Habitat fragmentation is a complex process that affects ecological systems in diverse ways, altering everything from population persistence to ecosystem function. Despite widespread recognition that habitat fragmentation can influence food web interactions, consensus on the factors underlying variation in the impacts of fragmentation across systems remains elusive. In this study, we conduct a systematic review and meta-analysis to quantify the effects of habitat fragmentation and spatial habitat structure on resource consumption in terrestrial arthropod food webs. Across 419 studies, we found a negative overall effect of fragmentation on resource consumption. Variation in effect size was extensive but predictable. Specifically, resource consumption was reduced on small, isolated habitat fragments, higher at patch edges, and neutral with respect to landscape-scale spatial variables. In general, resource consumption increased in fragmented settings for habitat generalist consumers but decreased for specialist consumers. Our study demonstrates widespread disruption of trophic interactions in fragmented habitats and describes variation among studies that is largely predictable based on the ecological traits of the interacting species. We highlight future prospects for understanding how changes in spatial habitat structure may influence trophic modules and food webs.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Global environmental change has altered patterns of biodiversity worldwide, and the loss and fragmentation of habitats are some of the leading drivers of such change (Vitousek 1994; Wilcove et al. 1998; Sala et al. 2000; Sih et al. 2000; Hanski 2005). Across multiple spatial scales, population dynamics, species richness and community composition vary greatly with habitat fragmentation and gradients in spatial habitat structure (Kareiva 1990; Saunders et al. 1991; Fahrig 2003; Ries et al. 2004; Tscharntke & Brandl 2004; Holyoak et al. 2005; Ewers & Didham 2006).
Alterations to spatial habitat structure also change the intensity of ecological interactions. In fact, this effect may be even more extreme than changes to single species abundances, as species interactions are affected by processes acting on more than one species (Holt 1996; Tylianakis et al. 2008; Dyer et al. 2010). For example, vertebrate and invertebrate pollination of plants is disrupted in more fragmented landscapes, leading to declines in seed production, viability and fruit set with increasing distance from natural sources of pollinators (Kremen et al. 2004; Aguilar et al. 2006; Ricketts et al. 2008; Kennedy et al. 2013). The loss of such mutualistic interactions may be more likely among species with specialised interactions and may even precede species extirpations on habitat patches (Aizen et al. 2012). Similarly, land use change can alter networks of antagonistic interactions, leading to greatly altered food web structure in intensively managed habitats (Tylianakis et al. 2007).
Overall, the literature on spatially mediated disruptions to trophic interactions is vast, but it has yet to be brought together and examined for commonalities in mechanism or outcome. Here, we employ quantitative meta-analysis to evaluate how variation in spatial habitat structure influences resource consumption in terrestrial arthropod food webs. Food webs are a common example of systems where habitat degradation acts on multiple species simultaneously, and efforts are underway to understand how quantitative food webs are altered in fragmented and patchy habitats (Amaresekare 2008; Bascompte 2009; Kaartinen & Roslin 2011). Thus far, a number of different mechanisms have been uncovered. For example, disruption to trophic interactions in arthropod food webs can uncouple specialist herbivores and their host plants, leading to a loss of herbivore populations and a reduction in the amount of leaf material consumed in small or isolated patches (Root 1973; Connor et al. 2000; Kéry et al. 2001; Colling & Matthies 2004). Similarly, fragmentation can uncouple herbivores and their natural enemies, reducing levels of predation and parasitism and leading to pest outbreaks in both natural and agricultural habitats (Kareiva 1987; Kruess & Tscharntke 1994; Roland & Taylor 1997; Anton et al. 2007). Ultimately, the consequences of altered spatial structure and its disruptive effect on species' trophic interactions can manifest in altered ecosystem structure and function (Didham et al. 1999; Terborgh et al. 2001; Tscharntke & Brandl 2004). Here, we convert measures of relative resource consumption in fragmented and continuous settings into standardised effect sizes to assess the consequences of trophic disruption in fragmented habitats.
Habitat fragmentation is the process by which habitat loss leads to the division of continuous habitat into smaller, isolated patches that are separated by dissimilar matrix habitats (Didham 2010). The set of spatial variables commonly altered in fragmented or naturally patchy landscapes includes the proportion of habitat in the landscape, the spatial contiguity of remaining habitat, altered matrix contrast or composition and spatial factors that vary at the patch scale (e.g. patch size, distance to patch edges, patch isolation, connectivity). For simplicity, we refer to these factors collectively as spatial variables. These spatial variables can have complex and interdependent effects on ecological communities (Didham 2010; Didham et al. 2012), and the (relative) importance of these variables may depend on the type of response variable studied. For example, Prugh et al. (2008) found that patch size and isolation were relatively poor predictors of patch occupancy, but Connor et al. (2000) found clear effects of patch size on animal density.
Several hypotheses pertain to how spatial habitat structure might alter resource consumption. For example, the trophic rank hypothesis of island biogeography predicts that, because consumers depend on their resources to persist, higher trophic levels should be more sensitive to altered spatial structure than lower trophic levels (Holt 1996; Holyoak et al. 2005; Holt 2010). Thus, we would expect that levels of resource consumption should decline in fragmented compared to more continuous habitats, especially for interactions involving species at higher trophic levels (i.e. parasitism of herbivores).
