Interactions between native diversity and invasive species can be more complex than is currently understood. Invasive ant species often substantially reduce diversity in the native ants diversity that act as natural control agents for pest insects. In Indonesia (on the island of Sulawesi), the third largest cacao producer worldwide, we show that a predatory endemic toad (Ingerophrynus celebensis) controls invasive ant (Anoplolepis gracilipes) abundance, and positively affects native ant diversity. We call this the invasive-naivety effect (an opposite of enemy release), whereby alien species may not harbour anti-predatory defences against a novel native predator. A positive effect of the toads on native ants may facilitate their predation on insect vectors of cacao diseases. Hence, toads may increase crop yield, but further research is needed on this aspect. Ironically, amphibians are globally the most threatened vertebrate class and are strongly impacted by the conversion of rainforest to cacao plantations in Sulawesi. It is, therefore, crucial to manage cacao plantations to maintain these endemic toads, as they may provide critical ecosystem services, such as invasion resistance and preservation of native insect diversity.
Cacao is the largest international trade commodity after petroleum and coffee, and currently covers 8 million ha in the tropics . Cacao plantations, despite their structural similarities to natural tropical forests, harbour reduced biodiversity, especially when shade trees are removed to maximize crop yields [2,3]. The effects of land-use change can be further intensified when these agricultural areas are subsequently colonized by invasive species, particularly on remote islands with high endemism [4,5]. Structural simplification in plantations can also compromise the ecosystem services provided by native biodiversity. Natural ant diversity, for example, has been shown to regulate insect vectors of cacao pests and decrease the risk of disease outbreaks . Introductions of the yellow crazy ant (Anoplolepis gracilipes) are known to affect native biodiversity and ecosystem processes on islands [7,8]. As a result, A. gracilipes is ranked among the top 100 invasive species worldwide .
A ubiquitous (yet expensive and often unsuccessful) approach to controlling invasive ant species in poor developing countries is the heavy application of pesticides, which can also damage native biodiversity and human health . Theoretically, native diversity could provide free services by controlling invasive ants and mitigating pesticide use . This requires that native predators become effective predators of invasive ants (figure 1). Further, the native predator should be abundant, ground-living and, ideally, an ant specialist. The common celebes toad (Ingerophrynus celebensis) and a skink (Eutropis grandis) are the most abundant ground-living vertebrates in the cacao plantations of Sulawesi . To assess their roles as invasive ant control agents, we first determined dietary preferences of these two species and then experimentally excluded ant-predating specialists (the toads) from cacao plantations. We tested the hypothesis that an endemic amphibian species can control an invasive ant and thereby alleviate its impact on natural ant diversity.
2. Material and methods
(a) Amphibian, reptile and ant sampling
To predict toad and lizard abundances by different ant species abundances and richness, we collected data from 43 cacao plots (40×40 m; 21 cacao agroforest and 22 cacao monoculture plantations) around the Lore Lindu National Park in Central Sulawesi (Indonesia). Amphibians and reptiles were sampled three times during day and night with visual and acoustic encounter surveys . Ants were sampled with observation plates (10 on the ground and 10 in the trees) on each plantation, baited with tuna and sugar solution. For ant recruitment rates (hereafter we refer to ‘ant abundance’ for clarity) and species richness, we counted the number of ant individuals and species on each plate every 15 min for 1 h. We used the mean maximum number of all plates as a measure of ant abundances per plot. For ant diversity, we used the total number of species encountered in the plot .
(b) Biomass analysis: faeces and reference data from the cacao plantations
We collected faeces samples of amphibians and reptiles between December 2007 and August 2009, and analysed arthropod species composition. All individuals captured in the 43 plots were kept in plastic containers for 12 h and faeces were collected. We identified the prey items found in faeces samples to taxonomic order whenever identifiable species parts were available, and where possible we quantified individual numbers (e.g. counting the number of heads, thoraxes etc.). We equated hymenopterans with ants, because we mainly identified ant remnants in the faeces; thus, we assume that bees and wasps are less likely prey items for these lizard and toad species. We used five Barber traps randomly placed within a plot equipped with a 1 : 1 ethylene : glycol solution. The traps were emptied twice (once every four weeks) over a period of two months. We identified the samples to taxonomic order and measured dry weight. The summed biomass of all traps within a plot was used for standardization of the faeces biomass.
(c) Toad exclosure
The toads migrate through the plantations to form calling choruses in the rice fields roughly every three months. They stay in the plantations for an average of one week, where their abundance can reach up to 300 toads per hectare (T. C. Wanger & I. Motzke 2008, personal observation). In contrast to the control plots, we expected that by excluding the toads from treatment plots, Anoplolepis abundance would increase owing to a lack of toad predation. If this is a long-lasting effect, the removal of the exclosure would not change ant abundance patterns when few or no toads are present in the plantations. Therefore, exclosures were removed.
