Leaf-Cutting Herbivory in Successional and Agricultural Tropical Ecosystems Author(s): Chantal M. Blanton and John J. Ewel Reviewed work(s): Source: Ecology, Vol. 66, No. 3 (Jun., 1985), pp. 861-869 Published by: Ecological Society of America Stable URL: http://www.jstor.org/stable/1940548 . Accessed: 01/11/2012 15:22

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp

. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

.

Ecological Society of America is collaborating with JSTOR to digitize, preserve and extend access to Ecology.

http://www.jstor.org Ecology,66(3), 1985, pp. 861-869 ? 1985by the Ecological Society of America

LEAF-CUTTING ANT HERBIVORY IN SUCCESSIONAL AND AGRICULTURAL TROPICAL ECOSYSTEMS1

Chantal M. Blanton and John J. Ewel Departmentof Botany, Universityof Florida, Gainesville,Florida 32611 USA

Abstract. Herbivory by cephalotes was measured in four plant communities of different complexityin Costa Rica. The fourcommunities were a monocultureof cassava {Manihot esculenta) and three diverse assemblages, each 1.5 yr old: (1) successional vegetation,unmodified by the in- vestigators;(2) imitationof succession, a communityof investigator-introducedspecies designed to mimic the unmodified succession; and (3) enriched succession, a successional vegetation that the investigatorshad augmentedby propagule inputs. Each ant colony had access to all fourcommunity typessimultaneously. The average herbivoryrate (daily cuttingof leaf per square metreof ground,all treatmentscombined) was ~ 150 mg (or 38 cm2). In communitieswith greater leaf area index, structuralcomplexity, and richness,A. ceph? alotes cut lower portionsof total leaf area. Beforeharvest of the cassava monoculture,the herbivory rate of A. cephalotes was 87.9 cm2m~2d_1 in the monoculture,21.4 in the imitation, 14.7 in the succession, and 6.8 in the enrichedsuccession community.These amounts represented0.3% of total leaf area in the monocultureand a mean of 0.03% of total leaf area in the threecomplex ecosystems. Cassava, which occurred in threeof the fourcommunities, was attacked most heavily (per unit leaf area) in the imitationsuccessional community,least heavily in the enrichedsuccession, and at inter- mediate intensityin the monoculture.In response to loss of theirpreferred forage (cassava) through harvestingby humans, the cut more leaf tissue in the threespecies-rich communities, especially the imitation. As cassava resproutedin the monoculture,A. cephalotes' rate of attack on the three diverse treatmentsreurned to preharvestlevels. Atta cephalotes cut only 17 of 332 available plant species. They cut proportionallymore woody than herbaceous species, more introducedspecies than natural colonizers, and species with below- average water contents. Plant relative abundance alone did not determinehost plant selection, but most of the attacked species were cut in proportionto theirtotal leaf area. Key words: ants; Atta cephalotes; Costa Rica; herbivory;Manihot esculenta; succession; tropical agriculture;tropical succession.

