893 Negative Effects of Azteca Ants on the Distribution of the Termite Neocapritermes braziliensis in Central Amazonia by Pedro A. C. Lima Pequeno1* & Pauline O. Pantoja1 ABSTRACT Termites play important roles in tropical ecosystem functioning, and their evolutionary success has been linked to their defense mechanisms. However, microhabitat overlap with potential aggressors may constrain their distribu- tion and thus, their environmental impacts on an ecological timescale. We investigated a possible negative effect of abundant generalist ants (Azteca sp.) on the termite Neocapritermes braziliensis. Both taxa frequently build their nests attached to trees. We determined the densities of their active nests in 10 plots (250 x 10 m) systematically distributed over 5 km2 in central Amazonia, Brazil, and recorded their co-occurrence in individual trees. Using general- ized nonlinear modeling in a Bayesian framework, we found good support for a negative effect of Azteca’s nest density on N. braziliensis’. This effect conformed to a power law, and accounted for more than half of the variation in the termite’s nest density (r2 = 0.56). Additionally, of all counted N. bra- ziliensis mounds, only 1.08 percent was attached to trees also hosting Azteca. Such patterns may have arisen due to N. braziliensis’ inability to establish new nests within Azteca territories, or predation by ants on established colonies of the termite. We suggest that even non-strictly termitophagous ant species may have important impacts on termite populations and, consequently, on their roles in nutrient cycling and ecosystem engineering. Key Words: Bayesian inference; generalized nonlinear modeling ; popula- tion limitation; power law; social insect; species distribution. INTRODUCTION Population size is subject to different constraints, which is the basis for spatial variation in species abundance (Krebs 2002). There has been a long Instituto Nacional de Pesquisas da Amazônia (INPA), Programa de Pós-Graduação em Ecologia. Av. André Araújo 2936, Petrópolis, Caixa postal: 478. Manaus, AM, Brasil. CEP: 69011-970. *Corresponding author. Email: pacolipe@gmail.com 894 Sociobiolog y Vol. 59, No. 3, 2012 debate on whether populations are mainly driven by bottom-up (resources) or top-down forces (predation) (Hairston et al. 1960; Hunter & Price 1992; Polis & Strong 1996; Borer et al. 2006). It has been recently argued that under most circumstances animal populations are limited by food quality or quantity (White 2008). However, while resource availability may set an upper limit to prey populations, predation may prevent these populations from reaching carrying capacity (McIntosh et al. 2005), especially under the presence of generalist predators ( Jiang & Morin 2005). This might be reinforced when a predator is both generalist and more abundant than a particular prey, so that the predator may limit this prey’s populations without being limited in turn. Nonetheless, it is likely that any such effect depends on the spatial scale under consideration (Levin 1992). Ants (Hymenoptera: Formicidae) and termites (Blattaria: Isoptera) are well known for their convergent social evolution, but they are also closely connected by trophic interactions. In fact, ants are generally more abundant than termites and, although few ant species are strictly termitophagous, most ants prey upon termites if allowed to (Hölldobler & Wilson 1990). This, coupled with the diverse array of termite defense strategies, has been taken as evidence for an evolutionary arms race between both taxa (Noirot & Darlington 2000). Nevertheless, current ant-termite interactions remain poorly explored, especially in tropical rainforests, where they are remark- ably abundant and diverse (Scholtz et al. 2008). Given the acknowledged roles of social insects in the functioning of these ecosystems, understanding interactions between these groups can contribute to the conservation of the environmental services they provide (Chapman & Bourke 2001). In the Amazon rainforest, nests of several ant and termite species are con- spicuous. Namely, arboreal ants of the genus Azteca (Formicidae: Dolichoderi- nae) are dominant in some areas. They build large carton, stalactite-like nests on trees and attend coccid hemipterans for honeydew (Blüthgen et al. 2000), although they can be predatory too (Dejean et al. 2009). Similarly, Neocapri- termes braziliensis (Termitidae: Termitinae) is a common, although poorly studied, mound-building termite from the same areas. It feeds on rotting wood, and its mounds are often attached to the base of tree trunks (Constantino 1992). Therefore, encounters between N. braziliensis and Azteca are likely. 