1. Invasive species pose ecological threats in many areas, but attempts to control invaders by introducing other exotic species may cause further unanticipated problems. If we can use predators native to the introduced range to assist in control of the invader, the risks of collateral damage are lower.
2. In tropical Australia, high desiccation rates restrict newly-transformed (metamorph) cane toads Bufo marinus to the margins of waterbodies, rendering the metamorphs vulnerable to predatory ants (Iridomyrmex reburrus). By adding bait (catfood) to selected areas, we increased ant densities (and thus, toad mortality) more than fourfold.
3. Over 50% of attacks by ants in the field were immediately fatal to the metamorph toads, and most ‘escapee’ toads (88%) died of their injuries within 24 h after the attack.
4. When we increased ant densities by artificial baiting, 98% of metamorph toads were encountered, and 84% attacked, within the two-minute observation period. Collateral damage to native fauna appears to be low, but warrants closer examination.
5.Synthesis and applications. Manipulating the foraging locations of native predatory ants can substantially increase their off-take of invasive toads. More generally, vulnerabilities of invasive species to predators native to the introduced range may facilitate control of invader numbers with little collateral damage to the rest of the fauna.
Globally, invasive species pose a major threat to biodiversity (Mack et al. 2000). Despite occasional stories of partial or complete success (e.g. Dodd 1940; Ratcliffe et al. 1952), most attempts to control invasive species have failed (Saunders et al. 2010). Many such failures reflect a tendency for control measures to be implemented with little understanding of the biology of the invader, the nature and magnitude of its impacts, or indeed, the control agent itself (Gurr & Wratten 1999). Ideally, pest management should be based upon an understanding of the invader’s ecology in its new environment; new approaches are urgently needed (Molloy & Henderson 2006). One possibility involves identifying ecological and behavioural ‘mismatches’– ways in which the characteristics of the invading species render it vulnerable to some mortality source operating within the introduced range. Presumably, such ‘mismatches’ are common in invading species, reflecting the myriad traits evolved in one context (the native range) but maladaptive in the new environment. We can explore such potential ‘mismatches’ in the hope of identifying novel control opportunities. Clearly, we also need to assess any effects of manipulations on native fauna as well as the invader, to identify potential collateral damage from any attempts at control.
Surveys of the Australian public typically rank cane toads Bufo marinus, Linnaeus 1758 [see Pramuk (2006) for suggested nomenclatural change to Rhinella marina] as among the most significant invasive species in that continent, and regard the animal’s introduction as an ecological catastrophe (Low 1999). Cane toads are large anurans native to South and Central America, brought to Australia in 1935 in an unsuccessful attempt to control insects pests of commercial sugar cane crops (Lever 2001). Perhaps reflecting the lack of bufonids (toads) among the native anuran fauna of Australasia, many native predators are sensitive to the toad’s toxic chemical defences (primarily bufadienolides: Chen & Kovarikova 1967). The expansion of cane toads across tropical Australia thus has killed native anurophagous predators including marsupials, crocodiles, snakes, lizards and anurans (Covacevich & Archer 1975; Letnic, Webb & Shine 2008). Massive effort has failed to slow the rate of toad invasion.
Mathematical modelling of cane toad life-history traits led Alford et al. (1995) to conclude that toads in their Australian range experience high mortality rates (> 98%) prior to breeding, and hence that small additional increases in mortality might significantly reduce toad abundance. In an attempt to develop novel approaches to toad control, we have examined the feasibility of exploiting a ‘mismatch’ of the type identified above: a set of traits that presumably enhance fitness of cane toads in their native range, but produce a vulnerability in tropical Australia. This ‘mismatch’ involves the morphology, behaviour and ecology of metamorph toads, the first terrestrial stage of the life cycle. After transforming from the tadpole stage, toad metamorphs show a series of traits not seen in native frogs: the toad metamorphs are small (usually < 1 g), diurnally active, and restricted to the open margins of their natal waterbody (Freeland & Kerin 1991). Large, highly aggressive and behaviourally dominant ants (Iridomyrmex spp.) are ubiquitous in tropical Australia, but no comparable functional group occurs in the cane toad’s home range. Metamorph toads tend to ignore the approach of predatory ants, and to rely on crypsis rather than attempting to escape if seized (Ward-Fear et al. 2009; Ward-Fear, Brown & Shine 2010). As a result, metamorph toads often are consumed by ants (Clerke & Williamson 1992; Ward-Fear et al. 2009, 2010).
