Empirical landscape studies must consider four major questions, which are addressed here.
1What is the appropriate spatial scale for the studied organisms? The appropriate scale probably depends on the dispersal abilities or foraging radius of the species involved and may vary greatly within or between species groups ( Roland & Taylor, 1997 ; Roslin & Koivunen, 2001 ; Steffan-Dewenter et al., 2001 ). Spatial effects can be expected only for selected spatial scales depending on species- or group-specific dispersal rates ( Kareiva, 1990 ). As a consequence, it seems necessary to analyse multiple spatial scales to find potential interactions with landscape complexity ( Keitt et al., 1997 ; Jonsen & Taylor, 2000 ).
2How can landscape structure be quantified? Landscape ecology provides a great variety of parameters to quantify landscape structure (e.g. Turner & Gardner, 1991 ), however most parameters are inter-correlated and difficult to interpret ( Gustafson, 1998 ). Therefore, the use of simple and ecologically relevant parameters such as the proportion of suitable habitats and the diversity of habitat types was preferred in this study.
3How can species richness, abundance, and biotic interactions of a species group be estimated simultaneously at large spatial scales with sufficient independent replications? In order to answer this question, above-ground nesting bees and wasps were used, which provide the opportunity to monitor their species richness, abundance, and interactions with natural enemies on a landscape scale using standardised trap nests ( Tscharntke et al., 1998 ). Bees and wasps are an important ecological group due to their functions as pollinators and natural enemies respectively ( LaSalle & Gauld, 1993 ). In central Europe, habitat conditions for these species groups have been changed greatly by human activity for several thousands of years and, apart from forests, there are almost no natural, primary bee habitats such as moors, inland dunes, or river flood-plains ( Klemm, 1996 ; Küster, 1999 ). Today, native bees depend mainly on man-made, semi-natural habitats such as calcareous grasslands, meadows, and fallow, which have developed through extensive land use. Many bee and wasp species have specific requirements with respect to food resources, nest site conditions, and building material, and are expected to profit from a variety of habitats within their foraging range ( Westrich, 1996 ; Cane & Tepedino, 2001 ). Therefore, not only the proportion of suitable habitat may be important for the occurrence of a bee species in a local landscape sector, but also the diversity of habitat types at a particular spatial scale.
4How can local effects of habitat quality be separated from regional landscape effects? The effects of landscape structure can be studied best when local habitat effects can be standardised, but surrounding landscapes change greatly. Therefore, in this study, grassy field margins in an agricultural landscape were used because they provide a generally uniform habitat structure of only marginal significance for above-ground nesting bees and wasps.
The aim of the study was to analyse the effects of landscape complexity on community composition and predator–prey interactions of bees and wasps in spatially separated landscape sectors that represented a gradient from structurally poor to complex landscapes. For each landscape sector, spatial scales from 0.25 to 3.0 km were analysed to address the question of species-specific reactions at different spatial scales. Landscape complexity was quantified in two ways: as the proportion of suitable habitat and as habitat diversity in a landscape sector.
The following hypotheses were tested. (1) Species richness and abundance of above-ground nesting bees and wasps increase with increasing landscape complexity. (2) Species richness of natural enemies and rate of parasitism increase with increasing landscape complexity. (3) Different species or species groups respond to landscape complexity at different spatial scales.
Material and methods
Study region and study sites
The study was conducted in 1997 around Göttingen (51.5°N, 9.9°E) in southern Lower Saxony, Germany. The region is characterised by intensively managed agricultural areas with annual crops, and patchily distributed fragments of forests and different semi-natural habitat types. Fifteen spatially separated landscape sectors were selected, which represented different levels of landscape complexity in the study region (Steffan-Dewenter et al., 2001). The mean distance to the nearest neighbouring landscape sector was 4.47 ± 0.31 km. There was no geographical gradient (south–north or east–west) in landscape complexity. For each sector, landscape structure was quantified at radii of 250, 500, 750, 1000, 1500, 2000, 2500, and 3000 m, which represented a nested set of eight spatial scales. The proportion of semi-natural habitats and habitat diversity were used as indicators of landscape complexity.