895 Pequeno, P.A.C.L. & P.O. Pantoja — Effects of Azteca Ants on Termites Being a cryptic forager, N. braziliensis should be inaccessible to predators most of the time and the combination between cryptic behavior and the physical structure of the mound may allow the termite to avoid predation by ants (Buczkowski & Bennett 2008). However, termite nests are potential targets for predation because they represent concentrated food resources. Also, Neocapritermes species have rather low soldier/worker ratios (around 1:100; Krishna & Araujo 1968), a trait likely to affect termite susceptibility to predation by ants (Wells & Henderson 1993). Because Azteca and N. braziliensis overlap in their nest substrates and can be locally abundant, their point interactions may result in agonistic patterns at larger scales. Here, we investigate whether N. braziliensis and Azteca are negatively associated at the levels of individual trees and nest densities across the landscape. MATERIAL AND METHODS The study was undertaken in a lowland rainforest area of 10 km2 between the Purus and Madeira rivers, Amazonas State, Brazil (3°41'9.85"S, 60°19'9.65" W). Local topography is weakly undulated, and much of the substrate is waterlogged in the wet season. Fieldwork was carried out from May to June 2009. We surveyed ten plots of 250 x 10 m, distributed along two parallel 5-km trails 1 km apart from each other; both trails contained five sampling plots also 1 km apart from each other. Plots were oriented along topographic contour lines so as to reduce soil variation (as well as its correlates) within sampling units, thus maximizing between-plot variation (Costa & Magnusson 2010). In each plot, we counted active nests of both taxa, which were recognized by their typical architecture and residents. We also recorded whether each nest co-occurred with a nest of other taxon in the same substrate. Samples were taken for taxonomic confirmation and deposited in the Entomological Collection of the National Institute for Amazonia Research (INPA), Manaus, Brazil. Since all surveyed ants were equivalent in terms of their nesting and foraging modes, they were pooled for analysis. Following graphical inspection of the data, we modeled N. braziliensis nest density (Y) as a power function of Azteca nest density (X), assuming a log link (i.e. log(Y) = β 0 .Xβ1) and Poisson distributed errors given the count 896 Sociobiolog y Vol. 59, No. 3, 2012 response. The model was fit in a Bayesian framework, which allows direct probabilistic statements about parameters and produces exact results even for small sample sizes (Link & Barker 2010). We used uninformative priors (normally distributed with zero mean and standard deviation of 1000) so that inference was mainly driven by the data themselves (i.e. likelihood). Three parallel Markov Chain Monte Carlo simulations (MCMC) were iterated 300,000 times each using the Gibbs sampler, starting from random values and with a thinning rate of 10. The first 1000 realizations of each chain were discarded as burn-in. Chain convergence was checked with trace plots, autocorrelation plots and the Brooks-Rubin-Gelman statistic (BRG). Once validated, chains were pooled for inference (n = 89,700), and the means of the resulting posterior distributions were used as point estimates for model parameters. We assessed model adequacy by plotting residuals against fitted values and by squaring the correlation between model predictions and the observed response (r2) (Zheng & Agresti 2000). We refrained from formal inference regarding nest substrate overlap since our sample was biased towards the presence of nests. Modeling was carried out in R 2.10.1 (R Core Development Team 2009) coupled to WinBUGS 1.4 (Spigelhalter et al. 2003), with the R package “R2WinBUGS” (Sturtz et al. 2005) as interface between the two pro- grams. RESULTS We counted 30 nests of Azteca and 54 of N. braziliensis. These values cor- respond to local densities of 12 nests/ha and 21.6 nests/ha, respectively. The model for the relation between nest densities was fit as log(Y) = 2.35 . X-0.49 (Fig. 1) and accounted for more than half the variation in the response (r2 = 0.56). MCMC diagnostics indicated chain convergence (BRG = 1) and no autocorrelation, whereas residual analysis revealed no clear patterns, thus validating the model. The 95% credible interval (i.e. within which there is a 95% chance the true parameter value is found) for the constant β 0 ranged from 1.98 to 2.69, and from -0.25 to -0.76 for the exponent β 1 . Also, only two nests of N. braziliensis (1.08% of this species’ total) were attached to trees hosting Azteca nests. 