Habitat Complexity Modiwes Ant–Parasitoid Interactions: Implications for Community Dynamics and the Role of Disturbance

Oecologia (2007) 152:151–161 DOI 10.1007/s00442-006-0634-6

Habitat complexity modiWes ant–parasitoid interactions: implications for community dynamics and the role of disturbance

Elliot B. Wilkinson · Donald H. Feener Jr

Received: 4 November 2006 / Accepted: 27 November 2006 / Published online: 22 December 2006 © Springer-Verlag 2006

Abstract Species must balance eVective competition itat complexity was also studied, and demonstrated with avoidance of mortality imposed by predators or that the immediate negative impact of Wre on habitat parasites to coexist within a local ecological commu- complexity can persist for multiple years. Our Wndings nity. Attributes of the habitat in which species interact, indicate that habitat complexity can increase dominant such as structural complexity, have the potential to host competitive success even in the presence of parasi- aVect how species balance competition and mortality toids, which may have consequences for coexistence of by providing refuge from predators or parasites. Dis- subordinate competitors and community diversity in turbance events such as Wre can drastically alter habitat general. complexity and may be important modiWers of species interactions in communities. This study investigates Keywords Competition · Formicidae · Fire · whether the presence of habitat complexity in the form Trade-oVs · Indirect interactions of leaf litter can alter interactions between the behaviorally dominant host ants Pheidole diversipilosa and Pheidole bicarinata, their respective specialist dipteran Introduction parasitoids (Phoridae: Apocephalus sp. 8 and Apoceph- alus sp. 25) and a single species of ant competitor Trade-oVs resulting from variability in resource acqui- (Dorymyrmex insanus). We used a factorial design to sition, competition and mortality are central to our manipulate competition (presence/absence of competi- understanding of coexistence in local ecological com- tors), mortality risk (presence/absence of parasitoids) munities (Tilman 1994; Kneitel and Chase 2004). Spe- and habitat complexity (presence/absence of leaf cies coexist by trading oV the beneWts of one strategy litter). Parasitoid presence reduced soldier caste forag- for those of another in ways that minimize competition ing, but refuge from habitat complexity allowed and mortality while maximizing energy intake. Species increased soldier foraging in comparison to treatments with parasites or predators must therefore balance two in which no refuge was available. Variation in soldier conXicting demands: competing eVectively for limited foraging behavior correlated strongly with foraging resources and escaping parasitism/predation. Contin- success, a proxy for colony Wtness. Habitat complexity ued coexistence of such species in the community allowed both host species to balance competitive suc- depends on successfully balancing demands imposed cess with mortality avoidance. The eVect of Wre on hab- by competition and predation/parasitism. A mechanis- tic understanding of how organisms balance both competitor and predator-related trade-oVs is needed if we Communicated by Andrew Gonzales. are to understand forces controlling community structure and composition. E. B. Wilkinson (&) · D. H. Feener Jr A variety of factors have been posited to aVect how Department of Biology, University of Utah, X 257 South 1400 East, Salt Lake City, UT 84112, USA species balance these con icting demands, including e-mail: [email protected] organismal attributes such as host energy state (Alonzo 123 152 Oecologia (2007) 152:151–161

2002), competitive environment (LeBrun and Feener and leads in some cases to loss of resources to competi- 2002), and variation in parasitoid abundance through tors (LeBrun and Feener 2002; Orr et al. 1995). As a space and time (Morrison et al. 2000). However, physi- result, parasitoid presence can render some dominant cal attributes of the habitat such as structural complex- competitors ineVective, which may lead to heightened ity can work alongside organismal attributes to species diversity or altered community structure inXuence species’ abilities to balance the conXicting (Feener 2000; LeBrun 2005). The coincidence of indi- demands of competition and escape from parasitism/ rect eVects on competition that dwarf the direct eVects predation. Examples from many diVerent systems high- of mortality from parasitoids, and the fact that parasi- light the importance of mortality risk in inXuencing toids specialize on a single species, make ant–parasit- prey/host behavior (Fraser and Huntingford 1986; oid interactions especially likely to inXuence ant Lima and Dill 1990; Rothley et al. 1997; Ripple and community coexistence (Chase et al. 2002). However, Beschta 2004). In some systems, habitat complexity such trait-mediated indirect interactions are wide- enhances mortality risk by aggregating prey or provid- spread and thought to be important in the dynamics of ing favorable habitats/hiding places for predators most communities (Werner and Peacor 2003). Small- (Gosselin and Bourget 1989; Finke and Denno 2002). scale habitat complexity in the form of leaf litter has Habitat complexity may also oVset mortality risk by the potential to provide refuge for host ants and ame- increasing prey/host ability to escape from predators or liorate negative eVects from parasitoids, thereby allow- parasites. We focus on the role of habitat complexity in ing hosts to continue competing eVectively while also enhancing prey/host escape because habitat complexity avoiding parasitism. is likely to impede movement of the Xying predators/ This study investigates whether small-scale habitat parasitoids used in this study. complexity alters ant–parasitoid interactions in ways Habitat complexity may facilitate escape by provid- that can impact ant community structure. For two ing a refuge to which parasites/predators do not have diVerent host–parasitoid systems, we use a multi-fac- access. Examples of such refuge eVects on predator– tor experiment to determine the eVects of parasitoids, prey dynamics include the protective services that aca- habitat complexity and competitors on the number of rodomatia provide for non-plant feeding mites (Norton soldier ants defending and harvesting resources. Sol- et al. 2001), lowered aphid accessibility to predacious diers are the best choice to measure eVects of habitat lacewing larvae on architecturally complex grasses complexity on host ant–parasitoid–competitor interac- (Clark and Messina 1998), lowered intraguild preda- tions because parasitoids only attack soldiers tion in salt marshes with higher vegetation complexity (although release of alarm pheromone after attack (Finke and Denno 2002) and lower mortality of thysan- soon induces colony-wide response) and because sol- opteran prey species in complex shaped arenas (Hod- diers are critical in the defense and harvest of large dle 2003). Limited evidence suggests that habitat resources. Within the context of our multi-factor complexity can provide lepidopteran crop pests with experiment, we ask Wve questions regarding the rela- refuge from hymenopteran parasitoids (Andow and tionship between habitat complexity and ant–parasit- Prokrym 1990). In addition, multiple ant hosts of the oid interactions: (1) can habitat complexity beneWt the genus Pheidole may exhibit behavioral responses to colony by providing refuge from parasitoids and parasitoid presence by hiding in nearby leaf litter allowing soldiers to harvest resources when parasi- (D. H. Feener, personal observation). toids are present, or (2) does habitat complexity have Ant communities are model systems for studying a positive eVect on soldiers even in the absence of par- community ecology, yet little attention has been given asitoids? To address whether habitat complexity to the potentially important role of predation in such allows hosts to balance the conXicting demands of communities because of the common view that ants are competition and escape from parasitism, we ask: (3) relatively immune to predators (Hölldobler and Wil- can hosts in simple (no complexity) habitats balance son 1990; Holway 1999). However, a growing body of the conXicting demands of competition and parasitoid knowledge has provided support for top-down inXu- avoidance by competing at the same level as treat- ence on ant community coexistence and structure in ments with no parasitoids and no habitat complexity, the form of behavioral responses to Dipteran parasi- (4) can hosts in complex habitats balance competition toids, which specialize on a single, usually dominant, and parasitoid avoidance by competing at the same ant species (Feener 1981; Morrison 1999; LeBrun and levels as treatments with no parasitoids and no habitat Feener 2002; LeBrun 2005). Behavioral response to the complexity, and (5) do hosts compete any diVerently threat of mortality from parasitoid attack usually con- in complex and simple habitats when parasitoids are sists of colony-wide alteration of foraging behavior, absent? 123 Oecologia (2007) 152:151–161 153

