Introduction

Diversification of agro-ecosystems is often advocated as a tool to provide essential nectar and pollen resources to parasitoids and predators and thereby enhance the efficacy of biological pest control (Gurr et al. 2005; Heimpel and Jervis 2005). Under laboratory conditions, nectar supply often has a great impact on longevity (Zoebelein 1955; van Lenteren et al. 1987; Idris and Grafius 1995; Wäckers 2001), fecundity (Idris and Grafius 1995; McEwen and Liber 1995; Baggen and Gurr 1998; Winkler et al. 2006) and activity of parasitic wasps (Takasu and Lewis 1994; Wäckers 1994). Under field conditions, the presence of flowers may (locally) increase the abundance of beneficial insects (White et al. 1995; Patt et al. 1997; Bianchi and Wäckers 2008) and rates of parasitism (Stephens et al. 1998; Lee and Heimpel 2008).

While conservation biological control programs have largely focused on nectar and pollen use by beneficial insects, the fact that many herbivores also feed on these floral resources has received relatively little attention (McEwen and Liber 1995; Baggen and Gurr 1998; Romeis et al. 2005; Wäckers et al. 2007). For a number of natural systems it has been shown that oviposition by herbivores can be increased in the vicinity of adult food plants (Grossmueller and Lederhouse 1987; Murphy 1983; Murphy et al. 1984; Karban 1997). Also in agricultural systems habitat diversification may result in elevated pest numbers (Latheef and Irwin 1979; Andow 1991; Zhao et al. 1992; Romeis et al. 2005; Bianchi et al. 2006).

Whereas nectar and pollen feeding has often been implicated as the underlying mechanism for these adverse effects, the experiments were usually not laid out to establish causality. Instead, experiments frequently addressed other questions like the impact of biodiversity (Andow 1991; Bianchi et al. 2006) or repellence by companion plants (Latheef and Irwin 1979). Zhao et al. (1992) reported that Pieris rapae L. (Lepidoptera: Pieridae) eggs and larvae were more abundant in crops interplanted with a mix of nectar plants, as compared to monocultures, but did not underpin this with statistics. Luna (personal comm.) found a higher abundance of P. rapae larvae in broccoli due to Fagopyrum esculentum Moench (Caryophyllales: Polygonaceae) borders, while Lee and Heimpel (2005) found no increase in P. rapae densities in cabbage adjacent to F. esculentum. In order to single out nectar feeding as a potential mechanism underlying increased herbivore levels, Winkler et al. (2009b) studied the impact of flowering margins on herbivore and parasitoid carbohydrate reserves. Using HPLC sugar analyses, it was shown that the average overall sugar content of Plutella xylostella L. (Lepidoptera: Plutellidae) and Diadegma semiclausum Hellén (Hymenoptera: Ichneumonidae) collected in fields bordered by Anethum graveolens L. (Apiales: Apiaceae) and Lobularia maritima (L.) Desv. (Brassicales: Brassicaceae) was significantly higher than those of individuals collected from Lolium perenne-bordered control plots.

Here we study whether floral resources affect herbivore and parasitoid spatial dynamics with an experimental design based on our previous work in which we established visitation rates and survival of Brassica herbivores and/or their natural enemies on a range of flowering herbs. Field observations showed clear differences in flower visitation (Winkler et al. 2005a) and laboratory studies indicated that nectar plants are not equally suitable in providing accessible nectar and prolonging lifespan of cabbage herbivores and their parasitoids (Winkler et al. 2009a). These results were used to identify and test a parasitoid selective/herbivore selective scenario which is unique in allowing us to address both benefits of flowering vegetation to biological control agents as well as risks in enhancing pest pressure.