However, we expect that factors outside the trophic rank hypothesis will be important as well. The intensity of some trophic interactions, such as nest predation of forest birds, increases with fragmentation or proximity to a patch edge (Chalfoun et al. 2002; Batáry & Báldi 2004). In such settings, the habitat and feeding requirements of the interacting species determine the outcome of spatial effects (Connor et al. 2000; Rand et al. 2006; Ryall & Fahrig 2006). Consequently, a second prediction is that habitat fragmentation should increase levels of resource consumption for consumers with broad niche breadths, either in terms of diet or habitat affinity. In contrast, resource consumption of consumers with narrow niche breadths should decline with increasing habitat fragmentation.
We also expect that the disruption to resource consumption in fragmented habitats may depend on other characteristics of the component studies, such as habitat, location, and the particular species studied. For example, Sala et al. (2000) found that land use change was one of the most important drivers of global environmental change, but its impact and relative importance varied between tropical and temperate forests, grasslands and other major habitat types. Finally, the phylogenetic relationships among interacting species may influence their joint response to habitat modification. In particular, phylogeny may govern body size (Roland & Taylor 1997), movement and searching behaviour (Hambäck & Englund 2005; Hambäck et al. 2007), and other key life history traits predicted to influence species' susceptibilities to spatial habitat structure (Ewers & Didham 2006; Öckinger et al. 2010).
In compiling the literature on how variation in spatial habitat structure influences resource consumption in terrestrial arthropod food webs, we draw upon a data set of 419 studies, representing diverse taxa, trophic interactions and habitat types. Across such variation, we uncover strong and robust signals of the disruption of trophic interactions in fragmented systems via formal meta-analysis; these signals depend strongly on study characteristics and species traits. We point to the most sensitive types of consumptive interactions, and discuss future prospects for understanding how complex dynamical systems such as food webs can be understood as habitat fragmentation and other changes in spatial structure accelerate.
Materials and methods
Literature search and compilation of data set
To identify relevant papers, we searched Web of Science for papers published between January 1945 and May 2012, using combinations of key words related to spatial habitat structure (patch AND size, isolation, connectivity, arrangement; edge effect; habitat fragmentation; shape complexity; landscape AND matrix, fragmentation, connectivity) in combination with those related to trophic interactions (food web, trophic interaction, detritivory, herbivory, predation, parasitism, disease). To this extensive list, we added papers with which we were already familiar and those citing recent reviews of habitat fragmentation.
For a publication to be included, it had to empirically evaluate how habitat fragmentation or natural spatial structure affected levels of resource consumption. Response variables included the proportion of plant material consumed, herbivores parasitised or detritus processed. Studies were included if resource consumption was measured (1) as a function of the proportion of natural habitat in a landscape; (2) in continuous habitats vs. habitat fragments (3) in contiguous vs. fragmented landscapes (fragmentation per se; sensu Fahrig 2003); (4) with respect to other measures of landscape fragmentation (e.g. amount of edge, number of patches, average connectivity); (5) as a function of patch size or degree of connectivity or isolation or (6) at the centres vs. edges of patches. To make the broadest inference possible, we include each of these spatial variables as measures of habitat fragmentation in the broad sense (Lajeunesse & Forbes 2003; Didham 2010; see also van Nouhuys 2005; Aguilar et al. 2006; Tylianakis et al. 2008). Studies were limited to those in which at least one trophic level involved terrestrial arthropods.
For each study, we recorded details of study location and design, the type of spatial variable and trophic interaction, and details about the interacting species. To compare the sensitivity of resource consumption to fragmentation across habitats or biomes, we recorded the location of each study (locality, country, latitude, longitude) and the biogeographical realm (Olson et al. 2001). We also recorded whether each study was conducted on an island or a mainland, whether it used experimental or observational methods, whether the cause of the spatial variation was natural (e.g. naturally patchy host plants), anthropogenic (e.g. due to land use change) or experimental in nature, and how long the habitat had been fragmented. Finally, we recorded the type of habitat in which the trophic interaction was evaluated, whether the habitat was composed of a single species or many, and the composition of the matrix, when possible. We categorised studies comparing large to small patches as well as those comparing continuous habitat to habitat fragments jointly as studies of patch size.