Can we increase those rates of predation to provide an additional weapon to assist in the control of numbers of this unwelcome invasive species? In this paper we describe field experiments to increase the mortality rate of metamorph toads by manipulating the distribution (and thus, local densities) of predatory ants around waterbodies. This approach may offer a logistically feasible, inexpensive, low technology approach to complement other methods being developed to reduce recruitment rates of the invasive cane toad.
Materials and methods
The predator-prey system
Despite the large size of adult cane toads (up to 2 kg and 24 cm long), their metamorphs are small (0·07–0·15 g, 8–15 mm long) relative to those of other anurans (Werner 1986). The small size of the metamorphs (i) renders them vulnerable to cannibalism by larger, nocturnal conspecifics and thus favours diurnal activity (Freeland & Kerin 1991; Child, Phillips & Shine 2008), and (ii) increases their surface area to body ratio and thus desiccation rate, which in turn restricts them to the moist margins of natal ponds during the tropical dry-season (Child, Phillips & Shine 2008). Cane toads preferentially spawn in ponds surrounded by large expanses of open ground, rather than vegetation (Hagman & Shine 2006). In combination, these factors render toad metamorphs susceptible to predatory ants (Ward-Fear et al. 2009,2010).
Although many native Australian vertebrate predators cannot tolerate the toxins of the cane toad (e.g. Letnic et al. 2008), invertebrates tend to be less sensitive to bufadienolides than are many vertebrates (e.g. Toledo 2005). Dolichoderine ‘meat ants’ of the genus Iridomyrmex (purpureus group) can prey on cane toad metamorphs without ill effect (Clerke & Williamson 1992). Ants of this group occur across most of continental Australia (Andersen 2003), and hence have the potential to contribute to cane toad control over a broad area. Iridomyrmex species generally rely on scavenging, but frequently take live prey as well (Carrol & Janzen 1973; Clerke & Williamson 1992). They lack a sting, relying on mechanical force and pygideal secretions to subdue their victims (Shattuck 1999). Once captured, prey is carried back to the nest prior to consumption, typically along foraging trails (Mobbs et al. 1978). A single colony of I. reburrus (Shattuck 1999) may have several active nests spread over a wide area.
The study area
We studied ant-toad interactions on the Adelaide River floodplain 60 km east of Darwin, Northern Territory, in the Australian wet-dry tropics. Our work was conducted during the dry-season, the primary period of cane toad breeding in this area (M.R. Crossland, pers. comm.). Initially we selected six isolated waterbodies that contained toad tadpoles, or had been used as toad breeding sites in previous years. These waterbodies were small (100–1700 square metres surface area) ponds among agricultural (buffalo-grazing) land. Highly seasonal precipitation regimes in this area (> 75% of annual rainfall is monsoonal, from January to March: Brown & Shine 2007) result in rapid drying-out of these ponds, such that some were completely dry by the end of our experimental period (thus leaving three waterbodies around which baiting trials were feasible). Edges of the waterbodies were smooth and open, but soil cracks appeared as the mud dried out at the receding edge of the water. All sites had at least one meat ant colony within 10 m of the waterbody; single colonies typically monopolised the water’s edge at any one time. Each day the ants commenced foraging in the morning by leaving their nest to circumnavigate the waterbody.
Spatiotemporal overlap of ants with toads, and effective baiting period
To identify the most effective period in which to apply baits, we surveyed the spatiotemporal overlap of ants and toads, and the magnitude of ant response to baiting at three waterbodies (Farm Dam – 12°38′24·80″S/131°18′57″E, Lily – 12°34′39″S/131°19′05″E and Ropehead – 12°36′15 16″S/131°18′77″E). First, we established a 4 m transect between the nest entrance of a meat ant colony, and the edge of the waterbody. At each 0·5 m interval (starting at the nest and ending at the water’s edge) we counted the numbers of ants and toads present within 0·3 × 0·3 m quadrats at four times of day (08:00, 12:00, 16:00 and 20:00 h). We repeated this procedure the following day, after applying 20 g of catfood bait (see below) in the centre of the transect (i.e. 2 m from both water’s edge and nest).