The “best-case scenario” of our study consisted of a Brussels sprout plot bordered by A. graveolens. This species provides accessible nectar for the parasitoids Cotesia glomerata L. (Hymenoptera: Braconidae) and D. semiclausum, while being unsuitable for the herbivore P. rapae. The “worst-case scenario” consisted of a Brussels sprout plot bordered by Centaurea jacea L. (Asterales: Asteraceae). This species is an attractive and suitable nectar source for P. rapae, while being unsuitable for the parasitoids. In addition, we also tested the plant species F. esculentum, which takes an intermediate position providing accessible nectar for both P. rapae and the parasitoids. All three plant species provide accessible nectar for P. xylostella, the host of D. semiclausum. As a control we used field margins with the grass Lolium perenne L. (Poales: Poacea), which does not provide nectar.

Materials and methods

Experimental setup

Small fields (8 × 10 m in size) with Brussels sprouts (Brassica oleracea var. gemmiferra cv. Maximus) were established at two experimental locations, Wageningen Hoog (WH) in 2002 and Achterberg (AB) in 2002 and 2003. These locations are situated near Wageningen, The Netherlands. Brussels sprouts were transplanted as seedlings to the field in week 21 in 2002 and week 22 in 2003. Planting distance was 80 cm between the rows and 50 cm within the rows. The distance between the plots was 35 m.

Along the two long sides of each Brussels sprout plot a 1 m wide field margin was established. Field margins contained one of the following plants: L. perenne (grass, control), C. jacea, A. graveolens, or F. esculentum (Table 1). Each treatment was replicated twice in each site. In 2002, C. jacea was sown in the greenhouse and planted into the field in week 18. L. perenne (week 20), A. graveolens (week 17 in WH and week 18 in AB), and F. esculentum (week 20) were sown directly into the field. A. graveolens was resown in week 20 to ensure good establishment of the plant. This resulted in flowering periods as indicated in Fig. 1.

Table 1 Plant species used, their floral colour, nectar position and nectar provision for Pieris rapae and parasitoids
Fig. 1
figure 1

Flowering and sampling periods in the two experimental years

In 2003, L. perenne and F. esculentum were sown directly into the field in week 20. A. graveolens was sown in the greenhouse and planted into the field in week 22. As C. jacea is a perennial plant it was not replanted in the second year.

Plots and field margins were hand-weeded. The area between the plots was sown with L. perenne, which was mown regularly during the season.

Monitoring adult insects

In 2003, we monitored naturally occurring adult P. rapae, P. xylostella and D. semiclausum both within the field margins and in row 3–4 and row 7–8 of each Brussels sprout plot. To count the easily visible P. rapae, we walked once around each field margin as well as through the Brussels sprout rows and recorded any visible adults. Small species like P. xylostella and D. semiclausum were sampled by sweep netting and checked for their sex. A standardized number of 12 sweeps was taken in each field margin as well as in row 3–4 and row 7–8 of each Brussels sprout plot. Sampling was done only in 2003 from week 29 until week 32 on sunny and dry days between 10:00 h and 15:00 h. Between 15 and 19 sweep net samples were taken per treatment.

Monitoring eggs and larvae

Brussels sprout plants were randomly monitored in a non-destructive way for P. rapae eggs and larvae and for P. xylostella larvae, pupae and cocoons (parasitized pupae). Monitoring was done on a weekly basis. In 2002, sampling took place from week 26 until week 36. Sampling started as soon as the first flower species started blooming and lasted until two weeks after the first flower species had terminated blooming. Sampling was done earlier in 2003, from week 24 until week 34, due to exceptionally warm weather and the resulting faster development of the crop. With increasing plant size the number of monitored plants per field decreased. In 2002, 20 plants per field were sampled in week 26–28, 15 plants in week 29–30 and 10 plants in week 31–36. In 2003, 15 plants per field were sampled in week 24–25 and 10 plants in week 26–34.

Data evaluation

Adult insects

Per species we totalled the number of observed individuals for each plot and sample date to attain one figure for the field margins and one for the crop. To check for differences among treatments the data were totalled over sampling days and experimental locations. As data were not normally distributed, non-parametric statistics were used. When the Kruskal–Wallis test indicated differences among treatments, the Mann–Whitney U test was used to compare the control treatment with each of the other treatments (α = 0.05).