Introduction ants had an opportunity to select forage from a broad of all within easy access of their Developing nations, particularly those with limited range ecosystems, nests: old forest, forest, pasture, plantations of fossil fuel resources, face difficultdecisions in selecting young exotic trees and Pinus), and four kinds of appropriate agricultural ecosystems to keep pace with {Gmelina established as part of the re? food and fiber needs of their rapidly expanding pop? experimental ecosystems search The latter from monocultures ulations. Monocultures are notorious for pest out- project. ranged to even more diverse than successional breaks (Gibson and Jones 1977), whereas mixed crop- ecosystems ping systems sometimes do and sometimes do not afford vegetation. Atta a member ofthe more protection to their component plants (Perrin and cephalotes, myrmicine fungus- tribe Attini, is a widespread her? Phillips 1978, Risch 1981, Ewel et al. 1982). One growing conspicuous, bivore whose increase under human stew- promising line of research suggests that useful char? populations 1967,Haines 1975). acteristics of natural ecosystems may be incorporated ardshipof ecosystems (Goncalves This forest-dweller invades sub- into agroecosystems (Hart 1980). We are part ofa team gap-loving frequently sistence and from to Bra- of researchers that has been working in Costa Rica since plantation agriculture zil and Cherrett Over 1978, testing this idea. An integral part of such research (Cherrett Peregrine 1976, 1981). this A. is a is to study how herbivores interact with host plants in geographic range cephalotes generalist (sensu it attacks of several chemi- simple and diverse ecosystems (van Emden and Way Feeny 1976); i.e., species unrelated local 1972, Root 1973, Feeny 1976, Rausher 1981). cally plant families, although popula? tions have a diet From the time our study plots were cleared and may very specialized (Rockwood When substrate for their planted in early 1979 (Ewel et al. 1981), it was obvious 1976). harvesting fungal gar- leaf-cutters remove a select of the that the leaf-cutting ant Atta cephalotes L. was one of den, very portion total of leaves, flowers, and fruits the most important herbivores on the site. This impres- standing crop (Lugo et al. Rockwood by A. ceph? sion was reenforced by observations made by Brown 1973, 1978). Herbivory alotes can reduce potential (Rock? (1982) while measuring overall herbivory rates. The plant reproductive wood 1972) and undoubtedly affects net primary pro? 1 Manuscript received 16 February 1984; revised 25 June ductivity as well. Furthermore, the ants' refuse dumps 1984; accepted 26 June 1984. concentrate nutrients and their nest-building disturbs 862 CHANTAL M. BLANTON AND JOHN J. EWEL Ecology, Vol. 66, No. 3 soil profiles, creating point sources or sinks of nutrients tuber. After the first2 mo of this study it was harvested (Haines 1975, 1978, 1983, Alvarado et al. 1981). (September 1980), but the aboveground parts were re- Many agricultural ecologists propose that structural turned to the plots after weighing and subsampling. and floristic complexity can reduce the probability of The twigs continued to sprout for several months. severe pest outbreaks in plant communities (see review Six blocks were established, each containing the four by Altieri and Letourneau 1982). Because of our in- treatments in randomized block design (Fig. 1). Each terest in the attributes, including losses to herbivores, treatment occupied one quadrant of each block and of successional and agricultural communities of differ- thus had two sides contiguous with other treatments ing degrees of complexity, the following question was and two sides contiguous with the second-growth forest of key importance to our research: Do the ants pref- that surrounded each block. Because the 16 x 16 m erentially forage in certain of the four experimental quadrants were far smaller that the area foraged by a ecosystems? A related question was of agricultural sig? A. cephalotes colony, each colony had access to all four nificance: How do the ants respond to a dramatic re? experimental ecosystems. The ants' preferred forage duction in availability of their preferred forage? The species, cassava, was available in all six replications of answer to this question was obtained by measuring the monoculture, imitation, and enriched treatments. herbivory before, during, and after harvest of the Measurement monoculture, which contained >85% (by leaf area) of offoraging the ants' preferred forage species, cassava. Finally, we From late July through mid-November 1980 (rainy asked a third question: What are the characteristics of season months), experimental plots were searched sev? those species and ecosystems that are especially at- eral times weekly for leaf-cutting ants. Plots were vis? tractive to the ants? This question was answered by ited in a predetermined random order to ensure that comparing grazed and ungrazed plant species in terms all received an equal number {n = 33) of surveys, about of their importance in each plant community, their life half during the day and half at night. Because daytime forms, and the moisture content of their leaves. surveys normally began after dawn and ended in late afternoon, and nighttime surveys began after dusk and Methods ended before daybreak, dawn and dusk were under- sampled. Foraging on a given plot was measured only site Study on trails leading from plants on that plot. To determine This study was carried out in the Florencia Norte the area and mass of leaf tissue removed, laden ants Forest, near Turrialba, Cartago Province, Costa Rica. passing an observation point on each active trail were The site lies on lightly undulating terrain located at counted for 5 min. The mean area and mass of each 9?53' N, 83?40' W, and at 650 m elevation. Mean an? ant's burden were estimated by collecting >50 leaf nual rainfall approaching 2700 mm is concentrated fragments from each active trail during each obser? between April and December. The area is in the Trop? vation. The combined area of these fragments was ical Premontane Wet Forest life zone (Tosi 1969). measured to within 1 mm2 with an area meter; the Four experimental ecosystems were established sub- composite sample was then dried at 70?C for 24 h and sequent to slashing (in January 1979) and burning (in weighed to within 0.1 mg. Leaf fragments were iden- March 1979) 9-10 yr old second-growth (Ewel et al. tified to species using indumentum, color, texture, or 1981; species richness data from Brown 1982). other unique characteristics. 1) Succession: natural regrowth from seedlings and In addition to the weekly observations, total daily sprouts, unmanipulated by the investigators and con? impacts of A. cephalotes foraging on seven represen- taining 121 species. tative trails were measured at approximately 1-h in? 2) Enriched succession: natural regrowth augmented tervals over 24-h periods on three dates: 16 August, 8 by propagules {> 10 000 seeds of >20 species per rep- September, and 12 October 1980. This was done to lication per month) of additional taxa and containing provide a second, independent estimate of A. cepha? 159 plant species. Introduced seeds were harvested op- lotes herbivory, one that included fluctuation in for? portunistically from off-sitesources by researchers; they aging activity in response to time of day, rain showers, included cultivars and noncultivars, indigenous and and sun-fleck patterns. Observations and 5-min counts exotic species. This self-design ecosystem was estab? of laden foragers were as described above. Counts were lished to test hypotheses concerning susceptibility to converted to leaf area and plotted over time. The area invasion, facilitation, and functional attributes of beneath the curve thus represented total leaf area car- species-saturated communities. ried over 24 h. 3) Imitation of succession: 82 nonindigenous species, characteristics both wild and cultivated, chosen to mimic the physi- Vegetation ognomy ofthe natural successional vegetation. Natural To estimate the plant resource available to A. ceph? colonizers were weeded out. alotes, Leaf Area Index value (LAI, the area of leaf 4) Monoculture: mature, weed-free cassava {Man- tissue [one side of leaf] per unit ground area) was mea? ihot esculenta), which is a shrub that produces an edible sured using a plumb-bob method (detailed analyses in June 1985 LEAF-CUTTING ANT IMPACTS 863

