At what spatial scale do high-quality habitats enhance the diversity of forbs and pollinators in intensively farmed landscapes?

Authors

  • Florian Kohler,

    1. Nature Conservation and Plant Ecology Group, Wageningen University, Droevendaalsesteeg 3a, 6700 AA Wageningen, the Netherlands;
    2. Ecole Polytechnique Fédérale de Lausanne EPFL, Laboratory of Ecological Systems ECOS, Station 2, CH-1015 Lausanne, Switzerland;
    Search for more papers by this author
  • Jort Verhulst,

    Corresponding author
    1. Nature Conservation and Plant Ecology Group, Wageningen University, Droevendaalsesteeg 3a, 6700 AA Wageningen, the Netherlands;
      Correspondence author: E-mail: Jort.Verhulst@gmail.nl
    Search for more papers by this author
  • Roel Van Klink,

    1. Nature Conservation and Plant Ecology Group, Wageningen University, Droevendaalsesteeg 3a, 6700 AA Wageningen, the Netherlands;
    Search for more papers by this author
  • David Kleijn

    1. Department of Aquatic Ecology and Environmental Biology, Nijmegen University, Toernooiveld 1, 6525 ED Nijmegen, the Netherlands; and
    2. Alterra, Centre for Ecosystem Studies, PO Box 47, 6700 AA Wageningen, the Netherlands
    Search for more papers by this author

Correspondence author: E-mail: Jort.Verhulst@gmail.nl

Summary

  • 1Over the last decades, biodiversity in agricultural landscapes has declined drastically. Initiatives to enhance biodiversity, such as agri-environment schemes, often have little effect, especially in intensively farmed landscapes. The effectiveness of conservation management may be improved by scheme implementation near high-quality habitats that can act as a source of species. We evaluated up to what distance high-quality habitats (nature reserves and artificially created flower-rich patches) affect the diversity of forbs and pollinators in intensively farmed landscapes of the Netherlands.
  • 2We surveyed forbs, inflorescences, bees and hover flies and estimated pollination services in transects along ditch banks extending 300 m from four nature reserves forming small islands in landscapes dominated by agriculture.
  • 3In a separate experiment, we surveyed inflorescences, bees and hover flies in 1500 m long transects on farmland adjacent to five newly introduced flower-rich patches and in five control transects.
  • 4Species density of forbs declined over the first 75 m and species density and abundance of hover flies declined over the first 125 m beyond the nature reserves. Beyond these distances, no further declines were observed. The effects of flower-rich patches were spatially limited. The species density and abundance of bees and hover flies were significantly enhanced in the flower-rich patch, but only the abundance of hover flies was enhanced up to 50 m beyond the patch.
  • 5Synthesis and applications. In intensively farmed areas, remnant high-quality habitats sustain more abundant and diverse pollinator and forb communities than the surrounding countryside. They do enhance biodiversity on nearby farmland but increases are spatially restricted (< 150 m) and relatively small. These habitats may therefore function only as dispersal sources for ecological restoration sites or agricultural fields under extensification schemes that are located in close proximity. Habitat restoration in intensively used farmland should therefore be implemented preferentially in the immediate vicinity of high-quality habitats. In the short term, newly created flower-rich habitats are no alternative to pre-existing seminatural habitats for the promotion of pollinators on nearby farmland.

Introduction

Over the last decades biodiversity in agricultural landscapes of western Europe has declined dramatically (Krebs et al. 1999; Gregory, Noble & Custance 2004). Agricultural intensification has led to considerable losses in biological diversity at various spatial and temporal scales (Benton, Vickery & Wilson 2003). Fragmentation of habitats and high application rates of fertilizers and pesticides have had a significant impact on the levels of plant and animal diversity (e.g. Baur & Erhardt 1995; Robinson & Sutherland 2002; Tscharntke et al. 2002, 2005). The maintenance of biodiversity is not only desirable from a conservation viewpoint. Biodiversity is often correlated positively with the provision of ecosystem services sensu Kremen (2005) (Chapin et al. 2000; Kremen & Ostfeld 2005), and biodiversity conservation or restoration may therefore have economic benefits (Thomas, Wratten & Sotherton 1991; Ricketts et al. 2004).

