Apparent Competition Drives Community-Wide Parasitism Rates and Changes in Host Abundance

Apparent Competition Drives Community-Wide Parasitism Rates and Changes in Host Abundance Across Ecosystem Boundaries

Carol M. Frost1,2*, Guadalupe Peralta1,3, Tatyana A. Rand4, Raphael K. Didham5,6, Arvind Varsani1,7,8,9, and Jason M. Tylianakis1,10 *

Affiliations:

1 Centre for Integrative Ecology, School of Biological Sciences, University of Canterbury, Private Bag 4800, Christchurch 8140, New Zealand.

2 Current address: Department of Forest Ecology and Management, Swedish University of Agricultural Sciences, Skogsmarksgränd, 90183 Umeå, Sweden.

3 Current address: Instituto Argentino de Investigaciones de las Zonas Áridas, CONICET, CC 507, 5500, Mendoza, Argentina.

4 USDA-ARS Northern Plains Agricultural Research Laboratory, Sidney, MT 59270, USA.

5 School of Animal Biology, The University of Western Australia, 35 Stirling Highway, Crawley WA 6009, Australia.

6 CSIRO Land & Water, Centre for Environment and Life Sciences, Underwood Ave, Floreat WA 6014, Australia.

7 Biomolecular Interaction Centre, University of Canterbury, Private Bag 4800, Christchurch, 8140, New Zealand.

8 Structural Biology Research Unit, Department of Clinical Laboratory Sciences, University of Cape Town, Observatory, 7700, South Africa

9 Department of Plant Pathology and Emerging Pathogens Institute, University of Florida, Gainesville, FL 32611, USA

10 Department of Life Sciences, Imperial College London, Silwood Park Campus, Buckhurst Road, Ascot, Berkshire SL5 7PY, United Kingdom.

*Correspondence to: Carol Frost, and Jason Tylianakis,

Abstract

Species have strong indirect effects on others, and predicting these effects is a central challenge in ecology. Prey species sharing an enemy (predator or parasitoid) can be linked by apparent competition, but it is unknown whether this process is strong enough to be a community-wide structuring mechanism that could be used to predict future states of diverse food webs. Whether species abundances are spatially coupled by enemymovement across different habitats is also untested. Here, using a field experiment, we show that predicted apparent competitive effects between species, mediated via shared parasitoids, can significantly explain future parasitism rates and herbivore abundances. These predictions are successful even across edges between natural and managed forests, following experimental reduction of herbivore densities by aerial spraying over 20ha. This result shows that trophic indirect effects propagate across networks and habitats in important, predictable ways, with implications for landscape planning, invasion biology and biological control.

Introduction

Communities frequently experience population reductions (e.g., via harvesting or native species decline) or species additions (during invasions, biological control or species range shifts). These changes can directly impact consumers or prey of the affected species1. However, population changes may also indirectly affect the entire community via longer pathways across the interaction network2,3. Unfortunately, these many subtle indirect interactions4,5 limit our ability to predict the community-wide consequences of changes in abundance of a focal species, because they can significantly alter population growth and persistence6,7, and even affect species abundances and distributions as strongly as direct feeding interactions8. In particular, the dynamics of different prey species can be linked via shared enemies (predators or parasitoids), even if the prey never compete directly for resources9. That is, an increase in the population of one prey species can cause a decrease in the population of other prey species by driving increases in shared enemy abundance and attack rates or changes to enemy behavior6. This phenomenon, termed ‘apparent competition’9, can be common in food webs10, where it may be the most important indirect interaction affecting pairwise species dynamics11.

There are several mechanisms by which enemy responses to prey population growth can cause apparent competition. The timescale over which these mechanisms occur can vary from within one prey generation, through aggregative12 or functional responses13, to between prey generations, through numerical responses6, though distinguishing between them for an entire community under field conditions is unfeasible, and we do not attempt to do this here. Apparent mutualism is equally possible, if the shared enemy can be satiated or switch to the most abundant prey species in the short term9, thereby releasing less-abundant prey from consumer pressure. Apparent mutualism could occur over the longer term if the population of one prey species cycles, such that it repeatedly satiates an enemy, and thus repeatedly alleviates consumer pressure on another prey species that shares the enemy14. However, fewer empirical examples of apparent mutualism have been documented13,15,16. Despite numerous isolated examples of apparent competition between species pairs7,13,17, it remains untested whether the simultaneous pairwise effects of apparent competition across all species within a food web are strong enough to detectably affect population dynamics of all species, even amidst the network of direct interactions among them. The ability to predict the effects of changing population densities on interactions among all other species would be invaluable in addressing some of the most pressing questions in ecology and global environmental change, such as how ecosystems will respond to native species decline or species invasions.

