Apparent competition drives community-wide parasitism rates and changes in host abundance across ecosystem boundaries

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 enemy movement 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 of insecticide over 20 hectares. 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.

Figure 1. Summary of methods.

Sampling at training sites produced a measure of shared parasitism for each species pair among all species found in the region. (a) Lepidopteran larvae were collected along transects within native and plantation forest at eight forest edge sites. (b) Larvae were identified and reared to determine parasitism rates and parasitoid (Hymenoptera, Diptera and Nematoda) identities. (c) Data were pooled across sites and sampling dates, but kept separate by forest type, in order to produce one regional quantitative food web (host–parasitoid matrix), in which host habitat was explicit. (d) The potential for apparent competition, d_iAjB, was calculated for each host pair within and across habitats. (e) This d_iAjB value was the regional measure of shared parasitism used to calculate Expected parasitism rate in equation (3) (f,i). (g) We sampled at control and herbivore reduction validation sites at two time steps (g,h), before and after the aerial spray herbivore reduction treatment, which occurred between time steps (h). This allowed measurement of the initial attack rates and initial host abundances (g) and changes in host abundances (g,h) necessary to calculate expected parasitism rate (i), as well as measurement of the final attack rates and final host abundances (h) necessary to calculate observed parasitism rates (j) and change in focal host abundances (k). We used generalized linear mixed models to test whether Expected parasitism rate significantly predicted observed parasitism rate and change in focal host abundance (l) at validation sites, and thus whether apparent competition structures host–parasitoid assemblages in a predictable manner.

Figure 2. Regional metaweb built from quantitative food-web data.

Data were collected at habitat edges between native forest and exotic plantation forest and are pooled across eight training sites and seven sampling dates. Bars on the left represent herbivore host species in native forest (blue) or plantation forest (red), and bar thickness is proportional to number of parasitism events (358 total parasitism events from 2725 individual hosts reared). Bars on the right represent parasitoid species, and lines connecting parasitoids and host species denote parasitism, with parasitoids attacking hosts in native forest only (green), plantation forest only (purple) or both forest types (yellow). Parasitoid bar thickness is proportional to attack rate. Numbers adjacent to host names correspond to the host numbers in Figure 2.

Figure 3. Parasitoid overlap graph for the region.

Data compiled from training sites (358 total parasitism events from 2725 individual hosts reared). This figure shows the potential for apparent competition between all parasitized herbivore host species in native forest (blue) and plantation forest (red). Numbers represent host species (see Fig. 2 for names), some of which are found in both habitats. Host circle size is proportional to the number of parasitoids recruiting from that host species, and circle fill represents the proportion of self-loops, that is, parasitoids attacking the same species from which they recruit. Lines between hosts represent sharing of parasitoids between those host species, with line thickness proportional to export of parasitoids. Blue lines denote sharing of parasitoids by hosts within the native forest, red lines denote sharing of parasitoids by hosts within the plantation forest, and grey lines denote sharing of parasitoids between hosts in different habitats (the latter allowing cross-habitat apparent competition).

Figure 4. Community average effects of experimental herbivore reduction.

Experimental herbivore reduction (a) reduced caterpillar abundance in treated plantation forest relative to in control plantation forest (GLMM; z=−3.2, P=0.002) and (b) may have affected caterpillar parasitism rates in adjacent native forest (though the effect of experimental herbivore reduction on parasitism rates was non-significant, as tested with a GLMM in which the interaction between herbivore reduction treatment and collection was removed during model selection). H is herbivore reduction treatment, C is control. Solid orange lines indicate a contrast significant at α=0.05. Error bars are s.e.m. In a and b n=8 pairs of treatment and control sites as replicates.

Figure 5. Predictions of parasitism rate and change in host abundance.

(a) Expected parasitism rate predicted observed parasitism rate (GLMM; z=2.7, P=0.007, R²_(GLMM(m)=0.16), whereas (b) initial parasitism rate did not predict observed parasitism rate (GLMM; z=1.1, P=0.266). Observed parasitism rate was also significantly predicted by expected parasitism rates calculated with only (c) data from within the same forest as the focal host (GLMM; z=2.8, P=0.005, R²_GLMM(m)=0.20), or (d) data from the forest adjacent to the focal host's habitat (GLMM; z=2.1, P=0.040, R²_(GLMM(m)=0.14). (e) Expected parasitism rate also significantly predicted change in host abundance (linear model; z-value=−3.0, P=0.007, R²=0.31), but (f) initial parasitism rate did not predict change in host abundance (linear model; initial parasitism rate was removed as a predictor during model selection). In a–d, residuals of the best model excluding the predictor on the x axis are plotted, and raw data are plotted in e and f. Residuals are deviations of the logit-linked data from the best model, and show the unexplained variation in the data remaining once variation due to the other fixed and random effects in the model have been accounted for. That is, they show the variation in the data that we are hoping to explain with the predictor on the x axis. Each point represents a species within a site that was collected and successfully reared at both time steps, and was parasitized in the first time step (see equation (3); Fig. 1i). In a and b closed circles (red) represent hosts in plantation forest and open circles (blue) represent hosts in native forest, though only one fitted line is presented because there was no significant habitat_A × expected parasitism rate interaction (in a the interaction term was removed during model selection; (b) interaction z=1.9, P=0.060). In all panels, n=8 pairs of treatment and control sites as replicates.

Figure 6. Magnitude of the potential for indirect effects (d_iAjB) among host species.

N and P refer to native forest and plantation forest, respectively, such that, for example, PN refers to the situation where host i (the species that is affected) is in plantation forest and host j (the species that generated the parasitoids attacking species i) is in native forest. The potential for hosts in plantation forest to affect hosts in native forest (NP) was not significantly greater than the potential for hosts in native forest to affect hosts in plantation forest (PN) through apparent competition (linear model: t=1.9, P=0.065). Letters above the bars are post-hoc mean comparisons. Error bars represent s.e.m. Data for this analysis were from the regional metaweb (Fig. 2), so sites were pooled, and habitat-specific species pairs were replicates (PN: n=178; NN: n=255; NP: n=178; PP: n=155).