INTRODUCTION
Rain forest and savanna are contrasting closed and open vegetation types respectively that dominate the world's tropics, and support contrasting biotas. The boundaries between them are determined primarily by an interaction between moisture availability and fire frequency, and their dynamics are a key issue in tropical biogeography (Bowman Reference BOWMAN2000, Bullock et al. Reference BULLOCK, MOONEY and MEDINA1995, Furley et al. Reference FURLEY, PROCTOR and RATTER1992). However, studies of rain forest/savanna dynamics have focused primarily on vegetation, with less attention paid to contrasting rain-forest and savanna faunas (Lacher & Alho Reference LACHER and ALHO2001).
The rain forests and savannas of tropical northern Australia support remarkably disjunct ant faunas, with rain forests featuring shade-tolerant taxa of Indo-Malayan origin, and savannas dominated by autochthonous, ‘sun-loving’ taxa centred on arid Australia (Andersen et al. Reference ANDERSEN, PARR, LOWE and MÜLLER2007, Reichel & Andersen Reference REICHEL and ANDERSEN1996, Taylor Reference TAYLOR and Walker1972). The faunas also have contrasting functional composition, with rain-forest communities featuring high proportions of litter-dwelling (cryptic) and arboreal species, and savannas featuring ground-nesting behaviourally dominant species of Iridomyrmex (belonging to the functional group Dominant Dolichoderinae; Andersen Reference ANDERSEN1995) and highly specialized thermophiles and granivores (Hot-Climate Specialists) (Andersen Reference ANDERSEN, Agosti, Majer, Alonso and Schultz2000a). Further, although tropical rain forests generally are regarded as supporting the world's richest ant faunas (Brühl et al. Reference BRÜHL, GUNSALAM and LINSENMAIR1998, Longino et al. Reference LONGINO, CODDINGTON and COLWELL2002, Verhaagh Reference VERHAAGH, Veeresh, Mallik and Viraktamath1990, Wilson Reference WILSON1959), Australia's rain-forest ant fauna is relatively depauperate (Taylor Reference TAYLOR and Walker1972). In contrast, the ant fauna in Australian savannas is exceptionally rich (Andersen Reference ANDERSEN2000b).
This faunistic disjunction is especially pronounced in monsoonal north-western Australia, where rain forest occurs as isolated patches within a vast savanna landscape (Bowman Reference BOWMAN2000). Here, rain forest-savanna boundaries are remarkably abrupt, and the contrasting ant communities occur within a few metres of each other (Andersen & Majer Reference ANDERSEN, MAJER, McKenzie, Johnston and Kendrick1991, Andersen & Reichel Reference ANDERSEN and REICHEL1994). However, in the humid tropics of north-eastern Queensland there is often an extended gradient between rain forest and savanna, typically involving a band of eucalypt-dominated tall open forest (Ashton Reference ASHTON and Groves1981). It is not known if ant community structure varies continuously along this gradient, or, if there is a major disjunction, where it occurs.
Here we address this issue by describing ant community structure along an extended environmental gradient from rain forest to savanna in North Queensland. We examine patterns of ant richness, species composition and functional composition along the gradient, and test three competing hypotheses: (1) there is a continuous gradient between rain-forest and savanna ant communities; (2) a major disjunction occurs between rain forest on one hand, and open eucalypt forests and savanna woodlands on the other (i.e. a mesophyll-sclerophyll disjunction); (3) a major disjunction occurs between forests, both mesophyll (rain forest) and sclerophyll (tall open forest), and savanna woodland (i.e. a forest-savanna disjunction).
METHODS
Study sites
The study was conducted at ten sites (Table 1) distributed along a 6-km environmental gradient encompassing the crest and slopes of Mt. Lewis, in the Australian Wildlife Conservancy's Brooklyn wildlife sanctuary (16°59′S, 145°25′E) 100 km north-west of Cairns. The sanctuary covers 600 km2, and is remarkably diverse biologically, providing habitat for about 40% and 30% of all Australian bird and mammal species respectively (http://www.iucn.org/themes/wcpa/pubs/pdfs/Figgis_Parks2006.pdf). Previous studies of its ant fauna have been limited to opportunistic collections, but it includes one species that is known only from Mt. Lewis (Monomorium draculai; Heterick Reference HETERICK2001).
Summary descriptions of the ten study sites. Habitat type follows Groves (Reference GROVES1981), and plant species nomenclature follows the Australian Plant Name Index (http://www.anbg.gov.au/cpbr/databases/apni-search-full.html).
