SEARCH

SEARCH BY CITATION

Keywords:

  • community structure;
  • connectivity;
  • gall wasp;
  • leaf-miner;
  • metacommunity;
  • oak;
  • parasitoid;
  • quantitative food web;
  • Quercus robur;
  • species richness

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

1. With habitat fragmentation spreading around the world, there is a pressing need to understand its impacts on local food webs. To date, few studies have examined the effects of landscape context on multiple local communities in a quantitative, spatially realistic setting.

2. To examine how the isolation of a food web affects its structure, we construct local food webs of specialist herbivores and their natural enemies on 82 individual oaks (Quercus robur) growing in different landscape contexts.

3. Across this set of webs, we find that communities in isolated habitat patches not only contained fewer species than did well-connected ones, but also differed in species composition.

4. Surprisingly, the effects observed in terms of species composition were not reflected in the quantitative interaction structure of local food webs: landscape context had no detectable effect on either the interaction evenness, linkage density, connectance, generality or vulnerability of local webs.

5. We conclude that the quantitative structure of food webs may be stable in the face of habitat fragmentation, despite clear-cut impacts on individual species. This finding offers hope-inspiring news for conservation, but should clearly be verified by empirical studies across both naturally and more recently fragmented systems.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

Habitat fragmentation is a leading threat to global biodiversity (Hanski 2005). Fragmentation entails not only a reduction in the overall amount of original habitats, but also in the connectivity of remaining animal and plant populations. That the spatial configuration of landscapes has important implications for processes at both the levels of populations (Hanski 1999; Tscharntke et al. 2002) and communities (Holt 1997; Holyoak, Leibold & Holt 2005) has been convincingly demonstrated. Yet, to our knowledge, no studies have addressed how the structure of fragmented landscapes affects trophic interactions within multi-species food webs in a spatially realistic setting (sensu Hanski & Simberloff 1997). As food web structure may affect multiple aspects of ecosystem functioning, including pollination services (Fontaine et al. 2006) and herbivory pressure (Petermann et al. 2010), such data are critically needed. Testing fundamental theories of spatial impacts on food web structure has therefore been identified as a key priority for modern community ecology (Amarasekare 2008).

Habitat fragmentation is expected to alter the species richness at different trophic levels, but also the distribution of trophic interactions among the species present. At the level of individual species and guilds, an influential hypothesis posits that species at different trophic levels in the web should react differently to landscape structure. In particular, species tied to a given resource (such as a given host or prey) can only utilize the parts of the landscape where this resource occurs. Sensitivity should then increase with increasing specialisation, and with increasing trophic level. Termed the ‘trophic rank hypothesis’ (Holt 1996; Holt et al. 1999), this notion has gained empirical support from some studies (e.g. Kruess & Tscharntke 1994; Davies, Margules & Lawrence 2000; Tscharntke et al. 2002; Steffan-Dewenter 2003), but been challenged by others (Mikkelson 1993; Doak 2000; Starzomski & Srivastava 2007).

At the level of interaction structure among the species present, quantitative food webs have been identified as promising tools for pinpointing the impacts of habitat change (Memmott 2009). Being quantitative in nature, these webs describe not only which species occur where and interact with whom, but also offer concise metrics of the relative frequencies of different interactions. A seminal study by Tylianakis, Tscharntke & Lewis (2007) showed how habitat modification may change the quantitative structure of interactions between species without changing species diversity. Whether similar effects will follow from habitat fragmentation has been suggested, but not fully explored. Work by Valladares, Salvo & Cagnolo (2006) suggests that the strength of key interactions, such as parasitism, may differ across fragments of different size, and between the fragment interior and the edge. In another study related to fragment area, Cagnolo et al. (2009) reported that the position of species within tritrophic food webs consisting of plants, leaf-miners and parasitoids may affect their vulnerability to habitat loss, but only if the species was specialized. Other effects of habitat area have been reported by Sabatino, Maceira & Aizen (2010) who – working on mutualistic webs of pollinators and host plants – found a lower number of species and especially of insect-plant interactions on smaller than on large habitat patches.

Such empirical findings have recently been explored from more theoretical perspectives. In a modelling study, Fortuna & Bascompte (2006) observed an altered structure of plant-mutualist networks under habitat loss: after a critical point in the degree of destroyed habitat, species in the community quickly went extinct. By contrast, Ashworth et al. (2004) suggested less dramatic effects, based on the assumption that specialist species are generally more sensitive to the detrimental effects of habitat fragmentation than are generalist species. If interactions between plants and generalist pollinators are typically more common than are interactions between plants and specialist pollinators, then the loss of a few specialist pollinators will have relatively little impact on the total pollination service. Such a pattern, they argue, will reduce the functional consequences of habitat loss.

In this study, we use a species rich community of monophagous herbivores and their natural enemies to study the effects of landscape context on the composition and quantitative structure of host-parasitoid food webs. Based on samples of 82 local insect food webs in different spatial settings, we measure the effects of habitat connectivity on the classical metric of species richness as compared to recent quantitative and qualitative metrics of food web structure. Specifically, we ask: Do more isolated food webs have fewer species than do food webs that are well-connected in space? If so, do all species and all functional groups of species respond to landscape context in the same way? Does the landscape context alter the species-specific interaction patterns in the local communities? We show that both quantitative and qualitative measures of food web structure are less affected by habitat fragmentation than are species richness and species composition, a finding with important implications for both biological conservation and for spatial community theory.

Materials and methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

Study system

Even small habitat patches may harbour a huge set of microhabitats and a substantial number of species. This makes it difficult to characterize the local community at any one site, and to compare community structure across a larger set of sites. To examine the effect of landscape context, we therefore chose a model system based on a natural, yet easy-to-sample network of local communities in a spatially well-defined setting: communities of specialist insect-herbivores and their natural enemies on the pedunculate oak, Quercus robur L., 1753. This is the only naturally-occurring oak species in Finland, where the genus meets its northern range limit. The current, fragmented distribution of Qrobur in Finland is primarily dictated by man (see Appendix S1, Supporting Information).