897 Pequeno, P.A.C.L. & P.O. Pantoja — Effects of Azteca Ants on Termites DISCUSSION This study provides evidence that arboreal Azteca ants act as limiting factors of N. braziliensis in the study area. Not only nests appear to strongly segregate among substrate trees, but there is also a negative, nonlinear relationship be- tween nest densities across the landscape. Presumably, these patterns emerge from ant-termite behavioral interactions at the level of individuals. Hence the observed patterns suggest a scalable negative effect. We did not observe aggression by Azteca against undisturbed termites during the present study, though we did observe the ants attacking both N. braziliensis workers and soldiers that were removed from their nests. Impacts of ants on termites have been recorded several times in the field. For instance, Jutsum et al. (1981) detected a negative association between Fig.1. Mean response (solid line) of Neocapritermes braziliensis’ nest density to Azteca’s, estimated as log(Y) = 2.35 . X-0.49. Dotted lines are 95% credible limits. 898 Sociobiolog y Vol. 59, No. 3, 2012 the presence of Azteca sp. and Nasutitermes corniger (syn. N. costalis) on trees in Trinidad; Leponce et al. (1999) argued that the presence of the ant Crematogaster irritabilis was a limiting factor for arboreal termites in New Guinea, causing a two-fold decrease in their local abundance; Gonçalves et al. (2005) recorded a strong negative effect of predatory ants on arboreal termite activity; and Dejean et al. (2007a) found that the nest densities of arboreal Nasutitermes were about three times lower in plots occupied by the invasive ant Pheidole megacephala than in those ones without this species. We note, however, that Azteca has neither the morphological nor behavioral traits known from specialized, termite-hunting species (e.g. Leal & Oliveira 1995). Thus, it is improbable that it normally raids termite nests. Nonetheless, most ant taxa are known to prey on exposed termites in central Amazonia (Bandeira 1979), and our results suggest that even ant species that do not specialize in termitophag y (such as Azteca spp.) may have strong negative ef- fects on natural termite populations. The fact that this observation involves a termite species presumably well protected by the mound structure such as N. braziliensis, somewhat contradicts previous suggestions (Mill 1982; Buczkowski & Bennett 2008). We propose two hypotheses to account for the observed patterns. Azteca could affect N. braziliensis populations by inhibiting newly mated reproduc- tive pairs from founding new colonies within its territories. Different colonies of the same termite species can liberate their alates in synchronized events, and these attract many opportunist predators – including ants (Mill 1983). In this scenario, Azteca would prey upon N. braziliensis only sporadically, due to random encounters during swarming. This would suffice to produce segregation between nesting substrates. If the probability of encounters were a function of nest density, then higher densities of Azteca nests would also reduce N. braziliensis’ nest founding success and thus, the termite’s nest densities themselves. Our second hypothesis is that Azteca could shift its foraging at least par- tially towards termites if they become locally abundant. Arboreal ants exhibit strong preference for protein-rich baits in comparison to litter-dwelling ones, suggesting that their colonies are limited by nitrogen availability (Kaspari & Yanoviac 2001). Thus, Azteca may take advantage of the protein supply repre- sented by N. braziliensis colonies. Termite mounds can be cracked by rainfall, 899 Pequeno, P.A.C.L. & P.O. Pantoja — Effects of Azteca Ants on Termites tree fall or specialized predators (e.g. anteaters, armadillos, nest-raiding ants), which may allow generalist ants to raid nests. These two hypotheses are not mutually exclusive, though. Why would N. braziliensis nest density relationship with Azteca’s follow a negative power function? Ants are territorial insects and thus are expected to compete for space (Parr & Gibb 2010). For instance, experimental work has revealed that higher nest densities intensify intraspecific conflict in the fire ant Solenopsis invicta (Adams & Tschinkel 2001). Accordingly, a shift in the strength of interactions from inter- to intraspecific ones as ant nest density increases may dampen the predatory impact of ants on termites and thus, the rate of decline in termite abundance. Dominant arboreal ants engage