The study sites varied in altitude from 640 to 1060 m asl, with mean annual rainfall ranging from <900 mm at the lowest site to about 4000 mm at the top of Mt. Lewis (http://www.iucn.org/themes/wcpa/pubs/pdfs/Figgis_Parks2006.pdf). Vegetation ranged from savanna woodland (canopy cover <30%) dominated by eucalypt species at low elevation, through tall open forest (canopy cover about 50%) dominated by other eucalypt species at mid-elevation, to complex notophyll vine forest (rain forest) at highest elevations (vegetation nomenclature follows Groves Reference GROVES1981). Some of the rain forest at lower elevation is regrowth from extensive historical logging that ceased 30 y previously (S. McKenna, pers. comm.), so that we recognized four vegetation types: rain forest (RF, three sites); rain-forest regrowth (RFR, two sites); tall open forest (TOF, three sites); and savanna woodland (SW, two sites) (Table 1, Figure 1).
Photographs of the four habitat types occurring along the study gradient: Rain forest (site RF1) (a), Rain-forest regrowth (site RFR1) (b), Tall open forest (site TOF3) (c) and Savanna woodland (site SW2) (d).
Sampling
Ants were sampled during April (late wet season) 2007 using pitfall traps located in the ground (for ground-active species) and on the trunks of trees (for arboreal species). Sampling of leaf litter for cryptic species was also attempted using Winkler sacs (Agosti et al. Reference AGOSTI, MAJER, ALONSO and SCHULTZ2000); however, there was insufficient leaf litter for collection at the tall-open-forest and savanna sites, and so this was abandoned.
Pitfall traps were 4-cm-diameter plastic containers partly filled with ethylene-glycol as a preservative. At each site, 15 ground traps were established in a 5 × 3 grid with 5-m spacing, buried in the soil with their rims flush with the soil surface. An arboreal trap was taped to the tree nearest to each ground trap at 1.7 m height, following Andersen et al. (Reference ANDERSEN, HERTOG and WOINARSKI2006). Arboreal traps had their inner rims smeared with fish paste as an ant attractant, whereas ground traps were not baited. Each trap was opened for a single 48-h period. There was no substantial rain during the sampling period.
All ants collected in traps were sorted to species, and where possible named, with species nomenclature following Bolton (Reference BOLTON1995). Species that could not be confidently named were identified to species-group following Andersen (Reference ANDERSEN2000b), and assigned number codes (sp. 1, sp. 2, etc.) if they had been recorded as such in published studies from the Top End of the Northern Territory (Andersen et al. Reference ANDERSEN, HERTOG and WOINARSKI2006, Reference ANDERSEN, PARR, LOWE and MÜLLER2007). They were otherwise assigned letter codes (sp. A, sp. B, etc.) that apply only to this study. Voucher specimens of all species are held at the CSIRO Tropical Ecosystems Research Centre in Darwin.
Data analysis
Rarefaction curves, plotting the cumulative number of species recorded as a function of sampling effort (Gotelli & Colwell Reference GOTELLI and COLWELL2001), were used to compare species richness among the four habitat types and to assess sampling completeness. The curves were based on combined ground and arboreal traps from all sites from a particular habitat type (i.e. four curves, derived from either 60 or 90 traps), and were generated using EstimateS ver.7.5.0 (Colwell & Coddington Reference COLWELL and CODDINGTON1994). Mean species richness and abundance per trap were compared among habitat types using one-way ANOVA, with abundance data square-root transformed to meet the assumption of normality. A Tukey test of post hoc comparison was used to determine statistically significant differences between habitat types (Zar Reference ZAR1999).
Patterns of ant species composition were investigated at the site level using multidimensional scaling (MDS) in two dimensions, on species presence/absence data. MDS was based on a Bray-Curtis dissimilarity matrix, and performed using the software Systat 10. The extent of clustering according to stratum (ground vs arboreal) and vegetation type was then assessed by analysis of similarity (ANOSIM; Clarke & Warwick Reference CLARKE and WARWICK2001), using Primer 5.0 (Clarke & Gorley Reference CLARKE and GORLEY2001). Functional composition was examined by assigning species to one of nine functional groups based on global responses of their species-groups to environmental stress and disturbance (Table 2). Finally, the biogeographic affinities of the faunas from each habitat type were examined by assigning species to one of four biogeographic classes based on the distribution of their species-group within Australia, following Andersen (Reference ANDERSEN2000b): Torresian – occurring primarily in the tropical north; Bassian – occurring primarily in the cool-temperate south; Eyrean – occurring primarily in the arid zone; and widespread – well-represented throughout Australia.
