Oecologia (2003) 135:442-450 DOI 10. 1007/s00442-003-1212-9

Valerie Fournier * Jay A. Rosenheim a Jacques Brodeur *Lee 0. Laney *Marshall W. Johnson Herbivorousmites as ecological engineers: indirecteffects on arthropodsinhabiting papaya foliage

Received: 28 October 2002 / Accepted: 4 February2003 / Published online: 11 March 2003 ? Springer-Verlag2003

Abstract We examined the potential of a leaf roller to Introduction indirectly influence a community of . Two mite species are the key herbivores on papaya leaves in Ecologists are increasingly appreciating the importance of Hawaii: a spider mite, Tetranychus cinnabarinus Boisdu- indirect interactions in structuring communities val, and an eriophyid mite, Calacarus flagelliseta, which (Strauss 1991; Wootton 1994; Polis and Winemiller 1996; induces upward curling of the leaf margin at the end of Janssen et al. 1998). For communities in the summer when populations reach high densities. A particular, indirect interactions can be mediated by survey and three manipulative field experiments demon- shelters created by one species and subsequently used strated that (1) leaf rolls induce a consistent shift in the by another species (Damman 1993). spatial distribution of spider mites and their predators, the Shelter building through leaf rolling, folding, or tying coccinellid Stethorus siphonulus Kapur, the predatory is a common behavior among several arthropod taxa, mites Phytoseiulus spp., and the tangle-web building including caterpillars (Lepidoptera), sawflies (Hymenop- spider Nesticodes rufipes Lucas; (2) the overall abun- tera: Pamphiliidae), weevils (Coleoptera: Attelabidae), dance of spiders increases on leaves with rolls; (3) the and herbivorous mites in the superfamily Eriophyoidea specialist predators Stethorus and Phytoseiulus inhabit the (Acari) (Frost 1959; Keifer et al. 1982; Castagnoli 1996; rolls in response to their spider mite prey; and (4) the Westphal and Manson 1996). By protecting residents spider inhabits the rolls in response to the architecture of from natural enemies, providing a suitable microenviron- the roll itself. This study shows the importance of indirect ment for development, or reducing effects of plant effects in structuring a terrestrial community of herbi- defenses, leaf shelters may have strong impacts on the vores. survivorship, abundance, distribution, and diversity of other plant-dwelling arthropods(reviewed in Fukui 2001). Keywords Herbivory Leaf rolling * Plant architecture* For example, leaf rolling by early-season lepidopterans Spider facilitates later colonization and survival by late-season lepidopterans (Cappuccino 1993). Leaf rolls made by lepidopteran larva on cottonwood trees enhance arthropod abundance sevenfold and biodiversity fourfold (Martinsen V. Foumier (X) - J. A. Rosenheim et al. 2000). Leaf rollers exemplify "bioengineers" or Departmentof Entomology, "physical ecosystem engineers," organisms that directly University of California, CA 95616 Davis, USA or indirectly control the availability of resources to other e-mail: [email protected] Fax: 1-530-7521537 organisms by causing physical state changes in biotic and abiotic materials (Jones et al. 1997). V. Foumier * L. 0. Laney * M. W. Johnson Here we report a study in which a herbivorous mite Departmentof Plant and EnvironmentalSciences, modifies the architecture of its host plant, papaya, by University of Hawaii at Manoa, HI 96822 Honolulu, USA inducing leaf rolls. We conducted a survey and three manipulative field experiments to address the following J. Brodeur questions: (1) Do leaf rolls modify the spatial distribution Departementde Phytologie, of foliar arthropods in the community? (2) Do leaf rolls Gl K 7P4, Canada Universite Laval, Ste-Foy, Quebec, change the population densities of arthropods in the Present address: community? (3) Do predators colonize leaf rolls in M. W. Johnson, Departmentof Entomology, response to the architecture of the rolled leaf itself or in University of California, response to spider mite prey? Riverside, CA 92521, USA

