ECOLOGY, POPULATION BIOLOGY & ANIMAL BEHAVIOR ANIMAL CELLS AND SYSTEMS 2018, VOL. 22, NO. 4, 267–272 https://doi.org/10.1080/19768354.2018.1497708 Directional raids by army ants as an adaption to patchily distributed food: a simulation model a b,c a,d a,e Woncheol Song , Ho-Young Kim , Sang-Im Lee and Piotr G. Jablonski a b School of Biological Sciences, Seoul National University, Seoul, South Korea; Department of Mechanical and Aerospace Engineering, Seoul c d National University, Seoul, South Korea; Institute of Advanced Machines and Design, Seoul National University, Seoul, South Korea; School of Undergraduate Studies, Daegu-Gyeongbuk Institute of Science and Technology, Daegu, South Korea; Museum and Institute of Zoology, Polish Academy of Sciences, Wilcza, Warsaw, Poland ABSTRACT ARTICLE HISTORY Received 28 May 2018 A typical colony of Neotropical army ants (subfamily Ecitoninae) regularly raids a large area around Revised 17 June 2018 their bivouac by forming a narrow directional column that can reach up to one hundred meters in Accepted 18 June 2018 length. The raid is ﬁnished and then relaunched 12–17 times, each time toward diﬀerent orientation. After completing all bouts the colony relocates to a new area. A hypothetical KEYWORDS alternative to this foraging mode is raiding radially and symmetrically by expanding the search Army ant; simulation; raid; front in every direction like a circular bubble. Using an existing agent-based modeling software foraging that simulates army ants’ behavior, we compared the two possible modes of foraging in diﬀerent food distributions. Regardless of the food patch abundance, the radial raiding was superior to the directional raiding when food patches had low quality, and the directional raiding was favorable when the patches were rich. In terms of energy eﬃciency, the radial raiding was the better strategy in a wide range of conditions. In contrast, the directional raiding tended to yield more food per coverage area. Based on our model, we suggest that the directional raiding by army ants is an adaptation to the habitats with abundance of high-quality food patches. This conclusion ﬁts well with the ecology of army ants. Introduction However, from the exploratory viewpoint, their The army ants are specialized collective predators that columnar raid formation is an unusual choice. The always forage in large groups (Kronauer 2009). Their search front is narrow (as short as one tenth of the colony can form a swarm of many thousands of column length; Couzin and Franks 2003), and only the hunters, advancing in a column over a hundred meters minority of the foragers are exposed to the novel long (Couzin and Franks 2003). In the ‘nomadic phase,’ environment. The remaining majority run over the they move their camp every day, but when a colony same path as their predecessors did, contributing enters ‘statary phase,’ it launches multiple successive almost nothing to the search. The successive raids will ‘raids (Willson et al. 2011).’ The raids occur about once increase the ﬁnal coverage area, but still it seems inferior a day, and they avoid the recently exploited direction to non-directed search patterns. For example, an Eciton (Willson et al. 2011). After depleting a region with burchelli raid can employ 200,000 individuals (Couzin about 14–17 raids, the colony relocates to a new area and Franks 2003), and if they were to radiate uniformly (Willson et al. 2011). Their nomadism depends on their from the colony in every direction, they could form an skills to form a ‘bivouac,’ a huge ball of ants that tempor- unbroken ring as large as 63,000 body-lengths in diam- arily shelter the young and the queen (Anderson et al. eter. Even with minor workers, this would be nearly 2002). This set of behaviors has been evolutionarily con- 200 meters wide, and it would not miss any single food served in more than 200 species of this clade over two item within the expanding ‘bubble’. This would provide continents (Brady 2003). much better coverage than the column raiding does. CONTACT Sang-Im Lee email@example.com, firstname.lastname@example.org School of Biological Sciences, Seoul National University, Seoul 08826, South Korea; School of Undergraduate Studies, Daegu-Gyeongbuk Institute of Science and Technology, Daegu, 42988, South Korea; Piotr G. Jablonski email@example.com, firstname.lastname@example.org School of Biological Sciences, Seoul National University, Seoul 08826, South Korea; Museum and Institute of Zoology, Polish Academy of Sciences, Wilcza, 64, Warsaw, Poland; Ho-Young Kim email@example.com Department of Mechanical and Aerospace Engineering, Seoul National University, Seoul, 08826, South Korea; Institute of Advanced Machines and Design, Seoul National University, Seoul, 08826, South Korea © 2018 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. 