In Europe, initiatives to restore or conserve farmland biodiversity have focused mainly upon the extensification of farming practices and creation of seminatural habitats and are usually stimulated by agri-environmental payments (Kleijn & Sutherland 2003). These ‘agri-environment schemes’ have mixed biodiversity effects (Kleijn & Sutherland 2003; Kleijn et al. 2006), with most effects being restricted to common and widespread species (Kleijn et al. 2006; but see Peach et al. 2001). In intensively farmed areas such as the Netherlands these initiatives often fail to deliver biodiversity benefits (Kleijn et al. 2001; Bradbury & Allen 2003; Kohler et al. 2007). An important impediment to biodiversity conservation on farmland may be the limited dispersability of most species of plants and arthropods (Kleijn et al. 2001). Seed banks of (former) intensively managed agricultural fields are very poor, so that seeds from outside the field are required to increase plant diversity (Bakker & Berendse 1999; Blomqvist, Bekker & Vos 2003a). Conservation practices may succeed in improving the quality of farmland habitats, but nevertheless do not result in biodiversity benefits if species are not able to colonize the restored sites. The effectiveness of nature conservation measures in agricultural areas may therefore be enhanced by implementing them on fields in the vicinity of high-quality habitat, which can act as a source of biodiversity (Blomqvist et al. 2003b).

In intensively farmed north-western European landscapes, the last remaining areas with high biodiversity are seminatural habitats such as woodlots, ponds, riparian fragments, unimproved grasslands, heathlands or nature reserves (Kleijn & van Langevelde 2006). Farmland biodiversity and the associated ecosystem services may be related positively to the proportion of (semi)natural habitats in agricultural landscapes (Steffan-Dewenter et al. 2002; Kremen et al. 2004; Kleijn & van Langevelde 2006). Duelli & Obrist (2003) showed that the presence of over 60% of the arthropod species observed in a Swiss agricultural landscape depended on the occurrence of nearby seminatural habitats. Steffan-Dewenter & Tscharntke (1999) and Öckinger & Smith (2007) demonstrated that plant–pollinator interactions declined with distance to the nearest seminatural habitat. These results suggest that (semi)natural habitats do indeed act as a source of biodiversity and may facilitate the colonization of agricultural fields under extensification schemes.

An important remaining question is up to what distance the positive effects of species-rich habitat extend into the intensively farmed landscape and whether this varies per species group. This information is essential if we want avoid situations where agri-environment schemes or other ecological restoration projects in the agricultural landscape fail to deliver biodiversity benefits because they are located beyond the dispersal range of target species. The dispersability of plants in agricultural landscapes is poorly studied, but the few available studies suggest a very low mobility (e.g. Geertsema & Sprangers 2002). Pollinators are obviously more mobile. Nevertheless, the maximum foraging distance of the relatively well-studied bees is generally less than 1 km (Greenleaf et al. 2007). Concerning pollination, Steffan-Dewenter & Tscharntke (1999) showed that seed set per plant was halved at 1000 and 250 m from a species-rich grassland for mustard Sinapis arvensis L. and radish Raphanus sativus L., respectively. Thus, most studies show that forbs and pollinators are generally mobile at relatively small spatial scales.

The aim of this study is twofold. First, we aim to estimate the distance at which species density and abundance of plants and pollinators in intensively farmed areas are enhanced by the presence of seminatural habitats. To this end, we examined the distribution of three species groups relative to seminatural habitats. These species groups, dicotyledonous plants (forbs), bees (Hymenoptera: Apiformes) and hover flies (Diptera: Syrphidae), have different mobilities and are linked to different ecosystem services (e.g. pollination, pest control). Simultaneously, we evaluated pollination success at increasing distances from the seminatural habitats. The examined seminatural habitats consisted of small- to medium-sized nature reserves (few to a few tens of hectares) containing mosaics of grassland, heathland, woodlots and occasional pools, thereby providing potential nesting sites and foraging habitat for a wide range of bee and hover fly species. Secondly, artificial flower-rich field margins are widely used agri-environment schemes in arable land and positive effects on pollinators have been observed (e.g. Pywell et al. 2005; Carvell et al. 2007). In contrast to pre-existing seminatural habitats, flower-rich margins are easy to establish and could be used as an alternative to seminatural habitats where these are missing. Newly created flower strips initially provide only additional resources, whereas pre-existing seminatural habitats also provide nest sites. It is unclear whether they have positive effects on biodiversity of pollinators beyond the flower strip itself. This study therefore additionally tests whether newly created flower-rich patches enhance pollinator abundance and species density in the neighbouring farmland. We established flower-rich patches in intensively farmed areas and compared the spatial distribution of hover flies and bees in areas with and without flower-rich patches.

We addressed the following specific research questions: (1) up to what distance do nature reserves enhance biodiversity (abundance and species density of forbs, bees and hover flies) and pollination rate into the surrounding countryside; (2) does this distance differ between species groups with different life history traits such as reproduction or dispersal type; and (3) do newly established flower-rich patches enhance diversity and abundance of flower-visiting insects in the nearby countryside in a similar way to nature reserves?