Landscape composition was quantified by two methods. (1) For a radius up to 750 m, detailed field surveys were made to quantify the total area of each habitat type in the landscape sectors (Thies & Tscharntke, 1999). Habitat types were classified according to von Drachenfels (1996) as non-natural habitats (arable land, intensively used grasslands, water area, and settlement area) and semi-natural habitats (fallow fields and ruderal areas, including field margins, extensively used grasslands, including orchard meadows and calcareous grasslands, hedgerows, rock habitats, and vegetation along inshore waters). Forests were divided into a core area and forest margins (10 m deep boundary) adjacent to non-forest habitats. Only the forest margins were classified as semi-natural habitats. (2) For a radius up to 3000 m, a less-detailed method was used to quantify landscape structure using commercially available digital maps (ATKIS®-DLM 25/1, Landesvermessung + Geobasisinformationen Niedersachsen, Hannover, Germany, 1991–1996), classifying the habitat types arable land, grassland, hedgerows, garden land, forest, and settlement area. For each of the eight nested spatial scales, the proportion of semi-natural habitats (grasslands including intensively and extensively used grasslands, calcareous grasslands and orchard meadows, hedgerows, and garden land) was quantified using Geographical Information Systems ARC/View 3.1 (ESRI Geoinformatik GmbH, Hannover, Germany) and Topol 4.506 (Gesellschaft für digitale Erdbeobachtung und Geoinformation mbH, Göttingen, Germany). The Shannon–Wiener index of habitat diversity (Hs) was calculated for each spatial scale using the percentage (pi) of each habitat type (Hs =–Σ pi × log pi; Krebs, 1989). Habitat diversity and the proportion of semi-natural habitats were correlated significantly and positively for small scales (< 1000 m) but not for larger scales (Steffan-Dewenter et al., 2002).
In order to evaluate the effects of landscape composition on the diversity, abundance, and interactions of bees, wasps, and their natural enemies under standardised nest site conditions (Tscharntke et al., 1998), nest sites for above-ground nesting bees and wasps were established experimentally in each landscape sector. In the centre of each landscape sector, two wooden posts were erected at a distance of 10 m, each fitted with four trap nests for bees and wasps at a height of 1.0–1.2 m. Nesting traps were placed on grassy field margins along field tracks and adjacent to cereal fields. Each trap consisted of 150–180, 20 cm long inter-nodes of common reed Phragmites australis, which were put into 10.5 cm diameter plastic tubes. The reed inter-nodes had a range of diameters from 2 to 10 mm. The traps were set up in the field in mid-April and collected in October. In the laboratory, all reed inter-nodes with nests from bees or wasps were removed and opened using a scalpel. For each nest, the genus (and species if possible), the number of intact brood cells, and the number of brood cells attacked by natural enemies were identified. The nests were then reared separately in the laboratory to get the imagines of bees, wasps, and their enemies for species identification. In some cases, no adults emerged or all brood cells were parasitised, so that only the genus could be identified. These nests were only included in the analyses as additional species for a study site if no other species was found for this genus. Bees and wasps were identified using the key of Gathmann and Tscharntke (1999) and further taxonomic literature. Ichneumonidae were identified by K. Horstmann (Würzburg) and Chalcididae by O. Niehuis (Bochum).
Statistical analysis was performed using Statgraphics plus for Windows 2.1 (Statgraphics, 1995). All data were tested for normality and transformed if necessary. Arcsine transformation (arcsin √p, where p is a proportion) was used to achieve normal distribution for percentages (Sokal & Rohlf, 1995).
