Katharine L. Stuble, 569 Dabney Hall, 1416 Circle Drive, Knoxville, TN 37996, U.S.A. E-mail: email@example.com
Abstract 1. This correlational study examines the relationship between the red imported fire ant (Solenopsis invicta) and native ants in a longleaf pine savanna. Fire ants are frequently associated with a decline in native ants throughout the invaded range, but fire ant invasion is often coupled with habitat disturbance. Invasion of fire ants into the longleaf pine savanna provides an opportunity to examine the structure of the ant community in the absence of habitat disturbance.
2. Pitfall trapping was conducted within the longleaf pine savanna as well as across a naturally occurring soil moisture gradient, in plots that had been artificially watered.
3. Species richness did not vary as a function of fire ant density. There was an inverse relationship between native ant density and fire ant density, but this abundance pattern does not necessarily imply a causal link between fire ant invasion and native ant decline. For individual species, fire ant densities were negatively correlated with the densities of only two native ant species, including Solenopsis carolinensis, a native species that potentially limits the invasion of fire ants. Additionally, fire ants and native ants respond differently to soil moisture, with native ants favouring drier conditions than fire ants.
4. The possible exclusion of fire ants by some native ants, as well as differences in habitat preferences, provide alternative explanations for the frequently observed negative correlation between fire ants and native ants.
Invasive ants threaten ecosystems worldwide (Williams, 1994). Due largely to the extremely high densities reached in their invaded ranges, invasive ants alter the composition of terrestrial communities through a variety of mechanisms including competition and predation (Holway et al., 2002; Allen et al., 2004). One of the more conspicuous exotic ants in North America is the red imported fire ant (Solenopsis invicta Buren), but the impact of its extensive invasion on native ant assemblages in North America is currently unresolved. This study examines patterns of fire ant and native ant abundance in a native North American ecosystem.
Native to South America, fire ants invaded the United States in the 1930s (Wojcik et al., 2001). The range of the red imported fire ant has since expanded to include the majority of the southeastern United States, as well as parts of the west coast (Callcott & Collins, 1996; Korzukhin et al., 2001; Tschinkel, 2006). Typically, invaded areas are highly disturbed habitats such as lawns, pastures, and agricultural fields, but may also include native habitats, such as the endangered longleaf pine–wiregrass savanna (Carroll & Hoffman, 1997).
Fire ants have displaced two species of native fire ants, Solenopsis geminata and Solenopsis xyloni (Wilson, 1951), and may also be displacing a wider array of native ant species throughout much of the invaded range (Morris & Steigman, 1993; Jusino-Atresino & Phillips, 1994). Native ant abundance has been found to drop by 90%, and species richness by 70% immediately following invasion of the polygyne form of the red imported fire ant (Porter & Savignano, 1990). Further evidence that fire ants may be displacing native ants has been presented from a regional perspective, where ant species richness along the east coast of the United States was found to peak in Virginia, near the northern range limit of the red imported fire ant, rather than in areas of lower latitude, as was expected (Gotelli & Arnett, 2000).
Although these findings could indicate competitive displacement by fire ants, evidence of alternate mechanisms is emerging. An alternative explanation for the negative association between fire ants and native ants is that fire ants may preferentially invade areas that are highly disturbed (Stiles & Jones, 1998). Such sites may have already lost much of their native ant fauna (Zettler et al., 2004; Tschinkel, 2006). Thus, rather than depressing levels of native ants, fire ants may simply be capitalising on the reduction of native ants caused by habitat modification (King & Tschinkel, 2008). In forest clear cuts and along roads, native ants have been found to decline dramatically relative to undisturbed areas, while fire ant densities increase, despite fire ants remaining virtually absent in adjacent forest interiors (Tschinkel, 1988; Zettler et al., 2004).
