SEARCH

SEARCH BY CITATION

Keywords:

  • biological control;
  • diamondback moth;
  • disturbance;
  • ecosystem service;
  • grey cabbage aphid;
  • habitat diversity;
  • hyperparasitoid;
  • primary parasitoid;
  • resource availability;
  • trophic interactions

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

1. Agricultural land use threatens ecosystem services such as biological control by natural enemies because of simplification of habitat structure and intensification of disturbance and agrochemical inputs. Low parasitism rates of agricultural pests have typically been attributed to a lack of resources for parasitoids in highly simplified landscapes, but this could be confounded by the nearly ubiquitous correlation between landscape complexity and the cover of intensively farmed agricultural crops.

2. Here, we disentangle the mechanisms driving landscape-scale effects on host–parasitoid interactions by taking advantage of a landscape modification gradient in which the diversity of habitat types and annual crop cover in the landscape were uncorrelated. We quantified herbivore densities and parasitism and hyperparasitism rates on two important crop pests (aphids and Plutella xylostella) across 30 landscapes. We used structural equation modelling (SEM) to test whether land-use intensity (insecticide application and habitat disturbance) or resource availability for parasitoids (floral resources and alternative host plants) was mediating the effects of habitat diversity and annual crop cover on the landscape.

3. Rates of primary- and hyperparasitism of aphids and primary parasitism of P. xylostella decreased with increasing annual crop cover, whereas decreasing habitat diversity in the landscape had little effect on parasitism rates. These effects were mediated almost entirely by greater habitat disturbance and greater frequency of insecticide application, rather than by changes in resource availability.

4. Parasitoids were more sensitive to intensive farming practices than were their herbivore hosts, and in turn hyperparasitoids were more sensitive than were primary parasitoids. This supports the theoretical prediction that higher trophic levels should be increasingly sensitive to the disturbances associated with land-use change.

5.Synthesis and applications. Our work suggests that increased land-use intensity (e.g. higher insecticide inputs and greater levels of disturbance associated with increasing area of annual crops) has been underestimated as a driver of landscape effects on host–parasitoid interactions. These findings have important implications for the maintenance of ecosystem services such as biological control. The promotion of low-intensity farming practices that limit the extent and frequency of agrochemical inputs and habitat disturbances will be essential for the maintenance of effective biological control by parasitoids in agroecosystems.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

Increased agricultural production since the 1950s has modified agricultural landscapes in several ways, including the destruction and fragmentation of natural habitats, reduction in habitat diversity, and increases in habitat disturbance and agrochemical application (Tilman et al. 2001). Agricultural expansion is therefore considered a major driver of global biodiversity loss (Tilman et al. 2001). There is increasing evidence that this has important effects on species interactions (Tylianakis et al. 2008) and ecosystem services such as biological control (Bianchi, Booij & Tscharntke 2006). For example, it has been shown repeatedly that parasitism of crop pests is higher in complex landscapes with a high diversity of habitat types and a large proportion of semi-natural habitat cover (see Chaplin-Kramer et al. 2011). Interpretation of these effects is confounded by the fact that in most agricultural landscapes, habitat diversity and semi-natural crop cover are strongly negatively correlated with the extent of intensive agricultural land, making it difficult to tease apart the mechanisms driving the observed relationships between landscape structure, biodiversity and ecosystem processes. To date, there have been few studies that have successfully disentangled the drivers of landscape effects on species communities and ecosystem services in agricultural landscapes (Fahrig et al. 2011).

The commonly accepted mechanism underlying landscape effects on biological control is that complex landscapes provide parasitoids with resources, such as adult food and overwintering sites in proximity to the crop fields (Bianchi, Booij & Tscharntke 2006; Tscharntke et al. 2008). While there is explicit, small-scale evidence that parasitoids do use resources in non-crop habitats and then spill over into adjacent crops to parasitize pests (Lavandero et al. 2005; Rand, Tylianakis & Tscharntke 2006), the hypothesis that resource availability is the key driver of landscape effects on parasitism rates remains largely untested. An alternative explanation for lower parasitism rates in highly simplified landscapes dominated by agriculture is that parasitism is negatively influenced by increasing land-use intensity including ploughing, harvesting and insecticide application (Croft 1990; Kruess & Tscharntke 2002). While there have been many studies contrasting local-vs. landscape-level determinants of variation in natural enemy attack rates (e.g. Roschewitz et al. 2005; Geiger et al. 2010; Holzschuh, Steffan-Dewenter & Tscharntke 2010), there have been no previous studies that disentangle the relative importance of resource availability and land-use intensity as drivers of landscape effects on host–parasitoid interactions in agroecosystems.