This content downloaded from 169.237.77.249 on Thu, 4 Sep 2014 13:15:17 PM All use subject to JSTOR Terms and Conditions 443 Materialsand methods leaves (V. Fournier, unpublished data). These herbivorous mites tend to shift to the upper (adaxial) leaf surface when densities on Study system the lower surfacebecome high (>10,000 adults/leaf).When moving from the lower to the upper surface of the leaf, the mites first infest Papaya (Carica papaya L., Caricaceae) is a short-lived perennial the upper leaf margin and induce the edge to curl upward.Curling crop, native to Central America (Storey 1976). Papaya trees in may occur on some or all of the fingers of a leaf, with a heavily commercialorchards typically consist of a single erect stem with a infested leaf supportingfrom 25 to 45 rolled leaf segments. When terminalcrown of about 25-30 large leaves. The midrib of mature leaf rolling is very severe, the rolls may form complete tubes. Rolls leaves reaches 35-40 cm in length, and the total leaf area averages are about 10 cm in length and 2 cm in diameter. Usually, only the approximately 1,000 cm2. Leaves are held on long petioles (45- leaves located in the middle and lowest parts of the canopy crown 70 cm in length) that extend horizontally from the trunk. New are affected by rolling (V. Foumier, personal observation). leaves are producedyear-round, emerging from the growing point Calacarus infests papaya all year long, but leaf rolling is mostly at the tip of the trunk.They are palmatelylobed and usually possess observed in the late summer, when population densities peak (V. seven major fingers, each of which is subdivided into 3-8 minor Foumier, unpublisheddata). fingers, for a total of 40-50 fingers per leaf (Fig. 1). The life span of When papaya leaves are unrolled, the spider mite and its a papayaleaf varies from 75 to 125 days in a pesticide-freeorchard predators usually inhabit the lower leaf surfaces (Esguerra and (V. Foumier, unpublisheddata). Haramoto 1980). However, when the leaves are rolled by The community of arthropods inhabiting papaya foliage in Calacarus mites, preliminaryobservations suggested that members Hawaii includes two key herbivorous mites: the host-specific of the arthropodcommunity may colonize the upper surfaces of eriophyid mite, Calacarusflagelliseta Fletchmann,DeMoraes, and leaves within the leaf rolls and coexist there with the leaf edgeroller Barbosa (Acari: Eriophyidae),and the polyphagouscarmine spider mites. mite, Tetranychuscinnabarinus Boisduval (Acari: Tetranychidae). The most common predatorsin the system are the predatorymites Phytoseiulus macropilis Banks and P. persimilis Athias-Henriot Experimentalsite (Acari: Phytoseiidae), the Stethorus siphonulus Kapur (Coleoptera: ), and the tangle-web building spider Our researchwas conducted at the University of Hawaii Poamoho Nesticodes rufipes Lucas (Araneae: Theridiidae) (Esguerra and Experimental Station on Oahu, Hawaii, during the summers of Haramoto 1980; Rosenheim et al., unpublisheddata). The phyto- 2001 and 2002. Trees used in this study were from pesticide-free seiid mites and the coccinellid are specialist consumers of spider papaya plots (Solo variety, cv. X77). They were 8 months old and mites (Prasad 1973; Raros and Haramoto1974), whereas the spider approximately1.75 m tall in 2001 and 20 months old and 3.5 m tall Nesticodes, which spins a sparse and hardlydiscernable web at the in 2002. junction of leaf veins, is a generalistthat preys upon a wide variety of arthropods(Barreto et al. 1987; Cushing and LeBeck 1994; Fox 1998), including the carmine spider mite (Esguerraand Haramoto Survey 1980), Stethorus,and Phytoseiulus spp. (Rosenheim et al., unpub- lished data) (Fig. 2). No naturalenemies of the eriophyid mite are We measured arthropod distribution and abundance on papaya known to occur in Hawaii (V. Fournier,personal observation). leaves with and without rolls. We censused , mites, and The eriophyid Calacarus can colonize both the surfaces of arachnids on 13-16 August 2001, when the density of the leaf papaya leaves and the fruits (C. citrifolii in Jeppson et al. 1975; edgeroller mite population peaked and leaves began to roll. We Fletchmannet al. 2001), but it prefersthe lower (abaxial)surface of surveyed 50 mid-crown leaves, 25 with rolls and 25 without rolls,

F-qb-?.to

4 1~~~~~~-14 7

11r# s'

Fig. 1 Carica papaya leaf and a leaf roll. Papaya leaves are densities of Calacarusflagelliseta peak and trigger the edge of the palmately lobed and usually possess seven major fingers, each of leaves to curl upward,each one of the minor fingers can roll and which is subdivided into three to eight minor fingers. When form a complete tube

This content downloaded from 169.237.77.249 on Thu, 4 Sep 2014 13:15:17 PM All use subject to JSTOR Terms and Conditions 444 were randomly assigned to one of three treatments:(1) no rolls Nesticodes (unmanipulatedcontrol); (2) presence of wires or rubberbands but no rolls (wire/rubberband control); or (3) artificial rolls. Identical numbers of rubberbands or wire rings were used in treatments2 and 3. Experiment la comprised 30 replicatesof each treatment(total of 90 leaves) distributedover 22 trees. Experimentlb comprised51 Stethorus Phytoseiulus spp. replicates of each treatment(total of 153 leaves) over 37 trees. All experimental leaves were located at the mid-crown height of the trees to standardize leaf age and reduce the effect of leaf senescence. Experimentsran until the experimentalleaves started showing signs of senescence and/or naturalrolling. Experimentla Tetranychus Calacarus ran for 15 days, and experiment lb ran for 26 days (at 30 ?C the carmine spider mite requires6-7 days to develop from egg to adult [Hazan et al. 1973]). Arthropoddensities on upper and lower leaf surfaces were estimatedweekly using the non-destructivesampling a\/S proceduresdescribed above. Artificialrolls were sampledby gently Papaya foliage removing the rubberbands/wire rings, counting the arthropods,and then carefully replacingthe roll; this producedminimal disturbance Fig. 2 Simplified food web of the key herbivores and predators to the residentarthropods. Missing rubberbands or wire rings were occurring on papaya leaves replaced once a week.