in the defense of well-marked territories against conspecifics as well as other ant species, which can lead to the establish- ment of unoccupied zones or “no ant’s lands” between adjacent, competing colonies (Dejean et al. 2007b). Hence it may also be that these areas provide termites with refuges from the ant “crossfire” (e.g. by going unnoticed during earlier, more vulnerable stages of colony development). In addition, dominant ant species only have absolute control of the immediate nest neighborhood; there is plenty of within-territory space for opportunistic foraging (Andersen 2008) and, perhaps, termite nest founding and growth. The present study suggests a role for non-strictly termitophagous ants as effective limiting factors of natural termite populations at multiple scales. It has been observed that temporal gaps in termite activity can be a function of predation risk by ants (Korb & Linsenmair 2002; DeSouza et al. 2009). Thus, the limiting effects of ants on termite populations may further impair termite roles in ecosystem functioning such as nutrient cycling and ecosystem engineering. However, further study of the mechanisms creating the observed patterns is necessary. We encourage the application of Bayesian inference in future studies, as they could build on informative priors (i.e. posterior distri- butions as estimated here) to strengthen the robustness of their conclusions (McCarthy & Masters 2005). ACKNOWLEDGMENTS We thank Albertina Lima and Fabricio Baccaro for their support during field work. F. Baccaro also helped in identifying the ants and improving the 900 Sociobiolog y Vol. 59, No. 3, 2012 manuscript. William Magnusson provided valuable suggestions on an earlier version of the manuscript. P. A. C. Lima Pequeno received a POSGRAD scholarhip from the Fundação de Amparo à Pesquisa do Estado do Amazo- nas (FAPEAM) during this study, and P. O. Pantoja received a scholarship from the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES). REFERENCES Adams, E.S., & W.R. Tschinkel. 2001. Mechanisms of population regulation in the fire ant Solenopsis invicta: An experimental study. Journal of Animal Ecolog y 70(3):355– 369. Andersen, A.N. 2008. Not enough niches: non-equilibrial processes promoting species coexistence in diverse ant communities. Austral Ecolog y 33(2): 211–220. Bandeira, A.G. 1979. Ecologia de cupins (Insecta: Isoptera) da Amazônia Central: efeitos do desmatamento sobre as populações. Acta Amazonica 9: 481–499. Blüthgen, N., M. Verhaagh, W. Goitía, K. Jaffé, W. Morawetz & W. Barthlott. 2000. How plants shape the ant community in the Amazonian rainforest canopy: the key role of extrafloral nectarines and homopteran honeydew. Oecologia 125(2):229–240. Borer, E.T., B.S. Halpern & E.W. Seabloom. 2006. Asymmetry in community regulation: effects of predators and productivity. Ecolog y 87(11):2813–2820. Buczkowski, G., & G. Bennett. 2008. Behavioral interactions between Aphaenogaster rudis (Hymenoptera: Formicidae) and Reticulitermes flavipes (Isoptera: Rhinotermitidae): the importance of physical barriers. Journal of Insect Behavior 21(4):296–305. Chapman, R.E. & A.F.G. Bourke. 2001. The inluence of sociality on the conservation biolog y of social insects. Ecolog y Letters 4:650–662. Constantino, R. 1992. Abundance and diversity of termites (Insecta: Isoptera) in two sites of primary rain forest in Brazilian Amazonia. Biotropica 24(3):420–430. Costa, F.C.C., & W.E. Magnusson. 2010. The need for large-scale, integrated studies of biodiversity – the experience of the Program for Biodiversity Research in Brazilian Amazonia. Natureza & Conservação 8:3-12. Dejean, A., M. Kenne, & C.S. Moreau. 2007a. Predatory abilities favour the success of the invasive ant Pheidole megacephala in an introduced area. Journal of Applied Entomolog y 131(9-10):625–629. Dejean, A., B. Corbara, J. Orivel, & M. Leponce. 2007b. Rainforest canopy ants: The implications of territoriality and predatory behavior. Functional Ecosystems and Communities 1:105–120. Dejean, A., J. Grangier, C. Leroy, & J. Orivel. 2009. Predation and aggressiveness in host plant protection: a generalization using ants from the genus Azteca. Naturwissenschaften 96(1):57–63. 901 Pequeno, P.A.C.L. & P.O. Pantoja — Effects of Azteca Ants on Termites DeSouza, O., A.P.A. Araújo & J. Reis Jr.. 2009. Trophic controls delaying foraging by termites: reasons for the ground being brown? Bulletin of Entomological Research 99:603–609. Gonçalves, T.T., R. Reis Jr., O. DeSouza, & S.P. Ribeiro. 2005. Predation and interference competition between ants (Hymenoptera: Formicidae) and arboreal termites (Isoptera: Termitidae). Sociobiolog y 46(2):409–419. Hairston, N.G., F.E. Smith & L.B. Slobodkin. 