To understand patterns in species distribution and community dynamics in natural landscapes, we should clearly consider the fragmentation history of the system. Where a substantial fraction of species in recently disturbed landscapes may be on their way to extinction (Laurence 2008; Kuussaari et al. 2009), community structures in naturally fragmented landscapes may be more reflective of counter-balanced colonisation and extinction processes (Hanski 1999). In this context, our study system is clearly a naturally fragmented one. Over the last few 100 years, the distribution of oaks has remained fairly stable: current oak stands are evident on detailed distributional maps drawn a rough century ago (Skult 1956), and many oak trees themselves are likely hundreds of years old.

In our study, we specifically focused on the community of host-specific gallers and leaf-miners. For these taxa, the distribution of the host tree species will offer an adequate description of landscape configuration. Both gallers and leaf-miners host a species-rich and well-studied guild of natural enemies, consisting mostly of parasitoid wasps (Askew 1961; Askew & Shaw 1974; Askew 1980; Stone et al. 2001). Gall wasps are also attacked by inquilines, i.e. parasites which develop by feeding on the gall tissue and often kill the host directly or indirectly by consuming its resources (Askew 1961; Ronquist 1994). For further details on the study system, see Appendix S1 (Supporting Information).

Insect samples

Quantitative insect samples were collected from 82 oak trees in the archipelago of SW Finland (Fig. 1a). To cover the full range of variation in tree density and configuration, we sampled trees at two different scales: at the scale of different oak stands within a region of c. 100 km2 (Fig. 1b, this material is henceforth referred to as the large scale material), and at the scale of individual oak trees within a landscape of c. 5 km2, the island of Wattkast (Fig. 1c; this material is henceforth referred to as the landscape scale material). For clarity, all material names are formatted in small caps and the sampling sites of each material defined in Fig. 1.

image

Figure 1.  Empirical materials collected for this study. Panel (a) shows the location of the study area on the southwest coast of Finland (as indicated by a square). Panel (b) shows the sampling sites of the large scale material, with fragmentation zone indicated by colours. Panel (c) shows the details of the landscape scale material, as collected on the island of Wattkast. Here, black dots show the location of each 1868 naturally-occurring oak trees, of which 22 large trees (white circles) were selected for the big tree landscape scale material and 52 young trees (grey triangles) trees were chosen for the small tree landscape scale material. For the trees used to analyze fragmentation impacts on food web structure (the large scale material and the big tree landscape scale material), we show a quantitative representation of the local food web. In a web, each bar at the lower level represents a host species and each bar at the upper level a parasitoid or inquiline species. Inside the host bars, the black part indicates parasitized host individuals and the grey part unparasitized ones. Lines between hosts and parasitoids describe trophic interactions, with the width of the line proportional to the frequency of the interaction. Host species detected at a site but not involved in trophic interactions (i.e. not yielding a single parasite) can be distinguished as blocks from which no line emanates. The width of each web has been scaled to reflect the total number of individuals recorded in the field.

Download figure to PowerPoint

In practice, the landscape scale material formed a subset of the large scale material. Importantly, the two nested materials offer semi-independent tests of the effect of habitat fragmentation on local food web structure. Based on the large scale material, we may compare local communities from sites differing in the average density of host trees in the surrounding landscape (later referred to as ‘fragmentation zones’, see Fig. 1b). Based on the landscape scale material, we may compare the impact of the specific landscape context (connectivity, sensu Moilanen & Hanski 2001; Moilanen & Nieminen 2002) on the structure of individual food webs. In the context of the large scale material, we neglected the effects of variation in specific connectivity among trees within landscape, as these differences were secondary compared to the larger differences among fragmentation zones.

The large scale material encompassed 30 large trees: four trees within extensive and continuous oak stands on the Finnish mainland (the ‘non-fragmented zone’), four trees growing on isolated small islands with few or single oak trees (the ‘highly fragmented zone’), and 22 trees in between these two extremes (the ‘moderately fragmented zone’), growing on the island of Wattkast and equalling the big tree landscape scale material (see below and Fig. 1). Each local food web was constructed for a single tree, both at the large and landscape scales and for trees in all size classes.

The landscape scale material encompassed two sets of trees: 22 large trees (6–15 m in height; later referred as the big tree landscape scale material) and 52 small oak trees (1·2–2·2 m in height; henceforth the small tree landscape scale material). These trees were chosen from among the 1868 oak trees growing on the island of Wattkast (see Gripenberg & Roslin 2005) using a restricted randomisation procedure aimed at including trees from the full landscape. Here, the inclusion of two separate size categories was motivated by the different types of insect population dynamics observed on different-sized trees in a previous study (Gripenberg et al. 2008).

For the big tree landscape scale material and the rest of the large scale material, sampling was conducted three times in 2006: in May-June, in late July, and in September in 2006. During each sampling event, a standardized volume of foliage (30 half-meter branches per tree) was collected with the aid of a pole pruner, and all galls and leaf-mines present were recorded. For the small tree landscape scale material, sampling was implemented in September 2007. These trees were low enough to be searched through completely for every leaf mine and gall present. For additional details on and justification for the methods used, see Appendix S2 (Supporting Information).

To assess the number and identity of associated natural enemies (including parasitoids and inquilines, that are later referred as parasitoids for clarity), live herbivores were collected and reared. The taxa detected were categorised with respect to trophic level (hosts and parasitoids) and to feeding guild (leaf-miners, gallers, leaf-miner parasitoids and galler parasitoids). For details of the rearing procedure and species identification, see Kaartinen et al. (2010) and Appendix S2 (Supporting Information).

