1. Trophic structure within a guild can be influenced by factors such as resource availability and competition. While ants occupy a wide range of positions in food webs, and ant community composition changes with habitat, it is not well understood if ant genera tend to maintain their position in the trophic structure, or if trophic position varies across habitats.
2. We used ratios of stable isotopes of carbon and nitrogen to test for differences in the trophic structure and position of assemblages of ants among habitat types. We tested for differences between assemblages in replicate sites of the land use categories: (i) pastures with old large trees; (ii) recently revegetated pastures with small young trees; and (iii) remnant woodlands. Known insect herbivores and predatory spiders provided baselines for herbivorous and predaceous arthropods. Soil samples were used to correct for the base level of isotopic enrichment at each site.
3. We found no significant interactions between land use and ant genus for isotope enrichment, indicating that trophic structure is conserved across land use categories. The fixed relative positions of genera in the trophic structure might be re-enforced by competition or some other factor. However, the entire ant assemblage had significantly lower δ15N values in revegetated sites, suggesting that ants feed lower down in the food chain i.e. they are more ‘herbivorous’ in revegetated sites. This may be a result of the high availability of plant sugars, honeydew and herbivorous arthropod prey.
4. Surprisingly, ants in remnants and pastures with trees displayed similar isotopic compositions. Interactions within ant assemblages are thus likely to be resilient to changes in land use, but ant diets in early successional habitats may reflect the simplicity of communities, which may have comparatively lower rates of saprophagy and predation.
The realised trophic niche of organisms may be limited by a range of factors, including competition and resource availability (Brown & Wilson 1956; Hutchinson 1957). Trophic niche partitioning, driven by competition, is thought to be common within guilds of similar species (e.g. Hutchinson 1959; Pyke 1982; Luiselli 2008), with trophic shifts commonly reported in response to the introduction of competitors (Woodward & Hildrew 2001; Eastwood, Donahue & Fowler 2007). However, trophic niche partitioning within a guild may be historical, making it difficult to attribute causation (Connell 1980). Shifts in an organism’s trophic position can also occur in response to habitat change (Sierszen, Peterson & Scharold 2006; Stenroth et al. 2008), which can alter resource availability. Guilds of trophically similar species may thus respond to habitat change by shifting trophic positions as a result of factors such as competition and changes in resource availability.
Improved access of ecologists to stable isotope technology (Hobson & Wassenaar 1999; Dawson et al. 2002; Martínez del Rio & Wolf 2005; Newsome et al. 2007) has allowed closer investigation of the trophic links in assemblages of omnivores. The biologically important elements carbon and nitrogen have been particularly informative because enrichment of the heavier isotopes of nitrogen and carbon occurs with each trophic level (DeNiro & Epstein 1978, 1981). However, nitrogen enrichment occurs at a much greater rate than carbon enrichment, making nitrogen a superior indicator of trophic level, while carbon isotopes are used mainly to indicate whether food chains are based predominantly on C3 or C4 plants (Ponsard & Arditi 2000; Hood-Nowotny & Knols 2007). Although isotope ratios within organisms vary between sites, baseline measures of plants, litter or soil make inter-site comparisons possible (Ponsard & Arditi 2000; Nakagawa, Hyodo & Nakashizuka 2007). In this way, isotopes can provide a powerful tool for determining how food webs change in response to land management.
Ant assemblages provide an ideal model system in which to test how land use affects the trophic structure of omnivores. Ants are archetypal omnivores, consuming animal tissue, plant exudates, seeds and honeydew produced by herbivorous insects (Carroll & Janzen 1973; Hunter 2009). Competition is considered to be the key factor determining the structure of ant assemblages, with many species thought to compete for a shared pool of resources (Hölldobler & Wilson 1990). However, recent studies suggest that habitat, in terms of structure and resource availability, may regulate competition amongst ant species (Gibb 2005; Sarty, Abbott & Lester 2006; Wilkinson & Feener 2007; Parr & Gibb 2010; Gibb & Parr 2010). Thus, changes in habitat may lead to changes in partitioning of diets that reverberate through ant communities.