Pursuant to our focus on terrestrial arthropod systems, we categorised the studied trophic interactions into five broad interaction types (detritivory, herbivory, parasite–host, parasitoid–host and predator–prey), and recorded details of the specific measures used to quantify resource consumption. To test hypotheses about sources of variation among studies, we also recorded characteristics of the species involved in the trophic interactions. For each consumptive interaction, we defined a resource taxon and a consumer taxon and recorded trophic rank, diet breadth, habitat affinity, guild, taxonomic family and order, and taxonomic resolution (single species or entire assemblages). Trophic rank was assigned as follows: herbivores, pollinators and detritivores were considered primary consumers; parasitoids, predators and parasites of herbivores, as well as predators of unspecified trophic position were considered secondary consumers; hyperparasitoids and predators or parasites of secondary consumers were considered tertiary consumers. Omnivorous taxa in our data set were relatively rare (n = 11) and were omitted in analyses of trophic rank. Diet breadth for phytophagous insects was often given in the original publication as monophagous (feeding on one genus of plants), oligophagous (feeding on a limited number of species or genera) or polyphagous (feeding across several plant families). For predators and parasitoids, prey consumption or host use was considered specific to one or a few species, or generalist on many species. To achieve common currency across trophic levels, we considered species that were monophagous, oligophagous or specific to have a ‘narrow’ diet breadth, and species that were polyphagous, generalist or omnivorous to have ‘broad’ diet breadth. We recorded habitat affinity as ‘specific’ for species confined largely to the focal patch habitat and ‘general’ for species known to utilise both the focal patch habitat as well as the matrix surrounding the patch, according to the original publication or through literature searches. These two measures of niche breadth were highly related (χ2 = 90.14; d.f. = 1; P < 0.0001). We used habitat affinity as our measure of niche breadth for the analyses, as it was more frequently reported in the original studies. Guild was recorded as plant, herbivore, parasitoid, parasite, predator or detritivore; where available, details on subguild, such as galling herbivore or external parasite, were also recorded.
Finally, we recorded taxonomic family and order for each interacting species for two purposes: (1) to characterise whether trophic interactions for particular taxa are more sensitive to gradients in spatial structure, and (2) because many life history and behavioural traits are correlated with phylogeny. For example, aphids and butterflies use different mechanisms to locate host plant patches, and this can lead to contrasting density–area relationships for the two groups (Hambäck & Englund 2005).
We considered each combination of spatial variable and trophic interaction as a separate study in our data set. Most publications investigated the effects of multiple spatial variables on several consumer–resource pairs, and we had no a priori reasons to include particular spatial variables or consumptive interactions and exclude others. We also expected that trophic interactions within a publication may respond differently to the same spatial variable, as outlined in our hypotheses. We found that variation in effect size within publications was nearly as large as variation among publications (see Appendix S1). For example, the range of effect sizes (ES) within publications of patch size effects encompassed up to 78% of the range of ES among publications (for publications reporting > 3 ES: largest range in ES = 3.59, mean range in ES = 1.76, n = 10; for publications reporting 1 ES: range in ES = 4.61, n = 54). We therefore considered effect sizes to be independent observations for the meta-analysis (Gurevitch & Hedge 1999; see also Aguilar et al. 2006; Chaplin-Kramer et al. 2011; Magrach et al. 2014).
Calculation of effect sizes
We quantified the effects of variation in spatial habitat structure on resource consumption using the log response ratio, LRR = ln (Xe/Xc), where Xe is the mean response under the experimental condition and Xc is the mean response under the control condition (Hedges et al. 1999). For each spatial variable, we defined the control condition as the more spatially continuous condition (e.g. large patches, continuous landscapes, interior of patch) and the experimental as the more spatially disrupted condition (e.g. small patches, fragmented landscapes, edge of patch). Thus, negative effect sizes indicate that fewer resources were consumed in more fragmented settings, and positive effect sizes indicate that more resources were consumed. Where means were not reported, we digitised points from graphs and calculated means for the highest and lowest values of the spatial variable examined. When multiple comparisons were possible, we compared the most extreme treatment effects.
Effect sizes are unitless measures used to quantify the strength of the relationship between two variables (Arnqvist & Wooster 1995; Borenstein et al. 2009). The LRR is an appropriate effect size for this study because it represents the proportional change in resource consumption with fragmentation, regardless of the original unit of measurement (Hedges et al. 1999). This effect size can be calculated from studies reporting the mean responses alone, even if sample sizes and measures of variability are not reported (Lajeunesse & Forbes 2003). Furthermore, LRR is statistically well-distributed and weights deviations in the numerator and denominator equally (Hedges et al. 1999; Lajeunesse & Forbes 2003).
We also evaluated several other measures of effect size for our analysis. The relative interaction intensity effect size of Armas et al. (2004) and the difference ratio of Sorte et al. (2013) were both highly correlated with LRR (for both comparisons, r = 0.98; t417 = 95.71; P < 0.001) and were perfectly correlated with each other, as should be expected because these effect sizes are all mathematically related (Armas et al. 2004). Overall effect sizes and patterns of variation among potential explanatory variables were qualitatively similar regardless of the effect size used (data not shown), and we present all results using LRR as the effect size.
When possible, we also recorded the values for the control and experimental spatial variable, e.g. the size of the largest and smallest categories of patch size. We calculated the spatial extent of the study as the difference between ValC, the median value of the control spatial variable, and ValE, the median value of the experimental, or more fragmented, spatial variable. We categorised studies with spatial extents greater than or equal to the median value as having a ‘large extent’ and those less than the median as having a ‘small extent.’ Categorising studies based only on the maximum spatial extent (e.g. the largest patch size) typically provided the same designations of large vs. small spatial extent.