Manipulativeexperiment: mechanisms behind distributedover several trees. We sampled both rolled and unrolled the predatorresponses leaves located on the same tree. individuals of the following arthropods:(1) the We counted We conducted a second manipulativeexperiment (experiment 2) herbivores: the eriophyid mite Calacarus (all motile instars, from 2-11 July 2002 to determine whether the predatorsPhyto- subsampling technique described below) and the carmninespider Stethorus,and Nesticodes colonize rolls in response to the mite Tetranychus (adult females); and (2) the predators: the seiulus, of spider mite prey within rolls or in response to the roll Phytoseiulus spp. (adult females), the beetle presence predatory mites We set 15 replicates of each of the following including eggs), the tangle-web building architecture. up Stethorus (all stages, treatments:(1) no rolls, no spider mites on the upper leaf surface; Nesticodes (motile instars), and other spiders, including spider (2) artificialrolls, spider mites on the upperleaf surface (i.e., in the spiders (Araneae: Salticidae) (motile instars). primarily jumping and artificial rolls, no spider mites in the rolls. Thus the densities were assessed on upperand lower leaf surfaces. rolls); (3) Arthropod was an incomplete 2x2 factorial design in which the mite populationscan be very large (up to 100,000 mites/ experiment Eriophyid treatment"no rolls, spider mites on the upper leaf surface" was A. unpublished data). Thus, we subsampled leaf; J. Rosenheim, because spider mites are difficult to maintainon the upper a 2.5x2.5 cm grid (collapsible magnifier,Bioquip missing individualsusing leaf surface when rolls are absent. For treatments 2 and 3, 10 Gameda, Calif., USA) as a subsample unit. We Products Inc., leaf were rolled wire rings. In treatmentsI and 3, four subsamples on each side of the leaf and fingers per using randomly selected were removed from the upper surface daily by using a Calacarus. We also recorded the spider mites counted all motile instars of Densities of Nesticodes, Stethorus,and Phytoseiuluson fingers that were rolled on each sampled leaf. paintbrush. number of upper leaf surfaces were recorded every morning, and all spiders was when some but not all of the mid- Our survey performed were subsequentlyremoved to prevent predationon Stethorus. crown leaves of the orchardwere rolled. There are two potential problems associated with this non-manipulativeapproach. First, our to choose equivalent leaves in the mid-crown despite attempts Statistical analysis canopy, rolled leaves are likely to be older than unrolled leaves take some time to roll. Second, high densities of because leaves Data from the and experiments Ia and lb did not meet the cause the rolled leaves to senesce quickly (V. survey Calacarus mites of normality and homogeneity of variance. We data). Each of these factors may influence assumption Fournier, unpublished thereforeused Wilcoxon rank-sumtests to compare in addition to the rolled versus unrolled state. non-parametric arthropoddensities densities on rolled versus unrolledleaves (JMP 2000). In problems with such confounding variables by arthropod We circumvented la and Ib, we tested for main effects of leaf rolling on in which we created artificial papaya leaf experiments conducting experiments the density of arthropodsby summingacross all samples takenfrom rolls. the end of the first week until the end of the experiments. For experiment 2 we evaluated the effects of leaf rolls and mites on the mean density of Stethorus,Phytoseiulus, and distributionand abundance spider Manipulativeexperiment: spatial Nesticodes over the course of the experiment using a two-way ANOVA (JMP 2000) with rolls and spider mites as the main twice in differentpapaya fields: 3-18 We repeatedthis experiment effects. For the statisticalanalysis was performedon the and 16 July-9 August 2001 (experiment Stethorus, July 2001 (experiment la), mean of larvae, and pupae combined. We excluded the manipulativeexperiments just priorto density eggs, lb). We chose to conduct adultcoccinellids because they can be very mobile (J.A. Rosenheim of natural leaf rolling so that we could observe an the onset et al., unpublished) and were commonly found on upper leaf similar to that observed under naturalleaf rolling. arthropodfauna surfaces in all treatments during the course of this experiment We made artificial rolls using either rubberbands (experiment la; (F2,45=0.65,P=0.53). 12PCI, Continental Western Corporation, San Leandro, Calif., USA) or wire rings (experiment lb; Steelwire galvanized, 28 gauge, Elco, Textron Logistics Co., Rockford, Ill., USA). Prelim- inary tests indicatedthat rubberbands and wires were not dislodged by the wind and did not damage the leaves. The experimentalunit was a single leaf. Ten and 15 fingers were rolled per leaf in experiment la and experiment lb, respectively. We also controlled for a possible effect of rubberbands and wire rings by fitting leaf fingers with rubberbands or wires without forming rolls. Leaves