268 W. SONG ET AL. On the other hand, while being inferior in terms of the northeast grids. We modiﬁed the program to exploration, the directional column raiding enables accept arbitrary raiding direction and resized the simu- instant mass transportation after discovery. If the target lated world to 600 × 600 = 360,000 grids. The bivouac food source is far away, the colony may save a consider- was repositioned to the center. able time by skipping the return trip of the discoverers In order to simulate the foraging bout cycle, after 650 and the dispatch trip of the recruited transporters. simulated time steps all outside foragers were called However, to justify having hundreds of thousands of back to the bivouac. After extra 650 time steps in potential transporters following the search front, the return phase, the pheromone deposits were all set to 0, colony needs to ensure that the frontiers will ﬁnd a very and then the raid was re-launched. Fifteen bout cycles rich food patch. Otherwise, it could end up in a waste of were simulated before recording the performance, to time and energy for a very little gain. Therefore, the distri- imitate Eciton burchelli statary phases that launch 14– bution of food should be a major parameter aﬀecting the 17 raids before relocating (Willson et al. 2011). advantages of the directional column raiding. In the directional raid mode, the direction of the new To test the raiding performance under diﬀerent food raid was randomly chosen within the half–circle located distributions, we chose a simulation software (Brown opposite to the previous raid direction (Figure 1(a)). This 2008) aimed at modeling Eciton species, the popularly was to simulate the actual ant behavior of avoiding the studied new world army ants. We modiﬁed the recently raided direction (Franks 1989; Willson et al. program to enable comparison between the naturally 2011). In the radial raid mode, each individual had its occurring ‘directional raiding’ and the hypothetical own raid direction randomly chosen before leaving the ‘radial raiding’ strategies. We expected that the radial bivouac (Figure 1(b)). raiding would provide better coverage, but the direc- The food patch abundance values were 0.04%, 0.2% tional raiding would yield more food in a certain range and 1%, and the patch quality values were 1, 10 and of food distributions. 100. In AntSpace 1.1, the food patch abundance means the proportion of food-loaded grids among all 360,000 grids, and the patch quality is the number of visits Methods required to deplete the grid. This gave 3 × 3 = 9 food dis- tribution conditions. The number of ants were set to We used AntSpace 1.1 (Brown 2008), a NetLogo (Wilensky 2000, which is similar to the number used in the previous 1999) model of army ant raiding behavior. AntSpace simulations (Sole et al. 2000; Brown 2006). For all the combines many ﬁndings from the past (Deneubourg other parameters AntSpace requires, we used empirically et al. 1989; Franks et al. 1991; Sole et al. 2000; Couzin determined default parameter values provided with the and Franks 2003; Brown 2006) and faithfully simulates software (Table 1). the army ant behavior. In AntSpace 1.1, the ants were The ‘coverage area’ of the colony was measured by assumed to move northward by default, and to stochas- the count of grids visited by an ant at least once. In tically choose their direction by comparing pheromone order to represent collective energy use by the colony, concentration between the north, the northwest and Figure 1. Simulated examples of directional and radial raids. The beginning stages of both directional (a) and radial (b) raids are shown. In both examples, the colony (center) is launching the third raid out of 15 scheduled. The area covered by the previous raids is visualized with pale blue color. ANIMAL CELLS AND SYSTEMS 269 Table 1. Default parameter values set in AntSpace 1.1 (Brown 2008), extracted from the past mathematical models (Deneubourg et al. 1989; Sole et