Materials and methods mates is a pervasive phenomenon in ants and is known as alarm recruitment (Hölldobler and Wilson 1990). Study sites and systems Alarm recruitment typically consists of alarm phero- mones being released either alone or in conjunction This study focuses on two ant species of the Pheidole with odor trails when an enemy is encountered, and genus that host specialist Apocephalus parasitoids has been studied in detail with Pheidole dentata, which (Diptera: Phoridae), and was conducted at two sites in is also a host to Apocephalus parasitoids (Feener Chihuahuan desert habitat during July–September of 1981). Such a response is crucial to the foraging success 2003 and 2004. The host ants Pheidole diversipilosa and of dominant ants such as P. diversipilosa and P. bicari- Pheidole bicarinata coexist together in sites dominated nata, but can be severely inhibited by the presence of by oak, pine and juniper woodland near Portal, Ari- parasitoids (LeBrun and Feener 2002). D. insanus was zona. In 2003, P. diversipilosa was studied at 1,800 m in used because it is a common competitor around the National Forest land neighboring the Southwestern nest sites of P. diversipilosa and P. bicarinata, will chal- Research Station. In 2004, P. bicarinata was studied at lenge host ants for resources, but is not dominant to 1,700 m on land owned by the Southwestern Research either host and never wins competitive interactions Station. with hosts even in the presence of parasitoids (LeBrun 2003, 2005). By using D. insanus we could induce a pos- Experimental design itive competitive response in the host without fear of altering competitive outcome and thus confounding In 2003 we investigated how habitat complexity aVects our competition factor with parasitoid or habitat com- interactions between the host ant P. diversipilosa and plexity factors. its specialist parasitoid Apocephalus sp. 8 by forcing We measured the number of soldiers harvesting colonies to forage up into a plastic bin to retrieve cookie baits. We focused on the soldier caste because: cookie baits 50 cm from nest entrances. Placing baits (1) parasitoids only oviposit in soldiers, (2) soldiers are 50 cm away from colony entrances ensured that baits crucial to harvesting large resources that need to be were discovered and that colonies traversed a distance broken into smaller pieces for eYcient transport back during which they were susceptible to parasitoid to the colony by workers, and (3) soldiers are impor- attack. Foraging bins were 30 £ 60-cm Sterilite storage tant in the defense of a large and temporally stable containers, and had a 6-cm-diameter hole at one end resource from competitors. Cookie baits are an exam- that could be placed directly over colony nest ple of a large resource requiring participation from the entrances. Using foraging bins allowed us to: (1) mini- soldier caste to break resources into small pieces, and mize disturbance around nest sites and control exactly from the worker caste to carry small pieces back to the which resources P. diversipilosa was harvesting, (2) nest. We also weighed cookie baits before and after introduce or exclude competitors, and (3) introduce or their use in treatments and used these measures to exclude parasitoids from treatments using bridal veil to determine foraging success (total dry weight of 2 £ 2 cm cover the foraging bin. Our factorial design included pecan sandy cookie placed on a 10 £ 10 cm laminated two levels of habitat complexity (simple—no leaf litter, bait card 50 cm away from nest entrance minus remain- complex—approximately 5,000 cm3 oven-dried oak ing dry weight after 2.5 h foraging bout). These leaf litter), two levels of parasitoid abundance (parasi- response variables allowed us to quantify the eVect of toids excluded, two parasitoids introduced after at least treatments on both competitive response (number of one soldier had arrived at the cookie bait), and two lev- soldier ants recruited to cookie resources) and the els of competition (no competitors, Wve competitor resulting foraging success that is associated with it. Dorymyrmex insanus ants added every 10 min to Treatments were replicated on seven colonies. To mimic recruitment). control for change in cookie mass due to humidiWca- Our addition of leaf litter closely approximated the tion or evaporation on treatment days, all cookies were average leaf litter depth found around focal colonies. placed in a drying oven at 75°C for 72 h and then Introduced parasitoids were captured by aspiration at weighed to determine dry weight both before and after recruitment events instigated near unused host colo- treatments. Mean dry weight (§SD) before experi- nies. Introducing competitors allowed us to determine ments was 2.48 § 0.28 g. To control for environmental whether any eVects of habitat complexity on host–par- variation, colonies were randomly assigned the order asite interactions could alter colony response to com- in which they received treatments such that all colonies petitors (as seen by soldier build-up at the resource). on a given trial day received diVerent treatments. Response to competitors by alerting and attracting nest Colonies experienced 3-day treatment cycles: 1 day of 123 154 Oecologia (2007) 152:151–161 treatment implementation and 2 days of rest to control Southern Arizona (colony elevation 1,754–1,758 m). for colony energetic state after foraging on cookies. As Habitat complexity measurements were made using a it was not possible to monitor all seven colonies at once 50 £ 50-cm sampling grid placed 50 cm away from the due to time constraints and distance between colonies, colony nest entrance in each of the four cardinal direc- all replicates were divided roughly into two groups, tions. Sample grids were partitioned into twenty-Wve and replicate groups experienced treatments within 10 £ 10-cm squares, and measurements were taken at 24 h of each other to control for environmental condi- the center of Wve squares (the four corner squares and tions. All treatments were shaded to control for tem- the center square of the sample grid) in each direction. perature and humidity diVerences between colony We considered two aspects of habitat complexity: leaf locations. litter depth and structural complexity. At sample In 2004 we repeated the experimental design used in points, litter depth was measured by inserting a thin 2003 on P. bicarinata and its specialist parasitoid Apo- ruler through litter to the soil level, while structural cephalus sp. 25. To control for the surrounding compet- complexity was measured as the number of leaves itive community, the experiment was performed in skewered by a thin metal stake inserted through litter areas where P. bicarinata and Apocephalus sp. 25 co- and into the soil (modiWed from Smith 1985). Longer occur with dominant host P. diversipilosa and its term impacts of Wre on habitat complexity were mea- parasitoid Apocephalus sp. 8. Although hosts diVer in sured at 50 cm to the north of six sample points in dominance with respect to each other, they are both paired burned and unburned plots at four diVerent dominant to the rest of the ant community in which sites contained within the perimeter of two separate they co-occur (LeBrun and Feener 2002). Mean