This content downloaded from 169.237.77.249 on Thu, 4 Sep 2014 13:15:17 PM All use subject to JSTOR Terms and Conditions 445 Results * No rolls El Rolls A. Calacarus Survey 400- T On average,rolled leaves (n=25) had 60%of theirfingers rolled (range:22-100%), whereasunrolled leaves (n=25) 200 - had 4.4% of their fingers rolled. The high incidence of leaf rolling across the field preventedus from finding leaves with none of theirfingers rolled; therefore, we will 0 use the terminology"lightly rolled" leaves versus "heav- B. Tetranychus ily rolled"leaves. Heavily rolled leaves had significantly ** ** higher densities of Calacarus on their upper surface 20- (X2=10.4,dfrl, P=0.001, n=10; Fig. 3A) and significantly lower densities of Calacarus on their lower surface , 10 - compared to lightly rolled leaves (X2=16.3, dftl, U) P=0.0001; n=10; Fig. 3A). o - Heavily rolledleaves had significantlyhigher densities Z~. 0 _ of spider mites on their upper surface (X2=11.3,df=l, 6 P=0.0008; n=25; Fig. 3B) and significantlylower densi- n C. Stethorus N.S. ties of spider mites on their lower surface comparedto 0 N.S. lightly rolled leaves (X2=7.7, df=l, P=0.005; n=25; 0 Fig. 3B). Heavily rolled leaves had higher densities of 2-- the beetle Stethoruson their uppersurface (X2=3.7, dfrl, P=0.054; n=25), whereas similar densities of the beetle occurred on the lower surfaces of heavily and lightly 0 rolled leaves (Fig. 3C). Nesticodes dominatedthe spider community(99% of 9 D. Nesticodes ** all sampledindividuals [n=277]). Nesticodes were almost completelyabsent from the uppersurface of lightly rolled 6- N.S. leaves but achievedhigh densitieson the uppersurface of heavily rolled leaves (X2=21.8,dftl, P=0.0001, n=25; 3- Fig. 3D). Nesticodes densities on the lower surfaces of heavily rolled and rolled leaves did not and lightly differ, 0 thus, total Nesticodes density (upper + lower leaf Upper Lower Entire surfaces)increased substantially on heavily rolled leaves surface surface leaf (X2=7.7, df=l, P=0.006, n=25; Fig. 3D). Fig. 3A-C Survey of heavily rolled (rolls) versus lightly rolled leaves (no rolls). Shown are the mean densities (+1SE) of arthropodson the upper leaf surface, lower leaf surface, and both Manipulativeexperiment: spatial distribution leaf surfaces (entire leaf) of A Calacarus flagelliseta (motile and abundance instars);B Tetranychuscinnabarinus (adults); C Stethorussiphonu- lus (all instars combined); and D Nesticodes rufipes (motile Experiment la instars). The symbols above the error bars report the results of pairwise contrastsperformed with Wilcoxon rank-sumtests, where * P

This content downloaded from 169.237.77.249 on Thu, 4 Sep 2014 13:15:17 PM All use subject to JSTOR Terms and Conditions 446

Experiment 1a Experiment lb 0.3 - 100- A. Stethorus A. Tetranychus 100- B. Tetranychus * No rolls, no prey 7 No rolls E Rubber band/wire control 75- 0.2- Rolls, no prey L: Artificialrolls N.S. E3 Rolls, prey 802 L 250 N.S. 0.1 25 ~~~~~~~~~~25- _ N.S. 0 0 C C. Stethorus D. Stethorus N.S. N.S.

E. Nesticodes *** 15- T B. Phytoseiulus

(D 2- CZ C 20\L] L 10 CD 0.6-

T 0J 0 0 - ~ N.S. N.S. 0.3- <0 0

E. Nesticodes 15iF. Nesticodes CD e 7.5- (0 N.S. ~0 10 0L 0 5- C. Nesticodes

2.5 -5 N.S. 2- 0- 0 o g Upper surface Entire leaf Upper surface Entire leaf

Fig. 4A-F Influence of artificial leaf rolls on the summation of mean densities (?iSE) across 3-18 July (experiment la) and 16 July-9 August 2001 (experiment lb) of A, B Tetranychus cinnabarinus D Stethorus (adults), C, siphonulus (all instars 0 combined), and E, F Nesticodes rufipes (motile instars) on the leaf surface and entire leaf + upper the (upper lower leaf surfaces). Fig. 5A-C Mechanisms behind Treatments included no solid no rolls predatorresponses to leaf rolls: do (1) rolls, bars; (2) but the predatorscolonize rolls in of rubber bands or wires bands/wires response to their spider mite prey or presence (rubber control), in to the roll and response architecture?Shown are daily mean densities hatched bars; (3) artificial rolls, open bars. Each treatment was A 30 times in la and 51 times in (+1SE) of Stethorus siphonulus (eggs, larvae, and pupae), B replicated experiment experiment lb. Phytoseiulus The above the error bars of spp. (adults), and C Nesticodes rufipes (all motile symbols report the results pairwise instars) on the upper surface of contrasts with Wilcoxon rank-sum tests 2 (1) unmanipulatedleaves (no rolls, performed (treatments no prey; solid bars), (2) artificially rolled leaves with and 3 were compared to the control * P<0.05, ** no spider [1]); P<0.001, mites in the rolls (rolls, no prey; open and and n.s. not significant at a=0.05 bars), (3) artificially rolled leaves with spider mites in the rolls (rolls, prey; hatched bars). Each treatmentwas replicated 15 times Experiment lb densities on the lower surface of leaves (data not shown). We established the same three treatments as in experi- We do not know why the wires produced this positive ment 1a, with the exception that artificial rolls were effect on the lower surface of unrolled leaves (with wires) created using wires. We preferred wires to rubber bands and not on the lower surface of artificially rolled leaves. because we could produce artificial rolls that more closely Despite the direct effect of wires, the results obtained in mimicked the Calacarus-produced rolls. We found, experiment lb were congruent with those obtained in however, that the wires did have a positive, direct effect experiment la (Fig. 4). on spider mite and spider densities (Figs. 4B, F). Direct For all treatments, initial densities of spider mites were effects of the wires on arthropod abundance on the upper 10 adult females per leaf, and, with the exception of one surface of leaves were modest in magnitude (Fig. 4B); individual found on the upper surface, all spider mites thus, we expect that most of the effects of the artificial were located on the lower leaf surface. A substantial leaf rolls can be attributed to leaf architecture rather than population of spider mites became established on the the presence of wires. However, the magnitude of the wire upper leaf surface of artificially rolled leaves (X2=52.78, effect was larger at the level of the entire leaf (Figs. 4B, df=l, P=0.0001; Fig. 4B), generating an increase in the F), because the wires increased spider mite and spider total population of spider mites on rolled leaves (X2=3.78,