Full data set
All analyses were conducted in R (version 3.0.3, R Core Team 2014). Using the full data set of all studies from which an effect size could be obtained, we first calculated the mean and 95% bootstrapped confidence intervals (9999 permutations; boot package, Canty & Ripley 2014) for the overall effect of spatial habitat structure on resource consumption. Furthermore, we calculated the average effect sizes and confidence intervals for the following categorical variables: method, scale, tropical or temperate location, source of fragmentation, focal habitat type (crop, forest, grassland, laboratory study, open habitat, shrubland, urban or wetland), biogeographical realm, consumer taxonomic family and order, spatial variable, trophic interaction type, trophic rank and niche breadth. We considered effect sizes to be significant when 95% bootstrapped CI did not overlap zero. We examined the single effects of each of these variables with simple linear models.
We next wished to test hypotheses about how combinations of key explanatory variables might interact to influence effect size. For example, we hypothesised that the effect of spatial variable might depend on consumer habitat affinity or trophic interaction type. To understand which factors were most important to the overall patterns of variation among studies, we constructed a linear mixed effects model, specifying the fixed effects of spatial variable, type of trophic interaction, habitat affinity and two-way interactions among those variables, with a random effect of publication (lme function in nlme package, Pinheiro et al. 2014). To ensure adequate replication among treatments, we restricted this analysis to studies of matrix composition, patch size, connectivity, and edge effects and excluded studies of detritivory. We simplified this full model using the function stepAIC (in the MASS package, Venables & Ripley 2002), which retains the most important variables through iterations of forward and backward stepwise selection and identifies the best model as that with the lowest value of the Akaike Information Criterion (AIC; see also Prugh et al. 2008). We report significance of model terms based on likelihood ratio tests.
Well-studied spatial variables
We next investigated sources of variation in effect size separately for each of several well-studied spatial variables (proportion habitat, matrix, patch size, connectivity, edge; see Table 1). For each of these spatial variables, we constructed a candidate full linear mixed effects model with a set of fixed effects hypothesised to influence effect size and a random effect of publication. We evaluated model terms across separate subsets of the data for each well-studied spatial variable to ensure adequate replication across factor levels. Each data set was restricted to studies for which details of consumer habitat affinity were known and for which spatial extent (if included in the candidate full model) could be calculated. Studies of detritivory and parasite–host interactions could not be included in these analyses due to small sample sizes. We then performed model simplification based on AIC values (again using the stepAIC function, Venables & Ripley 2002) to derive a reduced model with only the most important terms retained.
Table 1. Number of studies included in the meta-analysis. For each spatial variable (columns), the categories of more and less spatially continuous conditions are given in parentheses
Proportion natural habitat (high vs. low proportion)
Includes studies comparing continuous habitat to habitat fragments.
The full data set is available in Appendix S1.
We constructed a data set of studies investigating how the proportion of natural habitat in the landscape affected levels of resource consumption. The full model included only the simple and interactive effects of spatial extent and consumer habitat affinity due to a limited sample size. Studies of predator–prey interactions were poorly represented in this data set and were excluded from analysis.
Matrix studies compared resource consumption in a focal habitat adjacent to contrasting matrix types (low contrast or more natural vs. high contrast or more disturbed). Because of a small sample size for matrix studies, we evaluated a full model that included only the fixed effects of type of trophic interaction, consumer habitat affinity and habitat (crop or natural habitats).
The full model for studies of patch size included the fixed effects of trophic interaction type, consumer habitat affinity, spatial extent and select two-way interaction terms (type of trophic interaction x habitat affinity, type of trophic interaction x spatial extent). We included studies conducted in natural habitats that compared resource consumption in large vs. small patches and continuous habitat vs. habitat fragments.
The full model for studies of patch connectivity included the fixed effects of trophic interaction type, consumer habitat affinity, spatial extent and all two-way interaction terms. The connectivity data set included studies reporting the extent of connectivity or isolation in terms of distance (converted to m when possible) so that spatial extent was comparable among studies. The data set was restricted to studies of herbivory and parasitoid–host interactions, as few studies of predator–prey interactions were represented in the data set.
Finally, to understand the factors most important to variation in effect size for studies comparing patch edges to interiors, we constructed a full model with the following fixed effects: type of trophic interaction, consumer habitat affinity, spatial extent and habitat (crop or natural), and two-way interactions among these variables. For measures of spatial extent to be comparable across edge studies, we restricted this data set to studies reporting the distance to the edge of the patch (m) for both edge and interior sites.
Full data set
We identified 169 publications containing a total of 419 unique studies of how resource consumption varied with spatial habitat structure (Table 1; see Appendices S1 and S2). Of the spatial variables investigated, patch size, connectivity or isolation and distance to edge were the most frequently studied in our data set; of the trophic interactions studied, herbivory, parasitoid–host and predator–prey interactions were the most frequently studied (Table 1). Studies were conducted in 33 countries from six continents (Appendix S1).