For the fence exclosure experiment, 14 plots (10×10 m; for environmental details, see electronic supplementary material, table S1 and figures S1 and S2) with Anoplolepis present were sampled four times in cacao agroforests. Ant abundances were determined once in the evening and morning over 7 days; on the first day before an exclosure fence had been established, the third and fifth day after the fence had been established and on the seventh day after the fence had been removed. Seven plots were randomly assigned to treatment and control groups. While the control plots were only disturbed on the edges for all sampling sessions, we built 50-cm-tall exclosures of 3 mm-wide plastic mesh on the treatment plots. Fences and exclosures were maintained every day and checked every night for toads inside the fence. In all plots and sampling sessions, ant diversity and Anoplolepis abundance—as the only invasive species present—were determined with four sampling plates per plot, baited with tuna and sugar solution . We only used four plates because increasing the number of plates may have resulted in unrealistic abundances owing to overabundant food sources. Time frames were chosen so that the evening sample was conducted before the toads moved through the experimental area and the morning sampling was after the toads had finished feeding in the area. For the analysis, we pooled abundance data from morning and evening sampling sessions for higher robustness of samples (e.g. climatic variation between morning and evening sessions may potentially lead to changes in ant abundances).
We sampled the plots for one week to minimize confounding climatic effects in the field rather than accounting for it in the models with limited sample size. Only 14 sites were sampled because each site had to fulfil a strict set of requirements for a robust sampling design: (i) toads had not moved through the cacao plot prior the first sampling session (all candidate sites were checked every night over a three-month period), (ii) toads were moving through the plots until the third sampling session was completed, and (iii) climatic conditions were suitable and similar for the ants (i.e. no rain or wet soil) to minimize confounding climatic effects on the results.
We used Bayesian linear modelling with multi-model inference to determine whether individual ant species abundances and general fossorial ant diversity were drivers for toad and lizard abundance. We defined three a priori main hypotheses that were challenged in a Bayesian multi-model inference approach: (i) cumulative abundance of large ants (i.e. A. gracilipes, Paratrechina longicornis, Pheidologeton sp., Philidris sp.), (ii) individual large ant abundances or (iii) ground-dwelling ant species richness drive abundance patterns in the most common amphibian and reptile species. For an introduction of Bayesian model evaluation, see  and the supplementary material of .
The exclosure experiment and its impact on ant diversity were analysed with a Bayesian repeated-measures ANOVA, where the first sampling session was specified as the reference level, and site was coded as a random effect to account for spatial differences not accounted for in the fixed effects. The treatment effect was modelled as an interaction with the sampling session, because we were interested in the exclosure effects over time and not a treatment effect per se. To reduce the predictors in the model, we used the relative abundance and relative native ant species richness (i.e. difference between treatment and control plots) as response variables.
Dietary analyses revealed that ants dominated arthropod biomass (74%; standardized by plantation arthropod biomass) in the faeces samples of the toad. In the skink faeces, ants played a minor role and prey items were evenly distributed between arthropod orders (electronic supplementary material, figure S3). A comparison of arthropod biomass removal between cacao agroforest and monoculture plantations revealed that I. celebensis had the highest impact on ant biomass (i.e. 45%) in both habitats. Eutropis grandis had the greatest impact on orthopterans and dermapterans (i.e. 26 and 65% in cacao agroforest and monocultures, respectively; electronic supplementary material, figure S4). Based on behavioural observations of toads feeding on ants (see also ), we are confident that results from the faecal analysis are not strongly biased (i.e. soft-bodied prey items are unlikely to be fully digested and thus overlooked).
Bayesian regression modelling showed that abundance of Anoplolepis ants was the strongest predictor of toad abundance; the model explained 19.4 per cent of the deviance in the data and was greater than 1 000 000 more likely than the null relationship (electronic supplementary material, table S2). Skink abundance was predicted by abundance of large ant species in general, but the relationship was weaker, with 7.7 per cent deviance explained. These results suggest that—as an Anoplolepis predator—the toads choose plantations with higher Anoplolepis abundance, and are abundant in cacao plantations during regular migration to their breeding grounds in rice fields at night.
A Bayesian repeated-measures ANOVA revealed a treatment (i.e. toad exclosure) effect over time relative to the reference sample (i.e. the first sampling session (session 1) before the exclosure fence was built; figure 2). We found measurable effects at all four sample sessions; the credibility intervals do not overlap with zero. While the relationship between treatment and control plots at the reference sample is negative (session 1: −11.9 (−22.5 to −1.2)), this relationship inverts over time (session 2: 3.5 (0.6–6.9)) and remains positive even after the fence is removed in the last sampling session (as suggested by the overlapping credibility intervals in the last two sessions, sessions 3 and 4: 33.6 (26.2–41.1)) and 28.7 (21.7–35.9), respectively). These effects were not confounded by the presence of the skink, because it is diurnal and not specialized on ants. In contrast, toads forage most actively at midnight, because their calling period has ended and their energy requirements are at their highest (T. C. Wanger & I. Motzke 2008, personal observation; ). Anoplolepis activity also peaked around midnight, but was constant throughout the day (electronic supplementary material, figure S5).