Previous studies of the trophic structure of ant assemblages have focused predominantly on assemblages within a single site (e.g. Davidson 2005; Ottonetti et al. 2008), so have not provided insight into how habitat differences might affect trophic interactions in ant assemblages. An interesting exception is Blüthgen, Gebauer & Fiedler (2003) who found higher δ15N values of green tree ants, Oecophylla smaragdina, in a regenerating forest than in mature forests. They hypothesised that a high abundance of honeydew and nectar in mature forests may make ants appear more herbivorous. Fiedler et al. (2007) showed that isotope analysis is applicable across extended ecological and geographical gradients. In this study, we extend this landscape approach to compare isotopic signatures of ant assemblages across replicate sites under different management. We use a between-site correction based on soil isotopic signatures. Specifically, we address the following questions:
1 Do ant genera alter their relative trophic positions in response to land use?
2 Does the ant assemblage undergo a shift in trophic position in response to land use?
Materials and methods
The study was conducted in south-eastern Australia, where the majority of native vegetation has been cleared for grazing, crops or development (Hobbs 2005). Sites were within 120 km of the town of Yass (34º 51′S, 148º 55′E) at altitudes ranging from 450 to 720 m above sea level. The region experiences a relatively dry continental climate with warm to hot summers and cold winters (Bureau of Meteorology, 2008: http://www.bom.gov.au).
We selected five study sites belonging to each of the following three categories: (i) pastures with trees (‘pastures’), (ii) revegetated pastures planted with tube stock (i.e. seedlings reared in a nursery) between 1998 and 2001 (‘revegetation’), and (iii) remnant areas (‘remnants’) that had never been cleared and had been protected from heavy livestock grazing for ≥10 years (Fig. 1). Isotope studies were performed at four replicate sites in each category. One of the remnant sites experienced occasional light grazing from cattle, and evidence of kangaroos and rabbits was present in all sites. Revegetated sites were planted predominantly with endemic species of Eucalyptus and Acacia, but species of Grevillea, Melaleuca and Allocasuarina were also present at some sites. Seedlings had typically been planted with inter-plant spacing between 3 and 5 m. Remnant sites were Eucalyptus-dominated woodlands, but the dominant species differed between sites, including mixes of E. rossii, E. albens, E. blakelyi, E. polyanthemos, E. melliodora and E. macrorhyncha. Pasture sites were also centred on Eucalyptus trees from this set of species. Species of Acacia were present in three of the four remnant sites. We selected remnants of a similar size to the revegetation patches, but pastures were unavoidably part of a continuous matrix of agricultural land and therefore did not have distinct site boundaries. Sites of different habitat categories were non-adjacent and spatially interspersed. Sites were located on a range of landforms, but landform was distributed evenly among land use types.
Sampling for isotope analyses
In March 2008, we collected ants, insect herbivores and spiders from each of the twelve study sites, in order to determine the relative trophic positions of different ant genera. All specimens were collected between 09.00 and 18.00 h. Plant hoppers and grasshoppers were used to provide a baseline for the isotopic composition of herbivores at each of the study sites. Spiders provided a baseline for predatory arthropods. We collected at least three individual herbivores and predators at most sites. Two or fewer herbivores were collected at four sites due to low grasshopper activity or difficulties in accessing the canopy for hemipterans.
Our goal was to collect a representative sample of the diurnally active ants in each site, but species composition differed considerably between sites. We used active searching and collected live animals to ensure that samples were uncontaminated by preservatives. We visited each site for a minimum of 4 h and searched by turning over logs and stones and following ants carrying food back to their nests. Up to 30 workers of between eight and seventeen ant species were collected by hand or aspirator at each site. For very small ant species, such as Tapinoma spp., 30 individuals were required for a single sample. Smaller numbers were collected for larger ant species, e.g. Myrmecia spp., but samples always consisted of at least two individuals. We calculated species means for each species collected at each site based on a sample size of 2–3 (each sample contained 2–30 ants, depending on their size) in most cases, but only 1 in a few cases. Ants were identified to genus using Shattuck (1999) and then to morphospecies. All arthropod samples were stored on dry ice, then transferred to a −20 °C freezer, where they were stored for several months before analysis. Abdomens were removed from all specimens before analysis to avoid complications related to gut contents (Blüthgen, Gebauer & Fiedler 2003; Tillberg et al. 2006; but see Davidson 2005).