Methods

distribution and pollination patterns near seminatural habitats

Study area and design

The study was conducted in 2005 in the south-eastern part of the Netherlands (Provinces of Gelderland, Limburg and Noord-Brabant). To examine the dependence of farmland forbs and flower-visiting insects on the presence of seminatural habitats, we selected four nature reserves on glacially deposited sandy soils. All reserves formed isolated remnants of seminatural habitats in an intensively managed agricultural landscape. Agricultural fields in the study areas consist of crop monocultures that are generally inhospitable to wild plants and flower-visiting insects. Field boundaries generally consist of highly disturbed and species-poor seminatural vegetation but still support communities of flower-visiting insects (Kleijn & van Langevelde 2006). We therefore selected two straight ditches bordering and perpendicular to the edge of each of the four nature reserves. The ditches were over 300 m long (except one, of 275 m) and were bordered by various types of intensively managed agricultural fields (corn, barley, carrots, silage, pastures). To minimize the influence of external factors on insect density, we avoided the vicinity of farmyards and gardens by selecting ditches at a distance of at least 800 m from these elements. Ditches were about 1 m wide at the surface level. Ditch banks were cut once in late summer and although the vegetation is not fertilized directly, indirectly the vegetation growth is enhanced considerably by the presence of the bordering agricultural fields (Kleijn 1996). The selected ditches were dug 15–40 years before our observation and none were under agri-environment schemes. On one side of each ditch, we established a 300 m long (in one case 275 m) transect, starting at the border of the nature reserve.

Plant and insect surveys

Each transect was subdivided into 12 (in one case 11) plots of 25 m long and about 1 m wide (exact width depending on the width of the ditch bank but fixed within each transect). During the first half of July, we recorded all forb species in each plot. Insects were sampled three times during July and August by catching all bees and hover flies observed in each plot during 10 min. All specimens were killed quickly with ethyl acetate and brought to the laboratory for identification. Surveys took place before the vegetation was cut and between 1000 h and 1700 h on sunny days with little wind and a minimum temperature of 18 °C. Each transect was surveyed on a single day and under similar climatic conditions. In different rounds, plots were sampled in different orders and after each survey the flower abundance was estimated by counting the number of inflorescences of each forb species in each plot.

Pollination experiment

Radish Raphanus sativus was used as the experimental plant. This plant is known to be self-incompatible (Steffan-Dewenter & Tscharntke 1999) and pollinated by insects (Klotz, Kühn & Durka 2002). Its reproductive success can therefore be related directly to pollination by insects (Steffan-Dewenter & Tscharntke 1999). Plants were grown in a greenhouse between April and June. Between 23 and 29 June, potted plants were put into the field. Along each of the eight transects, potted plants were placed at 0, 10, 25, 50, 100, 200 and 300 m from the nature reserve with two pots (each containing two plants) per distance. Unfortunately, due to an exceptional period of rain at the beginning of July a large number of plants were flooded. Consequently, only 27 of the 56 patches of experimental plants survived. Because the loss was distributed evenly along the transects and at least four patches at each position within the transect remained (except for the position at 10 m from the reserve, with only two patches), analyses were less powerful but could nevertheless produce meaningful results.

Each experimental plant was collected immediately after senescence, between 29 July and 22 August. We counted the number of fruits and the number of flowers that did not develop into fruits (but leave distinct scars on the stems) 2 weeks after the pots were collected to allow all fruits to develop fully. The total number of flowers was calculated by summing the number of fruits and the number of flowers that did not develop into fruits. To analyse the seed set per fruit, 20 fruits per plant were taken randomly. Reproductive success was measured using the number of fruits per flower, the number of seeds per fruit and the number of seeds per flower.

distribution patterns near experimental flower patches

To examine whether the effects of newly created flower-rich habitats on pollinators extend into the agricultural matrix at a similar scale as pre-existing nature reserves, we enhanced the food supply experimentally for pollinators at five locations in intensively farmed areas. This study was conducted in 2004 in the centre of the Netherlands (Provinces of Gelderland and Utrecht). In each plot of 10 × 10 m, a range of insect-pollinated plant species with different flowering periods were transplanted (Campanula rapunculus L., Centaurea jacea L., Hypochaeris radicata L., Leucanthemum vulgare Lam., Picris hieracioides L., Prunella vulgaris L., Salvia pratensis L., Silene latifolia spp. alba (Mill.) Greuter & Burdet, Trifolium pratense L., Trifolium repens L.) or sown (Daucus carota L., Echium vulgare L., Lotus pedunculatus Cav., Malva moschata L., Pastinaca sativa L., Rhinanthus augustifolius Gmel., Trifolium incarnatum L.). The plots were fenced to protect the plants against cattle grazing. The effect of the artificial flower-rich patch on species number and abundance of hover flies and bees was measured along ditches at increasing distance from the patch (0, 50, 100, 150, 200, 300, 500, 800, 1100 and 1500 m) using window traps, yellow water pans (Duelli, Obrist & Schmatz 1999) and catching by sight (same approach as in the first part of this study).