Description of isolation

To describe the level of isolation of each tree on the island of Wattkast, we used a connectivity metric modified from Hanski (1999). This metric can be seen as the opposite of isolation, being directly proportional to the expected immigration of insects to the focal tree i given maximal patch occupancy of all other trees j:

  • image

Here, dij is the distance between trees i and j, and Nj is the insect population size on tree j (assumed to be proportional to tree girth; for a justification, see Gripenberg et al. 2008). Following Tack et al. (2010), the value of parameter α was set to 0·004 (so that 1/α = 250 m; a rough estimate of the average dispersal distance of insects in our study system).

Analyses

Food web metrics

To characterise the structure of local communities, we calculated several metrics of food web structure. All metrics were first calculated in their quantitative form, following e.g. Tylianakis, Tscharntke & Lewis (2007). To describe the evenness of energy flow through different interactions in the food web, we used the measure of interaction evenness. To weight the number of trophic interactions by their respective abundance, we used the measure of linkage density. To depict the average, weighted number of trophic interactions per species, we used the measure of connectance. To describe the ratio of parasitoid species to host species, we used the measure of generality, and to describe the ratio of host to parasitoid species, we used the measure of vulnerability.

While quantitative networks measure changes in the frequency of different interactions, isolation may affect the ability of a parasitoid to locate one of its host species, such that interactions may not even be realised. This would be more obvious in binary (e.g. qualitative) webs than quantitative ones (where common interactions may dominate the structure). To complement quantitative measures of interaction structure, we therefore calculated the same metrics for binary (presence-absence) networks following Tylianakis, Tscharntke & Lewis (2007). For calculations, and for constructing visual representations of web structure, we used the bipartite package (ver. 1.02; Dormann, Gruber & Fründ 2008 and Dormann et al. 2009) of r (ver. 2.8.1; R Development Core Team 2008).

Statistical models

To analyse the influence of spatial connectivity on species richness, on species-specific presence-absence patterns and on different metrics of food web structure, we used generalized linear models (McCullagh & Nelder 1989). To examine how species richness varies across different levels of habitat fragmentation, we first used the landscape scale material to model local species richness as a function of the connectivity of the host tree. When examining the richness of all species, host species and host guilds, we used the complete material (both the bigtree landscape scale material and the small tree landscape scale material). To validate the patterns detected at the landscape level, we then used the large scale material. Here, we modelled species richness as a function of fragmentation zone (the non-fragmented, moderately and highly fragmented zone, Fig. 1b). Since based on the trophic rank and the specialisation hypotheses, we expected different trophic levels and feeding guilds to react differently to habitat fragmentation, we examined the total number of species and species within different species groups separately. As the response variable was a count (the number of species), we assumed Poisson-distributed errors and a log-link function.

To establish the basis of variation in species richness in species-specific distribution patterns, we used the big tree landscape scale material and the small tree landscape scale material. Here, we used the presence or absence of each individual species on each individual tree as a separate record (for the results of an alternative, repeated-measures model, see Table S1, Supporting information), modelling it as a function of tree connectivity and tree size (big vs. small), the identity of the guild to which the species belonged, and the identity of the species within the guild (a nested term). To test for a differential effect of tree connectivity in trees of different size, we added the interaction connectivity × size. Here, we a priori expected a stronger effect of connectivity on insect presence in smaller trees, given more pronounced extinction-colonization dynamics in the smaller insect populations inhabiting these trees (Gripenberg et al. 2008) To allow for variation in the response of different species and guilds to trees of different connectivity and size, we added the interactions connectivity × guild, size × guild, connectivity × species(guild), size × species(guild), size × connectivity × guild and size × connectivity × species(guild). Following the principles of model simplification (Crawley 2007), we started with a model including all the above interactions, then excluded the least significant highest-order interaction first, progressing with model simplification until the minimum adequate model was obtained. These models were fitted with sas for Windows v. 9.1.3 proc glimmix (Littell et al. 2006, SAS Institute Inc., Cary, NC, USA). For additional details, see Appendix S1 and Table S1 (Supporting information).

In order to examine how fragmentation affects the interaction structure of local communities, we used the big tree landscape scale material to model each quantitative food web metric as a function of tree connectivity. To support these analyses, we compared quantitative food web metrics from the moderately and highly fragmented zones of the large scale material. In both analyses, we took a liberal approach and analyzed all metrics as response variables in separate models – even though this will result in a large number of partly non-independent tests. By this approach, we explicitly aimed to identify any signal of connectivity on food web structure (i.e. to minimize the probability of a type I error, while being less concerned about type II errors). For each metric, we assumed normally distributed errors and an identity link function.

All analyses of parasitism rate and food web metrics were implemented in sas System 9.2 for Windows, proc genmod (SAS Institute Inc., Cary, NC, USA). In cases where the data showed signs of overdispersion (i.e. when the variance of the response significantly exceeded the variance expected under the assumption of the relevant error distribution), we estimated the dispersion parameter from the data. To test for another factor with the potential for generating overly liberal tests, we looked for signs of spatial autocorrelation among individual observations (Legendre 1993). Here, we evaluated spatial correlograms for each response variable and for the residuals of each statistical model. As a formal test of the patterns, we calculated Moran’s I coefficients and their respective probability values with proc variogram in sas 9.2. Neither raw variables nor residuals from the generalized linear models showed significant spatial autocorrelation (P ≥ 0·19 for all variables, P ≥ 0·23 for all residuals; cf. Diniz-Filho, Bini & Hawkins 2003). For additional details of statistical analyses, see Appendix S2 (Supporting Information).