Soil samples were collected in April 2008 at all twelve sites to use as a baseline for isotope values for the sites. Plant or litter samples are commonly used for baseline isotope values (e.g. Ponsard & Arditi 2000; Nakagawa, Hyodo & Nakashizuka 2007), but no single plant species was common to all sites, and not all sites had sufficient litter. In addition, no herbivore or carnivore species were common to all sites, so we did not feel confident that species composition would not affect the baseline if it was based on these groups. Soil thus presented the most consistent sample to use as our baseline. Grass, leaf litter and the top centimetre of soil were removed and a 10 cm deep plug of soil (diameter 43 mm) was collected using an auger. Three samples were collected from a transect at each site, with 6 m spacing between samples. Soils were stored in a −20 °C freezer, and dried at 40 °C for 48 h prior to analysis.
Stable isotope analysis was used to measure the ratio of the heavy to light isotopes of C and N. The analysis of samples was performed using a Sercon Hydra 20–20 isotope ratio mass spectrometer (made in Crewe, UK) with an ANCA (automated nitrogen and carbon analyser) preparation system at CSIRO Black Mountain Laboratories (Canberra, Australia). Samples were dried in an oven at 60 °C, then approximately 1·0 mg of each sample was weighed into a tin capsule. Test and reference samples, were used to correct for any drift or carry-over in the instrument. The references were calibrated for total C% and N%. 15N/14N was measured relative to atmospheric N, while 13C/12C was measured relative to Vienna Pee Dee Belemnite. Sample ratios were compared to this element-specific standard and reported as δX, where δX = [(Rsample/Rstandard)−1)] × 1 000. Rsample and Rstandard are the ratio of heavy to light isotopes for the sample and standard, respectively. We report δ values as ‘per mil’ deviation from standards for the 13C/12C and 15N/14N isotopic ratios of soil, herbivores, ants and predators.
Our initial statistical analyses tested if there were inherent soil differences between sites and whether it was a valid assumption that isotopic enrichment was additive. In order to determine whether different land use types differed in soil δ15N and δ13C levels, we used one way anova, with land use type as a fixed factor and no random effect. To examine the relationships between soil δ15N and δ13C values and organism δ15N and δ13C values, we calculated the site means for soil, ants, non-ant predators and the non-ant herbivores at each site. We tested for a relationship between organisms and soils using linear regression, with sites as replicates. Thus, each model had n = 12, except when non-ant herbivores were examined, in which case n = 11 because no herbivores were collected at one site. We also tested whether slopes differed between these different trophic groups using simple linear regression with groups.
The next set of analyses considered differences between ant genera and ant trophic status at different land use types using data corrected for soil isotope values. Analyses were performed at genus level because species composition differed between sites. Isotope values were considered to be additive (e.g. Ponsard & Arditi 2000; Nakagawa, Hyodo & Nakashizuka 2007), so we subtracted the soil isotope values from the organism values to obtain a corrected value for each species at each site. We asked if ant genera differed in δ15N and δ13C values, and whether different types of sites differed in the mean δ15N for ant genera. To conduct this analysis, we first reduced the data set so that it only included ant genera collected in at least 7 of the 12 sites (Camponotus, Iridomyrmex, Melophorus, Monomorium, Myrmecia, Pheidole, Rhytidoponera). We examined δ15N for ants, corrected for the soil δ15N in that site. We grouped ants by genus and calculated the mean (across species and sites) for soil corrected δ15N, and compared these means to the mean (across species and sites) soil corrected δ15N for the non-ant herbivores and the non-ant predators. We used a REML model in preference to anova because data were unbalanced, with not all genera detected in all sites. The model included genus (7 levels, fixed effect) and land use type (3 levels, fixed effect), the genus by type interaction, and site (i.e. 12 unique sites) as a random effect. Residuals from this model were normal, indicating no need for data transformation. Exactly the same approach was applied for analysis of genus and type effects on δ13C.
We also examined whether tissue-dependent variation had affected our results by testing for correlations between δ15N (both corrected and uncorrected data) and C : N ratio and % Nitrogen for the 44 species used in this study. All statistical analyses were conducted in GenStat (Lawes Agricultural Trust 2006).
