amazonia. Insectes Sociaux, 53, 108–114. , & (
- 2007) Polydomy in ants: what we know, what we think we know, and what remains to be done. Biological Journal of the Linnean Society, 90, 319–348. , , & (
- 2010) Geographic gradients. Ant Ecology (eds L.Lach, C.L.Parr & K.L.Abbott), pp. 38–58. Oxford University Press, Oxford. , , & (
- 2010) Biogeography. Ant Ecology (eds L.Lach, C.L.Parr & K.L.Abbott), pp. 18–37. Oxford University Press, Oxford. (
- 2000) Do ant mosaics exist in pristine lowland rain forests? Oecologia, 123, 129–137. & (
- 2001) Quantifying biodiversity: procedures and pitfalls in the measurement and comparison of species richness. Ecology Letters, 4, 379–391. & (
- 1993) Pollination ecology of Caryocar brasiliense (Caryocaraceae) in central Brazil cerrado vegetation. Journal of Tropical Ecology, 9, 199–211. & (
- 1990) The Ants. Harvard University Press, Cambridge, Mass. & (
- 2001) Differential associations with ants of three co-occurring extrafloral nectary-bearing plants. Ecoscience, 8, 325–335. , , , , & (
- 2001) The Unified Neutral Theory of Biodiversity and Biogeography. Princeton University Press, Princeton. (
- 1959) Homage to Santa Rosalia or why are there so many kinds of animals? The American Naturalist, 93, 145. (
- 1984) Ant distribution patterns in a Cameroonian cocoa plantation: investigation of the ant mosaic hypothesis. Oecologia, 62, 318–324. (
- 2010) Ant Ecology. Oxford University Press, Oxford. , & (
- 2006) Phenology of Caryocar brasiliense in the Brazilian cerrado region. Forest Ecology and Management, 236, 286–294. , , , & (
- 2009) The importance of niches for the maintenance of species diversity. Nature, 461, 254–257. & (
- 1991) Azteca ants in Cecropia trees: taxonomy, colony structure, and behaviour. Ant-Plant Interactions (eds C.R.Huxley & D.F.Cutler), pp. 271–288. Oxford University Press, Oxford. (
- 2002) The ant fauna of a tropical rainforest: estimating species richness three different ways. Ecology, 83, 689–702. , & (
- 1967) The limiting similarity, convergence, and divergence of coexisting species. The American Naturalist, 101, 377. & (
- 1972) The ant mosaic in Ghana cocoa farms. Bulletin of Entomological Research, 62, 151–160. (
- 1994) Arboreal ant community patterns in Brazilian cocoa farms. Biotropica, 26, 73–83. , & (
- 1997) The ecological function of extrafloral nectaries: herbivore deterrence by visiting ants and reproductive output in Caryocar brasiliense (Caryocaraceae). Functional Ecology, 11, 323–330. (
- 2002) Ant foraging on plant foliage: contrasting effects on the behavioral ecology of insect herbivores. The Cerrados of Brazil: Ecology and Natural History of a Neotropical Savanna (eds P.S.Oliveira & R.J.Marquis), pp. 287–305. Columbia University Press, New York. , & (
- 2002) Vegetation physiognomies and woody flora of the cerrado biome. The Cerrados of Brazil: Ecology and Natural History of a Neotropical Savanna (eds P.S.Oliveira & R.J.Marquis), pp. 91–120. Columbia University Press, New York. & (
- 2004) Wars of attrition: colony size determines competitive outcomes in a guild of African acacia-ants. Animal Behaviour, 68, 993–1004. (
- 2000) Short-term dynamics of an acacia ant community in Laikipia, Kenya. Oecologia, 123, 425–435. , , & (
- 2010) Competition and the role of dominant ants. Ant Ecology (eds L.Lach, C.L.Parr & K.L.Abbott), pp. 77–96. Oxford University Press, Oxford. & (
- 2005) Nest-site limitation in coffee agroecosystems: artificial nests maintain diversity of arboreal ants. Ecological Applications, 15, 1478–1485. & (
- 2008) Ecological specialization and the evolution of a specialized caste in Cephalotes ants. Functional Ecology, 22, 902–911. (
- 2009) How ecology shapes caste evolution: linking resource use, morphology, performance and fitness in a superorganism. Journal of Evolutionary Biology, 22, 1004–1013. (
- 2006) Biodiversity patterns of the woody vegetation of the Brazilian Cerrado. Neotropical Savannas and Seasonally Dry Forests–Plant Diversity, Biogeography, and Conservation (eds R.T.Pennington, G.P.Lewis & J.A.Ratter), pp. 31–66. Columbia University Press, New York. , & (
- 2002) Are all ant mosaics caused by competition? Oecologia, 131, 606–611. & (
- 2003) Tree heterogeneity, resource availability, and larger scale processes regulating arboreal ant species richness. Austral Ecology, 28, 305–314. , , & (
- 2008) Disintegration of the ecological community. The American Naturalist, 172, 741–750. (
- 1971) The relative distributions of ant species in Ghana’s cocoa farms. Journal of Animal Ecology, 40, 735–751. (
- 1975) Relative distributions of ant species in cocoa plantations in Papua New Guinea. Journal of Applied Ecology, 12, 47–61. (
- 2007) An ant mosaic revisited: dominant ant species disassemble arboreal ant communities but co-occur randomly. Biotropica, 39, 422–427. , , , & (
- 2004) Community ecology: is it time to move on? The American Naturalist, 163, 787–799. (
- 1977) The ant mosaic on cocoa and other tree crops in Western Nigeria. Ecological Entomology, 2, 245–255. (