MONOCULTURE ENRICHED 0 10 20 30 IMITATION METRES SUCCESSION TUNNEL OUTLET NEST TRAIL FOOT PATH

Fig. 1. Map ofthe studysite, showingexperimental communities and A. cephalotesnests and trails.

Brown 1982, Ewel et al. 1982). At 30 points in each values, which met the condition of normality for anal? plot, leaves touching a thin cord extended vertically ysis of variance (ANOVA) (Kolmogorov-Smirnov test: through the vegetation were identified to species and D = 0.037, P> .15). recorded by 25-cm height classes. The mean number The model for ANOVA was split-split plot in space of leaf-line intersections per point estimates overall and time, with replications as blocks subdivided by plot LAI. The percentage of all intersections attributed treatments, time of day (i.e., day or night), and state to a given plant species indicates its relative abundance ofthe cassava monoculture. The four states of cassava in the ecosystem. Cassava monocultures were mea? were 1) mature; 2) leaves still fresh (after having been sured on 5 September 1980, prior to harvest. LAIs of returned to the plot following harvest); 3) no leaves other treatments were measured 1 October 1980. (because of postharvest decay and herbivory); and 4) We also measured foliar moisture content of species resprouting. Ninety percent of the variation in trans- that A. cephalotes cut for substrate and/or species that formed mean herbivory rates was accounted for by the comprised >80% of the LAI of each experimental model. There were significant F tests for interactions treatment. Approximately 300 g of fresh leaves of each among treatments, states of the cassava monoculture, species were weighed to 0.1 g, dried for 24 h at 70?, and time of day, but not for any of these variables and reweighed. individually. Therefore, analyses of differences among levels of each interacting variable were performed at Results fixed levels of the others. Averaged over all vegetation types, the mean (?sd) Atta cephalotes herbivory rates based on three 24-h rate of leaf-cutting per square metre of ground surface sampling periods varied little: 21.4, 20.9, and 21.9 m2/ area per day was 37.6 ? 36.1 cm2m_2d_1 or 0.149 ? 24 h. Over all treatments, most leaf area was cut at 0.132 gm_2d_1. These rates are based on weekly ob? night, although daytime foraging predominated in the servations during the 5-mo study period. Leaf areas more heavily vegetated enriched and succession vege- cut by A. cephalotes varied greatly from one set of tations. This estimate of herbivory rate is lower than surveys to the next. In 506 out of 792 plot surveys, the the rate derived from weekly foraging measurements. ants cut no vegetation, which indicates that leaf-cutting Weekly measurements may overestimate the true mean was spotty through time (see Fig. 2). In the remaining because sampling seldom occurred at dawn or dusk, 286 surveys, there were, on average, 1.5 active trails which were times of low foraging activity. per plot. This activity pattern skewed the frequency distribution of herbivory rates to the right. A Kol- Foraging differences among ecosystems = mogorov-Smirnov test {D 0.35, P < .01) confirmed The ants consumed an average of 3.5 times more that the data were not normally distributed. Analyses leaf tissue per unit ground area per day in the cas? were therefore performed on \og{Y + l)-transformed sava monoculture and cultivar-dominated imitation 864 CHANTAL M. BLANTON AND JOHN J. EWEL Ecology, Vol. 66, No. 3

CO 20- LU O LU CL CO I 5-

rr LU oo ioH ** *

LU > h- 5H <

O JUL AUG SEP OCT NOV T T TIME Fig. 2. Plant species cut by A. cephalotesthrough time on experimentaltreatments. Arrows indicate durationof study.V above horizontalaxis indicates date of cassava harvest.Dots indicate each date on which a plant species was observed being cut and asterisksindicate both a cuttingobservation and the cumulative number of species cut since the beginningof the study.