dry Wres (96 total sample points, average elevation weight (§SD) of cookies used for baits in 2004 was 1,536 m). These low-intensity Wres took place in high- 2.18 § 0.27 g and treatments were replicated on eight elevation Chihuahuan desert in Sonora, Mexico, in colonies. habitat that was broadly similar in its vegetative com- To investigate natural sources of variation in habitat position to that of our Arizona study sites (common complexity, we assessed the immediate and longer- plant genera included Quercus, Juniperus, Yucca, term impacts of Wre on habitat complexity both around Nolina, and various grasses). Measurements were host P. diversipilosa colonies and at long-term Wre plots taken in July 2005, three years after the initial burn in the Chihuahuan Desert. Burned and unburned sites took place. Finally, in order to characterize the habitat were chosen based on their similarity in slope and complexity experienced by colonies used in our study, aspect, both of which can aVect the likelihood or Wre we measured habitat complexity around eight colonies spread in a particular location (Chandler et al. 1983). of P. diversipilosa at our study site in the Chiricahua Although we have no data on the similarities of burned Mountains using methods identical to those in the and unburned vegetative communities around P. diver- Santa Rita Mountains. sipilosa colonies, vegetative studies conducted at our long-term Wre plots in the Chihuahuan Desert indicate Analysis that our attempt to control for slope and aspect resulted in largely similar vegetative communities Treatment means were calculated by averaging within 1 m of sample points (data not shown, Jaccard recorded values of soldiers over the 2.5-h period of coeYcient of similarity = 0.81). None of the species observation to obtain mean values for each colony, that were unique to burned or unburned plots existed then averaging again across colony replicates to obtain on more than 25% of those plots and thus were not an overall mean for a particular treatment. Recorded widespread. In addition, tree and shrub cover as well as values were averaged from the time that colonies dis- plant species richness did not diVer between burned covered the cookie bait for all treatments without par- and unburned plots (P > 0.2 in all cases). Similar vege- asitoids, and from the point of parasitoid introduction tative communities likely support similar ant communi- for all treatments in which parasitoids were present. ties because of overlap in the availability of nesting Analysis was conducted using a three-way ANOVA, sites, microclimate and resources (Bestelmeyer and both planned and post hoc comparisons, and regres- Schooley 1999). sion in SYSTAT (2004). All means were transformed Immediate impacts of Wre on habitat complexity [log (mean + 1)] in an attempt to meet homogeneity of were measured around six P. diversipilosa colonies in variance and normality assumptions. Planned and post habitat that burned in July 2005 and around an addi- hoc comparisons were carried out in the context of the tional six colonies in nearby unburned habitat located three-way ANOVA using the Specify command in on the eastern Xank of the Santa Rita Mountains in SYSTAT 11 (2004). Despite data transformation, some 123 Oecologia (2007) 152:151–161 155 heterogeneity remained between levels of our parasitoid factor (Levene’s test, P = 0.007). Because of this heterogeneity of variance, interaction eVects involving the parasitoid factor cannot be tested reliably without using local error terms. We therefore used a combina- tion of local and global error terms in testing speciWc hypotheses as recommended by Kirby (1993). Compar- isons containing data cells from only one level of the parasitoid factor incorporated local error terms from that single factor level. All other comparisons utilized the experiment-wide error term because heteroscedas- ticity was not encountered between any other cells or factors for P. diversipilosa, and was not encountered at all for P. bicarinata. Planned comparisons one and two were orthogonal and used typical 0.05 -levels. Post hoc comparisons 3, 4 and 5 used sequential Bonferroni and Dunn–Sidak methods to obtain adjusted -levels (Sokal and Rohlf 1995). These adjustments are meant to limit the experiment-wide type-I error rate to 0.05 when conducting multiple comparisons, and resulted in an adjusted level of 0.009 for each comparison and the 99.1% conWdence intervals presented below. Variation in habitat complexity (depth and number of leaves) between experimental additions, natural Fig. 1 The relationship between the number of a Pheidole diver- habitat around focal colonies, burned and unburned sipilosa and b Pheidole bicarinata soldier ants harvesting and defending a resource and foraging success over a 2.5-h period for habitat around host colonies in the Santa Rita Moun- both hosts tains, and burned and unburned habitat in the Chi- huahuan Desert was analyzed using paired and two- sample t-tests. Burned and unburned sites in the eVect). Added habitat complexity resulted in a margin- Chihuahuan Desert were paired, and thus were ana- ally signiWcant increase in soldiers at resources for P. lyzed using a paired t-test. Comparisons between diversipilosa, but not for P. bicarinata (Table 1, Com- experimental additions and natural habitat around plexity main eVect). Although the tendency towards focal colonies, and between burned and unburned habitat increased numbers of soldiers in more complex habi- around host colonies in the Santa Rita Mountains were tats was true for both hosts, it is unclear whether these not paired and were analyzed using two-sample t-tests. increases were a result of refuge from parasitoids or response to altered competitive conditions. Planned comparisons below shed light on these questions. Results Introduction of competitors had a signiWcant positive eVect on P. diversipilosa soldiers, but not on P. bicari- Foraging success, eVects of parasitoids and competition nata soldiers (Table 1). Again, the tendency towards increased numbers of soldiers in the presence of com- In both species, increased numbers of soldiers harvest- petitors was true for both hosts, and indicates that ing and defending cookie baits resulted in higher forag- hosts engage in alarm recruitment as a competitive ing success (Fig. 1a, b). The number of P. diversipilosa response. soldiers at resources showed a very strong positive cor- V relation with foraging success (R = 0.854, F1,53 = E ects of habitat complexity 142.742, P < 0.001), while the number of P. bicarinata soldiers at resources showed a somewhat more scat- When host ants are under attack by parasitoids, habitat tered, but still highly signiWcant correlation with foraging complexity may beneWt colonies by providing refuge success (R = 0.649, F1,62 = 45.128, P < 0.001). Introducing for soldiers that are traveling to resources along a parasitoids had a pronounced negative eVect on the recruitment trail, or by providing refuge during number of soldiers harvesting and defending resources defense and harvest at the resource itself. In this exper- for both hosts (Fig. 2a, c; Table 1, Parasitoid main iment, refuge is provided by leaf litter only along the 123 156 Oecologia (2007) 152:151–161