Moreover, some native ant species may be able to persist in the presence of fire ants or to rebound following an initial decline as a result of fire ant invasion. A follow-up to the Porter and Savignano (1990) study, resampled the Texas site 12 years after the initial invasion and found that although fire ants remained the most abundant ant, the native ant community had largely recovered (Morrison, 2002), suggesting that ant community dynamics may shift as time progresses following invasion. Additionally, several studies in north Florida have found the abundance of fire ants in pastures to be positively correlated with the abundance of co-occurring ants (Morrison & Porter, 2003; King & Tschinkel, 2006), contradicting the perception that fire ants displace native ants. This suggests that native and invasive ant abundances may be controlled by the same environmental factors. Similarly, there is evidence that reducing fire ant densities may not be associated with a subsequent increase in native ant density, further implying that fire ants may not suppress native ant abundances in some situations (Stimac & Alves, 1994; King & Tschinkel, 2006).
Even less understood is the ability of fire ants to invade native ecosystems. Tschinkel (1988) found that fire ants occur in longleaf pine uplands, but only in highly disturbed areas or along pond margins. Fire ants have also been reported in low densities in undisturbed longleaf pine flatwoods with abundant herbaceous cover and periodically saturated soils (Lubertazzi & Tschinkel, 2003). Nevertheless, Carroll and Hoffman (1997) documented a high relative abundance of fire ants (composing 55% of ants) in an upland longleaf pine–wiregrass savanna, despite the intact nature of this native ecosystem.
Native remnant stands of longleaf pine–wiregrass that lack a history of soil disturbance, provide an opportunity to examine the role of fire ants on native ant declines, decoupled from the impacts of human-mediated disturbance. The objective of this study is to characterise the composition of the ant community relative to fire ant density in a fire ant-invaded longleaf pine savanna, approximately four decades post-invasion. Specifically, the following questions were addressed: (1) how does species richness, total ant abundance, and native ant abundance vary relative to fire ant density; (2) how does this relationship vary temporally; and (3) what is the role of soil moisture in structuring this relationship?
Materials and methods
This study was conducted on the property of the J.W. Jones Ecological Research Center (Ichauway). The 12 000 ha site, located in southwestern Georgia (Baker County), consists of remnant natural stands of longleaf pine (Pinus palustris Miller) with an understorey dominated by wiregrass (Artistida stricta Michx). This site has been managed for more than 70 years with frequent prescribed fire for game bird management. Currently, prescribed burns are conducted at approximately 2-year return intervals. The average daily temperature is 11 °C during winter and 27 °C during summer with an average annual rainfall of 132 cm year–1. These longleaf pine stands have been invaded by fire ants, which probably first appeared in the local area in the 1960s (Callcott & Collins, 1996).
Ant community composition
Species richness and abundance of ground-dwelling ants were sampled in nine 1 ha plots classified as ‘somewhat excessively drained upland terraces’, based on soil type, vegetation, and landscape position (Goebel et al., 2001). Soils consisted of loamy sands over sandy loams. All sites had an overstorey of longleaf pine and an understorey dominated by wiregrass. Wiregrass returns very slowly following soil disturbance, so its presence indicates that these sites had not been previously cultivated (Clewell, 1989). All plots were burned with prescribed fire in January 2006. Four plots were dropped from the study in 2007 due to unintended habitat disturbance.
Species richness and abundance of ground-dwelling ants were sampled using a standard pitfall trapping technique (Majer, 1978). Each pitfall trap consisted of a 15.3 cm long section of 2.1 cm diameter polyvinyl chloride (PVC) pipe, that was sunk into the ground. A test tube (15 cm long, 2 cm diameter) was inserted into the PVC pipe, such that the opening of the test tube was flush with the ground. Pitfall traps were arranged in arrays consisting of three pitfall traps positioned to form an equilateral triangle with a distance of 5 m between traps. Arrays of pitfall traps were distributed evenly throughout each plot, with nine arrays per plot arranged in a grid composed of three rows of three (27 pitfall traps per plot). Arrays were positioned 20 m from neighbouring arrays and the outermost arrays were 30 m from the plot’s edge.
Ants were trapped monthly between March 2006 and October 2007. To trap ants, a small amount of soapy water was added to each test tube and traps were left open for 24 h. Ants retrieved from the traps were stored in 70% ethanol until they could be identified to species. Voucher specimens are currently stored at the J.W. Jones Ecological Research Center. A rubber stopper was used to close the test tubes between sampling periods to prevent them from filling with soil and detritus. Ants were not sampled when rain was predicted during the 24-h trapping period.