The window traps and yellow water pans were opened for 2 days in June and 5 days in July. The first round of catching by sight was carried out in the beginning of August and the second at the beginning of September. Catches lasted 5 min and were executed along a transect 5 m long and 1 m wide at both sides of the trapping locations. At each of the four survey rounds, we counted the number of inflorescences of each forb species in a diameter of 6 m (approximately 100 m2) around the trapping location. Control transects were established in the same general area but at least 4 km away from the flower-rich patches. Species density and abundance of bees and hover flies at various distances from the flower-rich patches were compared with species density and abundance in transects without flower-rich patches (control treatment). The sampling protocol was identical over the 10 transects. Surveys on treatment and control transects within each region were collected on the same days.

analyses

We calculated the ‘flower abundance’ per plot by pooling the number of inflorescences of each species over the three or four survey rounds. Only plant species considered by Klotz et al. (2002) as insect-pollinated plants were taken into account. For the bees and hover flies we also pooled the number of caught individuals over the three or four survey rounds in transects next to nature reserves and flower-rich patches, respectively. The species density was determined as the total number of species caught in all survey rounds.

To determine whether the distribution patterns around nature reserves depended on species traits, we defined different functional groups. For plants we chose two traits linked to reproduction and dispersion: (1) pollination type (insect- or wind-pollinated, following Klotz et al. (2002)); and (2) seed dispersion type (anemochorous, zoochorous, hydrochorous or unspecialized, following Julve (1998)). For hover flies, again two traits were chosen: (1) type of main host plant (grass or forb, following van Veen (2004)); and (2) larvae type (aphidophagous or detrivorous, the latter consisting mainly of aquatic larvae, semiaquatic larvae and those occurring in liquefied faeces, following van Veen (2004)). Finally, for bees we distinguished only between bumblebee species and all bees. Because of the low number of observed bee species (see Results), no other subgroups could be defined.

Statistical analyses were performed using r version 2·4·1 (R Development Core Team 2006). Regression analysis was used to measure the effect of distance to the nature reserve on species diversity and abundance. Depending on the distribution of the dependent variables, multiple linear models or generalized linear models (GLMs) employing the Poisson distribution were used. When necessary, overdispersion in GLMs was accounted for by inflating the variance of the Poisson distribution with a constant factor (Breslow 1984). For linear models, logarithmic or square-root transformations were used to achieve normal distribution of the dependent variables. The models included site (i.e. nature reserves, random factor variable), transect nested within site (random factor variable) and distance to the nature reserve (continuous variable). For analyses concerning insects, flower abundance (continuous variable) was added to the model. Flower abundance and distance to the nature reserve were only slightly correlated (Pearson's coefficient = –0·17; P = 0·1; d.f. = 94), thus severe collinearity problems in the regression models are not expected.

To estimate up to what distance the response variables were affected by the presence of the nature reserve we analysed subsets of data. The first subset was constructed by removing the observations in plots bordering the nature reserve (0–25 m); and the second by removing observations of the first two plots (0–25 and 25–50 m). The next subsets were then selected sequentially to finally leave only values measured further than 200 m (leaving one-third of the original data set). This last subset, consisting of four subplots, was considered to be the smallest subset to have sufficient statistical power (21 residual degrees of freedom should facilitate reliable statistical testing of the examined relationship). For each of the eight selected subtransects we tested whether the slope of the factor distance in the regression model still differed significantly from zero using the same approach as for the full transect (obviously, the analyses on these subtransects should not be considered as independent tests and this way of testing may result in inflated Type I errors). The shortest subtransect at which the slope of distance was statistically significant was assumed to cover the maximum distance at which the influence of the nature reserves could be measured.

The impact of flower-rich patches on the abundance or species density of both insect groups was evaluated for each distance separately, with GLMs employing the Poisson distribution. For each analysis we used the value measured at one distance from the flower-rich patches (n = 1 × 5 locations) and the values of the control (n = 10 × 5 locations). The model included region (random factor variable), flower abundance (continuous variable) and the treatment (presence or absence of flower-rich patch in the vicinity). For the flower abundance, only plant species considered by Klotz et al. (2002) as insect-pollinated plants were taken into account. Furthermore, for measures on the flower rich patches, flowers counted on transplanted or sown plants were also excluded (they are included in the treatment effect).

Results

patterns near seminatural habitats

A total of 97 forb species was observed. Among them, 47 insect-pollinated plant species were observed in flower. A total of 1258 hover fly individuals of 34 species were caught. Finally, 161 individuals belonging to 11 different species of bees were observed. Sixty-four per cent of the caught individuals were Bombus species and 15% were Apis mellifera L., which are managed in the Netherlands. Only 34 individuals were solitary bees and half of them were caught on the same transect.