Ant functional groups based on global responses to environmental stress and disturbance (see Andersen Reference ANDERSEN1995, Reference ANDERSEN1997).
RESULTS
Species richness and abundance
In total, 3479 individuals representing 95 species, 35 genera and 10 subfamilies were collected in traps (Appendix 1). Seventy-six and 45 species respectively were collected in ground and arboreal traps, with 26 (27%) collected in both. The great majority of species collected in arboreal traps nest in the ground, with only eight known to nest in trees. The richest subfamilies were Formicinae (30 species from 11 genera) and Myrmicinae (30 species from nine genera), and the richest genera were Pheidole (11 species), Rhytidoponera (10), Camponotus (9), Monomorium (9) and Polyrhachis (7). There was a single record of an introduced species – Monomorium destructor from one of the savanna-woodland sites.
Total species richness per habitat was 17 each for rain forest (n = 3 sites) and rain forest-regrowth (n = 2), 35 for tall open forest (n = 3), and 58 for savanna woodland (n = 2). These differences in species richness were reflected in rarefaction analysis (Figure 2). Rarefaction curves approached asymptotes for rain forest, rain-forest regrowth and tall open forest, but not for savanna woodland. Mean species richness per trap also differed significantly between habitat types (ANOVA; F(3, 296) = 38.8, P ≪ 0.01); it was identical in rain forest and rain-forest regrowth, somewhat higher in tall open forest, and twice as high again in savanna woodland (Figure 3a). Ant abundance showed a similar pattern, although there was more of a continuum between habitat types (Figure 3b). Both site species richness and abundance showed very strong negative relationships with altitude (Figure 4).
Rarefaction curves for the number of ant species collected in combined ground and arboreal pitfall traps among the four habitats (RF = rain forest, RFR = rain-forest regrowth, TOF = tall open forest and SW = savanna woodland).
Mean (± SE) number of species (a) and square root-transformed abundance (b) per pitfall trap within rain forest (RF), rain-forest regrowth (RFR), tall open forest (TOF) and savanna woodland (SW) habitats. In each case, different letters indicate significant differences (P < 0.05) between habitat types according to the post hoc Tukey test.
Relationships between altitude and site species richness (a) and square root-transformed abundance (b). RF = rain forest, RFG = rain-forest regrowth, TOF = tall open forest, SW = savanna woodland.
Species composition
The great majority (78%) of the 58 species from savanna woodland were recorded only in this habitat type (Appendix 1). These savanna specialists were primarily from the cosmopolitan genera Monomorium (8 species), Rhytidoponera (6), Camponotus (4) and Crematogaster (3), but also included specialist savanna genera such as Melophorus (3), Opisthopsis (2) and Iridomyrmex (1). At the opposite end of the environmental gradient, a relatively modest 13 species were recorded only at rain-forest or rain-forest-regrowth sites, despite having a combined five sites compared with only two for savanna woodland. The rain-forest specialists included representatives of the specialist rain-forest genera Discothrea and Pristomyrmex, as well as specialist rain-forest species-groups within Cerapachys, Leptomyrmex, Rhytidoponera, Monomorium and Leptogenys. Most (65%) species recorded in rain-forest regrowth also occurred in tall open forest. A small number of species (Rhytidoponera sp. E (araneoides gp.), Notostigma carazii and Meranoplus hirsutus) were relatively common in tall open forest but not recorded elsewhere, and just two species (Rhytidoponera victoriae and Prolasius sp. nr. nitidissimus) were recorded in all four habitat types.
The MDS plot showed clear separation between habitat types along the primary axis (Global R = 0.67, P ≪ 0.01), and between ground and arboreal strata along the secondary axis (Global R = 0.41, P ≪ 0.01) (Figure 5a). For both ground and arboreal data, there was continuous variation along the first axis from rain forest through rain-forest regrowth to tall open forest, and then a discontinuity between these habitat types and savanna woodland (Figure 5a). This result was confirmed by within-group comparisons in ANOSIM, with the dissimilarity between savanna and tall open forest (R = 0.89, P < 0.01) being far higher than that between tall open forest and rain-forest regrowth (R = 0.41, P = 0.01). The discontinuity between the gradient from rain forest to tall open forest on one hand, and savanna on the other, was particularly marked when ground and arboreal data were pooled (Figure 5b).