of succession ecosystems (combined mean: 59 culture. As this leaf supply dwindled due to decom? cm2-m~2-d_1) than they did in the succession and en- position and A. cephalotes attack, the ants' daily leaf ? riched vegetation ecosystems (combined mean: 17 cm2 consumption per unit ground surface area continued m-2-d-!) (Table 1). to increase dramatically in the other three vegetation With mature cassava available in the monoculture, systems, reaching a maximum of 160 cm2-m~2-d_1 in treatment differences in herbivory rates were not sig? the imitation ecosystem. Furthermore, the total rate of nificant because of large within-treatment variances. ant foraging over all treatments combined temporarily These resulted in part from the ants' diel foraging pref? increased significantlyafter the cassava supply was ex- erences: A. cephalotes consistently foraged more at night hausted. Apparently, A. cephalotes activity was stim- than during the daytime in the monoculture. Though ulated by the cassava harvest; no other obvious en- maximum leaf removal by A. cephalotes occurred in vironmental changes occurred during this period to the monoculture, the ants were found foraging during account for the increased rate of leaf removal. a slightly (but not significantly) greater proportion of As cassava resprouted in monocultures in mid-Oc- the observations in the imitation ecosystem (55.6% tober, A. cephalotes herbivory in the other treatments compared to 51.4% in the monoculture) and slightly returned to rates not statistically different from pre- less in the succession (43.1%) and enriched (29.2%) harvest levels. systems. The plant resource During the 1-wk cassava harvest, mean A. cephalotes leaf consumption on experimental treatments more than During the study period, 332 plant species were doubled. Over 80% of their foraging effortwas con? growing on the experimental plots. Of these, A. ceph? centrated on the large mounds of leaves in the mono- alotes cut measurable amounts of only 17 (5.1%) (Table

Table 1. Herbivoryrates by A. cephalotesin the fourexperimental ecosystems.

Leaf tissue cut per unit ground area per day (amount m_2d_1)

* Means in the same row accompanied by the same superscriptare not significantlydifferent {P < .05). Significantlygreater (Duncan's Multiple Range Test, P < .05) day or nightherbivory rates within a mean are denoted by (d) and (n), respectively. June 1985 LEAF-CUTTING ANT IMPACTS 865

Table 2. Attributesof plants togethercomprising = of leaf area in each treatment.Asterisks indicate taxa cut by A. cephalotes.

t nat. = natural colonization; intro.= introducedby investigators. X vine = herbaceous climber (vs. liana = woody climber). ? Leaf area per unit ground surfacearea. || All treatmentscombined. IfComplex of species that are nearlyindistinguishable when sterile. #Includes Scleria pterotaand S. secans, and possibly some other sedges.

2). The ants cut 14 of 156 woody species on the plots cut. Total leaf area available on all experimental treat? (9.0%), but only 3 of 176 herbaceous species (1.7%). ments included 61% natural colonizers, 21% cassava, Mean leaf water content of plant species cut by A. and 18% other investigator-introduced species. cephalotes (80.7%) was significantly lower than back? The mean amount of leaf tissue available to the ants = ground levels (83.6%) calculated from all species to? was highest in the enriched (LAI 5.03) and slightly gether comprising >80% of leaf area {t test, P < .05). (but not significantly) lower in the succession and im? = Proportionally more introduced species (8 of 85, or itation treatments (LAIs 4.37 and 3.62, respective? 9.4%) than natural colonizers (9 of 247, or 3.6%) were ly). LAI in the monoculture was initially 2.91, but was 866 CHANTAL M. BLANTON AND JOHN J. EWEL Ecology, Vol. 66, No. 3

Table 3. Amount and frequencyof A. cephalotes attack on various species of plant. The number of observations (932) is greaterthan the number of plot surveys(792) because some surveysincluded measurementson more than one active trail.