Spatial manipulation of ant densities via baiting
We used the information gained in preliminary baiting trials (above) to examine effects of ant-baiting trials on metamorph toads at the three waterbodies described above. Outlines of 16·1 × 1 m quadrats were painted on the ground around each waterbody, comprising an inner ring of eight at the water’s edge and an outer ring of eight set 3 m back from the edge (directly behind the first). Based on pilot studies, we used 20 g of fish catfood (‘Ocean Platter’, by Whiskas – MasterFoods Australia, Wodonga, Victoria) as bait. Each bait (or a stone, in the control quadrats) was placed in a single mound in the centre of each quadrat, on a double layered 3 × 3 cm square of plastic food-wrap to prevent residual scent on the ground after the trial; all baits and controls were removed at the end of each day’s work. We alternated bait application among quadrats over two consecutive days at each waterbody, and repeated the trials two weeks later. Based on preliminary results, we ran trials between 08:00 and 11:00 h each morning, a period of high ant activity (Ward-Fear 2008; Ward-Fear et al. 2009, 2010). We made observations at 08:00, 08:15, 08:30, 09:00, 10:00 and 11:00 h. In each quadrat we counted the number of ants (in contact with the bait, and not in contact with the bait but on the surface of the quadrat); and the number of metamorph toads (on the ground surface, and within mud cracks). We recorded the numbers and locations of any attacks and fatalities of toads during 2-min periods.
Spatial manipulation of ant densities during toad recruitment events
Because densities of metamorphs were low during many of the baiting trials described above, we conducted additional trials at times when metamorphs were abundant. These trials ran for four consecutive days at each of the three waterbodies, with all quadrats ‘near’ to the waterbody (rather than split between ‘near’ and ‘far’, as above), and bait placed at the water’s edge (to maximise encounter rates between ants and toads). We added water to two ponds to prolong the period of inundation (simulating dry-season rainfall, which sometimes does occur: Taylor & Tulloch 1985), and released metamorphosing tadpoles into these waterbodies to standardize toad abundances. The extra water and toads resulted in conditions similar to those previously observed at these waterbodies (E. Cox and M.R. Crossland, pers. comm.).
We performed a two-way anova to compare mean densities of ants in baited vs. unbaited quadrats, and both ‘near’ and ‘far’ from the water’s edge. We analysed the changes in ant and toad density through time with repeated-measures anovas, using treatment as the factor and time as the repeated measure. We compared rates of attack by ants with a two-way anova, using treatment (baited vs. unbaited) and time of trial as the factors.
To examine potential effects of higher ant densities on other invertebrates, we added pitfall traps (containing 93% ethylene glycol) 20 cm from the treatment location in all quadrats, and changed them daily. We sorted the contents of these traps to order (or genus, for ants) and used contingency-table tests to compare baited and unbaited quadrats with respect to invertebrate abundance and species richness.
Ant predation on toads: assessing delayed effects and size-dependent vulnerability
Toads that appear to escape from ant attack might nonetheless be fatally injured. To gauge delayed mortality, we collected metamorphs that we saw escaping from ant attack in the field, as well as a control sample of metamorphs (not seen to be attacked by ants) taken concurrently from the same waterbody. We kept the young toads in the laboratory under favourable ambient conditions (28 °C and 45% humidity) and measured mortality over the following 24 h. Mortality rates were compared using chi-squared tests. To identify any size-dependent mortality, we measured snout-urostyle lengths (SUL) of random samples of live metamorphs in the trials conducted during toad recruitment events, to compare the size distribution of live toads with those killed. These distributions were compared with a repeated-measures anova using sample (attacked or random) as the factor and ‘day of collection’ as a repeated measure.
Effects of bait on behaviour of metamorph toads
Any effects of baiting on the rate that toads were attacked by ants might be influenced by reactions of toads to the presence of the bait. To determine whether metamorphs were attracted or repelled by the catfood baits, we placed 40 metamorph toads (body mass range 0·18–0·3 g) in individual enclosures (18 × 12 × 6 cm). Each enclosure had a 1 × 1 cm grid drawn onto the substratum, and contained a moistened paper towel. We introduced small trays that were empty (controls) or trays holding 20 g bait (treatment) into one end of the enclosure, and at intervals of 15, 30, 60, 120 and 180 min later we measured (over 1 min) the number of: (i) hops towards the tray; (ii) hops away from the tray; and (iii) whether metamorphs were sitting in the treatment half of the container. Data from these trials were analysed with a repeated-measures anova using treatment as the factor and time as the repeated measure. We used a one-way anova to analyse total numbers of movements within the enclosures and the number of times metamorphs were found in the treatment half.
Spatiotemporal overlap of ants with toads, and effective baiting period
Bait application induced a response from ants at all times except at night (20:00 h, Fig. 1). Ant densities increased at all baited locations, especially around the bait itself. The magnitudes of these increases were greatest during the morning (08:00 h). Toad densities were not affected by baiting, and were highest in the morning (08:00 h) and lowest at night (20:00 h). Thus, we concluded that baiting would be most effective during the morning hours.