This content downloaded from 169.237.77.249 on Thu, 4 Sep 2014 13:15:17 PM All use subject to JSTOR Terms and Conditions 447 df= I, P=O.OS; Fig. 4B). Artificially rolled leaves had to Calacarus.The leaf rolls do shield Calacarus mites significantly higher densities of the predatory beetle from directexposure to sun and desiccation.This may be Stethorus on their upper leaf surface compared to unrolled important,given that Calacarus prefers to inhabit the leaves (X2=32.4, df-=1, P=0.0001; Fig. 4D), but this shadedlower surfacesof papayaleaves and never builds increase did not translate into higher densities at the level to high populationson the upper surfaces of the leaves of entire leaf (upper + lower leaf surfaces; Fig. 4D). until rolls appear(V. Fournier,unpublished data). Once again, the spider community was dominated by Nesticodes, which constituted 96.8% of all sampled individuals (n=2,382). At the beginning of the experi- Use of leaf rolls by spidermites ment, 99.5% of the spiders were recorded on lower leaf surfaces. As in experiment 1a, Nesticodes' response to the We demonstratedthat leaf rolling consistently led to leaf-rolling treatment was strong: spiders colonized the spidermites rapidlycolonizing the upperleaf surfaceto upper leaf surfaces (i.e., artificial rolls) (X2=52.07, dftl, inhabitthe rolls, whereas unrolledpapaya leaves rarely P=0.0001; Fig. 4F), and the total population density harboredspider mites on their upper surface.Moreover, (upper + lower leaf surfaces) increased significantly overall densities of spider mites tended to be higher on (XW=5.02,df.=l, P=0.02; Fig. 4F). rolledleaves thanon unrolledleaves, suggestingthat such leaves may offer some advantagesover unrolledleaves. We discuss two non-exclusivehypotheses to addresswhy Mechanisms behind the predator responses spidermites move into leaf rolls. First, spider mites may colonize leaf rolls because The two specialist predators appeared to move into leaf these structures produce a microenvironmentthat is rolls in response to the presence of their spider mite prey similar to their normal habitat, the lower leaf surface. (Stethorus: spider mite effect: F1,45=3.17, P=0.08; roll Like spider mites residing on the lower leaf surface, effect: F1,45=0.35, P=0.56; Fig. SA; Phytoseiulus spp.: individualsinhabiting leaf rolls achieve protectionfrom spider mite effect: F1,45=17.96, P=0.0001; roll effect: directexposure to sunlight,rain, and wind. These climatic F1,45=0.17, P=0.68; Fig. SB). The small and non-signif- factors can be harmful to spider mites in the genus icant increases in the densities of Stethorus and Phyto- Tetranychus(Putman 1970; Tulisalo 1974;Barcelo 1981). seiulus on rolled leaves from which we attempted to Moreover, preliminarymeasurements suggest that leaf remove spider mites were likely in response to the spider rolls mightoffer an even more suitablemicroenvironment mites that colonized the leaf rolls between the daily for spidermites than is providedby the lower surfaceof removals. In contrast, the spider Nesticodes moved into leaves, becausetemperatures inside leaf rolls are slightly leaf rolls in response to the roll architecture, with no higher than on the lower leaf surface (inside rolls: additional response to the presence of spider mite prey 34.26?0.27 ?C; outside rolls on upper surface: (spider mite effect: F1,45=0.15, P=0.7; roll effect: 33.96?0.45 ?C; lower surface: 32.94?0.28 ?C; n=5; F1,45=14.92, P=0.0004; Fig. 5C). F=4.05, df=2, P=0.045; measured during the daytime). Spider mites develop more rapidly with increasing temperatureuntil an upper thresholdis reached (Hazan Discussion et al. 1973; Holtzeret al. 1988; Roy et al. 2002). Finally, leaf roll curvature also provides suitable attachment The eriophyid mite Calacarus causes community-level points for spidermite webbing. indirect effects mediated through its production of leaf Second, spider mites may move into leaf rolls to rolls. Both our survey and manipulative experiments escape specialist predators.The idea that leaf rolls may showed that leaf rolls altered the spatial distribution of act as a refuge from predationis supportedby some spider mites and their specialist predators, Phytoseiulus studies (Heads and Lawton 1985; Damman 1987; Alte- spp. and Stethorus, and enhanced the total abundance of grim 1992; Eubanks et al. 1997). Our study was not the tangle-web spider Nesticodes. Moreover, our study designed to assess the risk of predationexperienced by demonstrated that shifts in arthropodcommunity structure spider mites on rolled versus unrolledleaves. However, can take place over a very short period of time (i.e., other work conducted in this system suggests the within 15-26 days). Finally, both specialist predators possibilitythat the colonizationof leaf rolls by Nesticodes appeared to colonize rolls in response to the presence of may protect spider mites in rolls from predation by their spider mite prey, whereas the tangle-web spider Stethorus. Nesticodes are predators of Stethorus and can seemed to inhabit rolls in response to the architecture of disruptthe ability of Stethorusto suppressspider mite the roll itself. populations(J.A. Rosenheimet al., unpublished;Rosen- In this study we did not attempt to study the direct heim and Corbett2003). Additionalwork is necessaryto interactions between Calacarus and its papaya host plant. evaluate the hypothesis that spider mites experience We do not know, for example, whether leaf rolling is reducedrisk of predationin rolled leaves. actively initiated by Calacarus or whether it is simply a plant response to feeding damage generated by the mite. Furthermore,we do not know if the rolls provide benefits