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

In total, we recorded 31740 herbivore individuals of which 8143 were alive and at a stage suitable for rearing (Fig. 1; see Appendix S1, Supporting information). This material encompassed 25 host species of which 11 were leaf-miner and 14 galler species. Treating each generation of each gall wasp as a different taxon, we scored 18 gall wasp taxa (see Appendix S2, Supporting Information). The herbivores were attacked by 60 species of natural enemies, of which 32 mainly attack leaf-miners and 28 primarily attack gallers (Askew 1961, 1980; Noyes 2001).

Effects on food web structure

The most striking pattern to emerge from the data was an overall consistency in food web structure across host trees growing in a broad range of spatial settings. Among all sites, the structure of local food webs was relatively similar, with no single interaction being quantitatively dominant (Fig. 1). As a result, neither interaction evenness, linkage density, connectance, generality nor vulnerability varied detectably between the highly fragmented and middle zones (Fig. 2a–e; five metric-specific tests, P > 0·05 in all cases). Within the island of Wattkast, among the large trees, the same lack of effect was evident: none of the food web metrics examined was detectably altered by the connectivity of the focal tree (Fig. 2a–e; five metric-specific tests, P > 0·05 in all cases). These patterns were fully consistent among both quantitative and qualitative versions of each food web metric (compare Fig. 2a–e vs. Fig. S1, Supporting Information, respectively).

image

Figure 2.  Patterns in the measures of food web structure (a–e) and species diversity (f-h) vs. habitat isolation. Shown are the quantitative web metrics of (a) interaction evenness, (b) linkage density, (c) connectance, (d) generality, and (e) vulnerability, (f) the total number of species, (g) the total number of host species, and (h) the total number of parasitoid and inquiline species. For each response, the left-hand panel relates to patterns in the large scale material (i.e. differences between the highly-, moderately- and non-fragmented zones), and the right-hand panel to patterns in the landscape scale material (i.e. patterns across individual trees varying in connectivity). In all graphs, connectivity increases to the right and hence isolation to the left. Fitted curves refer to generalised linear models described in the text. The box plots on the left show medians, quartiles and ranges, whereas in the scatter plots on the right, data from the big tree landscape scale material are shown by black circles and thick lines, and data from small tree landscape scale material by open squares and thin lines. Note that the material from the small tree landscape scale material and the non-fragmented zone of the large scale material was excluded from some comparisons (panel h), since rearing success from these sites proved disproportionately variable and generally low (see Appendix S2, Supporting Information for details).

Download figure to PowerPoint

Species richness

Overall consistency in food web structure masked substantial variation in species richness. Within the island of Wattkast (Fig. 1c), increasing isolation significantly decreased the total size of local food webs. While the total number of species was significantly higher in samples from big than small trees (Fig. 2f; χ2 = 216·61, d.f. = 1, P < 0·0001), the imprint of fragmentation was consistent across both materials: isolated individuals of both small and large oaks hosted fewer species than did well-connected ones (Fig. 2f; χ2 = 20·96, d.f. = 1, P < 0·0001). Nonetheless, this pattern was not reflected in the comparison between different zones from the non-fragmented mainland areas to the highly fragmented islands, where the number of species did not differ detectably between zones (Fig. 2f; χ2 = 1·34, d.f. = 2, P = 0·51).

The overall pattern in species richness was largely driven by fragmentation effects on the hosts. The number of host species was significantly higher in the non-fragmented zone compared to moderately and highly fragmented zones (Fig. 2g; χ2 = 6·5, d.f. = 1, P = 0·01). There were fewer gall wasp than leaf-miner species on the isolated islands (χ2 = 4·9, d.f. = 1, P = 0·03; see Fig. S2, Supporting information), but otherwise diversity patterns did not differ between the two host guilds. This reduction in the species richness of hosts in the highly fragmented zone was supported by patterns within the island of Wattkast, where the number of host species increased with connectivity (Fig. 2g; χ2 = 11·4, d.f. = 1, P < 0·001). Again, samples from large trees included significantly more host species than did samples from small trees (Fig. 2g; χ2 = 168·3, d.f. = 1, P < 0·0001).

On the landscape scale across the island of Wattkast, parasitoid species number increased with connectivity (Fig. 2h; χ2 = 9·4, d.f. = 1, P = 0·002). There was a higher number of parasitoid species in samples from large than from small trees (Fig. 2h; χ2 = 366·9, d.f. = 1, P < 0·0001). When examining parasitoids of leaf-miners and gall wasps separately, the number of leaf-miner parasitoid species did not vary detectably with connectivity (χ2 = 0·5, d.f. = 1, P = 0·47; see Fig. S2, Supporting information), whereas the species richness of gall wasp parasitoids increased with connectivity (χ2 = 11·5, d.f. = 1, P = 0·001; see Fig. S2, Supporting information).

Species-specific distribution patterns

Patterns in species richness mirrored responses detectable at the level of individual species. Overall, the incidence of all guilds and species was consistently higher in samples from big trees than in samples from small ones (Table 1; main effect of Tree size), and the incidence of most species of leaf-miners, gall wasps, leaf-miner parasitoids and galler parasitoids increased with increasing connectivity (Fig. 3). Nonetheless, there was substantial variation in the exact response of both guilds and species within guilds (Table 1; two-way interactions – for similar results from an alternative analysis, see Table S1, Supporting information). The incidence of all gall wasp species consistently increased with increasing connectivity (Fig. 3a), as did the incidence of most of their parasitoids (Fig. 3c). Among the leaf-miners, the incidence also generally increased with increasing connectivity, but here there was more variation in species-specific responses (Fig. 3b). The most variable pattern was found among the leaf-miner parasitoids, where the incidence of half of the species actually decreased with an increasing connectivity of the host tree (Fig. 3d).

Table 1.   Generalized linear model of species-specific incidence as a function of connectivity, tree size, insect guild, species identity within guilds and relevant interactions
Independent variabled.f.FP
  1. The degrees of freedom given in the table refer to the nominator, whereas for the denominator, d.f. = 3091 for all terms in the model.