al. 2000) For detailed simulation algorithm which runs on these parameters, please see the publicly available code and information on AntSpace 1.1 (Brown 2008). For empirical basis and basic modeling principles underlying the choice of the parameters, please see Sole et al. (2000). Parameter Value Explanation maxPher-Return 540 threshold pheromone level at which returning ants refuse to deposit additionally maxPher-Out 51 threshold pheromone level at which outbound ants refuse to deposit additionally amtPherToRemove 0.005 amount of pheromone evaporated at each time step amtPherToDrop 47 amount of pheromone deposited by outbound ants at each time step, given that returning ants deposit 10 emptyNodeweightOut 24 basal pheromone level outbound ants perceive from a grid without pheromone emptyNodeweightIn 24 basal pheromone level returning ants perceive from a grid without pheromone antsPerStep 10 number of ants that can simultaneously depart from the bivouac per one time step a variable ‘total movement’ was incremented by 1 every the maximum available amount, represented the eco- time an ant moved to another grid. logical impact of the ants on the food resources. To measure the performance of the colony, the total ‘food collected’ was recorded. It was divided by the Results total movement or by the coverage area to demonstrate the diﬀerent aspects of foraging eﬃciency. Another For low-quality food patches (value 1 or 10) and regard- measurement, the proportion of the collected food to less of the food patch abundance, the radially raiding Figure 2. Colony foraging records after 15 raids are completed. White, directional raiding; gray, radial raiding. First column (a, e, i), total number of food obtained. Second column (b, f, j), number of food obtained per one million collective movements. Third column (c, g, k), number of food obtained per one million grids collectively discovered. Fourth column (d, h, l), the collected proportion among the initially available amount. Subpanels are organized in three rows, according to the food patch quality set in the simulation. Each sub- panel has horizontal axis for the abundance of food patches. 270 W. SONG ET AL. ants collected 29–63% more food compared to the direc- encounter the anteaters (Willson et al. 2011), and tionally raiding ants (Figure 2(e,i)). However, for high- unlike the Afrotropical Dorylus (Wilson 1971), inter-colo- quality food patches (100), the directionally raiding nial conﬂicts are easily resolved without much mortality ants collected more food overall (Figure 2(a)). A similar (Willson et al. 2011). They also have a set of behaviors general pattern could be seen in the two eﬃciency speciﬁcally tuned to access diﬃcult terrains, such as measurements (the second and third columns of the ‘living bridge (Reid et al. 2015; Graham et al. 2017)’ Figure 2). However, the two measurements diﬀered in or the ‘pothole plug (Powell and Franks 2007),’ implying details. In terms of the movement eﬃciency, the food that they gain beneﬁt by expanding their activity range. collected in one million collective movements, the Finally, they are nomadic species without permanent radial raiding was the better strategy in general (Figure shelter, and they frequently relocate to a newer area 2 (b,f,j)). In a wide range of conditions the radial raiding (Kronauer 2009; Willson et al. 2011; Garnier and Kro- outperformed or closely matched the directional nauer 2017) suggesting again that they do not pursue raiding, often winning by margin of almost 90% (Figure smaller coverage. 2(f,j)). Only in one condition (the middle plot of Figure Can selection in a foraging context explain why the 2(b)), the directional raiding was 13% better. army ants raid directionally? Our model demonstrated In terms of the coverage eﬃciency, the food collected that the directional column raiding was not a good fora- in one million explored grids, the directional raiding was ging strategy to search for scattered small food sources. a good strategy overall (Figure 2(c,g,k)). For low-quality The model parameters were determined from the obser- food patches (i.e. when the patch quality was 1; Figure vation, so the natural selection could have optimized 2(k)), the directional raiding was less eﬃcient, but the them for the directional raiding. In contrast, the radial gap was not greater than 11%. For higher-quality raiding behavior in our simulation did not involve