We aimed to identify the drivers of landscape effects on host–parasitoid interactions through a comparative experiment in which we quantified herbivore densities and parasitism and hyperparasitism rates across 30 landscapes varying in degree of agricultural modification. Our study system comprised two of the world’s most important insect crop pests, Plutella xylostella L. (diamondback moth) and aphids, mostly Brevicoryne brassicae L. (grey cabbage aphid), on a globally significant crop species, Brassica oleracea L. We used a landscape modification gradient in New Zealand in which habitat diversity and proportional cover of annual crops in the landscape were uncorrelated. We used structural equation modelling (SEM) to test different proximate mechanisms that might explain the observed landscape effects. We hypothesized that the effects of habitat diversity and annual crop cover on host–parasitoid interactions could be mediated by direct or indirect effects of altered resource availability (floral resources and alternative crucifer hosts) or land-use intensity (habitat disturbance and insecticide application).

Materials and methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

Gradients of habitat type diversity and annual crop cover in the landscape

We conducted the study in B. oleracea (kale) crops grown as winter feed for cattle and sheep in the Canterbury region of the South Island of New Zealand (Fig. S1, Supporting Information). The study was conducted at 30 randomly selected kale fields located at least 6 km apart with each field managed by a different farmer (Appendix S1, Supporting Information). Patterns of land use were quantified within a 500-m radius of the centre of each sampling area, distinguishing 14 land-cover classes using aerial photographs and ArcGIS 9·2 (ESRI, Redlands, California, USA) (Appendix S1, Table S1, Supporting Information). We selected a 500-m scale because parasitoid species have been found to respond strongly to landscape composition at this spatial scale (Thies, Roschewitz & Tscharntke 2005). Based on the land-cover data, habitat diversity and annual crop cover in the landscape were determined for each site. The Shannon diversity index was used as a measure of habitat diversity, and annual crop cover was calculated as the proportion cover of brassica crops, cereal crops, vegetable crops and recently harvested annual crops, combined.

Insect sampling

In each of 30 kale fields, a strip of land at least 120 m long by 15 m wide along one side of the field was left untreated with insecticide, and all insect sampling was conducted within this strip. In each field, P. xylostella pupae were sampled once between 19 February and 5 March 2007, and aphids (predominantly B. brassicae, but also Myzus persicae Sulzer, the green peach aphid) were sampled once between 19 March and 2 April 2007. Two sampling periods were necessary because P. xylostella abundances usually peak in late summer in New Zealand, whereas the aphid populations on brassica grown as winter feed generally peak in the autumn. Within each 14-day period, sampling commenced at the slightly warmer, northernmost sites in the region and ended at the southernmost sites to minimize potential phenological differences among sites. Sampling was conducted along a 100-m-long transect located 8 m from the field edge, in the middle of the unsprayed strip. The number of P. xylostella pupae (excluding empty pupal cases), live aphids and aphid mummies were counted on 25 kale plants randomly selected along each transect, with c. 4 m between sampled plants. The primary parasitism rate of aphids was estimated by dividing the number of mummies found on the plants by the number of live aphids plus mummies. Plutella xylostella pupae and aphid mummies were collected (max. five per plant) along each transect and taken to the laboratory for rearing and assessment of parasitism (for P. xylostella pupae) and hyperparasitism rates (for P. xylostella pupae and aphid mummies). The primary parasitism rate of P. xylostella was estimated by dividing the number of pupae from which primary- and hyperparasitoids emerged, by the total number of pupae from which any live host or parasitoid emerged (i.e. ignoring pupae that failed to rear to eclosion). Hyperparasitism rates of P. xylostella and aphids were calculated as the proportion of parasitized P. xylostella pupae and aphid mummies, respectively, from which hyperparasitoids emerged. We attempted to collect 100 P. xylostella pupae and 100 mummies at each site, but in some cases a lower number was collected because of low densities and time constraints. At a few sites, the number of emerged P. xylostella pupae, aphids or mummies found was below 10, and these sites were excluded from analyses. In all, 27 sites were analysed for P. xylostella parasitism, 29 for aphid parasitism and 26 for aphid hyperparasitism.

Proximate mechanisms of landscape effects on host–parasitoid interactions

From previous studies (Croft 1990; Kruess & Tscharntke 2002; Lavandero et al. 2005) and our own empirical observations, we identified four proximate mechanisms related to resource availability and land-use intensity that were most likely to mediate the landscape-level effects of habitat diversity and annual crop cover on host–parasitoid interactions, and we measured these in a portion of each 500-m radius landscape sector (for logistical reasons, proximate variables could not be measured in the whole landscape sectors) (Table S2, Supporting Information).