Soil isotope values
Soil isotope values were significantly affected by land use type (F(2,9) = 32·3, P < 0·001), with remnant sites having lower δ15N than pasture or revegetation sites (Tukey’s pairwise post hoc comparison, P < 0·05) (Appendix S2, Supporting Information). In contrast, there was no effect of type on soil δ13C (grand mean = −25·7 ‰, Appendix S1, Supporting Information). Differences in soil δ15N between undisturbed and agriculturally managed sites are well known (Evans 2007). They are thought to result from either the effects of nitrogen removal without replacement through grazing or by stimulating nitrogen cycling through the continual input of labile δ15N-rich urine and faeces (Frank & Evans 1997; Evans & Belnap 1999; Frank, Evans & Tracy 2004).
Relationship between soil and arthropod isotope values
Soil δ15N value was a significant predictor of organismal δ15N value for ants (90·5% explained, F(11) = 105·5, P < 0·001), for non-ant herbivores (42·4% explained, F(10) = 8·4, P = 0·018), and the non-ant predators (88·2% explained, F(11) = 83·0, P < 0·001) (Fig. 2). Notably, r2 was much higher for higher trophic levels. This is probably a result of the diet of species in higher trophic levels sampling more prey items, thus decreasing variation in isotopic signatures (see Bump et al. 2007). The δ15N values for ants and non-ant predators were similar but with the predators fractionally higher on average, whereas δ15N values for herbivores were clearly lower. The soil vs. organism relationships appeared to be linear. Although the equations differed in the y intercept, slopes did not differ significantly between organism types (simple linear regression with groups testing for difference from the slope for ants: herbivores P =0·178, predators P =0·099). The strong relationship between soil and organism δ15N value suggests that, all else being equal, this enrichment would pass through the food chain such that organisms in pasture sites would have greater δ15N values than those in remnants or revegetated sites. Given that soil δ15N differed between land use types and was correlated with organism δ15N (Fig. 2), it was important to correct for the site soil differences before comparing trophic positions of organisms in different habitats across multiple sites. Because the soil vs. organism relationships were approximately linear, we corrected organismal δ15N by subtracting the site soil δ15N.
The relationship between soil and organismal δ13C was generally weaker than for nitrogen, with at best marginal significance (ants 17·1% explained, F(11) = 3·3, P = 0·100; herbivores 5·3% explained, F(10) = 1·6, P = 0·243; non ant predators 21·1% explained, F(11) = 3·9, P = 0·075). Nevertheless, we corrected organismal δ13C values by subtraction, in the same way as for δ15N, because we felt the marginally significant regressions probably reflected a real underlying association and that correcting in this way was in fact more conservative than assuming no relationship.
Effects of land use and ant genus on isotope values
We grouped ants by genus and calculated the mean (across species and sites) for soil corrected δ15N, and compared these means to the mean (across species and sites) soil corrected δ15N for the non-ant herbivores and the non-ant predators. Non-corrected data is supplied in Appendix S1 and S2, Supporting Information. As expected, herbivores had the lowest δ15N, and the known non-ant predators had high δ15N. Most ant genera had mean δ15N values in between primary herbivores and non-ant predators. Only two ant genera had δ15N values that were higher than the known non-ant predators (Rhytidoponera and Tetramorium) (Fig. 3).
There was a significant effect of genus (Wald 6, 78·4 = 91·17, P < 0·001) and land use type (Wald (2, 9) = 8·65, P = 0·048) on mean δ15N of ants, but no significant interaction (Wald (12, 78·5) = 11·98, P = 0·458) (Fig. 4). Further, there was a significant effect of genus (Wald (6, 78) = 91·17, P < 0·001) but not land use type (Wald (2, 8·7) = 2·07, P = 0·393) on mean δ13C of ants, and no significant interaction (Wald (12, 78) = 12·5, P = 0·420). Whereas Myrmecia, Iridomyrmex, Melophorus, Pheidole and Rhytidoponera all had high δ15N, in the same range as known non-ant predators (Fig. 3), Camponotus and Monomorium had lower mean δ15N. Tukey’s pairwise post hoc comparisons, show that Camponotus and Monomorium are significantly different (P < 0·05) from each other in δ15N, and different from the cluster of five genera (Fig. 4). Camponotus is further distinguished by having δ13C greater than all other genera (Tukey’s pairwise post hoc comparison). Although many species in this genus are nocturnal or crepuscular, δ15N values did not differ between crepuscular and diurnal Camponotus species (t-test: t(1,18) = 0·32, P =0·75).