Forb species density declined significantly with increasing distance from the nature reserves (‘distance’ in the following; Table 1). However, most of the decline took place in the first 75 m (Fig. 1a, Table 2). Further from the reserves no obvious trend in forb species density could be observed and the species richness was about 65% of that observed at the edge of the nature reserve. Moreover, the amount of variance explained by distance was relatively small (R2 = 0·19; Table 1). Along the transect, plots had between 55 and 60% (mean = 57%, SE = 0·9) of the observed species in common with the plot bordering the reserves (Fig. 1a), indicating that species composition did not change considerably with increasing distance. The species density of forbs belonging to different functional groups all declined with increasing distance (Table 1). However, species with zoochorous seed dispersal were related most strongly to distance, while the declines of species with hydrochorous and anemochorous seed dispersal were not statistically significant. When considering inflorescences, we see that both abundance and species richness showed a slight but non-significant decrease with increasing distance from the nature reserves (Figs 1b and 2a, Table 1).

Table 1.  Results of regression analyses showing the influence of flower abundance and distance to nature reserves on different attributes of forb species density, on forb inflorescences and on different insect pollinator groups. The models include sites, transects within sites, flower abundance (only for insects) and distance to nature reserves. Flower abundance only includes insect-pollinated plants [following Klotz et al. (2002)] and was log-transformed prior to the regressions. Distances were included in the model in hectometres. Only coefficients for flower abundance (FA) and for distance to the nature reserves (Distance) are presented. R2 for FA and distance represent the proportion of variation explained by, respectively, flower abundance and distance which is not associated with sites and transects (partial r2 following Draper & Smith 1981: 265). For GLMs, the McFadden pseudo-R2 is given only for the full model (partial r2 cannot be rigorously calculated for this model type). Nsp: number of species for each attribute; Trans: transformation used (sqrt: square-root; log: log(y + 1)); Model: type of model (normal: linear regression, Poisson: Poisson regression with log link). *P < 0·05, **P < 0·01, ***P < 0·001
 NspTrans.ModelCoefficientsR2
FADistanceFull modelFADistance
Forb species density
 All forbs97sqrtNormal–0·220***0·620·19
 Pollination type        
  Insect pollinated62sqrtNormal–0·144**0·620·11
  Wind pollinated13Normal–0·365***0·650·16
 Seed dispersal type
  Zoochorous16logNormal–0·148***0·490·20
  Hydrochorous17logNormal–0·0590·560·02
  Anemochorous31sqrtNormal–0·0830·530·04
  Unspecialized seeds dispersal28sqrtNormal–0·126*0·550·06
Forb inflorescences
 Flower abundance47logNormal–0·2870·430·04
 Flower species richness47sqrtNormal–0·0350·470·01
Pollinator species density
 Hover flies34sqrtNormal0·275***–0·229***0·610·360·18
 Grass feeding hover flies 6Poisson0·077–0·1820·26
 Forb feeding hover flies28sqrtNormal0·284***–0·222***0·610·350·16
 Hover flies with detrivorous larvae15Poisson0·334***–0·1460·59
 Hover flies with aphidivorous larvae16sqrtNormal0·159***–0·162**0·420·170·11
 Bees11Poisson0·372**–0·2110·43
 Bumblebees 4Poisson0·295*–0·0450·35
Number of pollinator individuals
 Hover flies34logNormal0·345***–0·276***0·720·430·20
 Grass feeding hover flies 6Poisson0·053–0·306**0·35
 Forb feeding hover flies28logNormal0·385***–0·261***0·700·430·16
 Hover flies with detrivorous larvae15Poisson0·587***–0·370***0·64
 Hover flies with aphidivorous larvae16logNormal0·215***–0·195**0·510·220·12
 Bees11Poisson0·513***–0·435**0·55
 Bumblebees 4Poisson0·405**–0·2300·43
Frequent hover flies species (> 50%)
 Episyrphus baltheatusPoisson0·409***–0·1060·42
 Sphaerophoria scriptaPoisson0·515***–0·0650·44
 Melanostoma mellinumPoisson0·022–0·343*0·30
Figure 1.

Relationship between the distance to nature reserves and species density (means ± SE) of (a) all forbs, (b) forbs observed flowering at least once, (c) hover flies and (d) bees. Points represent centres of subtransects of 25 m (e.g. point of subtransect 0–25 m is at 12·5 m). Continuous lines: all species; dashed lines: species shared with the first plot (0–25 m).