reduced to 0 after harvest, then increased slightly as ofa species could not be predicted by its LAI. However, cassava twigs resprouted in mid-October. once attacked by A. cephalotes, the extent of attack was Cassava was removed most frequently (213 of 426 related to available leaf area. measurements) and extensively, accounting for >76% Discussion of total leaf area cut (Table 3). Other species that ac- losses counted for > 1% of total leaf area cut were the natural Ecosystem colonizers Bocconia frutescens, Ochroma pyramidale, The mean amount of leaf removed by A. cephalotes Cecropia obtusifolia, and Trema micrantha, plus the per square metre of ground per day in the young ex? investigator-introduced Dioscorea bulbifera. All of these perimental ecosystems, 0.149 g (or 37.6 cm2), surpasses were cut during the first month of observations (Fig. previous estimates of 0.081 g (Lugo et al. 1973) and 2) and consistently thereafterthroughout the study. All 0.027 g (Haines 1978) for in mature other species that the ants added to their cumulative lowland tropical forests. Other estimates in the liter- repertoire contributed a combined total of <2% ofthe ature are nest-specific and cannot be converted to leaf leaf area cut. consumption per unit area of ecosystem. Some species heavily cut by A. cephalotes were dom? Atta cephalotes accounted for a surprisingly large inant components of treatment and overall LAI; to? proportion ofthe total herbivory (measured by Brown gether the cut species made up 34% of the total leaf 1982) in the succession, enriched, and imitation treat? area available (sum of values of percent of total LAI ments: 25% of all leaf consumption per square metre * for species marked with in Table 2). Cassava ac- of ground (and 16% of all herbivory per square metre counted for the greatest proportion of overall LAI due of leaf) prior to the harvest ofthe cassava monoculture, to its abundance in the monoculture. B. frutescens made and nearly 50% of all leaf consumption per square up 7.7% ofthe total LAI; T. micrantha accounted for metre of ground (31% of all herbivory per square metre 1.5% ofthe total LAI (Table 2), or, expressed as leaf of leaf) that occurred in these three treatments through? area per square metre of ground, nearly lA m2/m2 in out the study period. Even though these high values the succession treatment. Not all common species were may, in part, reflect differences between the methods consumed, however. The exotic grass Panicum max? used by Brown (1982) to measure overall herbivory imum was the most abundant species (15.4% of total and those used by us, A. cephalotes was clearly the LAI) but was shunned by the ants. most important herbivore on the study site. There was a highly significant correlation (r = 0.97 The leaf-cutting ants consumed more leaf tissue in with cassava, 0.89 without; P < .01 in both cases) be? the monoculture than in the three species-rich com? tween leaf area cut (Table 3) and contribution of the munities. A. cephalotes' preference for the cassava cut species to total LAI (Table 2). The correlation was monoculture was predictable in light of the ants' pest not significant when abundant plant species that A. status in neotropical monocultures generally (e.g., cephalotes did not cut were included. Thus, selection Cherrett and Peregrine 1976, Fowler and Robinson June 1985 LEAF-CUTTING ANT IMPACTS 867

Table 4. Herbivoryby A. cephalotesas a fractionof leaf area available, based in part on estimatesof LAI of each ecosystem by Brown (1982).

* Conservative estimate based on LAI = 0.33.

