1. Habitat fragmentation is a major threat to biodiversity and can lead to the loss of both species and ecosystem services, but fragmentation effects vary greatly between studies and studied organisms. Understanding the distinct effects of habitat amount and isolation at the patch and landscape scale may account for some of this variation.
2. We studied biodiversity in 30 traditional orchards that were selected for independent variation in habitat amount and habitat isolation at the patch and landscape scale. We analysed species richness and abundance of snails, beetles, true bugs, spiders and breeding birds that avoid open farmland but occur in woody vegetation types. Additionally, the abundances of nine single species were analysed using specific habitat definitions.
3. Surprisingly, the effects of habitat isolation were more important than the effects of habitat amount. Effects at the patch scale were more frequent than landscape-scale effects.
4. Spider species richness decreased with increasing patch-scale habitat amount. Abundance of the weevil Phyllobius oblongus increased with landscape-scale habitat amount. Negative effects of patch isolation were greater for predatory birds and spiders, while the predominately herbivorous beetles, true bugs and snails were less affected. Species richness of birds, spiders and beetles, and abundance of birds, Cyanistes caeruleus, Parus major and Fringilla coelebs, decreased with increasing patch-scale habitat isolation. In contrast, species richness of spiders and beetles increased with increasing landscape-scale habitat isolation.
5.Synthesis and applications. The effects of habitat fragmentation differed between taxonomic groups, with stronger and more consistent responses in birds than invertebrates. Our understanding of fragmentation effects may be biased due to the dominance of bird studies in the literature, and further invertebrate studies are encouraged. Landscape management to improve biodiversity or ecosystem services requires a group-specific approach and coordinated priority setting. High habitat connectivity benefited wood-preferring birds, spiders and beetles, lending support to national initiatives for increased habitat connectedness. The negative effects of patch isolation were greater for natural pest regulators, birds and spiders than for herbivorous beetles and bugs.
Island biogeography and metapopulation theories assume that the matrix surrounding habitat fragments is hostile to the study organisms (Haila 2002). Utilization of the matrix by habitat generalists violates this assumption and contributes to the often unexpected outcomes of fragmentation studies (Haila 2002; Lindenmayer & Fischer 2007). This problem is enhanced by different usages of the term habitat which, strictly defined, refers to the resources and conditions present in an area that produce occupancy for a particular species (Hall, Krausman & Morrison 1997). However, habitat is widely used in synonymy with vegetation types that are inhabited by characteristic assemblages of animal and plant species (Ewers, Thorpe & Didham 2007; Radford & Bennett 2007; the European Environment Agency at http://eunis.eea.europa.eu) and this definition is adopted here. In this case, the island model applies to the species whose occurrence is associated only with the studied vegetation types and none other.
Habitat loss has large, consistently negative effects on biodiversity which are thought to exceed habitat fragmentation effects (Fahrig 2003; MEA 2005). However, habitat loss is usually associated with habitat fragmentation. A continuous habitat is transformed into many smaller patches – of less total area – which are isolated by a matrix unlike the original (Wilcove, McLellan & Dobson 1986). Many empirical studies have examined this combined effect of habitat loss and fragmentation. Habitat fragmentation per se, i.e. the splitting up and isolation of habitats without habitat loss, has been less examined. In a review of the effects of habitat fragmentation on biodiversity (Fahrig 2003), only 17 of over 100 studies examined habitat fragmentation per se. Most of these studies were micro-scale experiments (e.g. Collinge & Forman 1998, 10 × 10 m), which can lead to problems when scaled up to real landscape situations. We therefore worked in real landscapes, differentiating between the effects of patch- and landscape-scale habitat loss and habitat fragmentation per se on the fauna of apple orchards.
