- Top of page
- Materials and methods
It has long been recognized that the number of species in a community is linked to the availability of resources that limit the production of community biomass (Darwin 1859). Historically, studies have tried to understand this relationship by focusing on species diversity as a response variable, asking how the richness of species within and among trophic levels varies across experimental or geographical gradients of resource supply (reviewed by Rosenzweig & Abramsky 1993; Abrams 1995; Waide et al. 1999; Mittelbach et al. 2001). Over the past decade, an alternative perspective has emerged in which diversity is viewed as an independent variable that regulates how efficiently groups of organisms capture and convert available resources into new biomass (reviewed by Tilman 1999; Loreau et al. 2001; Naeem 2002; Hooper et al. 2005). These contrasting perspectives have engendered a lively debate about whether species diversity is the cause or the consequence of resource densities that limit the production of biomass (Grime 1997; Huston 1997; Fridley 2001; Loreau et al. 2001; Schmid 2002; Worm & Duffy 2003).
As researchers have tried to resolve these contrasting perspectives, new ideas have been formulated to explain how species diversity might simultaneously be a cause and a consequence of resource density. For example, several authors have converged on a similar hypothesis that spatial variation in the availability of key resources might drive variation in species diversity across sites while, at the same time, the diversity of species at any given site influences how efficiently available resources are captured and converted into biomass (Loreau et al. 2001; Schmid 2002; Worm & Duffy 2003). If correct, this idea could help resolve a seeming paradox by suggesting that species diversity and resource density exhibit reciprocal causal relationships, but the dependent and independent variables change as a function of spatial scale. Although this hypothesis has begun to take heuristic form, we are not aware of any study that has examined how species diversity and resource density might exhibit bi-directional causality in the same ecological system, or even demonstrated that reciprocal causal relationships are theoretically or empirically possible.
Here we report the results of a mesocosm study that illustrates how species diversity of a group of mobile consumers might respond to the availability of resources at one spatial scale, while species diversity at a different spatial scale controls how efficiently resources are consumed. Our research focused on a system of predatory ladybeetles and their primary resource, the pea aphid Acyrthosiphon pisum Harris. Our interest in reciprocal relationships between predator diversity and prey density was motivated by previous work in fields of alfalfa Medicago sativa L. In this system, predator diversity is strongly correlated with aphid density (Cardinale et al. 2003), which can vary by several orders of magnitude among fields (Gross, Ives & Nordheim 2005). This variation occurs because alfalfa is harvested three to four times per summer with 3–5 weeks between cuts. At harvest, pea aphid density is reduced by 100–1000 ×, but populations increase rapidly afterwards (Hutchison & Hogg 1985; Rauwald & Ives 2001). Because harvesting is asynchronous among fields there is considerable spatial variation in aphid density at any given time. Like most aphid predators, ladybeetles are highly mobile, and as they fly among fields in search of prey they aggregate in areas of high aphid abundance (Ives, Kareiva & Perry 1993). Within days, aggregation can lead to a strong association between aphid density and total predator abundance and, because predator abundance and diversity are also correlated, there is a corresponding association between aphid density and predator diversity (Cardinale et al. 2003).
While the observations described above lead us to suspect that spatial variation in aphid density is an important determinant of variation in predator diversity, there is also reason to believe that predator diversity in the regional colonist pool (i.e. the number of species able to colonize any given field) influences how efficiently predator assemblages control aphid populations. We have documented several types of nonadditive interaction among natural enemy species in this system (Snyder & Ives 2001; Cardinale et al. 2003), which can cause predator assemblages to impact aphid populations in a manner that is disproportionate to the total abundance of predators. We have become increasingly interested in such interactions because in the mid-western USA where our work is performed, the richness and composition of predators have been altered by species invasions. For example, ladybeetles are the most common generalist predators in our system (Snyder & Ives 2003). Of the four to six native species, Coleomegilla maculata Timberlake is the only one that remains common. The guild is now dominated by Harmonia axyridis Pallas and Coccinella septempunctata L., both of which were intentionally introduced into agricultural systems for aphid biocontrol (Angalet, Tropp & Eggert 1979; Koch 2003). Growing evidence suggests that native ladybeetle species are being displaced by antagonistic interactions with their introduced counterparts (Alyokhin & Sewell 2004; Snyder, Clevenger & Eigenbrode 2004).
