- Top of page
The abundance of specialist insect herbivores is often studied in relation to characteristics of their host plants, including patch size, patch density, and patch purity (from monospecific to diverse) (Kunin 1999). More specifically, the resource concentration hypothesis (Tahvanainien & Root 1972; Root 1973) predicts that specialist herbivores should achieve higher loads (density per unit mass of the host-plant species) when their food plants grow in dense patches in pure stands. A resource concentration effect occurs when specialist insects are more likely to find, remain and reproduce on their hosts when these plants grow in such stands. The resource concentration hypothesis has wide relevance with implications for both applied entomology (e.g. Altieri 1995; Andow 1991) and community ecology (e.g. Kareiva 1983; Strong, Lawton & Southwood 1984).
Whether or not herbivores respond to resource concentration will depend on various aspects of their biology and ecology, including their host-sensing ability, dispersal, food requirements and vulnerability to competitors and natural enemies (Kareiva 1983; Kunin 1999). For example, resource concentration effects are predicted for specialist insects with high population growth-rates and for those that are good at finding their hosts even at low density.
In this paper we report a test of the resource concentration hypothesis for nine species of specialist insect herbivores sampled from the experimental communities of the BIODEPTH project Silwood Park fieldsite (Hector et al. 1999, 2000, 2001). The resource concentration hypothesis mainly addresses the consequences of host-patch purity (e.g. agricultural monocultures vs. intercropping, see above). However, patch purity often also varies in concert with other factors such as plant density (abundance) and patch size. In our study, the experimental units were 4 m2 experimental grassland communities. A given host-plant species in our plots may differ simultaneously in its abundance, its patch size and its patch purity, with all three potentially reaching their maximum in monocultures. Our analysis follows the structure of the resource concentration hypothesis: we test for the effects of host-plant abundance (measured as above-ground mass) on (1) the presence/absence and (2) loads of specialist insects in our experimental communities, and for (3) additional residual effects of plant diversity (patch purity) after controlling for host-plant abundance.
- Top of page
The Silwood Park BIODEPTH experiment has already been described in detail (Hector et al. 2000, 2001). Briefly, the experiment comprised 66 experimental plant communities, each 2 × 2 m, varying in species and functional group richness and species composition. We established five levels of plant species-richness with one, two, four, eight or 11 species. These treatments were replicated 11, six, six, five and five times, respectively (allocating more replicates where variance was expected to be higher), forming 33 different experimental plant communities. We constrained the composition of these communities so that they had either one, two or three plant functional groups (grasses, legumes or other forbs) with all polycultures containing grasses. Finally, the 33 different communities were each present in two repeats to replicate species composition separately from richness, generating 66 plots in total. These diversity manipulations also generated gradients in the abundances of individual plant species that were highest in monoculture and generally declined with increasing diversity. Because individuals are both conceptually and practically hard to identify for clonal plants, we use plant mass as a measure of abundance.
Insects on the experimental plots were suction-sampled with a Burkard Vortis (Arnold 1994) during four sampling periods: May and August of both 1997 and 1998. During each sampling period, a series of four suction-samples, each of 30 s duration, was taken on different days over a period of 2 weeks, visiting plots in a different order each time to minimize potential biases of time of day and weather. The four suction-samples were combined to give a total area per sampling period of 0·076 m2. Because the specialists reported here were present only in May or August, reported densities and loads for each insect species are from a single sampling period from a single year. As the resource concentration hypothesis makes predictions about insect loads, these were estimated using the plant mass taken from the harvest of the appropriate year as:
- load = [specialist insect density (no. of insects m−2)]/[host-plant mass (g m−2)]
Plant masses were estimated from two 50 × 20 cm samples harvested from a central 1 × 0·5 m permanent quadrat (Hector et al. 1999, 2000, 2001). Our insect loads are therefore estimates from suction-samples that could be taken only immediately adjacent to the permanent quadrat where the vegetation samples were taken and not within it. We used all loads in the analyses reported here, as on log–log axes relationships were fairly linear across the whole range of estimates and values of Cook's distance statistic (Di) indicated that regressions were not influenced unduly by the more extreme values (see, e.g. Quinn & Keough 2002: 95).