Traditional orchards are considered to be species-rich High Nature Value Farming Systems in Europe (Cooper et al. 2007). Together with other woody habitats, they provide a refuge for arthropods and birds which otherwise cannot occur in modern, intensified agricultural landscapes (Herzog 1998). Whilst orchards were widespread in Europe a century ago, they have diminished strongly in the last decades (e.g. 80% loss in Switzerland between 1951 and 2001; BFS 2001). Continuous orchard landscapes have become fragmented, the remaining orchards being embedded in an open agricultural matrix of arable fields and grassland.
From the >12 000 orchards recorded on topographic maps in north-eastern Switzerland, 30 were selected, which varied as independently as possible in their degree of isolation and in the amount of woody vegetation in their surroundings. Breeding birds, snails, spiders, bugs and beetles were recorded in the tree crowns and ground vegetation. Analyses were focussed on species assemblages that avoid open farmland, but occur in woody vegetation such as orchards, hedgerows and forest. In addition, we examined the abundance of single species for which specific habitat definitions were applied according to their particular vegetation preferences. The following hypotheses were tested at the patch and landscape scale:
1 Abundance and species richness increase with habitat amount.
2 Abundance and species richness decrease with habitat isolation.
3 The effects of habitat amount outweigh the effects of habitat isolation.
Materials and methods
The study was undertaken in north-eastern Switzerland, a region characterized by traditional high-stem apple Malus domesticus orchards interspersed with forest and open farmland.
To distinguish between habitat amount and habitat isolation effects, 30 orchards were GIS-selected to vary as independently as possible in habitat isolation at the patch scale, and habitat amount at the landscape scale. Mature high-stem orchards were pre-selected using the Swiss Vector25 digital landscape model (http://www.swisstopo.ch). Habitat amount and habitat isolation were calculated within a 500 m radius surrounding each of the orchards present in the region, whereby all woody vegetation was regarded as habitat. The radius of 500 m represents an intermediate scale at which the investigated species groups respond to landscape composition in Europe (Schweiger et al. 2005; Billeter et al. 2008; Schmidt et al. 2008). Subsequently, orchards isolated from other habitat by 0 m and ≥ 70 m were examined in conjunction with orthophotos (spatial resolution <1 m). Orchards with ≥10 trees organized as a central group and ≤ 20 m apart were retained if not bordering settlements or large roads. Size (≥ 0·5 and ≤ 2·0 ha), elevation (≤ 650 m) and slope (≤ 20%) were restricted to limit variation in these factors. The remaining orchards were assessed in the field and orchards consisting predominantly of similar sized apple trees were retained, which were either spatially connected to or isolated from other woody habitats. These requirements rapidly reduced the number of potential sites from >12 000 to 59 isolated and 37 connected orchards.
Maps of potential study sites were digitized to calculate habitat amount in a 500 m buffer around the focal orchard. Woody habitat was defined as forest, hedgerows, scrub, tree rows, solitary trees and other high-stem orchards. To maximize independent variation in woody habitat amount in the 500 m radius, we formed two gradients of non-overlapping landscapes, one with 15 connected (12·7–46·7%) and the other with 15 isolated (7·9–38·6%) focal orchards (Fig. 1). In each orchard, tree number and trunk circumferences (at 1 m height) were quantified, deadwood noted and the farmer questioned regarding farming practice (See Table S1 in Supporting Information).
We assessed densities and species richness of breeding birds (Aves), snails (Gastropoda), spiders (Araneae), herbivorous beetles (Coleoptera: Chrysomelidae and Curculionoidea) and bugs (Heteroptera) from the tree crowns and ground vegetation of the orchards between April and October 2007. Sample size was standardised across orchards for all organisms except birds. Territories of breeding birds were mapped for the entire focal orchard during three visits between April and June. Surveys were undertaken in the early morning (05·30–09·30 h) and followed the breeding bird mapping scheme of the Swiss Ornithological Institute (http://www.vogelwarte.ch). For each orchard, the maps of the visits were combined and territories determined based on clustered observations and known species-specific territory size.