(Fig. 3a, b, closed triangle and open circle; Table 1, Comparison 4). In the absence of parasitoids, the number of soldiers defending against Dorymyrmex does not diVer between complex and simple habitat treatments for either host (Fig. 3a, b, open and closed circles; Table 1,Comparison 5). These results suggest that habitat complexity allows both hosts to balance competition with Dorymyrmex and escape from their specialist parasitoids. The increases in soldier numbers allowed by habitat complexity elevate soldier numbers to a level that is equivalent to foraging behavior when parasitoids are not present. Indeed, the conWdence intervals from both simple and complex no-parasitoid treatments (Fig. 3a, b, open and closed circles) include the mean of complex parasitoid treatments (Fig. 3a, b, closed triangles), but not the mean of simple parasitoid treatments (Fig. 3a, b, open triangles).

Fire as a source of variation in habitat complexity

Measurements of leaf litter depth and structural complexity of the leaf litter were used to estimate a Fig. 2 The impact of a, c parasitoids and b, d habitat complexity in the presence of parasitoids on the number of soldiers defending constant volume of leaf litter that could be added to a resource for two host ant species, P. diversipilosa and P. bicari- experimental foraging bins to mimic natural levels of nata. P. diversipilosa is attacked by Apocephalus sp. 8, while P. bi- habitat complexity experienced by colonies. Forag- carinata is attacked by Apocephalus sp. 25. Asterisks indicate ing bins with added habitat complexity (Fig. 4, Exp.) signiWcance determined by ANOVA (*P < 0.05, ***P < 0.001). Error bars are either experiment-wide or local SEs (see Materials approximate the habitat complexity found in natural and methods ) habitat where P. bicarinata and P. diversipilosa co-

occur (Fig. 4, Nat.; depth, t1,11 = 0.934, P = 0.370; no. leaves, t1,11 = 4.207, P < 0.05). Although the number recruitment path to the resource and not at the of leaves intersected by the sample stake is signiW- resource itself (the bait card is not covered by leaf lit- cantly higher in our experimental additions, the ter), which mimics natural conditions in that arthropod observed diVerence of approximately one leaf in the prey is often exposed. In the presence of parasitoids, vertical layer may not have biological consequences both hosts exhibited signiWcant increases in soldiers in terms of how eVectively leaf litter acts as a refuge harvesting and defending cookie baits when habitat from parasitoids. While foraging bins without added structure was more complex (Fig. 2b, d; Table 1, Com- habitat complexity are by nature very simple in their parison 1). In the absence of parasitoids neither host topography, Fig. 4 shows that plots burned in Santa maintained greater numbers of soldiers at baits in com- Rita Mountains [Burn (W)] and Chihuahuan Desert plex habitats (Table 1, Comparison 2). [Burn (D)] wildWres exhibit very low levels of leaf litter complexity immediately after a Wre and even Balancing competition with parasitism avoidance after recovering for 3 years [compare to Unburn (W) and Unburn (D)] [Burn (W)¡Unburn (W) compari-

When simultaneously competing with Dorymyrmex son for depth, t1,10 = 4.794, P < 0.001 and no. leaves, and under attack by parasitoids in simple habitats, t1,10 = 7.245, P < 0.001; Burn (D)¡Unburn (D) com- neither host can maintain the same number of sol- parison for depth, t1,7 = 4.907, P < 0.002 and no. diers as when there are no parasitoids in simple habi- leaves t1,7 = 5.473, P < 0.001]. The existence of tats (Fig. 3a, b, open triangle and open circle;Table 1, marked diVerences in complexity between burned Comparison 3). However, when simultaneously com- and unburned areas suggests that the diVerences in peting and under attack by parasitoids in complex complexity applied during our treatments are habitats, both hosts maintain soldier levels equivalent realistic proxies for the eVects of Wre on habitat to when there are no parasitoids in simple habitats complexity. 123 Oecologia (2007) 152:151–161 157