The overall effect of variation in spatial habitat structure on resource consumption was negative (−0.1786; n = 419), and 95% bootstrapped confidence intervals did not overlap zero [(−0.2749, −0.0807); Fig. 1a]. Resource consumption was reduced to an average of 83.6% in more fragmented compared to more continuous habitats (back-transformed LRR). Both experimental and observational studies exhibited negative effect sizes (Fig. 1b) and did not differ significantly in magnitude of effect (F1,417 = 0.33; P = 0.57). Effect sizes did not differ based on spatial scale (landscape vs. patch scale, F1,417 = 1.16; P = 0.28; Fig. 1c). Temperate and tropical studies did not differ significantly in effect size (F1,417 = 0.02; P = 0.88; Fig. 1d), nor did studies based on the source of fragmentation (F2,406 = 0.32; P = 0.72; Fig. 1e). Effect size also did not differ significantly among major habitat types (F7,411 = 1.77; P = 0.09; Fig. 1f), although grassland habitats reported overall negative effect sizes. Effect size did not differ significantly with biogeographical realm (F6,412 = 1.86; P = 0.09), or consumer taxonomic family (F71,189 = 1.16; P = 0.21) or order (F19,292 = 0.94; P = 0.53); degrees of freedom for these tests varied slightly due to inconsistent reporting of study details.
Effect size depended strongly on the type of spatial variable investigated (F6,412 = 4.91; P < 0.0001; Fig. 1g), with significant effects for patch-level variables but inconsistent effects for landscape-level variables. Specifically, resource consumption was depressed on small compared to large patches and on isolated compared to connected patches (i.e. negative effect sizes). In fact, average resource consumption on small patches was only 70.5% of that reported from large patches, and resource consumption on isolated patches was only 71.2% of that observed on well-connected patches (back-transformed LRR). On the other hand, the effect of patch edges was significantly positive, leading to 30.4% greater resource consumption at patch edges compared to interiors (back-transformed LRR). Among types of trophic interactions, parasitoid–host interactions were negatively affected by fragmentation, but the confidence intervals for other interaction types overlapped zero, and the simple effect of interaction type was not significant (F4,414 = 0.97; P = 0.42; Fig. 1h).
Variation in effect size with trophic rank was in the predicted direction: tertiary consumers had the lowest effect sizes and primary consumers the highest (Fig. 1i). Confidence intervals for tertiary consumers, however, overlapped zero due to high variability in life form (parasites of vertebrates, hyperparasitoids of insects). Therefore, only secondary consumers exhibited significantly negative effect sizes, and the simple effect of trophic rank was not significant (F4,412 = 0.51; P = 0.73). Finally, effect size depended on the niche breadth of consumers (F2,416 = 2.94; P = 0.054; Fig. 1j), with more negative effects (i.e. greater reductions in consumption) for habitat specialists than for habitat generalists.
Across the full data set, the effect of spatial variable on effect size depended on the niche breadth of the consumer (likelihood ratio test, LRT, for the two-way interaction: χ2 = 9.42; P = 0.02; n = 316; Fig. 2). Specifically, habitat generalists exhibited more positive effect sizes than specialists in studies of patch size, connectivity and edge effects. In studies of matrix composition, however, effect sizes were lower for habitat generalists than for specialists, leading to the significant interaction between spatial variable and habitat affinity (Fig. 2). On the other hand, we found no evidence for the importance of two-way interactions between spatial variable and type of trophic interaction (dropping two-way interaction from full model, χ2 = 6.61; P = 0.68), nor between type of trophic interaction and niche breadth (χ2 = 0.27; P = 0.96).
Well-studied spatial variables
Nearly all studies of the proportion of natural or semi-natural habitat in the landscape included in our analyses were conducted in crop habitats (19 of 23). The most important factors underlying the variation in effect size among these studies were the simple and interactive effects of spatial extent and trophic interaction type (LRT of interaction term, χ2 = 5.67; P = 0.017; Fig. 3). Overall, effect sizes for studies of parasitoid–host interactions were negative [-0.9876 (−1.2971, −0.6787), n = 14] and unrelated to spatial extent. On the other hand, the overall effect size for herbivory was neutral [0.7103 (−0.3218, 1.5652), n = 9] and varied significantly with spatial extent. That is, herbivore damage was higher in landscapes with little natural habitat (i.e. positive effect sizes) when the difference between the most and least natural landscape was greatest (i.e. large extents; Fig. 3).
The most important factors underlying variation in effect sizes for matrix studies were the simple effects of habitat affinity (χ2 = 12.67; P < 0.001) and type of trophic interaction (χ2 = 9.10; P = 0.011; Fig. 4a). Notably, resource consumption for habitat generalists was lower when the adjacent matrix was high contrast or more disturbed [i.e. a negative matrix effect: −0.8199 (−1.4301, −0.0528), n = 16]. The effect of matrix type had little effect on resource consumption among habitat specialists [0.1512 (−0.5046, 0.7012), n = 7]. Among types of trophic interactions, increased matrix contrast had a generally negative effect on parasitism and predation but little effect on the levels of herbivory (Fig. 4a).