Connectivity127·6<0·0001
Tree size1597<0·0001
Guild318·3<0·0001
Species (guild)413·8<0·0001
Connectivity × guild32·80·04
Connectivity × species (guild)411·80·002
image

Figure 3.  Estimated incidence vs. connectivity in four guilds: (a) gall wasps, (b) leaf-miners, (c) gall wasp parasitoids (also including inquilines), and (d) leaf-miner parasitoids. Each line describes a species-specific incidence function as derived from the generalised linear model presented in Table 1. Estimates for the big tree landscape scale material are shown by solid lines, whereas estimates for the small tree landscape scale material are described by dashed lines.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

This study reveals a striking consistency in the quantitative structure of local food webs surrounded by different landscapes. Such consistency occurred in the face of major variation in the number and identity of species present within individual habitat patches. Since the metrics examined by us describe (partly) different aspects of local species assemblages, we will first discuss isolation effects on each metric in turn. We will then finish by considering their potential use in understanding what fragmentation really does to natural communities.

Effects on quantitative food web structure

In our assessment of 82 local assemblages, we detected no imprint in terms of metrics characterising quantitative food web structure. This pattern was consistent between the two materials collected at different spatial scales, and occurred despite our deliberate focus on avoiding Type I errors. Hence, we find it more suggestive of a genuine lack of effect than of the lack of statistical power. Such a resistance in interaction structure to major variation in landscape configuration offers positive news for biological conservation. It suggests that key features of local food webs may prove robust in the face of habitat fragmentation.

While here reported for a specific system, our finding seems to agree with the few empirical assessments available to date: Robustness in network structure was also found in two unmanipulated, natural pollination networks examined by Petanidou et al. (2008) and Olesen et al. (2008). In these studies, stability was observed in time, not in space: Petanidou et al. (2008) followed Mediterranean networks of plants and pollinators over four consecutive years, observing striking changes in species composition and interaction patterns, but little variation in quantitative food web metrics. Focusing on a high arctic pollination web over two seasons, Olesen et al. (2008) found that while the web was characterized by pronounced seasonal changes over the summer, there were no differences in web structure between the 2 years. Such temporal constancy in web structure is consistent with the spatial consistency observed in our study. A key question is then whether stability is a general feature, or something which varies with the structure of specific webs. Here, theoretical studies suggest that opposite structural features may increase web stability in mutualistic (pollination) and trophic (herbivory) networks: In mutualistic networks, stability seems enhanced by increased connectance and nestedness, but decreased by modularity – i.e. by a pattern where a high number of interactions involve generalist species. In trophic networks, stability appears promoted by decreased connectance and nestedness, but by increased modularity – i.e. by a web structure where many interactions involve specialist species, each interacting with a small subgroup of the web (Thébault & Fontaine 2010).

The latter prediction seems directly applicable to a clear-cut difference in the patterns detected in the current study and those reported from recent work in Ecuador. Comparing food webs along a disturbance gradient, Tylianakis, Tscharntke & Lewis (2007) found a pronounced impact of habitat modification on the interaction structure of local food webs – a pattern in stark contrast with the lack of response reported here. Importantly, the Ecuadorian food webs were characterised by higher connectance than were our temperate webs. Thus, the difference in response seems indicative of a general coupling between decreasing connectance and increasing stability (Thébault & Fontaine 2010). Nonetheless, the current comparison between two case studies will clearly need added support from multiple studies to come. With only c. 20 insect-parasitoid webs published to date, the relationship between food web structure and stability remains an area of intensive investigation.

In addition to the general patterns examined above, one specific feature of our webs deserve particular notice: In our webs, both host and parasitoid species accounting for quantitatively abundant trophic interactions were also the least likely to disappear, whereas substantial turnover was observed among the rarest species (see Fig. 1). Indeed, current metrics of food web structure are rather insensitive to the presence or absence of rare species – as the rarer the host, the lower the probability of it being detectably involved in any trophic interaction at all (see Fig. 1). This consideration is clearly of vital importance when using quantitative food webs as a tool for evaluating environmental impacts on local community structure (cf. Memmott 2009), as it implies that metrics of relative interaction frequencies can be insensitive to changes in species composition (but for a counterexample, see Tylianakis, Tscharntke & Lewis 2007). Where a focus on structural metrics might suggest that local community structure remains unchanged, a tabulation of the species actually present could offer the opposite view – as in the current case. That rarer species are more redundant (cf. Ashworth et al. 2004) is clearly a moot point, since variation in the strength of links – that is, the spreading out of interactions among both rare and common species – and the sum of weak interactions contributed by multiple rare species may in fact have a strong stabilising effect on overall food web dynamics (Berlow 1999; Gross et al. 2009). Hence, whether we are prepared to trade rare species in the community for a summary descriptor of web structure is not so clear, suggesting that we should ultimately examine both who actually occurs in the local community and how they interact, not just one of these aspects. Nonetheless, recent approaches suggest that ecological analyses may be extended to address both food web structure and species identity (Laliberté & Tylianakis 2010).

Effects on species richness and on taxon-specific distribution patterns

Importantly, net consistency at the level of quantitative food web structure masked substantial variation in species richness and in the composition of local assemblages. Consistent with recent predictions (Öckinger et al. 2010), communities in isolated habitat patches not only contained fewer species than large and well-connected ones, but also different species. Our results complement earlier results by Cagnolo et al. (2009) who – focusing on the effects of fragment area – founder steeper species-loss rates at higher trophic levels and with increasing trophic specialization. In our study, we add another dimension of idiosyncratic responses to landscape configuration, by resolving differential responses of species guilds to the spatial context of equal-sized patches. Nonetheless, in our system, the specific patterns detected among taxa were only partly consistent with the general trophic rank hypothesis (Holt 1996; Holt et al.1999): While both host and total parasitoid species diversity increased with increasing connectivity, the pattern within the latter group was due to one particular feeding guild – while the species richness of galler parasitoids declined with increasing isolation, parasitoids of leaf-miners remained unaffected.