Community-wide predictions of indirect interactions are further hindered by species movement among habitats. Global land-use change creates mosaic landscapes of managed and remnant natural habitats, and consumer movement among habitats is predicted to drive resident prey species dynamics through direct and indirect effects18-20. Yet, it remains untested whether mobile enemies dynamically couple herbivore assemblages in multiple habitats via apparent competition, as predicted by theory18. Entire suites of enemies can ‘spill over’ across habitat boundaries21, and if they couple prey dynamics in the two habitats, this process could be an important mechanism by which anthropogenic habitats impact entire food webs in natural and managed areas throughout the landscape20.

Here we test: 1) whether apparent competition influences community-wide parasitism rates and changes in herbivore abundance in host-parasitoid interaction networks (food webs) at the interface between native and plantation forests. In so doing, we also test: 2) whether the future parasitism rate and abundance of each herbivore host species in the community can be predicted from quantitative food-web data on parasitoid overlap between hosts, as well as information about changes in abundance of all other hosts. We further test: 3) whether such predictions are possible across a habitat edge, or whether the edge hinders parasitoid movement or changes parasitoid host selection such that predicted apparent competitive linkages between herbivore populations on either side of the edge are not realized. We conducted a simultaneous study of adult parasitoid movement between the two forest habitats considered here21. That study showed that parasitoids of many of the same species considered here moved between habitats throughout the season. However, more individuals moved from plantation to native forest than in the other direction, likely due to the higher productivity of plantation relative to native forest21. Thus, it could be that apparent competitive effects are asymmetrical between habitats, with stronger effects from herbivores in plantation forest on herbivores in native forest.

Our approach (see overview in Fig. 1) was to first determine a regional measure of shared parasitism (i.e. the potential for apparent competition) between each pair of herbivore species (foliage-dwelling Lepidoptera larvae) in the system, by collecting quantitative food web data (i.e. numbers and identities of parasitoids attacking each host species) from a set of replicated training sites (Fig. 1a). Each site comprised samples from either side of a habitat edge between plantation Pinusradiata forest and native forest in New Zealand. To gather these data, we collected Lepidoptera larvae (caterpillars) and reared them to obtain and identify (morphologically and using DNA barcoding) the parasitoids (Hymenoptera, Diptera, and Nematoda) that had attacked them. We carried out seven sampling rounds over two summers, in order to observe as many host-parasitoid interactions as possible (Fig. 1b,c). We predicted the potential for apparent competition between each pair of herbivore host species, using Müller et al.’s index22 of the proportion of the parasitoids attacking one species that had recruited from the other species (Fig.1 d,e). This index quantifies the hypothesis that changes in abundance of each host species affect attack rates on every other host, proportionate to the number of parasitoid species they share and each host’s contribution to the parasitoid pool. That is, for every pair of species that shares parasitoids, the species that produces more shared parasitoid individuals should have a larger apparent competitive impact on the other species. Thus, it incorporates the frequent asymmetry in potential for apparent competition between host species17. Muller et al.’s index22, has been widely adopted to predict the potential for indirect interactions within a single location23-26, but its predictive success has rarely been tested, and never across all the species in an assemblage. It has also never been tested across habitats. Nevertheless, two experimental studies have shown that it holds great promise for predicting indirect interactions among herbivores. First, Morris et al.27 experimentally reduced the abundance of two leaf-miner species, and found reduced attack rates on other leaf miners with which they shared parasitoids. Similarly, Tack et al.15 used a quantitative food web to predict interactions among three leaf-miner species, then experimentally increased the abundance of each species. They found cross-generation indirect interactions between some species as predicted, except that the effects were positive (i.e., apparent mutualism9) rather than negative (apparent competition). Together these studies suggest that information on shared parasitoids can in principle be used to successfully predict indirect interactions between species. However, across entire food webs, many pathways of weak and strong, positive and negative indirect effects may render net outcomes unpredictable. Moreover, it remains unknown whether prey use by predators is consistent enough that food web information generated in one location can be used to make accurate predictions about another, or whether entire prey communities in one habitat indirectly affect attack rates on, and therefore abundances of, those in adjacent habitats via mobile predators21.