Ant community composition across a soil moisture gradient
An ongoing water-addition experiment across a natural soil moisture gradient was used to examine the influence of soil moisture on ant community composition. In this long-term study, eight 0.25 ha (50 m × 50 m) plots have been irrigated with approximately 825 mm of reverse osmosis-treated water per year (∼65% increase in yearly precipitation) to maintain soils at approximately 40% field capacity since 2002, while eight non-watered plots served as controls. Water was applied once a week for a 24-h period. Plots were split between opposite ends of a natural moisture gradient, with four plots of each treatment located in xeric conditions in sites classified as ‘excessively well drained’ and the remaining plots located in mesic conditions classified as ‘somewhat poorly drained’ (Goebel et al., 2001). All plots were burned on a 2-year return interval.
Ants were trapped in three randomly placed arrays within each of the 16 study plots. Arrays and pitfall traps were set up in the manner described in the ant community composition section above. Trapping was conducted twice in each plot between June and July of 2007.
Chao2 was calculated as an estimate of species richness per plot for each summer (EstimateS, version 8.0, Colwell, ). Means and standard errors were calculated for number of fire ants, native ants, and total ants per array, both over the course of the study and for the summer season (June through September) (PROC MEANS, SAS version 9.1). The relationship between fire ant density and native ant density and species richness, averaged by plot for each summer, was examined using a multiple linear regression analysis, in which year was used as a covariate (PROC REG, SAS version 9.1). Regression analysis was also used to examine the density of Solenopsis carolinensis as a function of native ant density, using year as a covariate. Correlation analysis was used to examine the density of individual native ant species, averaged by plot and summer, in response to fire ant density (PROC CORR, SAS version 9.1). A Spearman rank correlation (PROC CORR, SAS version 9.1) found no correlation among plots between 2006 and 2007 with respect to fire ant, native ant, and total ant densities within plots. Accordingly, data points for 2006 and 2007 within a plot were treated as independent samples in all regression analyses. Finally, ant densities and species richness of the ant community between summer 2006 and summer 2007 (PROC GLM, SAS version 9.1) were compared, and the relative abundance of fire ants monthly over the course of the study was calculated.
We tested for differences in mean fire ant density, native ant density, total ant density, and species richness (Chao2) in response to irrigation treatment using a general linear model analysis with water-addition treatment, site type (xeric vs. mesic), and a treatment × site interaction term as independent variables (PROC GLM, SAS version 9.1).
Ant community composition
Pitfall-trapping efforts resulted in the capture of 21 380 ants, representing 25 species, from May 2006 to October 2007 (Table 1). During this 16-month period, the most common ant collected was S. invicta, comprising 44% of the ants captured. Solenopsis carolinensis and Pheidole spp. were the next two most common ants, making up 18% and 16% of the captured ants, respectively. While Pheidole was identified only to genus due to the difficulty of identifying minor workers, this genus was comprised largely of P. tysoni, and to a lesser extent P. bicarinata. The relative abundance of fire ants varied seasonally, dropping during the winter months, although it remained relatively constant during the summer (Fig. 1).
Table 1. Percentage of each species trapped during the summers of 2006 and 2007, combined, and over the course of the experiment.
Considering ants trapped between June and September, there was a negative correlation between fire ant and native ant densities (t = −2.63; d.f. = 2, 11; P = 0.02; r2 = 0.4211) (Fig. 2). Species richness did not vary as a function of fire ant density (t = 1.27; d.f. = 2, 11; P = 0.23; r2 = 0.1193).
The mean density of ants trapped per array was lower in the summer of 2007 than 2006 (t = 2.85; d.f. = 358; P < 0.01). This decline was partially attributable to a decline in fire ant density between 2006 and 2007 (t = 2.39; d.f. = 358; P = 0.02). The mean relative abundance of fire ants per array declined from 53.9 (±2.7) % in 2006 to 45.8 (±3.0) % in 2007 (t = 2.04; d.f. = 349; P = 0.04). There was no decline in native ant density (t = 1.54; d.f. = 358; P = 0.12), but species richness declined from a mean of 3.7 (±0.11) species per array in 2006 to 2.6 (±0.10) species in 2007 (t = 7.52; d.f. = 356; P < 0.0001).