Two-dimensional multidimensional scaling (MDS) ordination of study sites (RF = rain forest, RFG = rain-forest regrowth, TOF = tall open forest, SW = savanna woodland) based on presence/absence data, considering ground and arboreal data separately (a; black and open symbols respectively) and pooled (b). Stress values are 0.15 and 0.07 respectively.
Functional groups and biogeographic affinities
Functional group composition varied substantially along the environmental gradient (Figure 6). Rain-forest habitat supported particularly high proportions of Specialist Predators, Cryptic Species and Cold-Climate Specialists, whereas Hot-Climate Specialists were absent. Hot-Climate Specialists were also absent from rain-forest regrowth sites, and were best represented in savanna woodland. The relative contributions of both Cryptic species and Cold-Climate Specialists decreased systematically from rain forest to savanna woodland. Generalized Myrmicinae and Subordinate Camponotini were particularly prominent in savanna woodland and tall open forest respectively.
Functional group profiles of ant samples from rain forest (RF), rain-forest regrowth (RFR), tall open forest (TOF) and savanna woodland (SW) habitats. Data are proportions of total species represented by each functional group: Dominant Dolichoderinae (DD), Generalized Myrmicinae (GM), Opportunists (O), Subordinate Camponotini (SC), Tropical-Climate Specialists (TCS), Cold-Climate Specialists (CCS), Hot-Climate Specialists (HCS), Cryptic species (C) and Specialist Predators (SP).
The biogeographic affinities of the fauna were remarkably consistent across habitat types, with about half the species representing Torresian taxa and about a quarter widespread in each case (Figure 7). However, Eyrean taxa were absent completely from rain forest and rain-forest-regrowth habitats, whereas they were equally as well-represented as Bassian taxa in savanna woodland.
Biogeographical composition of ant assemblages in each habitat type (RF = rain forest, RFR = rain-forest regrowth, TOF = tall open forest and SW = savanna woodland). Stacked bars show proportions of species with Torresian (T), Eyrean (E) and Bassian (B) affinities, or are from widespread species-groups (W).
DISCUSSION
We found a major disjunction in ant community structure along the environmental gradient from rain forest to savanna, and this occurred between savanna and forest (whether mesophyll or sclerophyll) rather than between rain forest and open eucalypt vegetation. The disjunction was evident in both species richness (which was far higher in savanna than in other habitats) and composition. The compositional disjunction occurred despite a very poor representation of species of Iridomyrmex from the savanna sites; such species are dominant members of most savanna ant communities throughout northern Australia, and epitomise the contrasting rain-forest and savanna ant faunas (Andersen et al. Reference ANDERSEN, PARR, LOWE and MÜLLER2007). Instead, the major behaviourally dominant dolichoderines at the savanna sites belonged to the shade-tolerant genus Anonychomyrma, which can be attributed to high grass density.
Rain-forest litter supports a very extensive cryptic ant fauna (Ward Reference WARD, Agosti, Majer, Alonso and Schultz2000), and given the lack of extensive litter in tall open forest and savanna, it is possible that the disjunction in this component of the fauna occurs between rain forest and tall open forest, rather than between tall open forest and savanna. Rain forest also supports a rich arboreal fauna (Blüthgen & Stork Reference BLÜTHGEN and STORK2007, Brühl et al. Reference BRÜHL, GUNSALAM and LINSENMAIR1998, Longino et al. Reference LONGINO, CODDINGTON and COLWELL2002, Wilson Reference WILSON1959), which was undoubtedly under-sampled in this study. However, it is clear that many arboreal taxa of rain forest origin extend into tall open forest, with most of such taxa recorded in the present study (e.g. Philidris sp., Polyrhachis sp. nr. mjobergi, Monomorium draculai) occurring in this habitat. No such taxa were recorded in savanna woodland. The under-sampling of the cryptic and arboreal faunas contributed to the low species richness recorded from rain forest, and therefore exaggerated the contrast with savanna.