Beyond general life-history traits assumed to create variation in species-specific responses to habitat fragmentation (e.g. Öckinger et al. 2010), two specific factors may contribute to the higher sensitivity to landscape context observed in parasitoids of gallers than in parasitoids of leaf-miners: First, cynipid galls are commonly characterised by hard shells and thick walls, and attacking such structures will call for specialised adaptations (Stone & Cook 1998; Bailey et al. 2009). Hence, parasitoids of oak galls are generally unable to attack any other hosts (Hayward & Stone 2005), whereas many parasitoids of oak mines are also known to attack mines on other species of trees and shrubs (Askew & Shaw 1974). Our measure of connectivity may then offer a more accurate description of the landscape as perceived by parasitoids of gallers than of leaf-miners. Second, the population dynamics of galling hosts differs from that of leaf-miners. Here, gallers are characterised by extreme fluctuations in density, frequently spanning several orders of magnitude between years – and between different generations within the same year (Schönrogge & Crawley 2000). Such boom-and-bust dynamics may increase the likelihood of local extinctions of the host, and in particular of its parasites. That galls are often lacking from small trees (Fig. 3a) – and their parasites even more often (Fig. 3b) – adds to the impression that small populations of these taxa will be subject to frequent extinction.

Also within feeding guilds, different species responded differently to landscape context (see Fig. 3). Part of this interspecific variation can be linked to the specific biology of the taxa involved. As an example, while the incidence of nearly all gall wasp parasites increased with connectivity, that of Aprostocetus aethiops (Zetterstedt, 1838) (Hymenoptera: Eulophidae) actually decreased (note the single downward sloping line in Fig. 3c). This species has been recorded from several alternative hosts, such as gall midges (Cecidomyiidae) feeding on maple (Acer) and birches (Betula) (Noyes 2001). Thus, the response of A. aethiops may be more reflective of the occurrence of alternative resources in the landscape than is that of other galler parasitoids. Given similar variation among other parasitoid species, individual taxa will likely span the full range from strict specialists of the oak habitat to more generalised species.

While variation among specific guilds and species in our system may seem like idiosyncrasies when described in full detail, all species will clearly come with their specific quirks and preferences. In fact, substantial variation in features like niche width and dispersal capacity appears typical of most or all natural communities (e.g. van Nouhuys 2005; Öckinger et al. 2010). Hence, if we want to predict the consequences of landscape change, we should clearly include such variation in models of food web and metacommunity dynamics – or else we will miss key features determining the realised response (see Pandit, Kolasa & Cottenie 2009).

Another conclusion emerging from the current patterns is that poorly connected food webs will be dominated by common species of particular characteristics. Consistent with recent observations of community homogenisation in homogenous habitats (Ekroos, Heliölä & Kuussaari 2010; Laliberté & Tylianakis 2010), this suggests that biotic homogenization may be widespread in natural communities and food webs, affecting multiple aspects of realized interspecific interactions.

Conclusions

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

At present, habitat fragmentation continues worldwide at an unabated rate. This has already resulted in significant loss of biodiversity (Hanski 2005), and there is substantial concern that the effects will also extend to changes in food web structure and associated function. In this context, our study offers hope by suggesting that key features of web structure may be resistant to the effects of habitat isolation. Nonetheless, our results were clearly derived from a particular study system, as characterised by natural and long-term habitat fragmentation. A priority for future work is determining whether the present results are contingent on the study system explored by us, or whether it offers a first glimpse of a phenomenon more widespread in nature. The few studies available to date suggest the latter. At the same time, we emphasize that quantitative measures of food web structure can and should not be used as a single silver bullet for measuring the structure and function of ecological communities. While an intact interaction structure may hopefully represent a first proxy of unimpaired function (McCann 2007; Memmott 2009), the differential responses to fragmentation detected among the metrics, guilds and taxa examined here suggest that no metric will, on its own, suffice to characterise the response of a community to human impact. Only by assessing multiple different measures may we discern the full imprint of anthropogenic change, and only by appreciating the diversity in taxon-specific response will we be able to model this imprint for predictive purposes.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

We wish to thank our many hardworking field and laboratory assistants who made this study possible, especially Fiia Berghem, Katja Bonnevier, Janne Haaksluoto and Helena Rosenlew. Mark R. Shaw identified all Ichneumonoids, whereas Christer Hansson and Richard R. Askew confirmed the identification of voucher samples of chalcidoid parasitoids. Sofia Gripenberg, Owen Lewis, Rebecca Morris, Saskya van Nouhuys and two anonymous referees offered helpful comments on an earlier version of the manuscript. This study was funded by the Academy of Finland (grant number 129636 to the Centre of Excellence in Metapopulation Research 2009–2011, and grant numbers 126296 and 129142 to TR) and by the Entomological Society of Helsinki (through personal grants to RK in 2006 and 2007).