We defined specialist insects as those feeding on a single host-plant species based on both published food preferences (Waloff 1980; Bullock 1992) and as confirmed by our abundance and occurrence data (Table 1 and Otway 2000). We required at least 70% of the individuals of a putative specialist to have been collected from plots containing its host species, with no more than one individual per plot collected from those without the host species. This classification is highly conservative, in that individuals found on non-host plots may simply be ‘tourists’ in transit, or reflect the temporary occurrence of the host plant as a weed in the wrong plot before removal (plots were weeded of unwanted species two to three times a year). Analyses were conducted only on abundant herbivore species with a total of more than 30 individuals. These selection criteria resulted in nine species of insect herbivore that we could be confident were specialist on a single host plant within our plots and which were abundant enough for analysis (Table 1). Seven of these were weevils (Coleoptera: Curculionidae and Apionidae) specializing on a variety of host plants, one a leafhopper (Errastunus occelaris, Fab.) on Holcus lanatus (L.) and one a psyllid (Craspedolepta nervosa, Förster) on Achillea millefolium (L.).
Table 1. Details of the nine specialist insect species and their host plants. The insects include weevils (Coleoptera: Curculionidae) (1–7), a leafhopper (Auchenorrhyncha) (8) and a psyllid (Psillidae) (9). ‘Present/absent’ gives the number of plots containing the host-plant species that were with and without specialists, respectively. ‘Density (range)’ gives the range of the total number of individuals of each species collected from four 30-s suction-samples with a total summed area of 0·076 m2. Host-plant associations observed in the field matched host-plant records from the literature except for the Rhinoncus species. According to Bullock (1992), R. castor attacks only Rumex acetosella (a species not present in our experiment) when it was found on Rumex acetosa in our plots
|Insect species (and authority)||Present/absent||Density (range)||Host plant|
|1 Gymnetron pascuorum (Gyllenhall 1913)||14, 4||(0) 1–17||Plantago lanceolata|
|2 Rhinoncus castor + R. pericarpius*||10, 8||(0) 1–4||Rumex acetosa|
|3 Apion dichroum (Beddel, 1886)|| 6, 8||(0) 1–15||Trifolium repens|
|4 Apion loti (Kirby, W., 1802)||13, 5||(0) 1–7||Lotus corniculatus|
|5 Apion curtirostre (Germar, 1817)||12, 5||(0) 1–3||Rumex acetosa|
|6 Ceutorhynchidius troglodytes (Fabricius, 1787)||14, 3||(0) 1–11||Plantago lanceolata|
|7 Mecinus pyraster (Herbst, 1795)||11, 7||(0) 1–8||Plantago lanceolata|
|8 Errastunus occelaris (Fallén 1806)||15, 3||(0) 1–4||Holcus lanatus|
|9 Craspedolepta nervosa (Förster)||12, 0|| 3–48||Achillea millefolium|
Individual species of plants can vary in their performance across the experimental gradients, sometimes overyielding in mixtures (Loreau & Hector 2001; Hector et al. 2002). Using plant mass in the denominator to calculate herbivore loads controls automatically for this. The different explanatory variables relevant to the resource concentration hypothesis and related ideas (e.g. plant patch purity, density, size, etc.) are not necessarily orthogonal and will often be colinear to some degree. As our analyses focus on testing the resource concentration hypothesis, we follow the hierarchy given by the hypothesis and test first for the effects of host-plant abundance and then additional residual effects that arise due to host-patch purity (i.e. the diversity of the background plant communities). Analyses were conducted using r (r Development Core Team 2004).
- Top of page
The responses of specialist herbivore populations across our plant diversity gradient were determined mainly by two contrasting effects of host-plant abundance with only a small additional effect of plant species-richness. The first effect worked in the direction predicted by the resource concentration hypothesis: specialist insect species located their host plants more frequently when their host plants were more abundant (Fig. 1). However, in plots where specialist herbivores were present, we found a resource dilution effect rather than the predicted resource concentration effect (Fig. 2). There was also an additional significant but weak effect of increasing plant species-richness, with reduced loads in high-compared to low-diversity plots as the resource concentration hypothesis predicts.