Table 1 Results of three-way Source P. diversipilosa P. bicarinata ANOVA and multiple com- a parison tests examining the df MS FPdfMS FP eVects of parasitoids (Parasit- oid), habitat complexity ANOVA (Complexity) and competitors Parasitoid (Para) 1 6.890 27.144 0.000 1 5.766 19.317 0.000 (Competition) on the number Complexity (Complex) 1 1.008 3.972 0.052 1 0.687 2.302 0.135 of soldiers at cookie bait re- Competition (Comp) 1 1.259 4.958 0.031 1 0.818 2.739 0.104 sources for the ant hosts Phei- Para £ Complex 1 0.001 0.003 0.957 1 0.546 1.829 0.182 dole diversipilosa, and Para £ Comp 1 0.043 0.170 0.682 1 0.039 0.131 0.719 Pheidole bicarinata Complex £ Comp 1 0.168 0.661 0.420 1 0.042 0.141 0.708 Para £ Complex £ Comp 1 0.055 0.218 0.643 1 0.015 0.049 0.825 Error 47 0.254 56 0.299 Comparisons Comparison 1 1 0.467 4.363 0.047 1 1.229 4.117 0.047 Error 25 0.107 56 0.299 a See Materials and methods Comparison 2 1 0.543 1.307 0.263 1 0.004 0.014 0.908 for justiWcation of error terms Error 26 0.416 56 0.299 used in planned comparisons. Comparison 3 1 2.360 9.296 0.004 1 2.991 10.021 0.003 Orthogonal planned compari- Error 47 0.254 56 0.299 sons were judged signiWcant at Comparison 4 1 0.514 2.025 0.161 1 0.612 2.050 0.158 = 0.05, while non-orthogo- Error 47 0.254 56 0.299 nal comparisons were judged Comparison 5 1 0.371 0.893 0.353 1 0.008 0.026 0.874 signiWcant at = 0.009 (See Error 26 0.416 56 0.299 Analysis)

Discussion Feener 2002; LeBrun 2005). The current study shows that habitat complexity can beneWt foraging ability (as The positive relationship between habitat complexity measured by soldiers harvesting and defending and species richness within local communities has been resources) in both host species when they are under demonstrated numerous times between both closely attack by their specialist parasitoids. related taxa (e.g., MacArthur and MacArthur 1961; One interpretation of these results is that soldiers Pianka 1966; Ellner et al. 2001) and distantly related are responding positively to habitat complexity in the taxa (e.g., Bell et al. 1991; Downes et al. 1998; Hansen presence of parasitoids because complex habitats alter 2000). Increased species richness in complex structured the interaction between hosts and their Dorymyrmex habitats is often attributed mechanistically to increases competitors rather than between hosts and their par- in resource partitioning (Williams 1943; Schoener asitoids. For example, increased complexity around the 1974) or availability (Heck and Wetstone 1977; O’Con- resource leads to an increased surface area that cannot nor 1991). Fewer studies address whether habitat com- be as easily defended by a given number of soldiers. If plexity can aVect species richness by altering a constant level of defense is necessary, the colony may competitive and predator–prey interactions in commu- recruit additional soldiers to resources in complex hab- nities (Heck and Wetstone 1977; Orth et al. 1984; itats. However, if the elevated soldier levels seen in Fletcher and Underwood 1987). Of these studies, none complex habitats in the presence of parasitoids are due focus on terrestrial organisms. to the eVect of habitat complexity on host–competitor This study demonstrates how habitat complexity interactions, there should be a signiWcant diVerence in aVects interactions among two species of host ants, soldier levels between complex and simple habitats in their specialist Dipteran parasitoids and a single com- the absence of parasitoids as well. There is no signiW- petitor ant species. The foraging ability (number of sol- cant diVerence between these treatments for either diers at resources) of both hosts was negatively aVected host (Fig. 3, open and closed circles). by the presence of their specialist parasitoids. Recent An alternative hypothesis, that elevated soldier lev- studies on P. diversipilosa and P. bicarinata have els in complex habitats with parasitoids are a result of shown that the negative eVect of parasitoids on host lowered host ability to detect parasitoids, is also foraging ability leads to resource loss, restructuring of unlikely. If soldier detection of parasitoids decreased the ant community dominance hierarchy, and a hypo- in complex habitats, parasitoid oviposition success thetical increase in the ability of subordinate competi- should go up. A separate study on the P. diversipilosa– tors to coexist within the community (LeBrun and Apocephalus sp. 8 system, in which the eVect of habitat