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information
  • Amarasekare, P. (2008) Spatial dynamics of food webs. Annual Review in Ecology, Evolution and Systematics, 39, 479500.
  • Ashworth, L., Aguilar, R., Galetto, L. & Aizen, M.A. (2004) Why do pollination generalist and specialist plant species show similar reproductive susceptibility to habitat fragmentation? Journal of Ecology, 92, 717719.
  • Askew, R.R. (1961) On the biology of inhabitants of oak galls of Cynipidae (Hymenoptera) in Britain. Transactions of the Society for British Entomology, 14, 237268.
  • Askew, R.R. (1980) The diversity of insect communities in leaf-mines and plant galls. Journal of Animal Ecology, 49, 817829.
  • Askew, R.R. & Shaw, M.R. (1974) An account of the Chalcidoidea (Hymenoptera) parasitizing leaf-mining insects of deciduous trees in Britain. Biological Journal of the Linnean Society, 6, 289335.
  • Bailey, R., Schönrogge, K., Cook, J., Melika, G., Csóka, G., Thuróczy, C. & Stone, G. (2009) Host niches and defensive extended phenotypes structure parasitoid wasp communities. PLOS Biology, 7, e1000179.
  • Berlow, E.L. (1999) Strong effects of weak interactions in ecological communities. Nature, 398, 330334.
  • Cagnolo, L., Valladares, G., Salvo, A., Cabido, M. & Zak, M. (2009) Habitat fragmentation and species loss across three interacting trophic levels: effects of life-history and food-web traits. Conservation Biology, 23, 11671175.
  • Crawley, M.J. (2007) The R Book. John Wiley and Sons, Ltd., Chichester, UK.
  • Davies, K.F., Margules, C.R. & Lawrence, J.F. (2000) Which traits of species predict population declines in experimental forest fragments? Ecology, 81, 14501461.
  • Diniz-Filho, J.A., Bini, L.M. & Hawkins, A. (2003) Spatial autocorrelation and red herrings in geographical ecology. Global Ecology and Biogeography, 12, 5364.
  • Doak, P. (2000) The effects of plant dispersion and prey density on parasitism rates in a naturally patchy habitat. Oecologia, 122, 556567.
  • Dormann, C.F., Gruber, B. & Fründ, J. (2008) Introducing the bipartite package: analysing ecological networks. R News, 8, 811.
  • Dormann, C.F., Fründ, J., Blüthgen, N. & Gruber, B. (2009) Indices, graphs and null models: analyzing bipartite ecological networks. Open Journal of Ecology, 2, 724.
  • Ekroos, J., Heliölä, J. & Kuussaari, M. (2010) Homogenization of lepidopteran communities in intensively cultivated agricultural landscapes. Journal of Applied Ecology, 47, 459467.
  • Fontaine, C., Dajoz, I., Meriguet, J. & Loreau, M. (2006) Functional diversity of plant-pollinator interaction webs enhances the persistence of plant communities. PLos Biology, 1, e1.
  • Fortuna, M.A. & Bascompte, J. (2006) Habitat loss and the structure of plant-animal mutualistic networks. Ecology Letters, 9, 278283.
  • Gripenberg, S. & Roslin, T. (2005) Host plants as islands: resource quality and spatial setting as determinants of insect distribution. Annales Zoologici Fennici, 42, 335345.
  • Gripenberg, S., Ovaskainen, O., Morriën, E. & Roslin, T. (2008) Spatial population structure of a specialist leaf-mining moth. Journal of Animal Ecology, 77, 757767.
  • Gross, T., Rudolf, L., Levin, S.A. & Dieckmann, U. (2009) Generalist models reveal stabilizing factors in food webs. Science, 325, 747750.
  • Hanski, I. (1999) Metapopulation Ecology. Oxford University Press, New York, NY, USA.
  • Hanski, I. (2005) The Shrinking World: Ecological Consequences of Habitat Loss. International Ecology Institute, Oldendorf, Germany.
  • Hanski, I. & Simberloff, D. (1997) The metapopulation approach, its history, conceptual domain, and application to conservation. Metapopulation Biology: Ecology, Genetics and Evolution (eds I.Hanski & M.E.Gilpin). pp. 526, Academic Press, San Diego, USA.
  • Hayward, A. & Stone, G.N. (2005) Oak gall wasp communities: evolution and ecology. Basic and Applied Ecology, 6, 435443.
  • Holt, R.D. (1996) Food webs in space: an island biogeographic perspective. Food Webs. Integration of Patterns and Dynamics (eds G.A.Polis & K.O.Winemiller). pp. 313323, Chapman and Hall, New York, NY, USA.
  • Holt, R.D. (1997) From metapopulation dynamics to community structure: some consequences of spatial heterogeneity. Metapopulation Biology: Ecology, Genetics and Evolution (eds I.Hanski & M.E.Gilpin). pp. 149164, Academic Press, San Diego, USA.
  • Holt, R.D., Lawton, J.H., Polis, G.A. & Martinez, N.D. (1999) Trophic rank and the species area relationship. Ecology, 80, 14951504.
  • Holyoak, M., Leibold, M.A. & Holt, R.D. (eds) (2005) Metacommunities. Spatial Dynamics and Ecological Communities. The University of Chicago Press, Chicago, USA.
  • Kaartinen, R., Stone, G.N., Hearn, J., Lohse, K. & Roslin, T. (2010) Revealing secret liaisons: DNA bar-coding changes our understanding of food webs. Ecological Entomology, 35, 623638.
  • Kruess, A. & Tscharntke, T. (1994) Habitat fragmentation, species loss and biological control. Science, 264, 15811584.
  • Kuussaari, M., Bommarco, R., Heikkinen, R.K., Helm, A., Krauss, J., Lindborg, R., Öckinger, E., Pärtel, M., Pino, J., Rodà, F., Stefanescu, C., Teder, T., Zobel, M. & Steffan-Dewenter, I. (2009) Extinction debt: a challenge for biodiversity conservation. Trends in Ecology and Evolution, 24, 564571.
  • Laliberté, E. & Tylianakis, J.M. (2010) Deforestation homogenizes tropical parasitoid–host networks. Ecology, 91, 17401747.
  • Laurence, W.F. (2008) Theory meets reality: how habitat fragmentation research has transcended island biogeography theory. Biological Conservation, 141, 17311744.