This content downloaded from 169.237.77.249 on Thu, 4 Sep 2014 13:15:17 PM All use subject to JSTOR Terms and Conditions 448 Use of leaf rolls by specialistpredators Knowledgeof the sourceof spidersinhabiting the rolls would contributeto understandingmicrohabitat selection We found that both specialist predatorsof the carmine by Nesticodesand the mechanismsleading to an increase spider mite, Phytoseiulus and Stethorus, moved into the in its total abundance.Nesticodes that colonized exper- rolls in response to their prey ratherthan in quest of a imentalrolls may have come from populationsfound on shelter.A similarquestion was addressedby Martinsenet (1) the lower surfaceof rolled leaves, (2) unrolledleaves al. (2000), who examined arthropodabundance and fromthe same tree, (3) unrolledleaves fromother papaya diversityin leaf rolls on cottonwoodinhabited by larvae trees, or (4) otherplant species in the habitat.Nesticodes of the leaf roller Anacampsis niveopulvella (Chambers) is commonin manyagricultural systems on Oahu,Hawaii (Lepidoptera:Gelechiidae). They foundthat experimental (Hooks and Johnson 2002). Leaf rolling might simply rolls supporteda similar numberof taxa and an even redistributespiders at the level of the papaya field, or greaterabundance of arthropodsthan naturalrolls. They mightactually generate an increasein the totalpopulation arguedthat the presence of arthropodsin leaf rolls was densityof Nesticodes.In the lattercase, leaf rollingmight motivatedby the protectionthey providedrather than the have consequences for herbivoryimposed on papaya. food (lepidopteranlarvae) they contained. However, a LargerNesticodes populations might increase spider mite more detailedanalysis revealed that while the abundance densities,as Nesticodescan disruptthe naturalcontrol of of herbivoreswas higherin artificialrolls than in natural spidermites by preyingupon Stethorus(J.A. Rosenheim rolls, the abundanceof predatorswas actually lower in et al., unpublished;Rosenheim and Corbett2003). artificial rolls, suggesting that some predatorsmay be attractedby food ratherthan shelter. Herbivorousmites as ecological engineers

Use of leaf rolls by spiders Jones et al. (1994) have suggestedthat the impactof an ecological engineerdepends upon the spatialand tempo- Our study showed a strong positive response of Nesti- ral scale of its actions.They listed severalfactors likely to codes to leaf rolling: leaf rolling not only modified the scale the impactof engineering:(1) the lifetimeper capita spatial distributionof Nesticodes but also enhanced its activity of individualorganisms; (2) the density of the total abundance. We also demonstratedthat spiders engineeringpopulation; (3) the spatialdistribution, both colonize leaf rolls in responseto the roll architecture. locally and regionally,of the population;(4) the lengthof Both prey availabilityand the physicalstructure of the time the populationhas been presentat the site; (5) the habitat can influence spider populationdensities (Wise durabilityof the engineeredconstruct in the absence of 1993). Spiders can display numericalresponses through the original engineer; (6) the number and types of either aggregationor reproduction(Olive 1982; Rypstra resourceflows that are modulatedby the constructs;and 1985; Wise 1993), and vegetation structurehas been (7) the numberof other species dependentupon these shown to influence species composition,abundance, and flows. The herbivorousmite Calacarusdoes reach large habitat selection by spiders (Robinson 1981; Rypstra populationdensities on papaya leaves (100,000 mites/ 1983, 1986; Greenstone 1984; Gunnarson1990; Wise leaf; J.A. Rosenheim,unpublished data) and is generally 1993;Halaj et al. 1998, 2000; Rypstraet al. 1999;McNett distributedrelatively uniformly in papayafields, produc- and Rypstra2000; Sunderlandand Samu 2000). Most of ing field-wide leaf rolling (V. Fournier, unpublished these studies suggest that the more complex the vegeta- data). However, the physical engineeringby Calacarus tion structure,the more diverse and/or abundantthe gives rise to structuresthat persist for only approximately arachnidfauna. McNett and Rypstra(2000) showed that 4-6 months per year in papaya.Furthermore, although the primaryfactor in the selectionof a new habitatby the leaf rolling seems to modulate resource flows by orb weaverArgiope trifasciata Forskal (Araneae: Aranei- increasingspider abundance, Nesticodes is not dependent dae) was the structureof the vegetationrather than prey upon these flows to colonize papayaleaves. Therefore, availability.More complex habitatsmay attractprey or with regardto Joneset al.'s (1994) criteria,the ecological provide shade, protection from the wind, spaces for and evolutionaryconsequences of engineering by the retreats,or attachmentsites for the snaresof web weavers eriophyid mite in the papaya agroecosystemmay be (Wise 1993). relatively modest. Nevertheless, leaf rolling, although Unexpectedly,the wire ringswe used to makeartificial transient,may have large, direct short-termeffects on rolls on papayaleaves significantlyincreased Nesticodes spidermites and Nesticodesas well as indirecteffects on densities, probably by providing additional web-site Stethorus and Phytoseiulus. In years when spider mite attachmentpoints. In contrast,the rubberbands did not outbreaksoccur, indirecteffects may also extend to the generate a direct effect. Therefore,it is likely that we performanceof the host plant. Thus, Calacarus, by overestimatedthe effects of leaf rolls on Nesticodes in producingleaf rolls, may have importantindirect effects experiment lb. The remaining community members on this plant-herbivore-naturalenemy community. showed consistentresponses to artificialrolls made with either rubberbands or wires. Acknowledgements We thank Jo Diez, Cerruti Hooks, and the staff at the Poamoho Research Station of the University of Hawaii