123 158 Oecologia (2007) 152:151–161

Fig. 4 Leaf litter depth (cm; dark bars) and leaf litter complexity (no. leaves intersected; light bars) in six habitats: experimental bins (Exp.), natural habitat of P. diversipilosa and P. bicarinata in the Chiricahua mountains (Nat.), burned habitat around P. diver- Fig. 3 Number of a P. diversipilosa and b P. bicarinata soldiers sipilosa colonies immediately after a low intensity Wre in Santa defending and harvesting a resource (cookie bait) in the presence Rita mountain woodlands [Burn (W)], unburned habitat around of competitors when parasitoids are absent (circles) or present (tri- P. diversipilosa colonies in Santa Rita Mountain woodlands [Un- angles) in complex habitat (Wlled symbols) or simple habitat (open burn (W)], upland Chihuahuan desert habitat 3 years after study symbols). Means and 95% conWdence intervals (adjusted for mul- plots burned [Burn (D)], unburned upland Chihuahuan desert tiple comparisons) are presented for P. diversipilosa—Wlled circle habitat near burned plots [Unburn (D)]. Means and 95% conW- 1.42 (0.68–2.94), empty circle 1.02 (0.49–2.12), Wlled triangle 0.70 dence intervals are presented (0.34–1.45), empty triangle 0.45 (0.22–0.93) and P. bicarinata— Wlled circle 4.24 (2.11–8.53), empty circle 4.06 (2.02–8.17), Wlled triangle 2.75 (1.37–5.52), empty triangle 1.71 (0.85–3.44) laboratory studies with simpliWed structural environments. complexity on parasitoid behavior was measured, The degree to which resource limitation in this showed that oviposition success was almost identical in experiment mimicked conditions found in our study is complex and simple habitats (E. B. Wilkinson, unpub- unknown. A number of studies have found that lished data). ground-nesting ant populations of the desert south- The most likely explanation for the observed results west are resource limited as opposed to nest site or is that, instead of utilizing the behavioral defense of predator limited (Davidson 1977; Wisdom and Whit- returning to the colony for protection, host ants can ford 1981; Davidson et al. 1984; Ryti and Case 1988). rely on habitat complexity for refuge, allowing them to Under such conditions of competition for limiting continue foraging in the presence of parasitoids. Bene- resources, decreases in foraging success caused by Wts to foraging ability translate to increased foraging parasitoids should lead to decreases in colony produc- success and should ultimately mean that host colonies tion of workers and reproductives, which aVects both foraging repeatedly in complex habitats will experience population size and community structure (Hölldobler higher Wtness than colonies foraging in simple habitats. and Wilson 1990). However, studies measuring the Refuge from parasitoids is likely to be important for a combined eVects of habitat complexity and parasi- variety of host ant species because: (1) numerous Weld toids at the population level in ant communities are studies conducted in natural habitats show large needed. negative indirect eVects of parasitoids on hosts (Feener 1981; Morrison 1999; LeBrun and Feener 2002; Community impacts LeBrun 2005), and (2) habitat complexity often varies to a large degree around colonies (see below). A recent To survive within the context of a community, ant laboratory study on the host ant Solenopsis invicta has species must be able to escape their predators/parasi- shown that lowered foraging success due to Pseudact- toids while continuing to compete eVectively for eon parasitoids does translate to lowered colony resources. For ants with specialist parasitoids, balanc- Wtness, as measured by worker biomass (Mehdiabadi ing parasitoid escape with eVective competition is and Gilbert 2002). However, it is important to note made easier by the presence of refuge in the form of that the impact of parasitoids may be exaggerated in habitat complexity. When under attack by parasi-