Spatial manipulation of ant densities via baiting during toad recruitment events
In our preliminary analyses, the effects of adding bait were similar under the two baiting designs (quadrats near vs. far from water, vs. all near to water). Thus for ease of interpretation, we present the results based on combined data from both methods.
Ant densities were higher in baited than unbaited quadrats, with the magnitude of this difference changing over time (repeated-measures anova with ant variables as factors, time as the repeated measure: Fig. 2, Table S1 in Supporting information). Despite significant interaction terms, the main treatment effect (F1,318 = 77·3, P < 0·0001) is interpretable because ant densities were always higher in the baited quadrats (Fig. 2). The distance of a baited quadrat from the waterbody did not affect ant density (F1,318 = 0·0016, P < 0·59).
The number of toads present decreased over time more rapidly in baited quadrats than in unbaited quadrats (repeated-measures anova with toad variables as factors, time as the repeated measure: Fig. 2, and see Table S1 for statistical results). Quadrats further from the water’s edge contained fewer toads (Table S1). Our pitfall traps captured few invertebrates, with no significant difference in species richness (χ2 = 0·072, d.f. = 1, P = 0·79) or invertebrate abundance (t = 0·89, d.f. = 18, P = 0·39) between baited and unbaited quadrats.
Metamorph toads were attacked and killed by ants at higher rates in baited quadrats than in unbaited quadrats (mean attack rates 18·0 vs. 4·5; mean mortality rates 8·2 vs. 2·1 respectively; all P < 0·0001; Table S2 Supporting information). The proportion of attacks resulting in immediate mortality of the metamorph did not differ significantly between the two treatments.
During fieldwork we observed > 300 attacks by ants on cane toads, but no attacks on other vertebrates, and only nine attacks on invertebrates (two grasshoppers, three dragonflies, four centipedes). Overall, then, > 97% of all recorded attacks were on cane toads.
Ant predation on toads: delayed effects and size-dependent vulnerability
In the laboratory, mortality rates of controls (i.e. metamorph toads not seen to have been attacked by ants) were zero over the 24:00 h observation period, whereas 56% of the ‘escapees’ from ant attack were dead within an hour post-attack, and 88% died within 24 h. Mortality rates of apparent ‘escapees’ thus were higher than those of the control sample (at 1 h –χ2 = 31·19, d.f. = 1, P < 0·0001; at 24 h –χ2 = 63·84, d.f. = 1, P < 0·0001).
Metamorphs that were killed in the field by ants were smaller than those from the random sample (F1,69 = 60·5, P < 0·0001) and this difference between survivors and victims was consistent within each day (F2,69 = 3·13, P = 0·05) despite significant day-to-day variation in mean body sizes of the entire sample (F2,69 = 4·56, P = 0·014).
Effects of bait on the behaviour of metamorph toads
The addition of cat food to the experimental containers did not significantly affect any of the behaviours of metamorph toads that we assessed. Also, none of these patterns changed significantly over time during our experiments (d.f. = 1, all P > 0·7; see Table S3 Supporting information).
Our data from tropical waterbodies identify an agent that could be manipulated to reduce toad survival rates in the metamorph stage: native meat ants of the genus Iridomyrmex. These predators consume many metamorph toads, across most of the cane toad’s current geographic range within Australia (e.g. Clerke & Williamson 1992 for south-eastern Queensland). Even in brief (2 min) trials, 98% of metamorph toads encountered at least one ant in trials at high ant density, and 87% at low ant density. Most of these toadlets were attacked (84% at high ant density, 37% at low ant density) and killed (overall, of animals attacked, 82·5% were killed at high ant densities, 51% at low ant densities: Ward-Fear et al. 2009, 2010). Given the massive (and largely ineffective) expenditure on cane toad control (Shine et al. 2006), it is surprising that the role of meat ants in killing toads has failed to attract significant attention.
We can increase mortality rates of metamorph toads by modifying the spatial location of predatory ants, and longer-term monitoring of ‘escapee’ metamorphs showed that mortality rates were far higher than would be inferred from observing the encounter. Our results thus are encouraging for the prospects of using meat ants as one component of an integrated approach to reduce the abundance of cane toads in their Australian range. Below, we consider the biological processes involved, then their relevance for toad control.