1979) and in cassava specifically (Weber 1972, Bellotti 1979). The sudden increase in A. cephalotes foraging and van Schoonhoven 1978). The crop was a concen- effortand leaf-cutting during and just after the cassava trated, spatially predictable source of palatable sub- harvest was such a response, comparable to Rock- strate. Before harvest the monoculture received two- wood's (1976) observation that foraging increased after thirds of all A. cephalotes attacks; daily losses (area of he mechanically excluded ants from their preferredhost. leaf lost per square metre of ground) were 88 cmVm2 Atta cephalotes' postharvest attack (leaf area cut per or 0.3% of its leaf area. This was an order of magnitude unit ground surface area per day) on plants in the im- greater than the mean percentage cut by the ants in the itation treatment reached 161 cm2-m~2-d_1; each day more complex ecosystems (0.03%). The cassava was the ants removed nearly 3% of the standing leaf area not devastated by these rates of herbivory, however, of palatable host plants. The regular arrangement of and maintained high LAI by continuous production of cassava plants and the vining habit of another preferred new leaves. host plant, D. bulbifera, appeared to facilitate the ants' The difference between A. cephalotes attack on sim- switch to increased exploitation of that ecosystem. Once ple and on complex ecosystems is further reinforced foragers had found an individual of one of these pal? when leaf-cutting is expressed as a fraction of leaf area atable species, the plants' interlocking crowns (cassava) available (Table 4). Leaf tissue in the succession and or large leaf area (often > 1 m2 of leaves per individual enriched treatments, for example, was consumed at a of D. bulbifera), regularly spaced along a trail-like vine, rate of ~0.02%/d during the interval when the ants provided maximal substrate for minimal search effort. consumed ~0.3%/d ofthe available leaf tissue in the mature cassava monoculture. Species abundances and characteristics Ofthe three species-rich communities, the imitation Six plant species constituted >97% of A. cephalotes' treatment was attacked most heavily. What was there total substrate. Individuals of other had about the imitation treatment that resulted in more many species one telltale semicircle clipped from a leaf, suggesting herbivory than in either the succession or enriched that these plants had been sampled by A. cephalotes vegetations? It was not that attacked species accounted and subsequently rejected. The same was true for in? for a greater proportion of total LAI in the imitation dividual leaves of palatable species. This selectivity has than in the other treatments. For example, >80% of been described elsewhere in relation to strat- LAI in both the enriched and imitation treatments feeding egies and relative numbers of host plants (Cherrett 1968, consisted of species that were not cut for substrate by Rockwood 1973, 1978, Fowler and Stiles 1980), phys- A. cephalotes, yet the imitation treatment was much ical and chemical deterrents or arrestants (Cherrett more heavily attacked. Differences in herbivory rates 1972a, b, Gamboa 1975, Littledyke and Cherrett 1978), among these three communities must therefore have the welfare of the Mudd and been due to characteristics other than the relative abun? fungus (Martin 1970, Bateman 1979), and ready availability of plant species dance of palatable species. One such factor might have to foragers early in the life ofthe colony (Fowler 1982). been the homogeneity of the resource (number of pal? The species most voraciously attacked by A. ceph? atable species involved and their spatial arrangement). alotes during this study was cassava. It is not surprising In the imitation treatment only eight species, most of that, before harvest of the cassava monocultures, the which were regularly spaced, were attacked, whereas absolute amount of cassava cut by the ants in the in the enriched succession the ants attacked 15 species, monoculture, imitation, and enriched treatments was most of which were irregularly spaced and not abun? correlated with its contribution to LAI of those dant. strongly systems {r = 0.99, P < .05). However, when the amount of cassava cut was expressed as a fraction of total cas? Response to loss offorage sava leaf area available in each treatment, herbivory Leaf-cutting ants thrive following disturbances of na- on cassava was greatest in the imitation (2.3%), fol- tive or agronomic ecosystems (Haines 1978, Jonkman lowed by the monoculture (1.8%) and enriched (0.9%) 868 CHANTAL M. BLANTON AND JOHN J. EWEL Ecology, Vol. 66, No. 3 treatments. Brown (1982) found the same trend in tation growth. After such expansion, humans and ants overall cassava herbivory (leaf area lost per unit avail? usually conflict. able leaf area): the proportion of leaves losing > 15 If the ants' major source of substrate is removed by cm2-m"2-d"' was 46% in the imitation, 13% in the crop harvest, they must find new sources of substrate, monoculture, and 0 in the enriched treatment. This perhaps neighboring fields or forests. If a palatable crop suggests that a cassava leaf located in the imitation sprouts, the young plants will probably be heavily at- treatment was more likely to be clipped