Table 2.  Estimating the limits of the influence of nature reserves on species diversity and abundance of the three species groups. The effects of nature reserves were analysed in subtransects perpendicular to reserves and starting at increasing distances from reserves (see details in the text). For each multiple linear regression, only the coefficient describing the effect of distance is presented. Therefore, each value of the table resulted from a different multiple linear regression. Statistical models and transformations are identical to those presented in Table 1. *P < 0·05, **P < 0·01, ***P < 0·001
 Distances relative to the nature reserve covered by the subtransects (m)
0–30025–30050–30075–300100–300125–300150–300175–300200–300
Species density
 All forbs–0·220***–0·147**–0·122*–0·075–0·011–0·003 0·040 0·097 0·219
 Hover flies–0·229***–0·240***–0·260**–0·267**–0·323**–0·247–0·157–0·091–0·116
 Bees–0·211–0·126–0·235–0·331–0·189–0·456–0·317–0·606–0·841
Number of individuals
 Hover flies–0·276***–0·280***–0·282**–0·275**–0·290*–0·219–0·173–0·067–0·071
 Bees–0·435**–0·163–0·317*–0·423*–0·403–0·366–0·118–0·295–1·103
Figure 2.

Relationship between the distance to nature reserves and abundance (means ± SE) of (a) flower abundance (sum of all the inflorescences of the insect-pollinated plant species), (b) hover flies and (c) bees. Points represent centres of subtransects of 25 m (e.g. point of subtransect 0–25 m is at 12·5 m).

The number of species and individuals of hover flies were related highly significantly to flower abundance and distance (Table 1). The amount of variation explained by distance was similar to that found for forbs (R2 = 0·18; Table 1). The relation with distance was determined mainly by the decline in the first 125 m for the species density and the number of individuals (Table 2). At the larger distances, species richness and abundance remained constant at a level of approximately 55% and 40% of that observed at the edge of the nature reserve (Figs 1c and 2b). Along the transect, on average 55% (SE = 2·8) of the observed species were also observed in the first plot (0–25 m, Fig. 1c). Hover flies feeding mainly on grasses were not related significantly to flower abundance but a significant relation with distance was observed for the number of individuals (Table 1). Of the three most frequently observed hover fly species, Episyrphus baltheatus DeGeer and Sphaerophoria scripta L. were related strongly positively to flower abundance but not to distance (Table 1). In contrast, Melanostoma mellinum L., a species feeding mainly on grass pollen, was not related to flower abundance but related negatively to distance (Table 1).

Bee species density showed no significant relation with distance (Table 1). In contrast, bee abundance was related positively to flower abundance and negatively to distance. Except for the subtransect including the distance from 25 to 300 m (Table 2), we observed a significant negative slope for distances up to 100 m. The absence of a significant slope for distance for the first subtransect is due to the clear drop we observed at 25–50 m (Fig. 2c). Additionally, we observed many more species and individuals (Figs 1d and 2c) at the edge of the reserve (0–25 m) compared to all other distances relative to the reserve. Bumblebees, on the other hand, were not related to distance to nature reserves (Table 1). Of the three species groups, bee species composition showed the greatest change along the transect with, on average, only 44% (SE = 3·5) of the observed species also present in the first plot (0–25 m) (Fig. 1c).

The relation between the reproductive success of the experimental R. sativus plants and distance to nature reserves was not as clear-cut. No significant relation was observed between fruit set and distance. The negative relation between the number of seeds per fruit and distance was only marginally significant (linear regression with square-root transformation of the dependent variable: coefficient = –0·107; t20 = –1·97; P = 0·06), but the number of seeds per flower was related significantly and negatively to distance (linear regression with square-root transformation of the dependent variable: coefficient = –0·110; t20 = –2·13; P = 0·04). In this case, we observed a drop at 25 m and, after 100 m, a decrease for this variable up to 300 m (Fig. 3). A linear model explaining seeds per flower with transect effect and bumblebee abundance at the experimental plants plots showed a significant positive effect of the bumblebee abundance (linear regression with square-root transformation of the dependent variable: coefficient = 0·098; t20 = 3·65, P = 0·002). This effect was not observed for bee and hover fly abundance.

Figure 3.

Relationship between the number of seeds per radish flower (R. sativus, mean ± SE) and the distance to the nature reserve.

patterns near flower-rich patches

A total of exactly 800 hover fly individuals of 30 species were caught. For bees, 290 individuals of 19 species were observed. Flower abundance in the flower-rich patches (mean = 4647, SE = 832) was, on average, about 10 times higher than the abundance observed normally in the studied farmland areas (mean = 478, SE = 65). Both species density and abundance of bees and hover flies were enhanced significantly by the flower-rich patches (0 m) compared to the controls (Fig. 4). Furthermore, abundance of hover flies was enhanced at 50 m from the flower-rich patch. For species density, the effect was only marginally significant (Z-value = 1·65, d.f. = 1, P = 0·10). Bees were not affected significantly outside the flower-rich patch. In fact, in contrast to the hover flies, the lowest values for species density and abundance were observed at 50 m from the flower-rich patch.