Other recent work only partially supports the resource concentration hypothesis. For example, the work of Knops et al. (1999), Koricheva et al. (2000) and Haddad et al. (2001) supports the hypothesis, at least for certain specialists. In contrast, Yamamura (2002; papers therein) finds evidence of resource dilution. Reviews of earlier literatures show evidence of both resource concentration and resource dilution (reviewed in Kareiva 1983; Andow 1991; Kunin 1999; Rhainds & English-Loeb 2003). Despite the undoubted occurrence of the resource concentration effect it is clear that it is far from a universal phenomenon. One limitation of some of these studies is that they analyse specialist insect densities rather than loads and therefore do not test directly the resource concentration hypothesis or control for variation in host-plant abundance.
As we describe in the Introduction the resource concentration hypothesis has several components, but it is not always clear which aspect of the hypothesis is being tested. In our study the increased colonization (presence/absence) of host-plant patches where hosts were more abundant was consistent with the resource concentration hypothesis. In contrast, when insects were present, the negative effect of increasing host-plant abundance on specialist herbivore loads runs counter to the predictions of the hypothesis. Different potential pathways by which resource concentration effects can arise − e.g. host-plant location vs. postcolonization population growth − can therefore show different responses to changes in host-plant abundance.
Our analysis (Fig. 1) shows that specialists could differentiate clearly between plots with high (monocultures) and low host-plant abundances. Under completely passive dispersal, densities of collected insects should also be constant and loads would then vary only when this constant insect density was divided by differing plant masses in different diversity plots. Insect density did increase with host-plant mass but less than proportionally (Joshi et al. 2004), so that insect loads nevertheless decline. It appears therefore that loads were not set only by initial colonization but that later population processes also play a role. However, we do not have data that would allow us to investigate this further. The natural enemies hypothesis proposes differential predation pressure on herbivores caused by differences in the nature of either the host-plant community (e.g. density, patch size) or of the background-plant community (diversity). There were no obvious general patterns in the abundance of predator or parasitoids in our data (Otway 2000), but we do not have the experimental data that would allow us to investigate this hypothesis further. In summary, we can rule out a scenario of completely passive dispersal and loads instead appear to be set by a mixture of active colonization and later population growth that together were not effective enough to generate resource concentration effects, at least at this stage of our experiment.
What are the implications of these resource dilution effects? This depends on whether the effects we observed are transient (created by the disturbance and establishment of the experimental plant communities) or whether the effects would have persisted. There were no obvious temporal patterns when individual sampling periods were compared for our data (Otway 2000 and unpublished) but we are limited to looking at 2 consecutive years.
Our resource dilution effects have clear implications for the interpretation of biodiversity experiments. Resource concentration effects on herbivores were one of the first mechanisms proposed for the differential performance of plant species in these experiments (e.g. Schulze & Mooney 1993) in which host plants in monocultures and low-diversity mixtures would experience higher specialist herbivore loads than those in high-diversity communities. Resource dilution effects, of the type we observe, should have the opposite effect by causing host plants in high diversity mixtures to experience greater herbivore pressure. If resource concentration effects take time to develop then early results from these experiments may underestimate the strength of the relationship between plant diversity and productivity and resource concentration effects may play a role in increasing the strength of diversity–productivity relationship as the experiments progress. There are few published data with which to assess this possibility. As we have already discussed, Mulder and colleagues provide evidence to support resource concentration effects early in their experiments (Mulder et al. 1999; Koricheva et al. 2000), although these patterns were not monitored in the longer term to see whether the effects persist or change. There is currently only one long-term (since 1994) biodiversity experiment (Tilman et al. 2001). The relationship between productivity and diversity has become more positive with time in this experiment and, as discussed above, data from 1997 provide some evidence for resource concentration effects on some specialist herbivores (Haddad et al. 2001). However, there are several possible explanations for the strengthening of the diversity–productivity relationship with time and there are currently no published data on whether resource concentration effects played any role.
Alternatively, resource dilution effects of the type we report here may be persistent. Such effects have been found elsewhere (Yamamura 2002) and may be more widespread than thought previously. If this is the case, then it means we will have to re-assess the extent to which the effects of resource concentration occur in nature, with all the accompanying implications for the ecology of both specialist herbivores and their host plants. This latter picture of a more complex set of relationships between herbivores and their host plants is in line with much of the recent literature (e.g. Kunin 1999).