For each sampling event we calculated, which proportion of the total number of P. rapae adults was observed in the edge and in the field. These proportions of individuals observed in margins versus crop within the same plot were arcsine-root-transformed and subjected to a General Linear Model procedure (SAS statistical software, version 8e, Cary, NC, USA). Contrasts between treatments and control were tested for significance at α = 0.05.

Eggs and larvae

The average number of eggs and larvae per plant was calculated per field and week. For each field we subsequently calculated the cumulative average number of eggs and larvae over the whole experimental period (integral of the curve).

Each field margin treatment had six replicates when pooling together both locations and years. To check for the effect of treatment, data were analysed using a General Linear Model and consequently tested for contrasts of the treatments with the control (SAS 8e) (α = 0.05). As C. jacea was poorly established in 2002 and primarily flowered in 2003, analysis was also done for 2003 data only.

We calculated the rate of parasitism by D. semiclausum as the proportion of pupae of P. xylostella containing a parasitoid cocoon out of the total number of pupae. This may underestimate the rate of parasitism, as cocoons, in which parasitoid larvae had not yet emerged and pupated, were counted as unparasitized. In addition, pre-pupal mortality of parasitoids was ignored. Average of parasitism per year and treatment were calculated from the rates of parasitism per plot and week.

Results

Presence of adult insects

We monitored visually a total of 54 P. rapae individuals. Numbers of adult P. rapae in the margins and/or in the crop did not significantly differ between treatments. However, when the crop was bordered by C. jacea, the proportion of adults recorded in the crop and in the margin was significantly different from the proportions of adults observed in the control (Table 2, Fig. 2a). Numbers of P. xylostella were too low to allow statistics (4 females and 17 males).

Table 2 P-values obtained from statistical tests used for adult and offspring data
Fig. 2
figure 2

Backtransformed average arcsine-root transformed proportions of a Pieris spp. adults in crop (light grey) and flower margins (dark grey), and average number per Brussels sprout plant of b Pieris rapae eggs and c P. rapae larvae for the four different treatments. Data were pooled over years and locations. An asterisk indicates significant differences at α = 0.05 in the proportions of Pieris spp. adults and in the average number of Pieris rapae eggs and larvae between the control and any treatment. Error bars indicate SE values. Acc accumulated

For D. semiclausum, we collected a total of 136 females and 469 males by sweep netting. Significantly more D. semiclausum females and males were collected within the Brussels sprout fields bordered by A. graveolens as compared to the control. The number of D. semiclausum males was highest in fields bordered by F. esculentum. Numbers of D. semiclausum females and males observed in the field margins were low (Table 2, Fig. 3b, c). In C. jacea margins we found significantly more male D. semiclausum than in Control margins. P-values for all statistical tests performed on adult data are listed in Table 2.

Fig. 3
figure 3

Average number of a P. xylostella caterpillars per Brussels sprout plant for the four different treatments and average number of b Diadegma semiclausum females and c males in crop (light grey) and edges (dark grey). Data were pooled over years and locations. Significant differences in the crop at α = 0.05 between the control and any treatment are indicated by an asterisk. Error bars indicate SE values

Presence of eggs and larvae

Overall, there were significantly higher numbers of P. rapae eggs per Brussels sprout plant in the field bordered by F. esculentum as compared to the control (Fig. 2b). In addition, the overall number of P. rapae larvae per Brussels sprout plant were higher in plots bordered by C. jacea and F. esculentum as compared to the control (Fig. 2b, c). When considering 2003 only, Brussels sprout plants bordered by C. jacea suffered higher P. rapae eggs and larvae loads as compared to the control.

We did not find significant differences in the number of P. xylostella larvae per Brussels sprout plant among treatments (Fig. 3a). The percentages parasitism of P. xylostella by the parasitoid D. semiclausum did not differ among treatments and reached on average 78 ± 3% in the year 2002 and 67 ± 3% in the year 2003. For P-values of all comparisons made see Table 2.