Figure 4.

Relationship between species density (a: hover flies, b: bees) and number of individuals (c: hover flies, d: bees) and the distance from the artificial flower rich patch. In grey: mean (continuous line) ± standard error (broken line) density of species or number of individuals in a landscape with no impact of an artificial flower rich patch (control, n = 50). The tests are performed for each distance separately (n = 5) against the control (n = 50) by taking the region and the number of inflorescences of insect-pollinated plants into account (model: region + number of wild flowers + treatment). The indicated P-values are those of the treatment effect. *P < 0·05, ***P < 0·001.

Discussion

up to what distance do nature reserves enhance biodiversity and pollination services?

Forbs

Nature reserves had a distinct positive effect on the species density of forbs in the surrounding agricultural landscape. The species number declined by about one-third over a distance of 300 m and most of this decline occurred in the first 75–100 m. Species present far from the reserve were largely a subset of the species present next to the reserve. In nutrient-rich habitats, extinction rate is an important driver of plant species diversity (Tilman 1993; Blomqvist et al. 2003b). Maintenance of diversity requires continuous colonization and our results suggest that beyond 75–100 m colonization of species from the nature reserves can no longer compensate extinction. Here, only species well adapted to the intensive management practices in the agricultural landscape survive.

Bees and hover flies

Diversity and abundance of flower-visiting insects were related strongly to flower abundance, which is in agreement with previous findings (e.g. MacLeod 1999; Pywell et al. 2005; Kleijn & van Langevelde 2006). Species number and abundance of hover flies and abundance of bees were all related negatively and significantly to distance to the nature reserves, suggesting that the observed pattern was caused at least partly by characteristics of the reserves themselves.

The 45% and 60% decrease that we observed over a distance of just 300 m for the number of hover fly species and individuals, respectively, is in sharp contrast to the results of Steffan-Dewenter & Tscharntke (1999). They found no relation between the number of hover flies visiting experimental plants and the distance to the nearest grassland over a spatial range of 1 km. On the other hand, by using pollen of Phacelia tanacetifolia Benth. as a marker, Wratten et al. (2003) observed hover flies up to just 200 m from the flower source in an agricultural landscape, which is in agreement with our findings.

On one hand, bees responded in a similar manner to hover flies, with a decline of about 70–80% in species density and abundance between the edge of the nature reserve and 300 m further in the agricultural landscape. On the other hand, we observed a distinct drop in bee abundance at 25–50 m beyond the reserve. The lack of favourable nesting sites in the intensively managed landscape (Westrich 1996) constrains bees to nest in seminatural habitat. This could explain the higher quantities of bees just next to the nature reserves in the direct vicinity of their nests. Moreover, the peculiar drop in bee species richness and especially abundance at 25–50 m from the reserves could be explained as a foraging response. Contrary to hover flies, which have a ubiquitous foraging behaviour (Pontin et al. 2006), bees are known to concentrate at and return to resource-rich patches on successive foraging trips (Osborne et al. 1999). However, resource richness is a relative measure and depends on the local context. Flower patches in the direct vicinity of flower-rich reserves may be perceived by bees as of lower quality than similar patches further away, where direct comparison with the nature reserve is not possible. Bees that are within 50 m from a nature reserve may therefore ignore the scattered flowers in the agricultural landscape and instead move directly to the larger quantity in the reserve. This hypothesis is corroborated by the distribution pattern of bees around the flower-rich patches and the relation between bumblebee abundance and the number of seeds per flower observed on our experimental plants. For the flower-rich patches, only abundance of floral resources was manipulated in the year of investigation. Nevertheless, the patterns are remarkably similar (compare Figs 1d, 4b, and 2c, 3, 4d).

Pollination services

The observed pattern of pollination success in relation to distance to the nature reserves was less clear than the patterns observed for forbs and flower-visiting insects. Behavioural responses of pollinators may have confounded the effects of nature reserves. Pollination efficiency can vary depending on pollinator taxa (Conner, Davis & Rush 1995). Bumblebee abundance, but not hover fly abundance, was related strongly to pollination success of R. sativus plants and was apparently a much better predictor of pollination success than distance. Bumblebee abundance did not decline significantly with increasing distance from reserves. Furthermore, the decrease in bee abundance near resource-rich patches showed a typical dip just outside the patches, possibly obscuring the linear relationships between pollination success and distance. Nevertheless, the seed number per flower at 300 m was about a third of the observed mean number in the first 100 m. This agrees with findings of Steffan-Dewenter & Tscharntke (1999), that the mean seed set per R. sativus plant was halved at a distance of 250 m from the nearest seminatural habitat.

do the relationships differ between species groups with different life history traits?