Three factors that potentially have an important influence on ant community structure co-vary along the environmental gradient studied: rainfall, temperature and vegetation structure. These are inextricably linked in the immediate study region, as rain forest distribution is driven by rainfall, and rainfall and temperature follow the same elevational gradient. However, the relative importance of these factors as direct drivers of the observed ant community patterns can be assessed through a broader analysis of patterns of ant community structure in rain forest and open sclerophyll habitats across Australia. Elsewhere in Queensland, similar disjunctions between rain-forest and open sclerophyll ant faunas have been documented in the absence of an altitudinal gradient, where variation in habitat type is determined by edaphic conditions rather than by rainfall or temperature (Greenslade & Thompson Reference GREENSLADE, THOMPSON, Gillison and Anderson1981). Savanna vegetation occurring under very high annual rainfall (up to 2000 mm) elsewhere in northern Australia still supports a characteristically savanna ant fauna (e.g. Tiwi Islands, Andersen et al. Reference ANDERSEN, WOINARSKI and HOFFMANN2004). Temperature clearly has a direct influence on ant species composition, and can explain, for example, changes in the relative contributions of Tropical- and Cold-Climate Specialists along the Mt Lewis gradient. Tropical-Climate Specialists tend to prefer high temperatures, and are known to be progressively replaced by Cold-Climate Specialists with increasing altitude (Andersen Reference ANDERSEN, Agosti, Majer, Alonso and Schultz2000a). However, the broader disjunction in ant community structure described here cannot be directly attributed to temperature, as characteristic rain-forest ant community structure is maintained throughout the rain forest that extends all the way to the coast on the eastern (higher rainfall) side of the mountain ranges that include Mt Lewis. Indeed, rain-forest ant communities have similar structure throughout northern Australia, regardless of temperature (Andersen Reference ANDERSEN, Agosti, Majer, Alonso and Schultz2000a). Similarly, ant community structure comparable to that recorded in savanna woodland in our study occurs in open sclerophyll habitats in much cooler habitats than Mt Lewis, including in the temperate zone (Andersen Reference ANDERSEN1995).
This leaves vegetation structure as the likely ultimate cause of disjunction between savanna and forest ant communities. The structure of the ground-layer is critical here, as this is where the great majority of savanna ant species nest and forage. Savanna has more open ground than forest, and therefore provides ground-foraging ants with greater direct insolation and unimpeded foraging surfaces, which are critically important factors driving ant community structure (Andersen Reference ANDERSEN1995).
The far higher local richness of savanna compared with rain-forest ant communities in Australia is opposite to the pattern found elsewhere in the tropical world. For example, in Brazilian Amazonia local ant richness in rain forest is twice that in savanna (Vasconcelos & Vilhena Reference VASCONCELOS and VILHENA2006). In Brazilian cerrado (savanna) landscapes, ant richness is highest in forest patches and there is more generally a positive relationship between ant species richness and tree density (Ribas et al. Reference RIBAS, SCHOEREDER, PIC and SOARES2003, Silva et al. Reference SILVA, BRANDÃO and SILVESTRE2004). This inter-continental contrast can be explained by a combination of two factors: the neotropics have a far richer forest-associated arboreal fauna (Andersen et al. Reference ANDERSEN, PARR, LOWE and MÜLLER2007), and Australian savannas are unusually rich, reflecting an exceptionally diverse arid-adapted fauna (Andersen Reference ANDERSEN2003). The species compositional contrast in savanna and forest ant faunas also seems to be greater in Australia than in the neotropics: in Brazilian Amazonia about two-thirds of the savanna species also occur in forest (Vasconcelos & Vilhena Reference VASCONCELOS and VILHENA2006), whereas in the present study nearly 80% of savanna species were not recorded in any forest type.
In conclusion, our study of ant community structure along an extended rain forest–savanna gradient in tropical Australia has shown relatively continuous variation between rain forest and tall open forest, but a major discontinuity between tall open forest and savanna woodland. These findings indicate that the contrast between rain-forest and savanna ant communities is an extreme manifestation of a broader forest–savanna disjunction.
ACKNOWLEDGEMENTS
We are most grateful to Alex Kutt for the opportunity to conduct this project, and for providing the logistical support for it. We are most grateful to the Australian Wildlife Conservancy for access to Brooklyn, and for its support of the project. We thank Brooke Bateman and Stephen McKenna for field assistance, Tony Hertog for laboratory assistance, and Jonathan Majer, Heraldo Vasconcelos, Alex Kutt, Kate Parr, Ben Hoffmann and two anonymous referees for their comments on the draft manuscript. Jeannette Kemp and Stephen McKenna kindly provided the environmental information given in Table 1.
Functional group, biogeographic affinity and total abundance in each habitat type of all species collected in traps. Functional groups (see Table 2): CCS, Cold-Climate Specialists; C, Cryptic Species; DD, Dominant Dolichoderinae; GM, Generalized Myrmicinae; HCS, Hot-Climate Specialists; O, Opportunists; SC, Subordinate Camponotini; SP, Specialist Predators; TCS, Tropical-Climate Specialists. Biogeographic affinity (following Andersen Reference ANDERSEN2000b): B, Bassian; E, Eyrean; T, Torresian; W, Widespread.
1Primarily from a single trap.