The significant effect of land use type on mean δ15N of ants occurred because the soil corrected mean δ15N for ants in revegetation sites was substantially lower than for pasture sites (Tukeys pairwise post hoc comparison, P < 0·05, Fig. 4). Ants from remnant sites had the highest mean δ15N but this was not different from pasture sites.
Reanalysis of the data, with ants allocated as ‘herbivore-omnivore’ (Camponotus and Monomorium) and ‘predator-omnivore’ (all other genera) or did not alter patterns. Although the more predatory species appear to cover a broader and lower range of δ15N values in revegetation sites (Fig. 4), with a drop in values particularly evident for Pheidole spp., we could not detect a significant interaction between land use type and trophic group in this reanalysis.
C and N concentrations
Dry mass C and N concentrations showed relatively little variation across workers of different ant species (Mean ± SE: Cdry mass = 50·5 ± 0·3%, Ndry mass = 11·8 ± 0·1%, C : N ratio = 4·29 ± 0·04; n = 44 species means). We found no significant correlations across ant species between Ndry mass and δ15N (uncorrected δ15N: r = −0·13, P =0·31; soil corrected δ15N: r = 0·12, P =0·45) or C : N ratio and δ15N (uncorrected δ15N:r = −0·16, P =0·42; soil corrected δ15N: r = −0·08, P =0·59), suggesting that our results were not affected by tissue-dependent variation.
Trophic flexibility of ant communities in response to land use
This study is the first to use stable isotopes to examine the effects of land use on the trophic structure of assemblages of omnivorous epigaeic arthropods. Ant genera maintained relatively constant trophic positions across systems, i.e. we found no interactions between genus and land use type. This is in agreement with Ponsard & Arditi (2000), which showed no shift in the relative trophic position of soil invertebrates between sites. Studies of the isotopic signatures of ants or other terrestrial invertebrates have not previously considered the effects of land management on relative trophic positions. However, widespread genera, such as Camponotus, Crematogaster, Pheidole and Paratrechina appear to maintain consistent relative positions across continents (compare Blüthgen, Gebauer & Fiedler 2003; Davidson 2005; Fiedler et al. 2007; Ottonetti et al. 2008, this study). Although there is some variation between species in these genera, this suggests that trophic roles within ant assemblages are relatively inflexible i.e. they are conserved across ecosystems and management regimes. This relative inflexibility provides some support for the hypothesis that competition (present or past) is critical in structuring assemblages because competitors constrain each others’ diets.
Although land management did not alter the relative positions of ant genera in the food chain, it did alter the position of the entire ant community. To clarify, all ant genera examined appeared to shift their position in the food chain in parallel, even after correction for soil isotope values. This is in agreement with previous studies, which have shown assemblage-wide trophic shifts in response to anthropogenic disturbances (Vizzini et al. 2005; Fox et al. 2009). The data suggest that the ant assemblages in recently revegetated sites fed lower in the food chain than those in pasture and remnant sites. We suggest this is driven from the bottom up, by plant age. Recently revegetated sites support young trees, which tend to be more vulnerable to herbivory than older trees (Larsson & Ohmart 1988). Herbivorous insects can provide resources to ants in the form of honeydew (mutualism) or protein (predation). In support of this idea, activity of the most abundant ant genus in our study (Iridomyrmex) on tree trunks in revegetated sites was significantly higher per hectare than in pastures or remnants in this study (unpublished data), probably because the availability of plant-based resources, particularly honeydew, is also higher (Gibb & Cunningham 2009). Similarly, a shift to greater ‘herbivory’ was identified in a longitudinal study of the invasion success of Argentine ants, Linepithema humile (Tillberg et al. 2007), and attributed in part to the species exploiting honeydew-rich early successional habitats at edges or in highly disturbed areas in the early stages of invasion. However, the less abundant genera did not increase activity on tree trunks in revegetated sites (unpublished data), possibly because other components of the habitat limited their abundances. Increased ant use of honeydew thus provides at best a partial explanation for the uniform change across ant genera. Another cause of lower δ15N values in revegetated sites (Fig. 4), could be an increase in herbivores as a proportion of the protein of ant diets. Indeed, the abundance of epigaeic herbivores was also higher in the young revegetated sites than pastures or remnants in the study area (Gibb & Cunningham 2010).