For the snail and arthropod sampling, four similar circumference trees were selected in each orchard. Snails were sampled during September and October in all orchards (apart from one felled orchard) according to the Swiss biodiversity monitoring methodology (Hintermann et al. 2002). Two sampling points were under the tree crowns of two of the four trees and two were between trees. At each location, eight soil samples were taken concentrically around the sampling point. The soil samples (11 × 11 × 5 cm) were washed and sorted with three filtering meshes (4, 2, and 0·7 mm). The snail shells were identified to species and the numbers of individuals counted. Only shells with signs of recently living individuals (body present, diaphragm recognizable, shell intact) were included in the analyses to focus on recent snail distribution patterns.
Arthropod sampling was undertaken in April, June, August and October. A beating tray was used to sample the tree crowns. Four similar branches were selected from each of the four trees and these were beaten with a stick. The fauna was caught in a textile-funnel (0·25 m2) with an attached jar filled with 70% ethanol. Each branch represented a different aspect. Herb layer arthropods were suction sampled using a handheld blower (Macleod, Wratten & Harwood 1994). Four 0·25 m2 samples were taken per tree, two under the crown and the others 2 m outside the crown’s border. The samples were frozen and later sorted for identification. Arthropod adults were identified to species and the number of individuals recorded. Immature spiders were not determined to species, except for Anyphaena accentuate (Anyphaenidae), the only representative of its family in the study region.
Species assemblages and individual species
In accordance with our research design, we restricted all analyses to species that prefer woody vegetation and avoid open land. Birds were classified as wood-preferring if they predominately forage on trees (Glutz von Blotzheim 1997; Zwygart 1983). Wood-preferring spiders had their niche position in the upper half of a gradient from open to woody vegetation (Entling et al. 2007). Herbivorous beetles and bugs were classified with respect to their host plant (Wachmann, Melber & Deckert 2008). Snails and non-herbivorous bugs were classified according to habitats mentioned in the determination literature (Falkner et al. 2001; Wachmann, Melber & Deckert 2008). Thus, our analyses focussed on the subset of the orchard fauna that depends on woody vegetation. The vast majority of these species are generalists for deciduous trees (e.g. Sackett, Buddle & Vincent 2008) and occur in all woody habitat types considered in the research design. Only a few species are specific to semi-open structures and thus avoid closed forest (e.g. Phyllobius oblongus, see below).
In addition to assemblages, we analysed one to three single species per group that combined a high frequency in the orchards with a particularly well-defined preference for woody vegetation. Of the breeding birds, we chose Cyanistes caeruleus (Linnaeus 1758), Parus major (Linnaeus 1758) and Fringilla coelebs (Linnaeus 1758). The Cyanistes and Parus species breed in cavities and forage in tree crowns whilst F. coelebs nests in tree crowns and feeds on and below trees. All three species are restricted to woody vegetation during the breeding season (Glutz von Blotzheim 1997). The snail species Perpolita hammonis (Ström 1765) has a clearly defined preference for woody vegetation (Falkner et al. 2001). Anyphaena accentuata (Walckenaer, 1802), a hunting spider, is common in tree crowns of both semi-open vegetation and closed forest. In contrast, the second spider species Lathys humilis (Blackwall 1855) builds webs on the leaves of a wide range of woody plants, but avoids forest interior (Hänggi, Stöckli & Nentwig 1995). The only abundant beetle with a clear woody vegetation preference was P. oblongus (Linnaeus 1758). Its larvae feed on plant roots, and adults on the leaves of shrubs and trees that are largely restricted to forest edge and other semi-open vegetation, including hedgerows and orchards (C. Germann, personal communication). The shield bug Pentatoma rufipes (Linnaeus 1758) feeds on plants and arthropods in the crowns of all major families of deciduous trees in the study region, thus occurring in both semi-open vegetation and forest interior (Wachmann, Melber & Deckert 2008). The nabid bug Himacerus apterus (Fabricius 1798) is flightless and lives on trees and their undergrowth (Wachmann, Melber & Deckert 2008).