Species groups with different life history traits demonstrated different relationships with distance to nature reserves. Traits related to mobility seemed to be most important in explaining differences between species. The most obvious example is that of the bumblebees, which showed no relation at all with distance over the studied range, whereas most other examined insect taxa or functional groups did. Steffan-Dewenter & Tscharntke (1999) observed only a slight decline of the highly mobile bumblebees, with increasing distance from natural habitats up to 1000 m. Another example that traits related to mobility seemed important is that the positive effect of nature reserves extended over greater distances for hover flies and bees than for sedentary plants. Even within plants this pattern holds if we compare the dissemination strategy. Colonization of plants is determined largely by seed dispersal (Levin et al. 2003). Our results indicate that in agricultural landscapes the occurrence of zoochorous species is more dependent upon the presence of nearby high-quality habitats than it is for anemochorous species. This corroborates the results of Devlaeminck, Bossuyt & Hermy (2005), who found that in the seed bank of arable fields zoochorous, but not anemochorous, species declined with distance to forest edges.

do temporary resource-rich patches enhance pollinators on nearby farmland?

Significantly more species and individuals of bees and hover flies were observed in the experimentally established flower-rich patches, which is in agreement with previous findings (Sutherland, Sullivan & Poppy 2001; Pywell et al. 2005; Carvell et al. 2007). The differences in pollinator density and abundance induced by the establishment of flower patches were in the same order of magnitude as those observed near reserves (60–80%). Nevertheless, there were no positive effects of the flower-rich patches on bees in the surrounding agricultural landscape and they may even have been affected negatively in the direct vicinity of the patches (Fig. 4b,d; see discussion in the previous section). Hover fly abundance was affected positively by flower-rich patches at distances of up to 50 m. There are a number of explanations for the differences in distribution patterns near nature reserves and flower patches. Bees may nest in high densities in reserves but use both reserves and farmland for foraging. This would result in declining densities from reserves but not from patches, as these did not provide nest sites. Hover flies are not tied to nest sites and are known to aggregate in patches with higher numbers of flowers (Sutherland et al. 2001). Additionally, well-fed individuals of several species groups live longer (e.g. Murphy, Launer & Ehrlich 1983; Smeets & Duchateau 2003). Both processes would lead to higher densities in and around any resource-rich habitat, be it a reserve or a flower patch. The effects would probably be stronger near reserves because these provide habitats for the larval stages of hover flies in addition to food. Mark–recapture studies are required to estimate the relative importance of these two mechanisms. Alternatively, the lack of effects we observed could also be explained by the relatively small size (100 m2) or young age of the flower patches we introduced.

implications for management

Our results highlight the importance of high-quality (semi)natural habitats for the diversity of forbs and flower visiting insects in the neighbouring intensively managed agricultural landscape. When aiming to conserve biodiversity on farmland, priority should therefore be given to conserve and maintain such fragments of species rich habitats (Öckinger & Smith 2007). An alternative approach to enhance biodiversity and ecosystem services in intensively farmed agricultural areas, the temporary or permanent establishment of high-quality habitats on former agricultural land does enhance pollinator diversity, but short-term effects are restricted mainly to the habitats themselves. It is worthwhile investigating whether the effects of establishing these habitats on neighbouring farmland increase with age or habitat size, and it emphasizes the important role of permanent vegetation types for pollinator diversity.

The intensity of farming in the Netherlands is very high and matched only in certain areas in the United Kingdom, Belgium, France and Germany. The spatial effects of seminatural habitats may depend on the suitability of the neighbouring agricultural landscape. Positive effects of (semi)natural habitats may be stronger and extend further into areas that are less inhospitable to forbs and flower-visiting insects. In the Netherlands, however, positive effects of nature reserves were relatively small and limited to distances of less than 150 m. Here, ecological restoration sites or fields with agri-environment schemes should be implemented preferentially in very close proximity to these habitats to avoid conservation efforts becoming constrained by colonization limitation.

Acknowledgements

We would like to thank the farmers, water boards and nature reserve managers for granting access to their land. The assistance of Mathilde Barreau, Frans Moller, Maurits Gleichman and Jan van Walsem in the field and the help of Nico de With and Ivo Raemakers for insect identification is greatly appreciated. We also thank Taylor Ricketts, Rachael Winfree, Wouter van Steenis and one anonymous reviewer for helpful comments on previous versions of the manuscript. This work was funded by the EU Project QLK5-CT-2002-1495, ‘Evaluating current European Agri-environment Schemes to Quantify and Improve Nature Conservation Efforts in Agricultural Landscapes’ (EASY) and for F. Kohler by WIMEK (Wageningen Institute for Environment and Climate Research) and the Swiss National Science Foundation (grant PBNEA-1102303).

Ancillary