1960. Community structure, population control and competition. The American Naturalist 94(879):421–425. Hölldobler, B. & E.O. Wilson. 1990. The ants. Massachusets: Harvard University Press 733 pp. Hunter, M.D. & P.W. Price. 1992. Playing chutes and ladders—heterogeneity and the relative roles of bottom-up and top-down forces in natural communities. Ecolog y 73(3):724–732. Jiang, L. & P.J. Morin. 2005. Predator diet breadth influences the relative importance of bottom-up and top-down control of prey biomass and diversity. The American Naturalist 165(3):350–363. Jutsum, A.R., J.M. Cherrett, & M. Fisher. 1981. Interactions between the fauna of citrus trees in Trinidad and the ants Atta cephalotes and Azteca sp. Journal of Applied Ecolog y 18(1):187–195. Kaspari, M. & S.P. Yanoviak. 2001. Bait use in tropical litter and canopy ants – evidence of differences in nutrient limitation. Biotropica 33(1):207–211. Korb, J. & K.E. Linsenmair. 2002. Evaluation of predation risk in the collectively foraging termite Macrotermes bellicosus. Insectes Sociaux 49(3): 264–269. Krebs, C. 2002. Beyond population regulation and limitation. Wildlife Research 29(1):1– 10. Krishna, K. & R.L. Araujo. 1968. A revision of the neotropical genus Neocapritermes (Isoptera, Termitidae, Nasutitermitinae). Bulletin of the American Museum of Natural History 138:84–138. Leal, I.R. & O.S. Oliveira. 1995. Behavioral ecolog y of the neotropical termite-hunting ant Pachycondyla (=Termitopone) marginata: colony founding, group-raiding and migratory patterns. Behavioral Ecolog y and Sociobiolog y 37(6):373–383. Leponce, M., Y. Roisin & J.M. Pasteels. 1999. Community interactions between ants and arboreal-nesting termites in New Guinea coconut plantations. Insectes Sociaux 46(2):126–130. Levin, S.E. 1992. The problem of pattern and scale in ecolog y: The Robert H. MacArthur Award Lecture. Ecolog y 73:1943–1967. Link, W.A. & R.J. Barker. 2010. Bayesian inference with ecological applications. London: Academic Press 339 pp. McCarthy, M.A. & P. Masters. 2005. Profiting from prior information in Bayesian analyses of ecological data. Journal of Applied Ecolog y. 42(6):1012–1019. 902 Sociobiolog y Vol. 59, No. 3, 2012 McIntosh, A.R., H.S. Greig, S.A. McMurtrie, P. Nyström & M.J. Winterbourn. 2005. Top- down and bottom-up influences on populations of a stream detritivore. Freshwater Biolog y 50(7):1206–1218. Mill, A.E. 1982. Faunal studies on termites (Isoptera) and observations on their ant predators (Hymenoptera: Formicidae) in the Amazon basin. Revista Brasileira de Entomologia 26:253–260. Mill, A.E. 1983. Observations on Brazilian termite alate swarms and some structures used in the dispersal of reproductives (Isoptera: Termitidae). Journal of Natural History 17(3):309–320. Noirot, C. & J.P.E.C. Darlington. 2000. Termite nests: architecture, regulation and defense. In: Abe, T., D.E. Bignell & M. Higashi (ed.). Termites: evolution, sociality, symbioses, ecolog y. Kluwer Academic Publishers, Dordrecht, The Netherlands: 121-139. Parr, C.L. & H. Gibb. 2010. Competition and the role of dominant ants. In: Lach, L., C.L. Parr & K.L. Abbott (ed.). Ant ecolog y. Oxford University Press, New York, USA: 77-96. Polis, G.A. & D.R. Strong. 1996. Food web complexity and community dynamics. The American Naturalist 147(5):813–846. R Core Development Team. 2009. R: A language and environment for statistical computing. R Foundation for Statistical Computing. Vienna, Austria. White, T.C.R. 2008. The role of food, weather and climate in limiting the abundance of animals. Biological Reviews 83(3):227–248. Scholtz, O.I., N. MacLeod & P. Eggleton. 2008. Termite soldier defence strategies: a reassessment of Prestwich’s classification and an examination of the evolution of defence morphology using extended eigenshape analyses of head morphology. Zoological Journal of the Linnean Society 153(4):631–650. Spiegelhalter, D., A. Thomas, N. Best & D. Lunn. 2003. WinBUGS User Manual. Version 1.4. Technical report, Medical Research Council Biostatistcs Unit, Cambridge. Sturtz, S., U. Ligges & A. Gelman. 2005. R2WinBUGS: A package for running WinBUGS from R. Journal of Statistical Software 12:1–16. Wells, J.D. & G. Henderson. 1993. Fire ant predation on native and introduced subterranean termites in the laboratory: effects of high soldier number in Coptotermes formosanus. Ecological Entomolog y 18(3):270–274. Zheng, B. & A. Agresti. 2000. Summarizing the predictive power of a generalized linear model. Statistics in Medicine 19(13):1771–1781.