Discussion

Our study demonstrates that floral resources can have a positive impact both on herbivore and parasitoid numbers in the field and shows unambiguously that this impact is insect- and flower-species specific. This finding underlines that the indiscriminate use of nectar plants in flowering field margins can enhance pest pressure and emphasizes the importance of floral composition in the optimization of ecosystem services.

The herbivore P. rapae showed higher egg and larval densities in fields bordered by C. jacea and F. esculentum, two flower species that had previously been shown to be suitable nectar sources for this herbivore (Winkler et al. 2009a). Moreover, adults of this species were recorded more frequently in plots bordered by C. jacea. The flower species that had been shown to be unsuitable as nectar source for P. rapae in our previous laboratory studies did not result in elevated pest numbers. The fact that P. rapae numbers are only enhanced by those plant species that had been selected on the basis of their suitability as nectar source for P. rapae supports the conclusion that nectar foraging and/or nectar feeding is the mechanism responsible for the enhanced herbivore pressure.

When the Brassica crop was bordered by C. jacea, the proportion of adult P. rapae observed in the margins was significantly higher as compared to the margin of the control treatment planted with L. perenne. This finding fits well with our prediction that C. jacea is especially attractive to Pieris sp. Winkler et al. (2005a) found significantly higher numbers of Pieris sp. visiting C. jacea and Oregano vulgare in comparison to F. esculentum and five other flowering plants when using 3 × 3 m single species plots.

For the pooled data over years there was a significant effect of F. esculentum on the number of P. rapae eggs. C. jacea had a significant effect on the number of P. rapae eggs in 2003 only. This may be explained by the fact that in 2002, C. jacea was newly established and started flowering four weeks later as compared to 2003 (see Fig. 1). The oviposition behaviour of P. rapae is characterized by laying eggs singly on plants, followed by a linear flight pattern during which the butterfly passes over many suitable hosts without laying eggs (Root and Kareiva 1984; Yamamura 1999). In our experiments P. rapae females may have left a plot after two to three ovipositions. This means that in large commercial fields the impact of a field margin with suitable flowers may be at least equally marked.

Unlike the clear results with P. rapae we did not see any significant impact of flower margins on P. xylostella. Even though all of the flowers tested extended P. xylostella longevity under laboratory conditions, neither adults nor larvae showed increased numbers in plots with flower margins, relative to the control. The relatively low numbers of P. xylostella adults collected in this study is probably due to the fact that this species is primarily active at dusk (Harcourt 1954). Using sticky traps, Bukovinszky et al. (2003) found that field margins dominated by flowering Sinapis alba, increased pest densities in adjacent fields. De Groot et al. (2005) found significantly more P. xylostella individuals on sticky traps in plots planted with Lobularia maritima than in Brussels sprout and S. alba plots. Both plant species, L. maritima and S. alba, are host plants for P. xylostella, which means that these effects need not have been food-mediated.

When host plants are abundant, P. xylostella does not fly far between ovipositions, but shows a strong arrestment response and stays on the plant or moves to directly neighbouring plants (Justus and Mitchell 1996; Bukovinszky et al. 2005). In contrast to P. rapae, P. xylostella lays more than one egg on a host plant. Food supply (0.5 M sucrose solution) increased longevity by a factor of 3 (Winkler et al. 2005b). Lifespan of P. xylostella was prolonged when given access to C. jacea, A. graveolens and F. esculentum under laboratory conditions (Winkler et al. 2009a). Previously we were able to demonstrate that P. xylostella exploits nectar sources under field conditions (Winkler et al., 2009b). The lack of significant differences among the treatments in this study leads us to conclude that food supply was not the limiting factor for the realized fecundity by P. xylostella. This assumption is supported by observations on the lifetime oviposition curve of P. xylostella. P. xylostella lays 90% of the total number of eggs within the first four days (Harcourt 1954; Cardona 1997). Under laboratory conditions, P. xylostella lives on average five days without food (Winkler et al. 2009a). As P. xylostella shows far less flight activity as compared to P. rapae, their energetic needs are likely relatively low. This may explain the limited impact of nectar availability in this species. In addition, P. xylostella might have utilized other food sources in the field like honeydew, produced by the cabbage aphid Brevicoryne brassicae or by the cabbage whitefly A. proletella. These species were observed in (very) low densities throughout the season in all of the four treatments.