123 Oecologia (2007) 152:151–161 159 toids, habitat complexity allows P. diversipilosa and Sources of variation in habitat complexity: disturbance P. bicarinata colonies competing for resources to maintain numbers of soldiers harvesting resources Disturbance regimes such as Wre can cause variation in that are indistinguishable from situations in which the structural complexity of habitats through time. We parasitoids are absent. In the absence of their respec- show that low-intensity Wre can lead to over 70% tive parasitoids, both P. diversipilosa and P. bicari- reductions in leaf litter depth and complexity, and that nata are near the top of their community dominance 50% reductions in depth and complexity can persist hierarchy (LeBrun 2005). Subordinate members in even after recovery for three growing seasons (Fig. 4). the community might eventually be extirpated were it The eVect of low-intensity Wre on habitat complexity not for the negative eVects that specialist parasitoids therefore has the potential to alter behavioral interac- have on the foraging ability of dominant hosts, also tions not just immediately after the Wre occurs, but known as trait-mediated indirect eVects (Werner and over longer time periods as well. In addition, the eVects Peacor 2003; LeBrun 2005). of Wre on habitat complexity in ecological time are In contrast, our work shows that habitat complexity likely to be magniWed because Wre is a widespread and in the form of leaf litter can reduce the negative eVects frequent occurrence in the Chihuahuan desert border of parasitoids on the competitive dominance of hosts region through evolutionary time (Swetnam and Bet- by allowing dominant host ants to forage even in the ancourt 1998), with Wres occurring approximately every presence of their specialist parasitoids. The role of 3–10 years in sampled locations [based on Weibull habitat complexity in enhancing dominant host com- median probability interval (Swetnam and Baisan petitive ability runs counter to the traditional under- 1996)]. standing that increased habitat complexity leads to It is unlikely that Wre would have negative direct increased species richness. This unexpected result is eVects on host ant populations. Most ant species in due to the previously unrecognized potential for habi- temperate zones construct nests in the soil (Hölldobler tat complexity to modify top-down eVects from parasi- and Wilson 1990). Although foragers caught above toids. Diel escape from Apocephalus sp. 8 and ground during a Wre will die, only a small proportion of Apocephalus sp. 25, which are only active during day- workers are active above ground at any one point in light hours, may also reinforce dominant host competi- time (Andersen 1985). The low-intensity Wres (as char- tive ability (e.g., Linepithema sp.; Orr et al. 2003). In acterized by Chandler et al. 1983) of this study typically the case of P. diversipilosa and P. bicarinata, diel do not lead to increases in soil temperature capable of escape may play only a minor role in enforcing their killing colonies (Andersen 1985). The eVects of Wre on dominance in the community because both hosts show ant communities are more likely to be mediated signiWcant (P. diversipilosa) or near signiWcant (P. through indirect eVects on physiology, resources and bicarinata) preferences for foraging during daylight predation rather than through direct mortality (Chan- hours (LeBrun 2005). dler et al. 1983). Thus, subordinate competitors may What then, is preventing behaviorally dominant have increasing negative eVects on dominant hosts if hosts protected by habitat complexity from forcing they are tolerant of the physiological conditions in other members in the community to extinction? Multi- burned habitats, and this negative eVect may be addi- ple trade-oVs can work in concert to maintain ant com- tive with negative eVects on hosts from predators or munity structure and coexistence. For example, ant parasitoids. Fire should aVect seed and insect resources species may vary in their thermal tolerance, with spe- utilized by the ant community equally, as recovery of cies that exhibit thermal tolerance escaping strong insect populations is directly associated with recovery competition by foraging during the hottest periods of of the Wre-adapted vegetative community (Whitford the day (Cerdá et al. 1997), or in their scout/recruit et al. 1995). Therefore, the majority of the ant commu- ratio, with species that maintain a large number of nity is likely to experience heightened levels of scouts Wnding resources quickly (exploitative competi- resource limitation until the vegetative community tors). Multiple trade-oVs help stabilize community recovers. dynamics because species that are dominant on one It is also unlikely that Wre would dampen parasitoid trade-oV curve may be subordinate on another. How- populations locally. Closely related Pseudacteon ever, multiple trade-oVs are not needed to explain parasitoids with similar life histories to Apocephalus coexistence. Disturbance by Wre can lead to reductions develop within hosts for »30 days, and live only 2– in habitat complexity suYcient for lower host ant 5 days as adults (Morrison 1997; B. V. Brown, behavioral dominance and an increased potential for personal communication). At any point in time, coexistence. most of the Apocephalus parasitoid population 123 160 Oecologia (2007) 152:151–161 would be developing in soldiers harbored within the Memorial Fund, Sigma Xi, the University of Utah Biology nest for purposes of recruitment. A single, low-inten- Department and the Associated Students of the University of W V Utah. A NSF Dissertation Improvement Grant to E. B. Wilkin- sity re would therefore have little e ect on local para- son partially funded this work. D. H. Feener Jr wishes to acknowl- sitoid populations. Even if a Wre were to completely edge support from NSF grant DEB 03-16524. These experiments extirpate parasitoids locally, the eVect of Wre on habitat comply with the current laws of the countries in which they were complexity persists so long that even small parasitoid performed. populations would be expected to recolonize a 2-km2 burn within a fraction of a year (extrapolated from References Morrison and Porter 2005). Alonzo SH (2002) State-dependent habitat selection games be- Conclusions tween predators and prey: the importance of behavioral interactions and expected lifetime reproductive success. Evol Ecol Res 4:759–778 This model system explores how host–parasite interac- Andersen AN (1985) Immediate eVects of Wre on ants in the semi- tions are aVected by habitat attributes. Reduction in arid mallee region of north-western Victoria. J Ecol 10:25–30 habitat complexity due to regular disturbance events Andow DA, Prokrym DR (1990) Plant structural complexity and W will expose dominant hosts to higher rates of parasit- host- nding by a parasitoid. Oecologia 82:162–165 Bell SS, McCoy ED, Mushinsky HR (1991) Habitat structure: the ism. If the behavioral response to increased parasitism physical arrangement of objects in space. Chapman and Hall, alters competition between dominant hosts and subor- London dinates throughout the community, regular distur- Bestelmeyer BT, Schooley RL (1999) The ants of the southern bance may be expected to maintain a community in Sonoran desert: community structure and the role of trees. Biodiversity Conserv 8:643–657 which behavioral dominance does not translate to eco- Cerdá X, Retana J, Cros S (1997). Thermal disruption of transi- logical dominance. However, changes in habitat com- tive hierarchies in Mediterranean ant communities. J Anim plexity caused by disturbance may also aVect dominant Ecol 66:363–374 host Wtness and competitive ability directly. The eVect Chandler C, Cheney P, Thomas P, Trabaud L, Williams D (1983) Fire in forestry vol 1. Forest Wre behavior and eVects. Wiley, of habitat complexity on ant–parasitoid interactions is New York likely independent of, but additive with, any direct Chase JM, Abrams PA, Grover JP, Diehl S, Chesson P, Holt RD, eVects of habitat complexity on host Wtness or competi- Richards SA, Nisbet RM, Case TJ (2002) The interaction be- tive ability. tween predation and competition: a review and synthesis. V Ecol Lett 5:302–315 Our use of two di erent ant species in this study Clark TL, Messina FJ (1998) Plant architecture and the foraging underscores the potential for habitat complexity to success of ladybird beetles attacking the Russian wheat aVect host–parasite interactions between a variety of aphid. Entomol Exp Appl 86:153–161 terrestrial organisms. Host–parasite and predator–prey Davidson DW (1977) Species diversity and community organiza- tion in desert seed-eating ants. Ecology 58:711–724 interactions are common within leaf litter communi- Davidson DW, Inouye RS, Brown JH (1984) Granivory in a des- ties, as well as at larger scales of habitat complexity. ert ecosystem: experimental evidence for indirect facilitation Our work demonstrates that Wre-induced changes to of ants by rodents. Ecology 65:1780–1786 habitat complexity can alter behavioral interactions Downes BL, Lake PS, Schreiber ES, Glaister A (1998) Habitat structure and regulation of local species-diversity in a stony, between hosts and parasites and potentially predators upland stream. Ecol Monogr 68:237–257 and prey. Further work is needed to answer the ques- Ellner SP, et al. (2001) Habitat structure and population persis- tions of how such changes in complexity aVect compet- tence in and experimental community. Nature 412:538–543 itive interactions in communities and community Feener DH Jr (1981) Competition between ant species: outcome controlled by parasitic Xies. Science 214:815–817 structure as a whole. Feener DH Jr (2000) Is the assembly of ant communities mediated by parasitoids? Oikos 90:79–88 Acknowledgements Ed LeBrun provided crucial training and Finke DL, Denno RF (2002) Intraguild predation diminished input at the inception of this study. This manuscript beneWted in complex-structured vegetation: implications for prey substantially from the comments of two anonymous reviewers suppression. Ecology 83:643–652 and Craig Osenberg. Jessica Pearce, Philipp Wiescher, Roxana Fletcher WJ, Underwood AJ (1987) InterspeciWc competition Arauco, Ryan Bixenmann, Arne Hultquist and Martin Moyano among subtidal limpets-eVect of substratum heterogeneity. also provided useful comments on the manuscript. Thanks to Ste- Ecology 68:387–400 fan Cover for identifying host ants and the Sperry lab at the Uni- Fraser DF, Huntingford FA (1986) Feeding and avoiding preda- versity of Utah for use of their drying ovens. This study would not tion hazard: the behavioral response of the prey. Ethology have been possible without the assistance of Joe and Valer Aus- 73:56–68 tin, the facilities of the Southwestern Research Station and the Gosselin LA, Bourget E (1989) Individual performance in relation help of Robert Minckely. Mary Wilkinson provided much needed to structural heterogeneity: the inXuence of substratum topog- support. We would like to acknowledge funding from the raphy on an intertidal predator, Thais lapillus. J Anim Ecol American Museum of Natural History Theodore Roosevelt 58:287–304