Taken collectively, our observations led us to hypothesize that predator diversity in this system is both a cause and a consequence of aphid density. At the larger scale where invasions and/or extinctions have altered the number and composition of species in the predator colonist pool, ladybeetle diversity may act as an independent variable that influences the efficiency of aphid biocontrol. Yet, at the smaller scale where predators aggregate in fields with high aphid density, predator diversity may be the dependent variable that responds to spatial variation in prey availability among fields. Ideally, we would test this hypothesis by manipulating aphid density and predator diversity at the ‘landscape’ scale where predators move among multiple fields. But of course, this is not possible, and we must settle for more limited insights that come from smaller scale models of this system. With this in mind, we set up a mesocosm experiment that attempted to capture certain key features of this system. Using a split-plot experimental design, we simultaneously manipulated predator diversity and aphid density in thirty 8-m3 field enclosures. Within each enclosure, we varied the initial density of pea aphids in 1 m2 patches of alfalfa. These patches were separated by physical barriers that could only be crossed by flying adult ladybeetles, which allowed us to assess how predator abundance and diversity respond to spatial variation in prey density at the scale of a patch. At the same time, we varied the number of species of ladybeetles that were added to the enclosures. By measuring the population growth rate of pea aphids in a cage, we determined how ladybeetle diversity impacted rates of predation at the scale of the whole enclosure. The mesocosms are only a caricature of this system, and important biological details are clearly simplified. Even so, the results are useful because they demonstrate that it is possible for species diversity to be both a cause and a consequence of resource density in the same ecological system, and they shed light on how this might occur for groups of mobile consumers.
- Top of page
- Materials and methods
Repeated measures anova confirmed there were significant differences in aphid density among enclosures stocked with differing numbers of ladybeetle species (F2,30 = 13·87, P < 0·01). When alone, each predator species reduced aphid density relative to control enclosures that had no predators (Fig. 1a, P < 0·05 for all F-tests comparing means). Yet, when all three predator species were together in the same enclosure, aphid densities were reduced less than would be expected from the effects of each predator species when alone (compare Obs with Exp, Fig. 1a). Indeed, the multipredator interaction term (coefficient b5 of eqn 1) was significantly less than zero on all but one date of the study (Fig. 1b), indicating that the effect of the three predator species when together was less than that expected from independent, additive effects. This suggests that when predator species were placed together in the same system, some form of antagonistic interaction reduced per capita predation efficiency and limited the ability of the predator guild to control aphid populations.
To investigate the nature of this interaction further, we performed a laboratory experiment in which we documented the foraging behaviour of predators on individual alfalfa plants when species were alone or together. When predators species were alone, each species foraged near the top and centre of plants where aphids tended to cluster (presumably to feed on new foliage, Fig. 1c). When predator species were together on the same plant, foraging H. axyridis and Co. maculata were displaced towards the plant margins, resulting in lower rates of predation by both species (t = 2·87, d.f. = 10, P = 0·01 and t = 2·40, d.f. = 10, P = 0·03, respectively, Fig. 1d). These results suggest that C. septempunctata is behaviourally dominant over Co. maculata and H. axyridis, and that interference competition may explain the nonadditive, antagonistic interaction among ladybeetle species observed in the field experiment.