any patches, the directional raiding performed better, and further optimization to the new foraging regime. Only when the patch quality was 100, the margin was as with the diversiﬁcation of the initial departing directions, large as 70–100% (Figure 2(c)). just one simple alteration of the model parameter, the Food exploitation eﬃciency (proportion of food col- colony gained substantial energetic reward in a wide lected) was lower in conditions of high-quality food range of test conditions. Compared to the radial patches, regardless of food patch abundance (Figure 2 raiding, the 15 directional raids were often inadequate (d,h,l)). In conditions where radial raiding yielded more to provide coverage over the full circle of range available food, 34–59% of the total available food was collected to the colony, and left many food patches unexploited. (Figure 2(h,l)). On the other hand, in conditions where However, if the food patches were of very high quality, directional raiding was superior, only 11–21% of the the directional raiding had advantages in various aspects total food was collected (Figure 2(d)). of eﬃciency. Unlike the radial raiding, the directional raiding could maintain the density of the search front even after a considerably long expedition. This would Discussion allow fast and instant concentration of the workforce Although the simulation showed that the directional into a resourceful patch, draining it within a short time. raiding is generally coverage-eﬃcient, this mode of fora- After that, the subsequent raids are unlikely to re-visit ging is not very energy-eﬃcient (the second and third the depleted patch. On the other hand, in a radial columns of Figure 2). These trends are likely to arise raiding, it was diﬃcult to recruit the remotely scattered when a large crowd of ants is concentrated in a small foragers to the discovered patch. The discoverers could number of food patches. In this situation, most of the lay a pheromone trail back to the bivouac, but the infor- individuals are active in the already visited area rather mation could not reach the majority of the outside fora- than a new unexplored territory, leading to a more gers until they come back home. This bottlenecked the thorough search and the higher coverage eﬃciency. transition from exploring to transporting jobs. However, the movement eﬃciency may be negatively The previous research on army ants support the adap- impacted by collisions between individuals due to the tive value of directional raids in habitats with high food high density. patch quality. Army ants, both the neotropical and the Why do the army ants raid directionally? We believe Afrotropical groups, are believed to have evolved from that the coverage eﬃciency is unlikely to be the ultimate a common Gondwanan ancestral clade that preyed on reason, because it is diﬃcult to ﬁnd a selective pressure social insect colonies (Berghoﬀ 2003; Brady 2003; Brady that may adaptively constrain the raid coverage. Neotro- et al. 2014), and numerous species still maintain the pical army ants are the top predator of the ecosystem diet (Berghoﬀ 2003; Ramirez and Cameron 2003; Powell (O’Donnell et al. 2007) except when they rarely and Clark 2004; Le Breton et al. 2007; Souza and Moura ANIMAL CELLS AND SYSTEMS 271 2008; Kronauer 2009; Powell 2011; Dejean et al. 2014). context it is diﬃcult to think of a situation where a Others have their diet diversiﬁed, but they also generally heavy investment in traﬃc infrastructure is more impor- opt for large preys or rich litter patches (O’Donnell et al. tant than increasing the coverage area. Second, as noted 2005; Kaspari et al. 2011). A study reported that some previously, the selective pressure from predation and army ant species generally cherry-pick higher quality competitive aggression is quite low for new world patches, only skimming the most convenient 25% of army ants (O’Donnell et al. 2007; Willson et al. 2011). the animal biomass and leave the rest intact (Kaspari Finally, although some species of army ants do hunt ver- et al. 2011). Interestingly, in our simulation, the con- tebrates (O’Donnell et al. 2005), most army ants primarily ditions favorable to the directional raiding were identical feed on social insect colonies (Berghoﬀ 2003). Social to the conditions of less exhaustive exploitations (Figure insect colonies are immobile food sources and success- 2(d,h,l)). To sum up, the