Of the nine selected species, seven preferred woody vegetation in general, whereas two avoided forest interior. To account for this, we generated two versions of landscape-scale habitat amount and isolation – either including or excluding the forest interior (see below). In accordance with the habitat preference of the assemblages and the majority of single species, woody vegetation is termed ‘habitat’ in the following unless otherwise stated.
Local, patch and landscape metrics
We tested relationships of animal abundance and species richness with explanatory variables representing local site, patch-scale and landscape-scale properties (Table 1). One metric represented habitat amount and the other isolation at both the patch and landscape scale. Two metrics represented the local site factors, orchard structure and management. At the landscape scale, we differentiated between woody vegetation (including forest interior) and semi-open vegetation (forest edge and other woody habitats, see below).
Table 1. Metrics used to examine the effects of habitat amount, habitat isolation and local site factors on biodiversity. For assemblages and species preferring woody vegetation in general, closed forest was included as habitat. For species preferring semi-open vegetation, closed forest was excluded, and the remaining woody vegetation types retained
PA: Orchard size (m2)
LAW:% Landscape covered by woody vegetation LAS:% Landscape covered by semi-open vegetation
PI: Distance to nearest woody vegetation patch (m)
LIW: Mean Euclidean nearest neighbouring patches of woody vegetation (m) LIS: Mean Euclidean nearest neighbouring patches of semi-open vegetation (m)
Local site factors
MA: Management STR: Structure
Euclidean distance was used to measure habitat isolation (Moilanen & Nieminen 2002). Many measures combine patch distance with patch size (Bender, Tischendorf & Fahrig 2003) and cannot distinguish between habitat amount and habitat isolation effects, our main study aim. Patch-scale habitat amount was represented by the orchard size (m2) and isolation through the Euclidean distance (m) to the nearest potential habitat. The percentage of habitat for the entire landscape (habitat amount) and the mean Euclidean distance (m) between woody habitats (isolation) were calculated using FRAGSTATS (McGarigal et al. 2002) for the wood-preferring organisms and then recalculated for the two species that avoid forest interior. Suitable vegetation for these two species was redefined as woodland edge (5 m width), hedgerows, scrub, tree rows, solitary trees and other high-stem orchards. The habitat gradient for the connected orchards was then 4·8–25% and the isolated 3·1–13·7%.
The local site factors used to represent orchard structure and management are listed in Table S1, Supporting information. We used principal component analyses to extract axes representing major variations in structure and management (CANOCO for Windows 4·5, ter Braak & Smilauer 2002). The principal component representing management indicated sulphur and copper applications, high mowing frequency, absence of deadwood and no grazing (loadings listed in Table S1, Supporting information). The principal component representing structure indicated high tree density and low tree circumference.
The relationships between animal abundance and species richness with local, patch and landscape factors were tested using generalized linear models with quasi–Poisson distribution and log link in R version 2.6.1 (R Development Core Team 2006). We used both forward and backward selection of explanatory variables according to their significance, because selection methods based on information criteria are unavailable for quasi–Poisson distributions. Forward and backward selection always resulted in identical models. In spite of the GIS-based selection of study sites, some explanatory variables showed correlations of up to r = −0·57 between landscape-scale habitat amount and isolation (Table S2, Supporting information). Thus, we used partial regression to calculate the independent effects of multiple explanatory variables in each model (Legendre & Legendre 1998). Explanatory variables were log10(X + 1) transformed when necessary to reduce skewness.
As birds were sampled in the entire orchard, sampling intensity was proportional to patch area. Following the definition of abundance as the number of individuals per unit area, a proportional increase of the number of bird territories with patch area represents no change in bird abundance. To correct for the variable sampling intensity, patch area was kept in all bird models regardless of its statistical significance. To test for an influence of patch area on bird abundance, the intercept of linear models between the number of bird territories and raw patch area was examined. A significant regression intercept indicates a significant deviation from proportionality between bird territory numbers and patch area, and thus a patch area effect on abundance. Sample size was standardised across orchards for all remaining groups. Thus, patch area effects on these groups indicate altered abundance or species richness as a consequence of higher local habitat amount.