Variation in effect size among patch size studies was a function of the simple effects of spatial extent, consumer habitat affinity and type of trophic interaction. Effect sizes were lower for studies encompassing larger spatial extents (χ2 = 5.33; P = 0.02; Fig. 5a) and among specialist consumers (χ2 = 4.03; P = 0.045; Fig. 5b). Effect sizes also varied among types of trophic interactions, with reductions to herbivory and parasitism but overall neutral effects on predation (χ2 = 6.80; P = 0.033; Fig. 5c).
The most important factors contributing to variation in effect size among studies of connectivity were the simple and interactive effects of consumer niche breadth and type of trophic interaction (LRT of interaction term, χ2 = 2.71; P = 0.10; Fig. 4b). Although parasitoid–host interactions were consistently disrupted with habitat isolation [−0.4125 (−0.6874, −0.1458), n = 27], the extent of herbivory depended on consumer niche breadth. Specifically, patch isolation led to moderate decreases in herbivory for habitat specialist consumers and moderate increases for habitat generalists (Fig. 4b).
The most important factor underlying variation in effect size among studies of habitat edges was the habitat affinity of the consumer (LRT for habitat affinity, χ2 = 6.80; P = 0.009; n = 71; Fig. 4c). Habitat generalists consumed more resources at habitat edges [0.4757 (0.2289, 0.7288), n = 44]. Resource consumption for habitat specialists, on the other hand, did not differ between the interior and edge of a patch [−0.1869 (−0.5136, 0.1679), n = 27].
The differential effects of habitat fragmentation among species are well-recognised. Traits such as rarity, niche specificity, high trophic position and population variability generally contribute to an increased extinction risk in fragmented habitats (Bender et al. 1998; Connor et al. 2000; Davies et al. 2004; Henle et al. 2004). Such trait-based analyses have allowed the characterisation and conservation ranking of the most sensitive species to habitat loss and have established that heterogeneity in sensitivity across species is largely predictable (Fagan et al. 2001; Ewers & Didham 2006; Pereira & Daily 2006). In this study, we demonstrate that habitat fragmentation also strongly influences the outcome of consumer–resource interactions in a quantifiable and predictable manner across a wide range of terrestrial arthropod food webs. Overall, the ecological traits of the interacting species and the nature of the spatial variables jointly influenced the magnitude of the effect of habitat fragmentation on resource consumption.
Theory predicts that species at higher trophic levels should be more sensitive to habitat loss and fragmentation than those lower in the food chain due to a necessary dependence on the presence of resources on a patch (Holt 1996, 2002), and empirical data from many systems support this species-level trophic rank hypothesis (Kruess & Tscharntke 2000; Post et al. 2000; Terborgh et al. 2001; van Nouhuys 2005). Here, we find that the trophic interactions in which these species engage are also strongly disrupted by habitat loss and fragmentation. On average, resource consumption in arthropod food webs was lower in more fragmented settings (Fig. 1a). The trophic rank of consumers in our study, however, was not a significant predictor of effect size. Although effect size tended to decrease with trophic rank and the mean effect size for parasitoid–host interactions was significantly negative, trophic rank itself did not significantly contribute to variation among studies. Instead, relative resource consumption depended on the simple and interactive effects of consumer habitat affinity, the types of spatial variables and trophic interactions studied, and the spatial extent of the studies.
The effects of habitat fragmentation depend on consumer niche breadth
Across the full data set and for studies of matrix composition, patch size, connectivity, and edges in particular, the magnitude of the effect of fragmentation on resource consumption differed between habitat generalists and specialists. Specifically, resource consumption was elevated in fragmented habitats for consumers with a broad niche breadth (utilising resources both inside a patch and in the matrix surrounding the patch), especially in studies of edge effects (Figs 2 & 4c). Rand et al. (2006) hypothesised that generalist natural enemies moving out of farm fields would have greater impacts in adjacent habitats than would specialists (see also Rand & Tscharntke 2007) and incursions of consumers from the matrix into the patch can generally influence resource consumption in a patch (Fagan et al. 1999; Cantrell et al. 2001; Denno et al. 2005; Rand et al. 2006; Martinson et al. 2012). We found overall higher levels of resource consumption at edges, and spillover of generalists consumers may contribute to these patterns in effect size. Notably, although parasitism was reduced across studies (Fig. 1h), parasitism by habitat generalists was higher at edges than interior sites (Fig. 4c).
In contrast to the findings for edge studies, the effects of matrix type were strongly negative for habitat generalists, but neutral for habitat specialists (Figs 2 & 4a). Matrix studies compared resource consumption in a patch that was surrounded by either a low contrast or more natural matrix (e.g. forest, natural vegetation, flowering border) or a high contrast, more disturbed matrix (e.g. water, bare ground, weeds). A high contrast or highly-disturbed matrix may provide few complementary resources for a habitat generalist, reducing densities, disrupting trophic interactions in the focal patch habitat and contributing to reductions in resource consumption. For example, Thomson & Hoffmann (2010) reported lower rates of egg predation by generalist predators in vineyards surrounded by bare ground compared to those surrounded by remnant vegetation. On the other hand, consumers deriving resources from only the patch habitat may be relatively less affected by the resources available in the matrix (Cook et al. 2002; Kupfer et al. 2006). Indeed, seed predation by monophagous Bruchophagus wasps differed little in experimental patches of clover surrounded by a grass matrix compared to those surrounded by bare ground (Diekötter et al. 2007). Together, these results provide strong support for the hypothesis that resource consumption in fragmented habitats depends on consumer niche breadth and access to complementary resources.