directional raiding is a trait fully exploitable with non-army-ant behaviors, e.g. by closely related to highly rich resources that are not the termite-eating Matabele ants (Villet 1990) and the easily exhaustible, both in the real world and in our slave-making social parasites (Alloway 1979; Hasegawa simulation. and Yamaguchi 1994). Therefore, we excluded the afore- Then, why the majority of other ant species that rely mentioned factors from the simulation and focused on on rich food patches, e.g. the leafcutter and honeydew- the eﬀect of the food distribution only. harvesting ants, do not utilize the directional column In summary, this study illustrates that food distri- raiding? In these species, a large number of reserve bution alone is suﬃcient to create ecological situations recruits waiting in the nest compensates the downside in which the natural selection may favor column of the undirected search (Jaﬀe and Deneubourg 1992). raiding over the radial searching. The future studies When a recruitment signal is given, the reserves follow should consider variations in diﬀerent movement par- the pheromone trail to the newly discovered food ameters as well as in the diet and the colony size to source, allowing massive and concentrated exploitation further investigate adaptive value of the raiding (Jaﬀe and Deneubourg 1992; Shaﬀer et al. 2013). Most behavior. ant species have highly varied recruitment strategies based on this principle, implying the evolutionary ﬂexi- Acknowledgements bility and universal utility of this behavioral scheme (Höll- dobler and Wilson 1990). In contrast, army ant foragers This work was supported by Seoul National University Research Grant in 2015 and BK21 program of National Research Foun- could not beneﬁt from this recruitment scheme, dation of Korea. because they leave the bivouac at a much faster rate and save less reserve in the colony (Deneubourg et al. 1989; Sole et al. 2000; Brown 2006). This extreme scout- Disclosure statement reserve imbalance is probably for overcoming the No potential conﬂict of interest was reported by the author(s). specialized prey defense mechanisms (Dejean and Corbara 2014; Dejean et al. 2014; Kessler et al. 2016). The model does not include some other possible Funding advantages of directional column raiding. First, the This work was supported by Brain Korea Programme to the large number may allow the ants to overcome physical School of Biological Sciences, SNU [BK21]; Seoul National Uni- obstacles collectively by forming self-assembled versity Research Grant in 2015. bridges (Reid et al. 2015; Graham et al. 2017). Second, their large number and density might serve as a protec- tion against predation or competitive aggression. Third, References large and mobile prey e.g. living vertebrates could be Alloway TM. 1979. Raiding behaviour of two species of slave- quickly overwhelmed right after the discovery. making ants, harpagoxenus americanus (emery) and lep- However, these additional beneﬁts from directional tothorax duloticus wesson (hymenoptera: formicidae). raiding by a large number of individuals may not Anim Behav. 27:202–210. always be high. First, the self-assembled structures are Anderson C, Theraulaz G, Deneubourg JL. 2002. Self-assem- blages in insect societies. Insect Soc. 49:99–110. costly commitments of many potentially active foragers, Berghoﬀ SM. 2003. Army ants: an evolutionary bestseller? Curr and their traﬃc enhancement is actually not very great Biol. 13:R676–R677. (Powell and Franks 2007; Brunelle 2011). During colony Brady SG. 2003. Evolution of the army ant syndrome: the origin migration, such bridges or rafts would be worth con- and long-term evolutionary stasis of a complex of behavioral structing because the queen, pupae, larvae and eggs and reproductive adaptations. P Natl Acad Sci USA. need to be transported. However, in the foraging 100:6575–6579. 272 W. SONG ET AL. Brady SG, Fisher BL, Schultz TR, Ward PS. 2014. The rise of army Kronauer DJC. 2009. Recent advances in army ant biology ants and their relatives: diversiﬁcation of specialized preda- (Hymenoptera: Formicidae). Myrmecol News. 12:51–65. tory doryline ants. Bmc Evol Biol. 14:93. Le Breton J, Dejean A, Snelling G, Orivel J. 2007.Specialized Brown TB. 2006. Biology and modeling of self-organization in predation on Wasmannia auropunctata by the army ant the new world army ant Eciton Burchellii [dissertation]. Salt species Neivamyrmex compressinodis. J Appl Entomol. Lake City (UT): University of Utah. 131:740–743. Brown T. 2008. AntSpace [computer program]. Version 1.1. O’Donnell S, Kaspari M, Lattke J. 2005. Extraordinary predation Available at: http://www.inﬁniteworld.org/ant/model/. by the neotropical army ant Cheliomyrmex andicola: impli- Brunelle D. 2011. Bridge building and foraging eﬃciency in the cations for the evolution of the army ant syndrome. army ant Eciton burchellii. Boulder (CO): University of Biotropica. 