The second question we asked is whether variation in aphid density among patches influenced the observed diversity of ladybeetle species. Within 24 h of adding predators to the cages, there was a positive relationship between observed predator richness and aphid density in a patch (Fig. 2a). By day 3 this positive relationship was statistically significant, and it persisted through day 6. By day 11 of the experiment, predators had reduced aphid densities to low levels across patches, which resulted in there being no significant variation in aphid density within enclosures during the latter portion of the experiment. From these results, we conclude that within a matter of days, predators had distributed themselves among patches in such a way that predator diversity was an increasing function of aphid density.
Figure 2. Ladybeetle diversity as a consequence of aphid density. (a) Richness of predators in a patch as a function of variation in aphid abundance (dark symbols = mean ± 95% CI for patches initialized with different aphid densities). The solid line gives the best fit from a mixed model anova with experimental unit included as a random effect (slopes ± 95% CIs are given at right). Grey symbols give the mean ± bootstrapped 95% CI of richness predicted from a Poisson regression of predator abundance (see Table 1). (b) Observed predator richness in a patch as a function of total predator abundance (mean ± 1 SEM for each patch averaged across all dates). (c) Rates of immigration to/emigration from patches for predator species combined (main) or individually (insets).
Download figure to PowerPoint
A potential explanation for the positive relationship between predator diversity and aphid density is that as predators aggregated in areas with high aphid density, a higher abundance of predators led to a greater probability that any given species would be observed in a fixed sampling effort. The potential for this mechanism to explain patterns in our study is clear from Fig. 2(b), which shows that predator richness in a patch was a monotonically increasing function of total predator abundance. To test this hypothesis explicitly, we used a Poisson regression to calculate levels of predator richness that would be expected based simply on how predator abundance covaried with aphid density (see eqn 2, Materials and methods). Parameter estimates for the most parsimonious Poisson regression (Table 1) indicate that: (1) predator abundance in patches differed among the three predator species (C. septempunctata > Co. maculata > H. axyridis); (2) predator abundance in a patch increased with aphid abundance in that patch; but (3) this response varied among dates. Using these parameter estimates to calculate expected predator richness, we found that observed levels of richness were no different than those predicted from variation in predator abundance alone (compare black/grey symbols, Fig. 2a). This result indicates that variation in predator diversity among patches can be explained simply by the accumulation of individuals of each predator species in patches with high aphid density.
Table 1. Maximum Likelihood estimates for a Poisson regression predicting predator abundance in patches (eqn 2 in text)
|Parameter||Estimate||SE||95% Confidence limits||χ2||Pr > χ2|
|Aphid density||0·29||0·12||0·06||0·53|| 5·82||0·02|
| C. septempunctata||2·39||0·33||1·74||3·04||52·27||< 0·01|
| Co. maculata||1·82||0·34||1·16||2·49||28·67||< 0·01|
| H. axyridis||0·00||0·00||0·00||0·00|| || |
| Day 1||−1·71||0·35||−2·39||−1·02||23·66||< 0·01|
| Day 3||−0·14||0·23||−0·59||0·31|| 0·38||0·54|
| Day 4||−0·09||0·23||−0·54||0·35|| 0·17||0·68|
| Day 6||0·00||0·00||0·00||0·00|| || |
|Aphid density × time|
| Day 1||−0·02||0·21||−0·44||0·40|| 0·01||0·92|
| Day 3||0·24||0·15||−0·07||0·54|| 2·35||0·13|
| Day 4||0·51||0·17||0·18||0·84|| 9·11||< 0·01|
| Day 6||0·00||0·00||0·00||0·00|| || |
Why did predators accumulate in patches with high aphid density? Predator movement on day 5 of the experiment suggests that predator emigration from a patch decreased as aphid abundance increased (E = 1·01 − 0·44*ln[aphids st−1], F1,16 = 8·23, P = 0·01), but rates of immigration into a patch were independent of aphid density (I = 2·47 + 0·02*ln[aphids st−1], F1,16 = 0·02, P = 0·88) (Fig. 2c). These results argue that ladybeetles did not differentiate among patches when searching for aphid prey, but upon finding a patch with high aphid abundance they tended to remain there for longer periods of time.