This content downloaded from 169.237.77.249 on Thu, 4 Sep 2014 13:15:17 PM All use subject to JSTOR Terms and Conditions 449 for their technical assistance. The manuscript was improved by Holtzer TO, Norman JM, PerringTM, Berry JS, Heintz JC (1988) comments from Cerruti Hooks, Sarina Jepsen, Gail Langellotto, Effects of microenvironment on dynamics of spider mite Teresa Leonardo, Chris Matthews, Ken Spence, and Andy Zink. populations.Exp Appl Acarol 4:247-264 Eriophyid specimens were identified by James Amrine. This work Hooks CRR, Johnson MW (2002) Lepidopteranpest populations was supportedby a USDA-ARS grant for Minor Crops Research and crop yield in row intercroppedbroccoli. Agric For Entomol (Agreement no. 59-5320-9-226) and scholarshipsfrom the Natural 4:117-125 Sciences and Engineering Research of Canada (NSERC) and the Janssen A, Pallini A, Venzon M, Sabelis MW (1998) Behaviour Fonds de Recherchesur la Natureet les Technologies du Quebec to and indirect interactions in food webs of plant-inhabiting V.F. arthropods.Exp Appl Acarol 22:497-521 Jeppson LR, Keifer HH, Baker EW (1975) Mites injurious to economic plants. University of California Press, Berkeley, Calif. References JMP User's Guide, Version 4 (2000) SAS Institute Inc., Cary, NC Jones CG, Lawton JH, Shachak M (1994) Organismsas ecosystem Altegrim 0 (1992) Mechanisms regulating bird predation on a engineers. Oikos 69:373-386 herbivorouslarva guild in boreal coniferous forests. Ecography Jones CG, Lawton JH, Shachak M (1997) Positive and negative 15:19-24 effects of organisms as physical ecosystem engineers. Ecology Barcelo JA (1981) Photoeffects of visible and ultravioletradiation 78:1946-1957 on the two-spotted spider mite, Tetranychus urticae. Pho- Keifer HH, Baker WE, Kono T, Delfinado M, Styer WE (1982) An tochem Photobiol 33:703-706 illustrated guide to plant abnormalities caused by Eriophyid Barreto M, Barreto P, D'Alessandro A (1987) Predation on mites in North America (Agriculture Handbook No. 573) Rhodnius prolixus (Hemiptera: Reduviidae) by the spider United States Departmentof Agriculture,Washington, DC Theridion rufipes (Araneae: Theridiidae). J Med Entomol Martinsen GD, Floate KD, Waltz AM, Wimp GM, Whitham TG 24:115-116 (2000) Positive interactions between leafrollers and other Cappuccino N (1993) Mutual use of leaf shelters by lepidopteran arthropodsenhance biodiversity on hybrid cottonwoods. Oeco- larvae on paper birch. Ecol Entomol 18:287-292 logia 123:82-89 Castagnoli M (1996) Omamental coniferous and shade trees. In: McNett BJ, Rypstra AL (2000) Habitat selection in a large orb- Lindquist EE, Sabelis MW, Bruin J (eds) Eriophyoid mites: weaving spider: vegetational complexity determines site selec- their biology, natural enemies and control. (World Crop Pest tion and distribution.Ecol Entomol 25:423-432 Series, vol 6) Elsevier, Amsterdam,pp 661-671 Olive CW (1982) Behavioralresponse of a sit-and-waitpredator to Cushing BS, LeBeck LM (1994) Foragingin cockroach sticky traps spatial variation in foraging gain. Ecology 63:912-920 by the spider Nesticodes rufipes Lucas (Araneae:Theridiidae): Polis GA, Winemiller KO (1996) Food webs: Integration of a super food resource. Acta Arachnol 43:49-55 patternsand dynamics. Chapmanand Hall, New York Damman H (1987) Leaf quality and enemy avoidance by the larvae Prasad V (1973) The role of Phytoseiulus macropilis (Banks) of a pyralid moth. Ecology 68:88-97 (Acarina:Phytoseiidae) in control of the spidermites in Hawaii. Damman H (1993) Patterns of interactions among herbivore Acarologia 15:401-405 species. In: Stamp NE, Casey TM (eds) Caterpillars:ecological Putman WL (1970) Effects of water and high humidity on the and evolutionary constraints on foraging. Chapman and Hall, Europeanred mite, Panonychus ulmi (Acarina:Tetranychidae). New York, pp 132-169 Can Entomol 102:955-961 EsguerraNM, HaramotoFH (1980) Distributionand populationof Raros ES, Haramoto FH (1974) Biology of Stethorus siphonolus