Divergent effects of fragmentation at the landscape level
With increasing land use intensification, the maintenance of key ecological services, such as biological control of crop pests, is a major goal (Cronin & Reeve 2005; Tscharntke et al. 2007; Chaplin-Kramer et al. 2011; Veres et al. 2013). Increased parasitism at edges suggests that these goals are being met to a certain extent, but when parasitoids are the biocontrol agents, such services may extend only partway into fields and may not be maintainable in intensively used landscapes. Indeed, parasitism in landscapes with little natural habitat was only 40.9% of that observed in more natural landscapes. Herbivory, on the other hand, increased in landscapes with little natural habitat when reported over large spatial extents (Fig. 3). Although the number of studies reporting resource consumption at the landscape scale was quite low, our findings are consistent with the hypothesis that natural enemies are more sensitive to habitat fragmentation than are herbivores (Tscharntke & Brandl 2004; Cronin & Reeve 2005; Ewers & Didham 2006). Similarly, Veres et al. (2013) found evidence that pest suppression was lower in landscapes with less natural habitat.
Divergent responses to habitat fragmentation also can be caused by differences in the scales at which species perceive and move in the landscape (Levin 1992; Holt 1996; Tscharntke & Brandl 2004). In a review of insect responses to landscape complexity, Chaplin-Kramer et al. (2011) found that specialist natural enemies responded to the landscape at smaller spatial scales than did generalists. Different mechanisms, ranging from local foraging behaviour at the smallest scales to migration at the largest, lead to scale-dependent responses of consumers to habitat fragmentation (van Nouhuys 2005). Here, we were not able to test for effects of spatial scale directly, but several lines of evidence suggest that the effects of fragmentation operate across a wide range of spatial scales and that certain types of consumptive interactions are more sensitive than others to spatial scale. First, we found disruptions to resource consumption at small spatial scales as a function of patch edges and the type of adjacent habitat, as well as at larger spatial scales as a function of patch and landscape characteristics. Second, among patch size studies, resource consumption depended on the spatial extent of the study, with more negative effects reported from studies with greater spatial extents (Fig. 5). Third, among landscape studies, the responses of herbivory and parasitism depended on the spatial extent of the study (Fig. 3). Together, these results suggest that the ecological consequences of habitat fragmentation depend at least in some cases on the spatial extent of studies and suggests that larger spatial extents should be employed when possible.
We found clear patterns for how patch size, connectivity and edges affected overall levels of resource consumption but no overall effect of the proportion of natural or semi-natural habitat in a landscape (Fig. 1g). Scale-dependent responses to habitat fragmentation both within (Chaplin-Kramer et al. 2011) and among (Cronin & Reeve 2005) trophic levels will likely preclude a simple conclusion about how landscape change affects overall levels of resource consumption. In addition, the spatial variables we investigated are interrelated, and the extent to which they act independently or interdependently is not known and may vary among study systems (Didham 2010; Didham et al. 2012). Thus, unmeasured variables at the landscape scale may drive, or at least contribute to, the patch-level effects on resource consumption. Although beyond the scope of our meta-analysis, the hierarchical structural equation modelling approach advocated by Didham et al. (2012) may provide a route forward for understanding how spatial variables may act together or independently to influence ecological responses to fragmentation.
Resource consumption is reduced on isolated and small patches
Several other mechanisms may contribute to the disruption of trophic interactions in fragmented habitats, including altered animal movement (Fagan et al. 1999; Wirth et al. 2008). Limitations to consumer movement among patches may have led to the overall negative effect of patch isolation on resource consumption (Fig. 1g). Specifically, isolated patches were characterised by a reduction in both herbivory and parasitism (Fig. 4b). On the other hand, fragmentation can alter movement behaviours in other ways. For example, Elzinga et al. (2005) found higher rates of seed predation in small populations of Silene latifolia in both experimental and observational studies. This increase in seed predation was attributed to higher rates of oviposition on edges and in small populations for this highly mobile herbivore, which may spend more time ovipositing per plant in small patches (Elzinga et al. 2005). Consumer or resource phylogenetic relationships were not important variables in our models, and details such as dispersal behaviours, or morphological correlates such as body size or wing loading, were seldom reported. Consequently, we cannot yet say how frequently behavioural changes might have contributed to the fragmentation-mediated disruption of consumptive interactions.