by A. cepha? tacked. lotes or other herbivores than a cassava leaf located in If, however, succession proceeds, the ants will forage the monoculture or enriched treatments. Cassava in on a subset of suitable plants. Our study indicates that the enriched successional vegetation, on the other hand, this is a small portion of available species, perhaps a was less vulnerable to herbivore attack. smaller portion than in old forests (cf. Cherrett 1968). Some species were very heavily cut in relation to Succession brings nearly 100% of the canopy (where their total available leaf area (e.g., O. pyramidale, Dios- Atta prefers to forage) to the ants' preferred foraging corea spp., and Nectandra sp.; Table 2). Brown (1982) range (<3 m above ground level; Cherrett 1968), so also found that some uncommon species on this site search and retrieval time for suitable substrate is min- were heavily attacked, although for most species in her imized. Net primary productivity of the vegetation is study the leaf area damaged per unit available leaf area high, providing a constant supply of the young leaves ? ? was < 5 cm2 m-2 d~l. The ants attacked plants of many Atta prefers (Cherrett 1972a). families, origins, and life-forms:a "susceptible-to-v4Mz" This study yields three conclusions useful in design- syndrome would be almost impossible to characterize. ing agroecosystems for areas prone to A. cephalotes However, some generalizations seem to hold. Atta outbreaks: cephalotes' observed predilection for woody and in- 1) It is more advantageous to plant many species vestigator-introduced species agrees with previous re- than few species. In the absence of information on ports (Weber 1972, Garling 1978, Mintzer 1979). specific plant defenses, choose species that are herba? Woody species typically offera more concentrated, less ceous and native. ephemeral supply of leaves than do herbs. Many ofthe 2) Atta cephalotes is a "switch-forager" and responds introduced species were cultivars and therefore might to a decrease in favored resources by expanding its have been genetically selected for features other than exploitation to alternate resources. Removal of one herbivore defense. crop is likely to lead to increased herbivory elsewhere, Plant water content, perhaps linked to phytochem- rather than to a diminution in overall consumption. istry (Mattson 1980), may also influence Atta's selec? 3) Ecosystems containing large amounts of leaf area tion of substrate in differentenvironments and during suffer lower losses, both relative and absolute, than different seasons (Mudd and Bateman 1979, Bowers equally accessible communities having less leaf tissue. and Porter 1981). The prime example of A. cephalotes' ACKNOWLEDGMENTS response to water depletion or physiological stress dur? We thankM. Artavia forfield assistance; D. Nickersonand our occurred when leaves of cassava ing study wilted, B. Brown forassistance with data analyses; G. Fuller for il- abscised, or were cut. The ants then attacked these lustrations;W. Judd,D. Hazlett, and K. Young forhelp with stressed leaves in preference to healthy leaves. In dam? plant-relatedproblems; J. Trager and the late W. Buren for withant-related and B. B. A. aged cassava tissue the membrane-bound enzyme lin- help problems; Brown, Haines, Lugo, F. Slansky,J. Stimac, K. and two amarase contacts and the lina- Young, anonymous hydrolyzes glucoside reviewersfor comments. This study is a contributionof a marin, liberating free hydrocyanic acid that can cooperative researchprogram being conductedjointly by the volatilize (Gondwe 1974). Perhaps the resulting de? Universityof Florida and the Centro Agronomico Tropical crease in hydrocyanic acid content of damaged cassava de Investigacion y Enseiianza (CATIE), and was supported National Science Foundation DEB leaves made them more attractive to ants than by grants 78-10721 and healthy DEB 80-11136. cassava leaves. LlTERATUREClTED Succession and agriculture Altieri,M. A., and D. Letourneau. 1982. Vegetationman? When humans slash and burn forests, Atta colonies agement and biological control in agroecosystems.Crop are likely to survive because their nests are deep in the Protection1:405-430. soil. In this study several A. cephalotes colonies sur? Alvarado,A.,C.W.Berish,andF.Peralta. 1981. Leaf-cutter ant {Atta cephalotes) influenceon the of an- vived the burn and thereafterdominated herbivory on morphology deptsin Costa Rica. Soil Science Societyof America Journal the site. Cultivators plant their seeds at the same time 45:790-794. that A. cephalotes disperse: the onset ofthe rainy sea? Bellotti,A., and A. van Schoonhoven. 1978. Mite and son. Atta queens that escape predation are likely to pests of cassava. Annual Review of Entomology23:39-67. M. and S. D. Porter. 1981. nest successfully in the moist soil. The claustral stage Bowers, A., Effectof foraging distance on water contentof substratesharvested by Atta of colony formation allows surrounding vegetation a colombica (Guerin). Ecology 62:273-275. head start while colonies are their and, small, substrate Brown, B. J. 1982. Productivityand herbivoryin high and requirements are low. Colony expansion parallels vege- low diversitytropical successional ecosystems in Costa Rica. June 1985 LEAF-CUTTING ANT IMPACTS 869