In total, we sampled 14 353 individuals of 240 species. Thereof, 85 species and 1756 individuals were specific to woody habitats (Table 2). Analyses were limited to these wood-specific species.
Table 2. Total numbers of individuals and species of the sampled animal groups, divided by their habitat preference. Only ‘wood-preferring’ species were analysed here. ‘Other’ species comprise habitat generalists and species preferring open habitats
The effects of patch area on species richness were absent in snails, beetles and bugs, and in an unexpected direction in spiders (Fig. 2a, Table 3). In strong contrast to our first hypothesis, species richness of spiders decreased with increasing patch area. The increase of the numbers of bird territories with patch area (Table 3) reflects the increased area sampled. The intercept of a linear model between territory numbers and (untransformed) patch area was non-significant (t1,28 =0·50, P =0·62). Thus, numbers of bird territories were proportional to patch area, and there was no significant effect of patch area on overall bird abundance. At the species level, the number of territories of the birds C. caeruleus and P. major increased significantly with patch area in accordance with the increased area sampled (Table 3). Examination of the linear model intercept revealed no significant effect of patch area on the abundance of these species (C. caeruleus: t1,28 =0·59, P =0·56; P. major: t1,28 =0·78, P =0·44). However, the model for F. coelebs had a significant positive intercept (t1,28 =2·65, P =0·013), meaning that territory numbers were less than proportional to patch area, and that abundance decreased with patch area.
Table 3. Effects of habitat amount, isolation and local site factors on the fauna of orchards. Generalized linear models with quasi–Poisson distribution and backward selection of non-significant explanatory variables. In the case of two or more significant explanatory variables, partial models were constructed to show only independent (conditional) effects of each variable (see methods)
Local site factors
Significant values in bold, for non-significant models, values for the explanatory variable with the lowest P-value are given. Shaded cells: relationships of bird species richness and abundance with patch area were retained in the models regardless of significance to correct for the uneven sampling intensity (see Results).
At the landscape scale, habitat amount showed no significant effects on species richness or abundance of the assemblages (Table 3). The weevil P. oblongus was significantly affected by landscape-scale habitat amount (Fig. 2b). In accordance with its habitat preference, its abundance increased with the amount of landscape-scale semi-open habitats.
In agreement with the second hypothesis, patch-scale habitat isolation reduced species richness of birds, spiders and beetles (Table 3, Fig. 3a–c). Also abundances of total birds, and of C. caeruleus, P. major and F. coelebs decreased with increasing isolation of the focal patch from other woody habitats (Fig. 3d–g). In contrast and unexpectedly, landscape-scale habitat isolation increased species richness of spiders and beetles (Fig. 3h,i). There were no significant effects of landscape-scale habitat isolation on any of the individual species.
Orchard structure and management intensity
Intensive management reduced the species richness and abundance of bugs, and the abundance of P. rufipes (Table 3, Fig. 4a–c). Cyanistes caeruleus abundance increased significantly in orchards with high densities of slender trees (Fig. 4d).
Both the species and assemblage approach revealed distinct responses to habitat fragmentation, particularly for the birds. Surprisingly, habitat isolation effects were more important than habitat amount effects. Patch-scale effects were more frequent than effects at the landscape scale. Intensive management had negative effects on bugs, corresponding with earlier studies (e.g. Schweiger et al. 2005).
Patch-scale habitat amount
The expected positive effects of patch area on abundance and species richness were absent. Contrarily, increasing patch area reduced species richness of wood-preferring spiders and the abundance of F. coelebs. Landscape-scale habitat amount effects were minor (see below). Thus, our first and third hypotheses, that species richness and abundance increase with habitat amount and outweigh habitat isolation effects, must be rejected. The negative response of F. coelebs abundance to patch area can be explained by its territoriality. If breeding pair abundance is limited by the distance to the next singing male, then even the smallest orchard (0·4 ha) can host one breeding pair of F. coelebs, which was indeed the case. However, the density of 0·5 breeding pairs per hectare in Central European semi-open habitats (Glutz von Blotzheim 1997) meant that none of the orchards (of up to 2·9 ha) were large enough to host >2 pairs. Thus, several small orchard fragments will host higher numbers of F. coelebs pairs than one large fragment of the same total area. From a conservation perspective, it is questionable if this is desirable as breeding success may be reduced in smaller fragments (Harrison & Bruna 1999).