Thus, we tested whether our calculated predictions of apparent competition (Fig. 1e) could be used, along with data on changes in abundance of the potential apparent competitors, to predict parasitism rate and change in abundance of each herbivore species (each separately as a focal host) in the system. For this test, we selected another set of replicated validation sites (Fig.1g). At these sites we again collected quantitative food web data by sampling, rearing, and identifying caterpillars and their parasitoids at two time steps: both before (Fig. 1g) and after a dramatic experimental herbivore reduction at half of the validation sites (Fig. 1h). This reduction was conducted by aerially spraying 2.5 ha per site with a selective insecticide that targets Lepidoptera larvae (the hosts in our study). The experimental herbivore reduction allowed us to test whether our predictions of apparent competition performed equally for small, typical variation in host abundance (at control sites) as well as more dramatic changes, such as may occur during pest outbreaks or at plantation harvest. We then predicted expected parasitism rates (Methods Eq. 3; Fig. 1i) on all host species at the next time step (Fig. 1h), based on three pieces of information: i) the regional measure of shared parasitism among hosts from training sites (Fig. 1d-f); ii) initial attack rates on each host species at a given validation site (Fig. 1g); and iii) the changes in abundance of all host species with which each focal host shares parasitoids (Fig. 1g-h). Finally, we used statistical models to test whether these expected parasitism rates predicted the observed parasitism rates at the time step after spraying (Fig. 1j), as well as changes in abundance between time steps (Fig. 1k). This test was conducted for each host species that occurred in the validation sites at the later time step (Fig. 1l).

We show that expected parasitism rate (our prediction of the final parasitism rate for each herbivore host species, based on the assumption of apparent competition with each other herbivore host species which occurred in the same site) significantly predicted both observed parasitism rate at the ‘after’ time step, and the change in abundance of a focal host between time steps. Expected parasitism rate predicted 31% of the variation in change in host abundance, and 15% of the variation in observed parasitism rate. Predictions worked equally well whether the focal host was in plantation or native forest, suggesting that in this system the habitat edge does not significantly hinder parasitoid movement or change host-selection behavior, and suggesting that herbivore population changes in one forest type can have rapid and important effects on herbivore populations across the habitat edge.

Results

Data description

Transect plus extra sampling at our training sites yielded 8321 caterpillars. Of these, 2725 individuals from 70 species in 13 families were successfully reared to moth or parasitoid emergence. These yielded 358 parasitism events by 46 species of Hymenoptera, Diptera, and Nematoda parasitoids on 44 Lepidoptera species. These 358 parasitism events made up the data from which the training metaweb was constructed (Figs. 1a-c, 2) and diAjB was calculated (Methods Eq. 3, Fig. 1d-f). The metaweb had a binary connectance of 0.057, which is within the range of connectance values exhibited in published quantitative food webs28.

Transect sampling at our validation sites yielded 5837 caterpillars that were identifiable to (morpho)species level, and included 67 species. These made up the data from which we calculated all abundance terms (njB, niA) in Eq. 3 and Eq. 6 (Fig. 1g-k). Of these caterpillars, 2067 individuals from 60 species were successfully reared to moth or parasitoid emergence, yielding 263 parasitism events by 36 species of parasitoid, in Hymenoptera, Diptera, and Nematoda on 25 species of Lepidoptera. Extra sampling yielded an additional 1121 identifiable caterpillars, of which 405 individuals from 65 species were successfully reared to moth or parasitoid emergence, yielding 40 additional parasitism events. The transect plus extra sampling total of 303 parasitism events by 37 species of parasitoids on 26 species of Lepidoptera made up the data from which αiAl(t) was calculated in Eqs. 3 and 5, from which αiAl(t+1) and niA(t+1) were calculated in Eq. 4, and from which niAt was calculated in Eq. 5 (Fig. 1g-k).