We expected higher parasitoid numbers and, consequently, increased rates of parasitism in plots neighboured by A. graveolens and F. esculentum. Laboratory as well as semi-field experiments demonstrated the positive impact of F. esculentum and A. graveolens on the longevity of D. semiclausum, as well as a substantial impact of F. esculentum on fecundity of D. semiclausum (Wratten et al. 2003; Winkler et al. 2006).

We did indeed find significantly higher numbers of D. semiclausum males and females in Brussels sprout fields bordered by A. graveolens. F. esculentum borders significantly increased the number of male D. semiclausum individuals in Brussels sprout. For female D. semiclausum, numbers of individuals were higher, but not significantly. During sampling we observed high numbers of bees and bumble bees on F. esculentum. Lee and Heimpel (2003) suggest that competition between bees and parasitoids on F. esculentum may restrict nectar access to parasitoids. This might explain the fact that this plant species did not have an impact on parasitoids as strong as expected in our study.

Independent of the treatment, the sex ratio of the collected D. semiclausum was strongly male-biased. Under natural conditions the sex ratio is approximately 1:1. Bukovinszky (personal comm.) found about as many D. semiclausum females as males on sticky traps in cabbage fields. Differences in flight behaviour might account for our male-biased catches. Direct observations on individual wasps in the field confirmed that males tend to fly more and fly above the canopy, while females fly within the canopy, close to the plant, while searching for host larvae.

Regardless of the treatment, we found high rates of parasitism by D. semiclausum: 78% in 2002 and 67% in 2003. The density of P. xylostella larvae did not differ among the treatments. We did not find any effect of A. graveolens or F. esculentum field margins on rates of parasitism by D. semiclausum in adjacent Brussels sprout fields. Neither did we find an effect on sugar levels of field collected D. semiclausum (Winkler et al. 2009b). Similarly Lee and Heimpel (2005) found no consistent effect of F. esculentum on Diadegma insulare parasitism in cabbage plots. This is in stark contrast with earlier results of Winkler et al. (2006), where presence of buckwheat plants near cabbage plants dramatically increased lifespan and fecundity of D. semiclausum. The finding of equal rates of parasitism in the present study may be explained by the fact that D. semiclausum might have exploited alternative food sources like honeydew. Wäckers and Steppuhn (2003) could demonstrate honeydew consumption for Cotesia glomerata and Microplitis mediator collected in a cabbage field where the cabbage aphid B. brassicae and the cabbage whitefly Aleyrodes proletella were present.

The high migration capacity of D. semiclausum might have further diluted the effect of the different treatments. As alternative food sources were extremely rare in the studies by Lee et al. (2006), they suggest between-treatment dispersal of a least 67 m for D. insulare. Thompson (2002) suggests that D. semiclausum travels more than 400 m within 12 h, with a net replacement distance of 55 m. Lavandero et al. (2005) found that D. semiclausum can move 80 m within four days. For the brachonid wasp Dolichogenidea tasmanica Scarratt et al. (2008) could demonstrate migration up to 30 m at least. Thus, the isolation distance of 35 m between our plots might have been too small to prevent nectar-fed wasps from dispersing among the plots. Lavandero et al. (2005) used larger distances of 60 m between plots and could demonstrate differences in parasitism rates due to plantings of F. esculentum.

We could demonstrate that previously reported findings from laboratory experiments regarding impacts of floral resources on herbivores and their parasitoids extend in part to distribution patterns under field conditions. The result that the indiscriminate use of nectar plants in flowering field margins can enhance pest pressure provides a mechanism that may explain the instances in which increased plant biodiversity has been linked to higher levels of herbivory (Andow 1991; Bianchi et al. 2006). Our finding that the impact of floral resources is a function of insect- and flower-species underlines the importance of an informed selection of non-crop vegetation for the optimization of conservation biological control programmes or other ecosystem services.