The decrease in spider richness with orchard area may be due to source-sink dynamics or trophic interactions. Negative relationships between species richness and habitat area can be expected if local population growth rates are negative. Sink populations are sustained by immigrants, but the number of immigrants per area usually decreases with patch size (Hambäck & Englund 2005; Englund & Hambäck 2007). Thus, the negative relationship of spider species richness with patch area may indicate that orchards are sink habitats for wood-related spider species. Alternatively, the reduced spider richness in large patches could be due to trophic interactions. Birds and wasps (Crabronidae and Pompilidae) are important spider predators (Wise 1993; Glutz von Blotzheim 1997). If natural enemies are more reduced in small fragments than their prey (Holt 1996), then their prey could benefit from reduced predation pressure and develop higher densities in smaller fragments. However, neither birds nor spider-eating wasps (J.D. Herrmann, unpublished data) had higher abundance in larger orchards, suggesting that top-down effects on spider species richness are unlikely to be responsible for their negative response to patch area. The alternative trophic pathway of bottom-up regulation requires higher density of spider prey in smaller orchards. Flying insects from the agricultural matrix may be more abundant in smaller orchards due to their higher circumference to area ratio. Thus, decreasing food availability with increasing patch size would be a plausible explanation for the observed spider pattern.
Landscape-scale habitat amount
Only the weevil P. oblongus increased in abundance with semi-open habitat amount, no other significant landscape-scale habitat amount influences were found. The lack of landscape composition effects contrasts with other studies, where habitat loss had consistent negative effects on biodiversity (Fahrig 2003). When investigating the effects of patch-scale and landscape-scale habitat amount, an important difference is if the effective area sampled increases with habitat amount, or if it is held constant, as in our study. In the first case, even a constant density within the studied habitat will lead to an increase in total numbers of individuals and (as a consequence) species with increasing habitat amount. In contrast, the second case will detect if the density of individuals or species within the habitat changes according to patch- or landscape- scale habitat amount. Recent studies suggest that landscape-scale habitat amount can have expected positive effects on within-habitat densities of individuals and species (e.g. Radford & Bennett 2007). For biodiversity conservation, total numbers of individuals and species in a landscape are of primary interest. For ecosystem functioning, within-habitat density of individuals and species is more important.
Patch-scale habitat isolation
In accordance with our predictions, species richness and abundance decreased with increasing patch-scale habitat isolation. This indicates that isolation has consistently negative consequences for wood-preferring species. Whilst Fahrig (2003) found the effects of fragmentation per se as likely to be positive as negative, our study is an example of where fragmentation affects species diversity and abundance negatively. Recent studies of wood-dependent bird species also found fragmentation per se to be consistently negative for species where fragmentation was influential (Radford, Bennett & Cheers 2005; Radford & Bennett 2007). Landscapes with greater habitat connectivity support more stable populations and greater bird species diversity (Taylor et al. 1993; Radford, Bennett & Cheers 2005). The abundance and richness of wood-preferring species is influenced by patch finding success which may decrease with isolation (Goodwin & Fahrig 2002) especially for insect groups with random movement (Grez, Zaviezo & Tischendorf 2004). Species mobility warrants investigation, low mobility being hypothesized to increase species vulnerability to isolation (Ewers & Didham 2006). However, even high mobility species (birds) have reduced breeding success (Villard, Martin & Drummond 1993) and lower nestling body weights (Lens & Dhondt 1994) in isolated habitats.