123 Oecologia (2007) 152:151–161 161

Hansen RA (2000) EVects of habitat complexity and composition O’Connor NA (1991) The eVects of habitat complexity on the on a diverse litter microarthropod assemblage. Ecology macroinvertebrates colonizing wood substrates in a lowland 81:1120–1132 stream. Oecologia 85:504–512 Heck KL, Wetstone GS (1977) Habitat complexity and inverte- Orr MR, Seike SH, Benson WW, Gilbert LE (1995) Flies sup- brate species richness and abundance in tropical seagrass press Wre ants. Nature 26:292 meadows. J Biogeogr 4:135–142 Orr MR, Dahlsten DL, Benson WW (2003) Ecological interac- Hoddle MS (2003) The eVect of prey species and environmental tions among ants in the genus Linepithema, their phorid complexity on the functional response of Franklinothrips or- parasitoids, and ant competitors. Ecol Entomol 28:203–210 izabensis: a test of the fractal foraging model. Ecol Entomol Orth RJ, Heck KL et al (1984) Faunal communities in seagrass 28:309–318 beds: a review of the inXuence of plant structure and prey Hölldobler B, Wilson EO (1990) The ants. Belknap Press of characteristics on predator–prey relationships. Estuaries Harvard University Press, Cambridge, Mass. 7A:339–350 Holway DA (1999) Competitive mechanisms underlying the Pianka ER (1966) Convexity, desert lizards, and spatial heterogeneity. displacement of native ants by the invasive Argentine ant. Ecology 47:1055–1059 Ecology 80:238–251 Ripple WJ, Beschta RL (2004) Wolves and the ecology of fear: Kirby KN (1993) Advance data analysis with SYSTAT. Van No- can predation risk structure ecosystems. BioScience 54:755– strand Reinhold, New York 766 Kneitel JM, Chase JM (2004) Trade-oVs in community ecology: Rothley KD, Schmidtz OJ, Cohon JL (1997) Foraging to balance linking spatial scales and species coexistence. Ecol Lett 7:69– conXicting demands: novel insights from grasshoppers under 80 predation risk. Behav Ecol 8:551–559 LeBrun EG (2003) When to lose your head: ant-phorid Xy inter- Ryti RT, Case TJ (1988) Field experiments on desert ants: testing actions, community level impacts and evolutionary signiW- for competition between ants. Ecology 69:1993–2003 cance. Dissertation. Graduate Degree Program in Ecology Schoener TW (1974) Resource partitioning in ecological commu- and Evolution, University of Utah, Utah nities. Science 185:27–39 LeBrun EG (2005) Who is the top dog in ant communities? Re- Smith DRR (1985) Habitat use by colonies of Philoponella resources, parasitoids, and multiple competitive hierarchies. publicanna (Araneae, Uroboridae) J Arachnol 13:363–373 Oecologia 142:643–652 Sokal and Rohlf (1995) Biometry. Freeman, New York LeBrun EG, Feener DH Jr (2002) Linked indirect eVects in ant- Swetnam TW, Baisan CH (1996) Fire histories of montane forests phorid interactions: impacts on ant assemblage structure. in the Madrean Borderlands. In: Folliott PF et al. (technical Oecologia 133:599–607 coordinators) EVects of Wre on Madrean Province Ecosys- Lima SL, Dill LM (1990) Behavioral decisions made under the tems. Proceedings. USDA Forest Service general technical risk of predation: a review and prospectus. Can J Zool report RM-GTR-289, pp 15–36 68:619–640 Swetnam TW, Betancourt JL (1998) Mesoscale disturbance and MacArthur RH, MacArthur JW (1961) On bird species diversity. ecological response to decadal climate variability in the Ecology 42:594–598 American Southwest. J Climate 11:3128–3147 Mehdiabadi NJ, Gilbert LE (2002) Colony-level impacts of para- SYSTAT, Inc. (2004) SYSTAT version 11. Richmond, CA sitoid Xies on Wre ants. Proc R Soc Lond [Biol] 269:1695–1699 Tilman D (1994) Competition and biodiversity in spatially struc- Morrison LW (1997) Oviposition behavior and development of tured habitats. Ecology 75:2–16 Pseudacteon Xies (Diptera: Phoridae), parasitoids of Solen- Werner EE, Peacor SD (2003) A review of trait-mediated indi- opsis Wre ants (Hymenoptera: Formicidae). Environ Ento- rect interactions in ecological communities. Ecology mol 26:716–724 84:1083–1100 Morrison LW (1999) Indirect eVects of phorid Xy parasitoids on Whitford WG, Forbes GS, Kerley GI (1995) Diversity, spatial the mechanisms of interspeciWc competition among ants. variability, and functional roles of invertebrates in desert Oecologia 113:122 grassland ecosystems. In: McClaran MP, Van Devender TR Morrison LW, Porter SD (2005) Testing for population-level im- (eds) The desert grassland. University of Arizona Press, Tuc- pacts of introduced Pseudacteon tricuspis Xies, phorid parasi- son, Ariz. toids of Solenopsis invicta Wre ants. Biol Control 33:9–19 Williams CB (1943) Area and the number of species. Nature Morrison WL, Kawazoe AE, Guerra R, Gilbert LE (2000) Eco- 152:264–267 logical interactions of Pseudacteon parasitoids and Solenop- Wisdom WA, Whitford WG (1981) EVects of vegetation change sis ant hosts: environmental correlates of activity and eVects on ant communities of arid rangelands. Environ Entomol on competitive hierarchies. Ecol Entomol 25:433–444 10:893–897 Norton AP, English Loeb G, Belden E (2001) Host plant manip- ulation of natural enemies: leaf domatia protect beneWcial mites from insect predators. Oecologia 126:535–542

123