Species richness represented the total number of species, and abundance the total number of brood cells of bees, wasps, and natural enemies reared from eight trap nests per landscape sector. Percentage parasitism was the number of parasitised brood cells divided by the number of all brood cells per study site. Each landscape sector represented one replication, i.e. each single data point was obtained from a separate landscape and therefore replicated at the landscape scale (Tischendorf & Fahrig, 2000).
Simple linear regression analysis was carried out to examine possible effects of the proportion of semi-natural habitats or habitat diversity on species richness and abundance of bees, wasps, and their natural enemies at each of the eight spatial scales. The resulting correlation coefficients were plotted considering landscape complexity at each of the eight scales (radius 250–3000 m) for each of the 15 landscapes. Statistics cannot be given for these scale-dependent patterns because the eight nested scales are not independent. Within each spatial scale, multivariate regression was used to test for the combined effects of habitat diversity and the proportion of semi-natural habitats. In addition, comparison of regression lines with the proportion of species (within each group) versus the proportion of semi-natural habitats with bee/wasp/natural enemy as categorical variable was used to identify whether the groups differed.
Altogether 1640 brood cells of 11 bee species (Hymenoptera, Apidae), seven sphecid wasp species (Hymenoptera, Sphecidae), and six eumenid wasp species (Hymenoptera, Eumeninae) were found in the trap nests. The most abundant and widespread species were the bees Osmia rufa and Hylaeus communis, and the wasps Trypoxylon figulus, Trypoxylon medium, and Symmorphus gracilis (Table 1). Eight species of natural enemy were found, which attacked a mean of 14.8% of all brood cells: 13.7% of the bees, 23.7% of the eumenid wasps, and 15.4% of the sphecid wasps respectively (Table 1).
Table 1. Trap-nesting bees, wasps, and their natural enemies from 15 study sites.
For small scales up to a radius of 750 m around the trap nests, data from field surveys were used to estimate landscape structure. The total species number of trap-nesting bees and wasps increased significantly with an increasing proportion of semi-natural habitats within a radius of 250 m (r = 0.656, P < 0.01), 500 m (r = 0.631, P < 0.05), and 750 m (Fig. 1). Habitat diversity did not explain additional variance in the multivariate regression. This pattern was determined mainly by the increasing number of wasp species in landscapes with higher proportions of semi-natural habitats (r = 0.679, n = 15, P < 0.01; 750 m radius), whereas for bees alone this pattern was not significant (r = 0.440, n = 15, P = NS). Comparison of regression lines showed significant differences between the proportion of bee and wasp species versus per cent semi-natural habitats for intercepts (F = 12.47, P < 0.01) but not for slopes, indicating that the presence of wasp species in landscapes with low habitat proportions was reduced significantly compared with bees.
Neither the total number of brood cells nor the number of brood cells of bees or wasps was correlated significantly with the proportion of semi-natural habitats or habitat diversity.
The species number of natural enemies tended to increase with the proportion of semi-natural habitats (r = 0.479, n= 15, P= 0.07, NS) but not with habitat diversity (r = 0.287, P = NS). There were no significant multiple models. The rate of parasitism did not respond to the proportion of semi-natural habitats (r = 0.048, n = 15, P = NS) or habitat diversity (r = 0.051, n = 15, P = NS).
Scale-dependent effects of landscape complexity
For the analysis of effects of larger spatial scales, data from digital maps were used because it was not possible to make field surveys at these spatial scales. In general, the results obtained from analyses with field data were confirmed, although the correlation coefficients were lower on average. Again, the species richness of wasps and natural enemies was correlated more closely with the proportion of semi-natural habitats than was the species richness of bees (Fig. 2). Furthermore, the analyses of multiple spatial scales showed that the importance of the proportion of semi-natural habitats decreased with increasing spatial scale for the species number of bees, wasps, and their natural enemies (Fig. 2). Comparison of regression lines showed significant differences among the proportions of bee, wasp, and natural enemy species versus per cent semi-natural habitats for intercepts (F = 7.6, P < 0.01) but not for slopes, and significant effects of the proportion of semi-natural habitats only for the 250 and 500 m radii. Using habitat diversity at multiple spatial scales as a measure of landscape complexity again indicated the highest significance at small scales for species richness of wasps (r = 0.672, P < 0.01; 250 m radius) but no dependence for bees or natural enemies.