Ant trophic positions in pastures were only slightly lower than in remnants. In remnant sites, food webs may be more complex, with omnivorous and predaceous species with higher δ15N values more common in the diet of ants. The pasture sites in this study were centred on large trees, which provided an island of structural complexity and diversity in the otherwise uniform pasture habitat. Pasture trees are known to sustain considerable diversity in agricultural landscapes (e.g. Majer & Delabie 1999; Fischer and Lindenmayer 2002a, 2002b, Oliver et al. 2006; Manning, Fischer & Lindenmayer 2006). Ant-hemiptera interactions were more common and total ant activity was higher on individual pasture trees than remnant trees in the study area (Gibb & Cunningham 2009). It is unclear how isotopic values of ants in pastures without trees would appear, but open pastures support more predaceous and saprophagous epigaeic beetles than remnants (Gibb & Cunningham 2010), which may provide prey with higher δ15N.
The absence of a land use effect on δ13C has two possible interpretations. It could be that there are no systematic differences between the land use types in the mix of C3 and C4 plant species. Whether or not this is so could be determined by a survey of the flora in each site. However, even if there are differences in the mix of C3 and C4 plants at the level of land use, these might be obscured by a tendency for ants to focus on resources linked to only a subset of the plants in the site. If, for example, most of the food resources are collected on the woody plants such as the Eucalypts (which are C3), then shifts in delta C are unlikely to be associated with land use.
Ant isotopic values
Consistent with previous studies on ant assemblages (Blüthgen, Gebauer & Fiedler 2003; Davidson et al. 2003; Davidson 2005; Fiedler et al. 2007; Ottonetti et al. 2008), ant species covered a spectrum of trophic levels, from highly predatory species with similar δ15N values to spiders, to species with isotopic signatures suggesting a diet that includes more plant-derived material. Our data suggest that no ant genus included in this study should be considered to be herbivorous, as δ15N values for all ant genera were considerably higher than for herbivores. This is in contrast to previous suggestions that the extreme abundance of ants in some ecosystems may be a result of ants effectively behaving as herbivores (Tobin 1991; Davidson 1997). Ant genus placement along the continuum was mostly as expected, with Camponotus more herbivorous than other species and ponerine ants more predatory. The isotopic values of Iridomyrmex placed it amongst the more predatory species, which was surprising, given its known high dependence on honeydew (Gibb & Cunningham 2009). Monomorium fed at unexpectedly low trophic levels.
Camponotus was very different from other ant genera, both in terms of δ15N and δ13C values. This is consistent with findings from previous studies (Blüthgen, Gebauer & Fiedler 2003; Davidson 2005; Fiedler et al. 2007; Ottonetti et al. 2008) that Camponotus and its close relative Polyrhachis have low δ15N values. Camponotus is thus more ‘herbivorous’ than other genera, being reliant on nutrient sources closer to the base of the food chain. Species from this genus diurnally utilised floral nectar and occasionally hemipteran honeydew in the study area (Gibb & Cunningham 2009). It is also likely that some Camponotus species tended Hemiptera nocturnally, when potential competitors for this resource, such as Iridomyrmex, were less active. A high dependence on honeydew and nectar might result in lower δ15N values (Blüthgen, Gebauer & Fiedler 2003; Davidson et al. 2003). It is also possible that predation on herbivorous insects, which might be expected in arboreal Camponotus, such as the C. gasseri species group, would maintain a lower δ15N value than predation on detritivorous or predaceous insects.
The low δ15N value of Monomorium is not consistent with previous studies in tropical forests that suggest a more predaceous role for this genus (Blüthgen, Gebauer & Fiedler 2003). However, Monomorium spp. are commonly seed harvesters (Shattuck 1999) and removed seeds from arboreal Eucalyptus fruits at the study sites (Gibb, pers. obs.). A reliance on plant-derived proteins would result in a lower δ15N value.