Dissertation. Universityof Florida, Gainesville, Florida, -. 1983. Leaf-cuttingants bleed mineralelements out USA. of rainforestin southernVenezuela. Tropical Ecology 24: Cherrett,J. M. 1968. The foragingbehaviour of Atta ceph? 85-93. alotes L. (,Formicidae). 1. Foraging pattern Hart, R. D. 1980. A natural ecosystemanalog approach to and plant species attacked in tropical rain forest.Journal the design of a successional crop systemfor tropical forest of Ecology 37:387-403. environments.Biotropica 12(Suppplement):73-82. -. 1972a. Chemical aspects of plant attack by leaf- Jonkman,J. C. M. 1979. Population dynamics of leaf-cut? cuttingants. Pages 13-24 in J. B. Harborne, editor. Phy- ting ant nests in Paraguayan pasture. Zeitschriftfur An- tochemical ecology. Academic Press, London, England. gewandteEntomologie 87:281-293. \912b. -. Some factorsinvolved in the selection of Littledyke,M., and J. M. Cherrett. 1978. Olfactory re? substrate Atta vegetable by cephalotes (L.) (Hymenoptera: sponses of the leaf-cuttingants Atta cephalotes (L.) and in rain Formicidae) tropical forest.Journal of Animal Ecol? Acromyrmexoctospinosus (Reich) (Hymenoptera:Formic- 41:647-660. ogy idae) in the laboratory.Bulletin of EntomologicalResearch -. 1981. The interactionof wild vegetationand crops 68:273-282. in ant attack. 315-325 in J. M. leaf-cutting Pages Thresh, A. E., E. G. Farnworth,D. Pool, P. Jerez,and G Kauf- editor. and Pitman Advanced Lugo, Pests, pathogens, vegetation. man. 1973. The impact of the leaf cutterant Atta colom- Boston, Massachusetts,USA. PublishingProgram, bica on the energyflow ofa tropicalwet forest.Ecology 54: Cherrett,J. M., and D. J. Peregrine. 1976. A review ofthe 1292-1301. status of ants and theircontrol. Annals of leaf-cutting Ap? M. biochemical basis ofthe 84:124-133. Martin,M. 1970. The fungus- plied Biology attineant Science 169:16-20. Ewel, J., C. Berish,B. Brown, N. Price, and J. Raich. 1981. symbiosis. W. in relation to Slash and burn impacts on a Costa Rican wet forestsite. Mattson, J., Jr. 1980. Herbivory plant Annual Review of and Ecology 62:816-829. nitrogencontent. Ecology System? Ewel, J., S. Gliessman, M. Amador, F. Benedict, C. Berish, atics 11:119-161. R. Bermudez, B. Brown, A. Martinez, R. Miranda, and N. Mintzer, A. 1979. Foraging activity ofthe Mexican leaf- Price. 1982. Leaf area, lighttransmission, roots and leaf cuttingant Atta mexicana (F. Smith), in a Sonoran desert damage in nine tropical plant communities.Agro-Ecosys- habitat (Hymenoptera,Formicidae). Insectes Sociaux 26: tems 7:305-326. 364-372. Feeny, P. 1976. Plant apparencyand chemical defense.Re? Mudd, A., and G. L. Bateman. 1979. Rates of growthof cent Advances in Phytochemistry10:1-40. the food fungusof the leaf-cuttingant Atta cephalotes(L.) Fowler, H. G 1982. Habitat effecton fungalsubstrate se? (Hymenoptera: Formicidae) on differentsubstrates gath- lection by a leaf-cuttingant. Bulletinof the New York En- ered by the ants. Bulletin of Entomological Research 79: tomological Society 90:64-69. 141-148. Fowler, H. G, and S. W. Robinson. 1979. Field identifi- Perrin,R. M., and M. L. Phillips. 1978. Some effectsof cation and relative pest status of Paraguayan leaf-cutting mixed croppingon the populationdynamics of insectpests. ants. Turrialba 29:11-16. Entomologia Experimentaliset Applicata 24:385-393. H. and E. W. Stiles. 1980. Conservativeresource Fowler, G, Rausher, M. D. 1981. The effectof native vegetationon the ants? The role of ter- managementby leaf-cutting foraging, susceptibilityof Aristolochiareticulata (Aristolochiaceae) ritoriesand and environmental Socio- trails, patchiness. to herbivoreattack. Ecology 62:1187-1195. 5:25-41. biology S. J. 1981. Insect herbivore abundance in G 1975. and ofthe desert Risch, tropical Gamboa, J. Foraging leafcutting monoculturesand an testof two ant versicolorversicolor polycultures: experimental gardening Acromyrmex (Pergande) 62:1325-1340. 20:103-110. hypotheses.Ecology (Hymenoptera:Formicidae). Oecologia (Berlin) on L. 1978. Selective Atta its Rockwood, L. L. 1972. The effectof defoliation seed Garling, foragingby cephalotes: tree 54: basis and effectson Thesis. of production of six Costa Rican species. Ecology fungalgrowths. University 1363-1369. Connecticut,Storrs, Connecticut, USA. Gibson, I. A. S., and T. Jones. 1977. Monocultures as the -. 1973. Distribution,density, and dispersion of two of in Guanacaste originof major forestpests and diseases. Pages 139-161 in species Atta (Hymenoptera:Formicidae) of Animal 42:803- J. M. Cherrettand G R. Sagar, editors. Origins of pest, Province, Costa Rica. Journal Ecology parasite,disease and weed problems. Blackwell Scientific, 817. Oxford,England. 1976. Plant selection and foragingpatterns in two Goncalves, C. R. 1967. As formigascortadeiras da Ama- species of leaf-cuttingants {Atta). Ecology 57:48-61. zonia, dos generos Atta Fabr. e AcromyrexMayr (Hym., 1978. Foragingpatterns and plant selectionin Costa Formicidae). Atas do Simposio sobre a Biota Amazonica Rican leaf-cuttingants. New York Entomological Society 5:181-202. Journal85:222-233. Gondwe, A. T. D. 1974. Studies on the hydrocyanicacid Root, R. B. 1973. Organization ofa plant-arthropodasso- contents of some local varieties of cassava (Manihot es- ciation in simple and diverse habitats:the fauna of collards culentaCrantz) and some traditionalcassava food products. {Brassica oleracea). Ecological Monographs 43:95-123. East AfricanAgriculture and ForestryJournal 40(2): 161- Tosi, J. A., Jr. 1969. Republica de Costa Rica, mapa eco- 167. logico. Tropical Science Center,San Jose, Costa Rica. Haines, B. L. 1975. Impact of leaf-cuttingants on vegetation van Emden, H. F., and M. Way. 1972. Host plants in the developmentat Barro Colorado Island. Pages 99-111 in F. population dynamics of . Pages 181-209 in H. F. B. Golley and E. Medina, editors. Tropical ecological sys? van Emden, editor. Insect/plantrelationships. Blackwell tems. Springer-Verlag,New York, New York, USA. Scientific,Oxford, England. -. 1978. Element and energyflow through colonies of Weber,N. A. 1972. The gardeningants: the Attines.Amer? the leaf-cuttingant, Atta colombica, in Panama. Biotropica ican Philosophical Society, Philadelphia, Pennsylvania, 10:270-277. USA.