The number of brood cells of bees and wasps did not correlate with landscape parameters at any spatial scale. Only the number of parasitised brood cells showed a significant, positive correlation with the proportion of semi-natural habitats (r = 0.563, P < 0.05) but not with habitat diversity at a small spatial scale (500 m radius).
The proportion of parasitised brood cells (rate of parasitism) did not correlate with the proportion of semi-natural habitats at any spatial scale, while habitat diversity was only correlated at a radius of 3000 m (r = 0.544, P < 0.05).
Studies on the effects of landscape complexity on insect communities and trophic interactions are very rare although they are needed to assess the loss of biological diversity due to declines in environmental and ecosystem quality. Here, it was shown that the proportion of semi-natural habitats and habitat diversity in the surrounding landscape were correlated positively with species richness of trap-nesting bees, wasps, and their natural enemies. The results indicate that wasps were influenced more strongly by landscape structure than were bees and natural enemies. Furthermore, the proportion of semi-natural habitats allowed for a better prediction of species richness of bees and natural enemies, whereas the richness of wasp species seemed to respond more to habitat diversity. This may be due to a greater dependence of wasps on the occurrence of different habitat types at a limited spatial distance. A positive relationship between landscape complexity and insect species richness has also been found for butterflies in Swedish agricultural landscapes (Weibull et al., 2000) and flower-visiting bees (Steffan-Dewenter et al., 2002).
The use of spatially separated landscape sectors as independent replicates is an approach that was suggested recently by theoretical ecologists to improve the power of empirical landscape studies (Tischendorf & Fahrig, 2000). Furthermore, multiple spatial scales were analysed to find potential interactions between landscape complexity and species or species groups with different dispersal rates (Kareiva, 1990; Keitt et al., 1997; Jonsen & Taylor, 2000). Bees and wasps showed the strongest response to the proportion of semi-natural habitats at small spatial scales up to 750 m radius. The higher significance of landscape structure for wasps than for bees may indicate more restricted dispersal rates of wasp species or a higher degree of specialisation for food resources, nesting places, or nest-building material. For two common bee species (Osmia rufa and Hylaeus communis), earlier studies showed that isolation of trap nests from semi-natural habitats did not decrease colonisation success (Steffan-Dewenter, 1998). The commonness of these species in the actual study may explain why no correlations between landscape complexity and the total number of brood cells were found. Unfortunately, the data set was too small to analyse species-specific responses to landscape parameters.
Other studies on a landscape scale have focused on plant–insect interactions (Steffan-Dewenter et al., 2001) and parasitism of herbivores (Roland & Taylor, 1997; Menalled et al., 1999; Thies & Tscharntke, 1999), but there has been no study of trophic interactions of aculeate Hymenoptera on a landscape scale. In contrast to expectations, there was no correlation between landscape complexity and rate of parasitism. It could be speculated that the diversity of habitats in a landscape sector should be a better predictor of species richness of natural enemies than the proportion of semi-natural habitats, however the number of species of natural enemy tended to increase at small scales with the proportion of semi-natural habitats but not with habitat diversity. In conclusion, the empirical patterns for species richness and trophic interactions underline the importance of analysing landscape complexity at different spatial scales but further studies are needed to identify causal relationships.
I greatly thank Sabine Eber, Teja Tscharntke and three anonymous referees for helpful comments on the manuscript, Christof Bürger for GIS analyses, Carsten Thies for field data of landscape structure, K. Horstmann (Ichneumonidae) and O. Niehuis (Chalcididae) for the identification of natural enemies, and the German Science Foundation (Deutsche Forschungsgemeinschaft) for financial support.