Despite the high dependence of ants of the ecologically dominant genus Iridomyrmex on hemipteran honeydew (Gibb & Cunningham 2009), this genus showed similar δ15N values to many ponerines and species of Myrmecia, which are thought to be highly predatory, e.g. Andersen (1990) classifies these groups as specialist predators. Other dominant honeydew-dependent ants, such as Oecophylla smaragdina and Formica spp. (Blüthgen, Gebauer & Fiedler 2003; Gibb & Johansson 2010), also appear more predatory than might be expected (Blüthgen, Gebauer & Fiedler 2003; Fiedler et al. 2007), given their considerable intake of honeydew. For ants, energy from honeydew may fuel active predation, with protein from prey allowing a more balanced carbon to nitrogen ratio in the diet (Davidson 2005). Honeydew- and extra-floral nectary-tending ants have previously been shown to benefit host plants by preying on herbivorous insects (Buckley 1987; Karhu & Neuvonen 1998; Styrsky & Eubanks 2007), suggesting that much of their prey should be herbivorous. However, orders that are predominantly herbivorous may make up less than 40% of the prey items collected by Iridomyrmex purpureus (Gibb 2003). In addition, as little as 22–40% of the nitrogen from the diet of another honeydew-dependent ant, Formica aquilonia, comes from invertebrate prey (Domisch et al. 2009). Given the high activity of Iridomyrmex on tree trunks and rates of tending of Hemiptera observed in tree canopies (Gibb & Cunningham 2009), it is likely that the diet of this genus also has a very high honeydew contribution. While extensive predation could be responsible for the high δ15N of Iridomyrmex spp., it is also possible that gut symbionts enrich δ15N in ants that have significant sugar components in their diets (Eisner 1957; Cook & Davidson 2006; Fiedler et al. 2007; but see Feldhaar, Gebauer & Blüthgen 2009).
Carbon isotopes are not known to be strong indicators of trophic level, although they are associated with the physiology of plants contributing to diets. Camponotus spp. had higher δ13C values, but we have no evidence to suggest that they fed more or less on herbivores of C4 plants than other ant genera. A previous study (Blüthgen, Gebauer & Fiedler 2003) showed that different sap-sucking hemipteran groups differ in their δ13C values, possibly as a result of differences in the plant tissues (i.e. xylem, phloem or parenchyma) on which they feed, or gut microbes. It is thus possible that the higher δ13C values of Camponotus spp. results from its relationship with sap-sucking Hemiptera.
Although ants are thought to be ecologically flexible and generalist in their diets, this work suggests that they are relatively conservative in their relative trophic niche positions. Ants may be inherently inflexible, but there may also be a role for competitive interactions in constraining the possible niche breadth of species. Previous studies on resource partitioning in ants suggest that this is likely (e.g. Kaspari 1993; Pfeiffer, Nais & Linsenmair 2006), although recent field experiments have found only a weak role of competition in structuring communities (e.g. Gibb & Hochuli 2004; King & Tschinkel 2006, 2008; H. Gibb & T. Johansson, unpublished). Ants in our study did not alter their relative trophic positions, but we observed differences in their δ15N, relative to base levels. This suggests that the ant assemblage as a whole is relatively flexible in its use of resources, with herbivore-dominated assemblages of arthropod prey supporting the same set of common genera as arthropod prey assemblages dominated by omnivores or saprophages.
We observed considerably lower δ15N values and thus shortening of the food web in revegetated pastures. Food web complexity may thus be much reduced in these herbivore-dominated sites. For revegetated sites, this is probably a necessary transitional stage from a treeless to a treed state and likely reflects a gain in plant-derived resource availability, rather than a loss of trophic diversity. The unlikely combination of structural conservatism and ecological flexibility of ant assemblages suggests that, in contrast to species composition, the trophic composition of ant assemblages may be relatively resilient to anthropogenic habitat alteration.
We thank Meredith Cosgrove for assistance with the lab work and Eliza Finlay for contributions to field work. We are also grateful to Greening Australia for help with locating the field sites and Environment A.C.T. and the landholders for their generosity in allowing access and sampling on their properties. Nico Blüthgen kindly provided comments on a draft manuscript. This project was supported through an Office of the Chief Executive CSIRO Postdoctoral Fellowship to H.G.