Fire ant density was positively correlated with the density of four ant species, Brachymyrmex depilis (r = 0.72; P < 0.01), Camponotus pennsylvanicus (r = 0.73; P < 0.01), Cyphomyrmex rimosus (r = 0.66; P = 0.01), and Proceratium silaceum (r = 0.73; P < 0.01). Fire ant density was negatively correlated with Crematogaster lineolata (r = −0.69; P < 0.01) and S. carolinensis (r = −0.64; P = 0.01). There was a positive relationship between the density of S. carolinensis and native ant density (t = 3.46; d.f. = 2, 11; P < 0.01; r2 = 0.5249).
Ant community composition across a soil moisture gradient
Pitfall trapping in the soil moisture treatment plots resulted in the capture of 2255 ants, representing 20 species. Fire ants composed 53% of the ants captured in these pitfall traps. Total ant density within plots was independent of treatment status as well as site (xeric vs. mesic) (Table 2). The response of native ant density to the water-addition treatment depended on site (Fig. 3). In the xeric site, greater densities of native ants occurred in the absence of water addition, whereas no differences occurred in response to treatment in the mesic site. The relative and actual abundances of fire ants were greater in the mesic site than in the xeric site, and no differences were attributable to treatment. However, a trend of increasing fire ant density with water addition was observed at both sites (Fig. 4). Greater species richness occurred on xeric sites than on mesic sites, and no difference in richness occurred in response to treatment.
Table 2. General linear models evaluating the effects of water-addition treatment and site soil moisture on ant community composition.
Mean ants per array
Site × Treatment
Mean native ants per array
Site × Treatment
Mean fire ants per array
Site × Treatment
Mean species richness per plot
Site × Treatment
Mean per cent fire ants per array
Site × Treatment
The red imported fire ant was found not only to be present in an intact longleaf pine–wiregrass savanna, but also to be the most common ant in this system. These results are similar to those of Carroll and Hoffman (1997), who also found fire ants to be dominant in the ecosystem. The presence of high densities of fire ants in an undisturbed, fire-maintained longleaf pine savanna interior counteracts most observations that fire ants are relegated to heavily disturbed areas and virtually absent in forest interiors (Tschinkel, 1988; Zettler et al., 2004). One explanation for the prevalence of fire ants within our study area may be the frequent fires used in the management of the land. Prescribed fires conducted on a 1–2-year return interval, maintain an open canopy of pines and a grassy understorey dominated by wiregrass, and reduce the abundance of hardwoods, potentially creating favourable conditions for fire ants. Additionally, wiregrass is not rhizomatous and thus fire exposes extensive areas of bare soil in this ecosystem, potentially generating recruitment sites for founding fire ant queens.
The relative density of fire ants varied over the course of the year, but densities were consistently high during the summer. Annual variation in fire ant densities was considerable, with fewer fire ants present in 2007. It is interesting that fire ants declined over this period while native ants did not. Variables that could have contributed to this decline include drought and fire. All plots were burned just prior to the beginning of this study and were not burned again in 2007. It is possible that fire ants declined as a result of accumulation of ground cover biomass 2 years post-burn. Moreover, the second year of study also coincided with a near-record drought (National Oceanic and Atmospheric Administration, 2008), and drier soil conditions potentially favoured native ants over fire ants (Tschinkel, 2006).
Our finding that species richness varied independently of fire ant density, may indicate resilience of an intact ant assemblage in its native ecosystem to fire ant invasion from the very beginning, or it may reflect the short-term nature of the influence of fire ants on native ants. Fire ants likely arrived in southwest Georgia in the 1960s (Callcott & Collins, 1996) and were abundant in Baker County, Georgia by the 1970s (Baker County Historical Society, 1991). We have no data on pre-invasion ant levels at our study site, nor do we know if native ant densities declined immediately following fire ant invasion. While we cannot determine early impacts on the ant community, nor any subsequent successional processes, our data do indicate that a negative relationship between fire ant densities and species richness does not exist, at least in this study system at this stage of the post-invasion process.