  • Legendre, P. (1993) Spatial autocorrelation: trouble or new paradigm? Ecology, 74, 16591673.
  • Littell, R.C., Milliken, G., Stroup, W.W. & Wolfinger, R.D. (2006) SAS System for Mixed Models. 2nd edn, SAS Institute Inc, Cary, NC, USA.
  • McCann, K. (2007) Protecting biostructure. Nature, 446, 29.
  • McCullagh, P. & Nelder, J.A. (1989) Generalized Linear Models. Chapman and Hall, London, UK.
  • Memmott, J. (2009) Food webs: a ladder for picking strawberries or a practical tool for practical problems? Philosophical Transactions of the Royal Society, 364, 16931699.
  • Mikkelson, G.M. (1993) How do food webs fall apart? A study of changes in trophic structure during relaxation on habitat fragments. Oikos, 67, 539547.
  • Moilanen, A. & Hanski, I. (2001) On the use of connectivity measures in spatial ecology. Oikos, 95, 147152.
  • Moilanen, A. & Nieminen, M. (2002) Simple connectivity measures in spatial ecology. Ecology, 84, 11311145.
  • van Nouhuys, S. (2005) Effects of habitat fragmentation at different trophic levels in insect communities. Annales Zoologici Fennici, 42, 433447.
  • Noyes, J.S. (2001) Interactive Catalogue of World Chalcidoidea 2001. Taxapad – Scientific names for Information management, Vancouver, Canada.
  • Öckinger, E., Schweiger, O., Crist, T.O., Debinski, D.M., Krauss, J., Kuussaari, M., Petersen, J.D., Pöyry, J., Settele, J., Summerville, K.S. & Bommarco, R. (2010) Life-history traits predict species responses to habitat area and isolation: a cross-continental synthesis. Ecology Letters, 13, 969979.
  • Olesen, J.M., Bascompte, J., Elberling, H. & Jordano, P. (2008) Temporal dynamics in pollination network. Ecology, 89, 15731582.
  • Pandit, S.N., Kolasa, J. & Cottenie, K. (2009) Contrasts between habitat generalists and specialists: an empirical extension to the basic metacommunity framework. Ecology, 90, 22532262.
  • Petanidou, T., Kallimanis, A.S., Tzanopoulos, J., Sgardelis, S.P. & Pantis, J.D. (2008) Long-term observation of a pollination network: fluctuations in species and interactions, relative invariance of network structure and implications for estimated of speciation. Ecology Letters, 11, 564575.
  • Petermann, J.S., Müller, C.B., Weigelt, A., Weisser, W.W. & Schmid, B. (2010) Effect of plant species loss on aphid-parasitoid communities. Journal of Animal Ecology, 79, 709720.
  • R Development Core Team (2009) R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria.
  • Ronquist, F. (1994) Evolution of parasitism among closely related species: phylogenetic relationships and the origin of inquilinism in gall wasps (Hymenoptera, Cynipidae). Evolution, 48, 241266.
  • Sabatino, M., Maceira, N. & Aizen, M.A. (2010) Direct effects of habitat area on interaction diversity in pollination webs. Ecological Applications, 20, 14911497.
  • Schönrogge, K. & Crawley, M.J. (2000) Quantitative webs as a means of assessing the impact of alien insects. Journal of Animal Ecology, 69, 841868.
  • Skult, H. (1956) Skogsbotaniska studier i skärgårdshavet med speciell hänsyn till förhållandena i Korpo utskär. Acta Botanica Fennica, 57, 1244.
  • Starzomski, B.M. & Srivastava, D.S. (2007) Landscape geometry determines community response to disturbance. Oikos, 116, 690699.
  • Steffan-Dewenter, I. (2003) Importance of habitat area and landscape context for species richness of bees and wasps in fragmented orchard meadows. Conservation Biology, 17, 10361044.
  • Stone, G.N. & Cook, J.M. (1998) The structure of cynipid oak galls: patterns in the evolution of an extended phenotype. Proceedings of the Royal Society of London series B, 265, 979988.
  • Stone, G.N., Atkinson, R.J., Rokas, A., Csóka, G. & Nieves-Aldrey, J.-L. (2001) Differential success in northwards range expansion between ecotypes of the marble gall wasp Andricus kollari: the tale of two lifecycles. Molecular Ecology, 10, 761778.
  • Tack, A., Ovaskainen, O., Pulkkinen, P. & Roslin, T. (2010) Spatial location dominates over host plant genotype in structuring an herbivore community. Ecology, 91, 26602672.
  • Thébault, E. & Fontaine, C. (2010) Stability of ecological communities and the architecture of mutualistic and trophic networks. Science, 329, 853856.
  • Tscharntke, T., Steffan-Dewenter, I., Kruess, A. & Thies, C. (2002) Characteristics of insect populations on habitat fragments: a mini review. Ecological Research, 17, 229239.
  • Tylianakis, J.M., Tscharntke, T. & Lewis, O.T. (2007) Habitat modification alters the structure of tropical host-parasitoid food webs. Nature, 445, 202205.
  • Valladares, G., Salvo, A. & Cagnolo, L. (2006) Habitat fragmentation effects on trophic processes of insect-plant food web. Conservation Biology, 20, 212217.

Supporting Information

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

Fig. S1. Patterns in binary (qualitative) measures of food web structure.

Fig. S2. Patterns in species richness vs. isolation.

Table S1. A repeated-measures model of species-specific response.

Appendix S1. Detailed description of the study system.

Appendix S2. Additional details on Materials and methods.

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials may be re-organized for online delivery, but are not copy-edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

FilenameFormatSizeDescription
JANE_1811_sm_AppendixS1.doc40KSupporting info item
JANE_1811_sm_AppendixS2.doc37KSupporting info item
JANE_1811_sm_FigS1.doc62KSupporting info item
JANE_1811_sm_FigS2.doc99KSupporting info item
JANE_1811_sm_TableS1.doc31KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.