Increasing ant densities enhanced toad vulnerability for three reasons (Ward-Fear et al. 2009, 2010). Firstly, higher ant densities increased encounter rates. Secondly, ant behaviour changes with density: at high ant densities, these predators are more likely to swarm onto a prey item (Ward-Fear et al. 2009, 2010). Thirdly, higher ant densities increase the body-size threshold at which metamorph toads shift from active escape tactics to the less effective tactic of crypsis (immobility; Ward-Fear et al. 2009, 2010). The end result is that higher ant densities kill more toads, and kill toads of a wider range of body sizes.
In the field, smaller metamorphs were killed more often than larger ones, because they respond less effectively to ant attack (Ward-Fear et al. 2009, 2010). Small body size at metamorphosis also renders the animals more vulnerable to lungworm parasites (Kelehear 2007) and desiccation (Child et al. 2008). Additionally, low toxin levels in metamorphs render them more vulnerable to predators (Hayes et al. 2009). Overall, the metamorph life stage is one in which these animals are at greatest risk. Such size-dependence in vulnerability is common among vertebrate populations (Caughley 1977; Nolan, Hill & Stoehr 1998; Reinhardt 2001).
Ideally, control efforts should use a multi-pronged approach (Zhang & Gu 1998; Liang & Zhang 2005). Manipulation of toad body sizes may be feasible, because larval toads metamorphose at smaller body sizes when exposed to pheromones of injured conspecifics (Hagman et al. 2009) and smaller metamorphs are more vulnerable to ant attack (Ward-Fear et al. 2009, 2010). Similarly, the attributes of waterbodies could be modified to concentrate anuran breeding (Hagman & Shine 2006), generating smaller size at metamorphosis in response to higher conspecific and heterospecific densities (Wilbur & Collins 1973; Crossland et al. 2008). However, meat ants do not provide a panacea for toad control. For example, they kill only the metamorphs, and are unlikely to be effective during wet-season recruitment events (when ants are less active, and the metamorph toads disperse from their natal waterbodies immediately: Child et al. 2008). Also, although intuition suggests that increased mortality of toads would reduce toad abundance, metamorph mortality rates may have little impact on toad recruitment because of density dependence in mortality schedules (e.g. Alford et al. 1995; Crossland et al. 2008; Pizzatto & Shine 2008; Bowcock, Brown & Shine 2009; Crossland, Alford & Shine 2009). Recruitment might even be increased by higher pre-recruitment mortality (Wilbur 1976; Millner & Whiting 1996). Field studies could evaluate whether or not elevating metamorph mortality rates affects total recruitment. Increased ant densities might provide food for adult toads (Bailey 1976), but diurnal meat ants are unlikely to be eaten by nocturnal adult toads.
Many attempts at biocontrol have inflicted collateral damage to native fauna (Snyder & Hise 1999); the cane toad itself is a good example (Low 1999). In the present case, however, the predator involved is already present (our only manipulation was to redirect foraging ants to an area virtually lacking other species except cane toads). Collateral impacts of ant-baiting appear to be minor, but further work is needed. For example, reduced densities of ants in unbaited areas may influence the fauna; or increased ant densities by the water’s edge may modify the behaviour of other species. Meat ants belong to the dominant dolichodorine group, which can competitively exclude ecologically similar ant species (Andersen 1990, 2003; Andersen & Patel 1994; Gibb & Hochuli 2003) and affect seed dispersal (Gibb & Hochuli 2003). Other habitats may contain diurnally active native taxa that would be affected by increases in ant density. Such issues warrant attention prior to implementing any toad-control methods based on increasing ant numbers.
In summary, our results demonstrate high rates of metamorph toad mortality, and identify a potential opportunity for biocontrol targeted at this vulnerable life stage. Clearly, the presence of meat ants across tropical Australia has not prevented cane toads from invading that region; but meat ants do kill many toads, and we can amplify the intensity of this naturally occurring predator-prey interaction to instigate even higher metamorph mortality rates. Thus, amplifying toad offtake by predatory ants might provide one component to a multi-faceted, ecologically-informed approach to reducing toad densities. Due to the spatiotemporal overlap between the invasive species and its predator (an overlap not shared by most native anurans: Ward-Fear et al. 2009, 2010), meat ant distributions can be manipulated to target areas with metamorphosing toads (Christian, Webb & Schultz 2003; Child et al. 2008; Ward-Fear 2008). The end result is an extraordinarily simple potential weapon to employ against invasive cane toads.
This study was made possible by the enthusiastic support of the staff at Beatrice Hill Farm, especially Eric Cox. Michael Crossland provided advice on experimental design and analysis, Miles Bruny provided logistical support, and Melanie Elphick helped with manuscript preparation. The Northern Territory Land Corporation provided critical infrastructural support. The Australian Research Council funded this work.