Landscape-scale habitat isolation
The positive effect of landscape-scale habitat isolation on species richness of spiders and beetles is surprising, especially as they showed the opposite response to patch-scale habitat isolation. It demonstrates that habitat isolation effects are scale-dependent, probably because different processes predominate at the patch and landscape scales (Wiens 1989). Enhanced species persistence in more isolated subpopulations may be due to refuge creation from competition (Atkinson & Shorrocks 1981) or predation (Hastings 1977), and/or by reduced synchrony in local population fluctuations, reducing the probability of overall population extinction (den Boer 1981). However, evidence of these processes from landscape-scale studies is limited (Fahrig 2003).
Species vs. species assemblage approach
The species and species-assemblage approach resulted in comparable observations for birds and bugs. In spiders and beetles, additional or different effects were observed for assemblages compared to single species. In particular, species richness showed stronger habitat fragmentation effects. This may be because species richness as an aggregate measure shows less sampling error. In addition, some of the contingency of single-species responses to habitat fragmentation may be averaged out when combined into species richness (Lawton 1999). Regardless of approach there was no significant response to any variables for snails. The landscape scale may be of limited relevance for low mobility groups (Harrison & Bruna 1999) although landscape effects have been observed for land molluscs (Gotmark, von Proschwitz & Niklas 2008; Knop, Herzog & Schmid 2010). As different species respond to landscape properties at different spatial scales (Cushman & McGarigal 2002) our sampling unit (500 m radius) may have been too large for snails.
The fauna of fragmented orchards showed specific responses to different aspects of habitat fragmentation depending on the taxonomic group. Birds showed strong and uniform responses to habitat fragmentation, whereas invertebrate responses were weaker and variable. Bird studies dominate the habitat fragmentation literature (Fahrig 2003), thus our current understanding of fragmentation effects may be biased. This underlines the need for further fragmentation studies on invertebrates, particularly in an agro-ecological context, where the relationship between pests and predators has a major economic impact. Here, beneficial orchard predators, the birds and spiders (Wyss 1995; Marc, Canard & Ysnel 1999; Mols & Visser 2002), showed greater sensitivity to habitat isolation than herbivorous beetles and bugs. This supports the trophic level hypothesis (Holt 1996) that herbivores and predators respond differently to habitat isolation. Vulnerability of predators to orchard isolation may reduce natural pest control of herbivore populations.
Surprisingly, habitat isolation was more important than habitat amount. Orchard connectivity was related to higher general species richness which may promote agro-ecosystem resilience (Tscharntke et al. 2005). Generally, increasing habitat connectivity is considered beneficial for biodiversity and ecological networks are promoted in Europe from the continental (Jongman, Kulvik & Kristiansen 2004) to the regional scale (Hepcan et al. 2009). Increasing orchard connectivity will benefit wood-preferring birds, spiders and beetles. This finding supports national initiatives to improve habitat connectedness (Bundesrat 2001; Gouvernement Français 2009). Further, it supports recommendations that the spatial arrangement of (restoration) habitats should be accounted for in agri-environmental schemes and their connectedness increased (Knop, Herzog & Schmid 2010). Our results also confirm that there are no general measures that will be advantageous for all species (groups) at all scales. Measures to stabilize and promote orchard biodiversity in particular and farmland biodiversity in general need to be tailored to the specific group or species to be promoted. In doing so, one has to accept that a specific measure in favour of a particular species (group) may be disadvantageous for other species (groups). This requires careful and co-ordinated priority setting, taking into account the regional context.
We are grateful to the farmers for allowing access to their land. We thank Susanne Mühlner, Karin Stämpfli, Karen Schubert, Erich Szerencsits and Jonas Winizki for their GIS support, Kim Stier, Susanne Riedel & Stephan Bosshart for the snail sampling and Christoph Germann (beetles) and Ambros Hänggi (spiders) for their taxonomic expertise. We acknowledge the comments of three anonymous reviewers. The study was supported by the Swiss National Science foundation under grant number 3100A0-114058.