papaya carmine spider mites Tetranychus cinnabarinus and Kapur(Coccinellidae: Coleoptera), a predatorof spider mites, their predators.Ann Trop Res 3:135-148 in Hawaii. Proc Hawaii Entomol Soc 21:457-465 EubanksMD, Nesci KA, Petersen MK, Liu Z, Sanchez HB (1997) Robinson JV (1981) The effect of architecturalvariation in habitat The exploitation of an ant-defended host-plant by a shelter- on a spider community: an experimentalfield. Ecology 62:73- building herbivore. Oecologia 109:454-460 80 Fletchmann CHW, De Moraes GJ, Barbosa FR (2001) A new Rosenheim JA, CorbettA (2003) Omnivory and the indeterminacy species of Calacaruskeifer (Acari: Eriophyiodea)on papaya in of predator function: can a knowledge of foraging behavior northeasternBrazil. Zootaxa 5:1-5 help? Ecology (in press) Fox I (1998) Predation on Aedes aegypti (Diptera: Culicidae) by Roy M, Brodeur J, Cloutier C (2002) Relationship between Theridionrufipes (Araneae:Theridiidae) in Puerto Rico. J Med temperature and development rate of Entomol 35:611-613 (Coleoptera:Coccinellidae) and its prey Tetranychusmcdanieli Frost SW (1959) life and insect naturalhistory. Dover, NY (Acarina:Tetranychidae). Environ Entomol 31:177-187 Fukui A (2001) Indirect interactions mediated by leaf shelters in Rypstra AL (1983) The importanceof food and space in limiting animal-plantcommunities. Popul Ecol 43:31-40 web-spider densities: a test using field enclosures. Oecologia Greenstone MH (1984) Determinants of web spider species 59:312-316 diversity: vegetation structuraldiversity vs. prey availability. RypstraAL (1985) Aggregationsof Nephila clavipes (L.) (Araneae, Oecologia 62:299-304 Araneidae)in relation to prey availability.J Arachnol 13:71-78 GunnarsonB (1990) Vegetation structureand the abundanceand Rypstra AL (1986) Web spiders in temperate and tropical forests: size distributionof spruce-living spiders. J Anim Ecol 59:743- relative abundance and environment correlates. Am Midl Nat 752 115:42-51 Halaj J, Ross DW, Moldenke AR (1998) Habitatstructure and prey Rypstra AL, Carter PE, Balfour RA, Marshall SD (1999) Archi- availability as predictors of the abundance and community tecturalfeatures of agriculturalhabitats and their impact on the organization of spiders in western Oregon forest canopies. J spider inhabitants.J Arachnol 27:371-377 Arachnol 26:203-220 Storey WM (1976) Papaya. In: Simmonds NW (ed) Evolution of Halaj J, Ross DW, Moldenke AR (2000) Importance of habitat crop plants. Longman, London, pp 21-24 structureto the arthropodfood web in Douglas-fir canopies. Strauss SY (1991) Indirect effects in community ecology: their Oikos 90:139-152 definition, study, and importance.Trends Ecol Evol 6:206-210 Hazan A, Gerson U, Tahori AS (1973) Life history and life tables SunderlandK, Samu F (2000) Effects of agriculturaldiversification of the carmine spider mite. Acarologia 15:414 440 on the abundance, distribution, and pest control potential of Heads PA, Lawton H (1985) Bracken,ant and extrafloralnectaries. spider: a review. Entomol Exp Appl 95:1-13 III. How herbivoresavoid ant predation.Ecol Entomol 10:29- 42

This content downloaded from 169.237.77.249 on Thu, 4 Sep 2014 13:15:17 PM All use subject to JSTOR Terms and Conditions 450 Tulisalo U (1974) Control of the two-spotted spider mite (Tetra- enemies and control. (World Crop Pest Series, vol 6) Elsevier, nychus urticae Koch) by high air humidity or direct contact Amsterdam,pp 231-242 with water. Ann Entomol Fenn 40:158-162 Wise DH (1993) Spider in ecological webs. CambridgeUniversity Westphal E, Manson DCM (1996) Feeding effects on host plants; Press, Cambridge gall formation and other distortion. In: Lindquist EE, Sabelis Wootton JT (1994) The natureand consequencesof indirecteffects MW, Bruin J (eds) Eriophyoid mites: their biology, natural in ecological communities. Annu Rev Ecol Syst 25:443-466

This content downloaded from 169.237.77.249 on Thu, 4 Sep 2014 13:15:17 PM All use subject to JSTOR Terms and Conditions