We found that overall mean ant diversity did not measurably differ between treatment and control plots in the reference sampling session, but decreased over time (0.8 (−1.8 to 3.2); −1.5 (−4.7 to 1.8); −2.1 (−5.1 to 1.1); −2.9 (−6.0 to 0.45); figure 3). As the credibility intervals of all sampling sessions overlap with zero, we found a strong trend but no effect of the manipulation (and, hence, Anoplolepis abundance) on ant diversity over time.
Island endemics are often claimed to be inferior competitors to invasive species, because they have been released from selective pressures . We show that an island endemic toad (I. celebensis), as a predator, negatively affects the noxious invasive ant A. gracilipes, and this, in turn, mitigates the impact of A. gracilipes on natural ant diversity. Thus, in contrast to an enemy-release effect, this is an invasive-naivety effect: the alien prey species may not harbour anti-predatory defence against the novel endemic predator. This effect may be widespread and so offer resistance to biotic invasions.
For an invasive-naivety effect to occur, the native species must possess certain defensive and dietary traits. The invasive success of the ant A. gracilipes is grounded in its tendency to aggregate in high densities, aggressiveness and chemicals used for attack [18,19]. Bufonids have a tough skin capable of producing potent toxins , and amphibian skin can be distasteful for ants . In addition, toads are sit-and-wait predators, sometimes exclusively feeding on ants . As endemic toads probably have been ant specialists before the invasion occurred, their prey search image may be ‘pre-set’ for an Anoplolepis type of prey. Density-dependent predation of the toads on the abundant invasive ants may be a likely explanation of the reduction of A. gracilipes abundance. In addition, there may be nutritious preferences for the invasive to the native ants such as essential amino acid composition. A preference for A. gracilipes may lead to reduction in their abundance owing to toad predation. Based on skin defences and potential predefined feeding preference of many toads, there may be a general potential for bufonids as biocontrol agents; this is an area in need of further research.
Intuitively, toads can only provide useful ecosystem services if they are abundant; yet (endemic) amphibians are the most threatened vertebrates on the planet . If toad populations are reduced as a result of land-use change and pesticide use [23,24], our results suggest that A. gracilipes can flourish as an ‘ecosystem transformer’ (figure 1). In Sulawesi, rapid expansion of these ants drastically depletes natural ant diversity . This may have a real economic impact because native ants in Southeast Asia have been shown to provide various ecosystem functions, including biocontrol of insect-mediated cacao pests (see table 1 in ). On other islands, A. gracilipes affects litter decomposition and nutrient cycling , trophic cascades  and even the tourism industry (via displacement of a bird species ). These negative impacts may, however, take several decades to unfold fully; on Christmas Island, yellow crazy ant populations exploded 60 years after its introduction .
The economic magnitude of the toad's biocontrol service in cacao agroforestry systems is, however, conditional on parameters affecting cacao yield and Anoplolepis's effectiveness as a predator of cacao pest insect vectors. In Sulawesi, both pest and disease occurrence are important determinants of cacao yield . Thus, the less effective Anoplolepis is as a predator on disease vectors relative to the native ant species, intuitively more economical it is to have the toads in cacao plantations. It is possible that the time period to detect an effect of invasive ants on the native ant species in our study may be too short. However, it has previously been shown that A. gracilipes depletes native ant diversity in Sulawesi's cacao plantations ; thus, we believe that our results indicate effects that will be evident over the long term.
Clearly, then, it is important to maintain toads and other amphibian populations in cacao plantations to sustain ecosystem services in the future. Plantation management for leaf litter cover and maintenance of intact canopy structure to buffer against temperature fluctuations will benefit amphibian diversity . This will also facilitate toad abundance in cacao agroforest and, hence, translate into ant biomass removal. In contrast, A. gracilipes invasions are facilitated by thinned canopy, which alters microclimatic conditions in cacao plantations and enhances the frequency of cacao disease outbreaks . It is through the beneficial effect for the toads in cacao agroforest that biomass removal of ants will increase. Managed sensibly, native biodiversity offers great potential to provide invasion resistance  and probably contributes to sustainable, pesticide-reduced crop production.
We thank our counterparts Shahabuddin Saleh and Damayanti Buchori. Akbar Saro helped in the field. RISTEK and LIPI issued the research permits. Research was funded by the German Science Foundation (Deutsche Forschungsgemeinschaft). T.C.W. is supported by an Endeavour International Postgraduate Research scholarship and a University of Adelaide scholarship. I.M. and A.C.W. are supported by a German Academic Exchange Service (DAAD) scholarship.
- Received July 16, 2010.
- Accepted August 20, 2010.
- © 2010 The Royal Society