Finally, one of the most striking results of the analysis of well-studied spatial variables was that plants in small patches experienced on average only 51.8% of the amount of herbivory of plants in large patches (back-transformed LRR). This reduction to herbivory in small patches is unlikely to be a result of generally increased natural enemy impacts. Indeed, parasitism in small patches was similarly reduced to only 45.6% of the levels recorded in large patches, and we found no evidence for an overall increase in predation on small compared to large patches (Fig. 5). Instead, reduced herbivore abundances and reductions to herbivory may consequently limit the abundance and impact of parasitoids on small patches. The size of habitat patches may therefore quite generally limit food chain length and the types of food web modules that will persist in fragmented habitats (Komonen et al. 2000; Melián & Bascompte 2002; Martinson et al. 2012).
Despite the measurable reductions in herbivory on small patches, the consequences for plant fitness and population persistence in habitat fragments will depend on the balance between herbivory, other negative interactions such as disease and competition, and positive interactions such as pollination and seed dispersal. For example, in small populations of Scorzonera humils, insect seed predation and fungal disease were reduced, but seed set was not consistently related to plant population size due to the concurrent reduction in the efficiency of pollination on small patches (Colling & Matthies 2004; see also Steffan-Dewenter et al. 2001). Recently, Magrach et al. (2014) found evidence that mutualistic interactions may be even more sensitive to fragmentation than antagonistic interactions. Because herbivory (Fig. 5c), pollination (Kremen et al. 2004; Aguilar et al. 2006; Kennedy et al. 2013), and seed dispersal (Magrach et al. 2014) are frequently disrupted in fragmented habitats, the existing data make it clear that the net outcome of changes in species interactions will be context- and species-dependent.
Meta-analysis allows us to critically evaluate what is known about a research question and what remains to be investigated. In this effort, we have explored results across diverse studies, and demonstrated that the magnitude of the effect of fragmentation on resource consumption is quantifiable, predictable, and in general, negative. Resource consumption and related properties of ecological communities may be even more sensitive to spatial habitat structure than are (perhaps easier to assess) measures such as occupancy, density or species richness (see Albrecht et al. 2007; Tylianakis et al. 2007; Prugh et al. 2008; Aizen et al. 2012; Magrach et al. 2014). Our meta-analysis has highlighted some persistent questions in the literature pertaining to the effects of fragmentation on terrestrial arthropod communities. Future studies, spanning multiple consumer–resource interactions and characterising responses across multiple levels of biological organisation are necessary to fill in the remaining blanks.
First, how might disruptions to trophic structure feed back to influence other ecological response variables? For example, spatial effects on the abundance of generalist predators may increase predation at edges or in small patches, leading to positive density–area relationships (Wimp et al. 2011; Martinson et al. 2012). If occurring over a prolonged period, changes in resource consumption may actually be one key mechanism that leads to changes in species distributions and occupancy patterns in fragmented habitats.
Second, when and where is locally depressed resource consumption tied to spatially dependent trophic cascades? Our analyses have shown that habitat fragmentation routinely alters the extent of resource consumption in herbivore–plant and parasitoid–host interactions at the patch level. However, few studies report results for resource consumption across multiple trophic levels, limiting our ability to test for spatially dependent trophic cascades (see also Chaplin-Kramer et al. 2011).
Third, how do the network properties of food webs mediate the impacts of habitat fragmentation on consumption? Studies show network properties vary with land use change (Albrecht et al. 2007; Tylianakis et al. 2007; but see Kaartinen & Roslin 2011). Might network traits such as interaction frequency or degree of generalisation influence the sensitivity of particular consumer–resource interactions to habitat fragmentation, in much the same way as fragmentation can differentially affect mutualistic interactions (Aizen et al. 2012)? When consumer–resource interactions in a system are redundant, fragmentation-induced changes in relative abundance may shift the flow of biomass to move through other interaction pathways. Our finding of increased consumption by generalists in fragmented habitats suggests such effects may regularly occur, but the magnitude of the effect would seem to hinge upon species' capacities to feed broadly. Consequently, the specific combinations of consumer–resource pairs within a system will determine whether the net effect on consumption is positive, negative, neutral or context-dependent.
Fourth, what are the proximate mechanisms by which fragmentation influences resource consumption? Fragmentation can directly affect resource or consumer abundances and population numerical responses. In addition, fragmentation can affect movement and searching behaviour, microclimate and vegetation structure (Wirth et al. 2008; Didham et al. 2012), all of which can alter resource consumption without directly changing resource or consumer abundances. Mechanistic studies focusing on the above outstanding questions provide a clear path towards developing a predictive framework for the effects of habitat fragmentation on terrestrial arthropod systems.
We thank members of the Fagan Lab Group for guidance on this project, and especially Leslie Ries for helpful discussions. Sharon Bewick, Leslie Ries, Tomas Roslin and three anonymous reviewers provided valuable comments on earlier versions of this manuscript. HM was supported by the Ann G. Wylie Dissertation Fellowship, the BEES program and an NSF-DIG (DEB-07100004).
Statement of Authorship
HM and WF designed the research, HM conducted the meta-analysis and wrote the first draft of the manuscript, and both authors contributed substantially to revisions.