37:706–709. Colorado at Boulder. O’Donnell S, Lattke J, Powell S, Kaspari M. 2007. Army ants in Couzin ID, Franks NR. 2003. Self-organized lane formation and four forests: geographic variation in raid rates and species optimized traﬃc ﬂow in army ants. P Roy Soc B-Biol Sci. composition. J Anim Ecol. 76:580–589. 270:139–146. Powell S. 2011. How much do army ants eat? On the prey intake Dejean A, Corbara B. 2014. Reactions by army ant workers to of a neotropical top-predator. Insect Soc. 58:317–324. nestmates having had contact with sympatric ant species. Powell S, Clark E. 2004. Combat between large derived C R Biol. 337:642–645. societies: a subterranean army ant established as a predator Dejean A, Corbara B, Roux O, Orivel J. 2014. The antipredatory of mature leaf-cutting ant colonies. Insect Soc. 51:342–351. behaviours of Neotropical ants towards army ant raids Powell S, Franks NR. 2007. How a few help all: living pothole (Hymenoptera: Formicidae). Myrmecol News. 19:17–24. plugs speed prey delivery in the army ant Eciton burchellii. Deneubourg JL, Goss S, Franks N, Pasteels JM. 1989. The blind Anim Behav. 73:1067–1076. leading the blind: modeling chemically mediated army ant Ramirez S, Cameron SA. 2003. Army ant attacks by Eciton raid patterns. J Insect Behav. 2:719–725. hamatum and E-rapax (Hymenoptera: Formicidae) on nests Franks NR. 1989. Army ants – a collective intelligence. Am Sci. of the Amazonian bumble bee, Bombus transversalis 77:138–145. (Hymenoptera: Apidae). J Kans Etomol Soc. 76:533–535. Franks NR, Gomez N, Goss S, Deneubourg JL. 1991. The blind Reid CR, Lutz MJ, Powell S, Kao AB, Couzine ID, Garnier S. 2015. leading the blind in army ant raid patterns: testing a Army ants dynamically adjust living bridges in response to a model of self-organization (hymenoptera: formicidae). J cost-beneﬁt trade-oﬀ. P Natl Acad Sci USA. 112:15113– Insect Behav. 4:583–607. 15118. Garnier S, Kronauer DJC. 2017. The adaptive signiﬁcance of pha- Shaﬀer Z, Sasaki T, Pratt SC. 2013. Linear recruitment leads to sic colony cycles in army ants. J Theor Biol. 428:43–47. allocation and ﬂexibility in collective foraging by ants. Graham JM, Kao AB, Wilhelm DA, Garnier S. 2017. Optimal con- Anim Behav. 86:967–975. struction of army ant living bridges. J Theor Biol. 435:184– Sole RV, Bonabeau E, Delgado J, Fernandez P, Marin J. 2000. 198. Pattern formation and optimization in army ant raids. Hasegawa E, Yamaguchi T. 1994. Raiding behavior of the Artiﬁcial Life. 6:219–226. Japanese slave-making ant Polyergus-samurai. Insect Soc. Souza JLP, Moura CAR. 2008. Predation of ants and termites by 41:279–289. army ants, Nomamyrmex esenbeckii (Formicidae, Ecitoninae) Hölldobler B, Wilson EO. 1990. The ants. Cambridge (MA): in the Brazilian Amazon. Sociobiology. 52:399–402. Harvard University Press. Villet MH. 1990. Division-of-labor in the matabele ant Jaﬀe K, Deneubourg JL. 1992. On foraging, recruitment systems Megaponera-foetens (Fabr) (Hymenoptera-Formicidae). and optimum number of scouts in eusocial colonies. Insect Ethol Ecol Evol. 2:397–417. Soc. 39:201–213. Wilensky U. 1999. Netlogo [computer program]. Version 5.3.1. Kaspari M, Powell S, Lattke J, O’Donnell S. 2011. Predation Evanston (IL): Center for Connected Learning and and patchiness in the tropical litter: do swarm-raiding army Computer-Based Modeling, Northwestern University. ants skim the cream or drain the bottle? J Anim Ecol. Willson SK, Sharp R, Ramler IP, Sen A. 2011. Spatial movement 80:818–823. optimization in Amazonian Eciton burchellii army ants. Kessler BJ, Kessler DJ, Calsbeek RG. 2016. Eciton burchelli Insect Soc. 58:325–334. army ants avoid Nasutitermes termites. Integr Comp Biol. Wilson EO. 1971. The insect societies. Cambridge (MA): Belknap 56:E109–E109. Press of Harvard University Press.
Animal Cells and Systems
– Taylor & Francis
Published: Jul 4, 2018
Keywords: Army ant; simulation; raid; foraging