Regional potential for apparent competition among species

From our regional quantitative food web data from training sites, pooled over all sampling dates and sites (Fig. 2), we found that most parasitoid species were reared from hosts in both plantation and native forest (yellow interactions in Fig. 2), and few parasitoid species were specific to one forest type (green and purple interactions in Fig. 2). From these regional species interaction data, our calculations of potential for apparent competition between all host species pairs (Methods Eq.1,2, Figs. 1d, 3) showed potential for apparent competition between 41 host species (circles connected by lines in Fig. 3). These calculations also showed potential for many cross-habitat apparent competitive interactions (gray lines in Fig. 3).

Effects of experimental herbivore reduction

Our experimental herbivore reduction at validation sites was successful in significantly reducing caterpillar numbers in treated plantation forests relative to control plantation forests (Fig. 4a, Supplementary Table 1). However, caterpillar numbers were also naturally lower in control plantation forests after the herbivore reduction treatment than before. The lower caterpillar numbers at control sites in the after time step was likely because of predation by invasive Vespula spp. wasps, which become very abundant in New Zealand plantation and native forests in the months during which the ‘after’ collection occurred21, and which predate heavily on caterpillars29. Following the experimental reduction in herbivore abundance in the treated plantation forests, which occurred when many parasitoids were in their adult phase (Supplementary Fig. 1), it is possible that attack rates could have been briefly higher in native forest next to treated plantations, due to low host availability in plantations. However, we found no significant interaction effect (Fig. 4b, Supplementary Table 2), indicating that, overall, hosts in native forests were not attacked disproportionately more or less when their adjacent plantation was sprayed. This result is not surprising. We would not expect apparent competition effects to be strongly detectable in such an analysis, because it includes ‘noise’ in the attack rates such as unchanged parasitism rates for species in the native forest that do not share parasitoids across the habitat edge, or that share parasitoids with species whose abundances did not change drastically in response to the spray. Thus, to truly test the importance of host reduction, we needed to weight this test by host species changes in abundance combined with expectations of which host species might be expected to exert apparent competition on one another due to sharing of parasitoids.

Apparent competition predicts community-wide parasitism

Our higher resolution test, in which we calculated the expected parasitism rates in the after time step for each host species (Methods Eq. 3), showed that apparent competition significantly structured host-parasitoid assemblages within and across habitats in a predictable way. Our expected parasitism rate explained 15.6% of the variation in observed parasitism rates across host species (z = 2.7, p = 0.007), and the whole model (fixed plus random effects) explained 24.3% (Fig. 5a; Supplementary Table 3). This predictive ability was equally good for small natural or larger experimental changes in host abundance (the interaction between expected parasitism rate and herbivore reduction treatment was removed during model selection; i.e. expected parasitism rate predicted observed parasitism rate equally well whether or not host abundance was drastically changed experimentally). Significant predictive ability remained even when within-habitat intraspecific effects were excluded from the calculation of expected parasitism rates (Supplementary Table 4). Therefore the indirect effects at work could not simply be explained by within-habitat delayed density dependence. Importantly, this predictive capacity did not depend on the specific habitat of the focal host (the host habitat by expected parasitism rate interaction was removed during model selection; Fig. 5a, Supplementary Table 3). Rather, predicted apparent competitive effects were realized equally for hosts in either habitat, such that the habitat edge did not filter parasitoids and prevent some of the predicted interactions from occurring. Furthermore, this same predictive ability could not be achieved by using just initial parasitism rate for each focal host, calculated from quantitative food web data from the initial time step at each site (z = 1.11, p = 0.266; Fig. 5 b; Supplementary Table 5. Supplementary fig. 2a,b provides equivalent figures to 5 a,b with raw data.). Rather, significant prediction required inclusion of the potential for apparent competition between each host species (the diAjB in Methods Eq. 3) and the changes in abundance of hosts with which the focal host shared parasitoids.