Nonetheless, overall abundance of native ants was found to be inversely related to the density of fire ants. This, we should note, is despite the fact that conditions such as canopy and understorey composition were very similar among sites, which were burned with the same frequency. As such, we are unable to identify a specific environmental factor that may vary among our plots, favouring fire ants while disfavouring native ants. Thus, we cannot eliminate the possibility that fire ants may have a long-term effect on the ant assemblage by displacing native ants, nor can we conclude from this evidence that fire ants are the cause of this reduction. Ant communities are highly structured by competition (Hölldobler & Wilson, 1990), and fire ants have been shown to be competitively superior to native ants with respect to collecting food resources (Porter & Savignano, 1990; Gibbons & Simberloff, 2005), providing a mechanism through which fire ants may displace native ants. This apparent competitive asymmetry, however, is probably driven by the disproportionately high abundances of fire ants (Morrison, 2000). However, it may also be influenced by the potential ability of fire ants to cheat the dominance discovery trade-off (Davidson, 1998).
Our observation of an inverse relationship between the ensity of native ants and fire ants might suggest that fire ant density limits native ants, but the opposite could also be the case (Rao & Vinson, 2004; Tschinkel, 2006). Other studies have suggested that some native ant species, including S. (Diplorhoptrum), may limit fire ants (Tschinkel, 1988) through their ability to eliminate small colonies (Tschinkel, 1988; Rao & Vinson, 2004; Vinson & Rao, 2004). The S. (Diplorhoptrum) group includes S. carolinensis, a native of the longleaf pine ecosystem and one of only two native species we found to be inversely related to fire ant density. This inverse relationship was largely driven by the near absence of fire ants from plots containing high densities of S. carolinensis. Further, we observed a positive relationship between S. carolinensis and native ant densities within our study area. Thus, it is possible that the reduced number of fire ants in areas with high densities of native ants may be a result of the ability of members of this group of species to exclude fire ants. We concur with others (Rao & Vinson, 2004; Tschinkel, 2006) in the suggestion that such interspecific exclusion of fire ants could be an explanation of the frequently observed negative correlation between fire ants and native ants in other studies. Potential explanations for the negative correlation between fire ants and Crematogaster lineolata are less clear and further examination of the ecological role of this species would be useful.
Additionally, the fact that fire ants and native ants responded differently to natural soil moisture variation and water-addition, suggests that fire ants and native ants may take advantage of different environmental conditions. Fire ants may thrive in more moist soils, a reasonable conclusion as they are native to the margins of seasonally flooded wetlands in South America (Tschinkel, 2006). These results are consistent with the findings of Tschinkel (1988), suggesting that fire ants are more likely to invade areas with wet–mesic soils. This propensity for fire ants and native ants to thrive in differing environmental conditions may be another factor leading to a negative correlation between fire ants and native ants.
It should be noted that red imported fire ant colonies in the study area were monogyne. The polygyne social form is known to reach higher densities than those of the monogyne form and may have the potential to be more detrimental to native communities (Macom & Porter, 1996). Thus, the presence of the monogyne form may play a role in moderating the impact of fire ants on the native ant community in this study.
This correlational study finds that fire ants can be negatively associated with native ant density in the absence of human-mediated soil disturbance, even 40 years post-invasion. While negative correlations, such as the one observed here, are often interpreted as evidence of the ability of fire ants to displace native ants, we suggest that the role of native ants in limiting fire ant invasion and the differing responses of fire ants and native ants to environmental variables, should be considered as potential explanations for this inverse relationship. However, to fully understand the impact of fire ants on native ants, it will be necessary to examine competitive mechanisms of aggression and exploitation, discovery and dominance of resources, and population measures of dispersal, establishment, colony demographics and turnover across longer time intervals. In addition, experimental removal and addition of fire ants is warranted to adequately decouple the role of environmental factors from that of competition in structuring native and fire ant populations.
We owe thanks to L.M. Conner for his advice on our statistical analyses. We acknowledge the helpful comments on this manuscript provided by R.E. Wyatt, N.J. Sanders, and two anonymous reviewers. Financial support was provided by the J.W. Jones Ecological Research Center, the R.W. Woodruff Foundation, and the University of Georgia.