Landscape effects may operate through resource availability for parasitoids and herbivores, in the form of availability of (i) floral resources and (ii) alternative crucifer hosts (Appendix S2, Supporting Information). The cover of flowering plants and of crucifers (predominantly the weed Capsella bursa-pastoris L. shepherds purse) was estimated within a 100-m buffer around each study field between 19 September and 30 November 2007 by the same observer (MJ). This time of year constitutes a bottleneck in terms of resource and host availability for specialist brassica herbivores in Canterbury, when the winter feed crops have been removed by grazing. The area surveyed for flowers and crucifers corresponded to on average 18% of the area in the full landscape sectors, and this comprised a good representation of land-cover types in these sectors (average correlation coefficient of proportion cover of the broad cover classes between the landscape subsample and the full landscape was 0·87).

Landscape effects may also operate through land-use intensity in the form of (iii) the frequency of insecticide applications and (iv) habitat disturbance (e.g. ploughing or harvest). The frequency of insecticide applications during one season within the brassica field outside the unsprayed sampling strip was ascertained through interviews with farmers. The study field comprised on average 70% of the total area of brassica crops within the full landscape sectors, and the majority of these fields were managed by the same farmer. We therefore consider our estimate of the frequency of insecticide application to be representative of brassica fields in the full landscape sectors. As most of the parasitoids in this study are relatively host specific (Appendix S2, Supporting Information), insecticide applications in annual crops other than brassica are likely to have a small influence on host–parasitoid interactions. All insecticides applied to the brassica crops were broad-spectrum products, including one or a mixture of the following active ingredients: deltamethrin, chlorpyrifos, imidacloprid and lambda-cyhalothrin. Insecticides with these ingredients are known to have significant effects on natural enemies, with the effects on the beneficial arthropod component of the Environmental Impact Quotient (EIQ) being 22·2, 23·3, 39·3 and 47·5, respectively (Kovach et al. 2010). We constructed a simple habitat disturbance index (H) reflecting the frequency of soil and vegetation disturbance within 100 m of the outer perimeter of each study field. The proportion of cover (pi) of each of the 14 land-cover classes was weighted by its estimated level of disturbance (k, Appendix S2; Table S3, Supporting Information), and the weighted cover scores were summed across classes to derive the index (equation 1):

  • image(eqn 1)

We conducted a sensitivity analysis to test whether the assumption about specific differences in disturbance levels between land-cover classes used for the habitat disturbance index affected the results of statistical analyses (Appendix S3, Supporting Information).

Discriminating the direct and indirect effects of habitat type diversity and annual crop cover on host–parasitoid interactions

We used structural equation models (SEM) to discriminate the relative direct, indirect and total effects of habitat diversity and annual crop cover in the landscape on herbivore density and parasitism rate, using Amos v19·0 (Amos Development Corporation, Crawfordville, Florida, USA). First, we hypothesized that indirect effects of habitat diversity in the landscape would be mediated by resource availability (floral resources and crucifer hosts; Fig. 1a). Both crucifer weeds and flowering plants are likely to be common along borders between habitat types (Bianchi, Goedhart & Baveco 2008), and these resources could therefore be expected to increase in abundance with increasing habitat diversity. Secondly, indirect effects of annual crop cover in the landscape were hypothesized to operate through both altered land-use intensity (habitat disturbance and insecticide application) and altered resource availability (Fig. 1a). Increased annual crop cover may lead to an increase in the intensity and frequency of habitat disturbance and in the frequency of insecticide application (Meehan et al. 2011), and floral resources might be more abundant in non-crop habitats (Steffan-Dewenter, Münzenberg & Tscharntke 2001). Crucifer weed abundance is likely to be highest in the more disturbed habitats, and we therefore hypothesized that any effects of annual crop cover on host–parasitoid interactions that were mediated by crucifer availability would occur indirectly via the effect of annual crop cover on habitat disturbance. In the full SEM model (Fig. 1a), the direct effects of habitat diversity and annual crop cover on herbivore density and parasitism were also tested, representing unexplained associations between these variables. After inspection of goodness-of-fit of the full model (see Appendix S4, Supporting Information for details of model fitting procedures), it was evident that the hypothesized relationships among variables (solid arrows in Fig. 1a) were inadequate to describe the observed covariance structure of the data. Therefore, we added two further paths to the full model (dashed arrows in Fig. 1a), with one path from alternative crucifer hosts to flowers and another path from habitat diversity to habitat disturbance. In the first case, this makes biological sense as the effect of crucifer plants on parasitism rates might operate through increased host-plant availability for herbivores, or via their contribution to increased flowering resources for parasitoids. In the second case, we believe that there is no strong a priori conceptual basis as to why an increase in habitat diversity should necessarily lead to changes in habitat disturbance in certain landscapes, but it is easy to see how a significant statistical relationship between variables could arise through a ‘selection frequency’ effect, in which increases in habitat diversity lead to a greater likelihood of including land-use types with lower average disturbance frequency or intensity.

image

Figure 1.  Structural equation models discriminating the direct and indirect effects of habitat type diversity and annual crop cover on host–parasitoid interactions, showing (a) the full model, and the most parsimonious models for (b) parasitism of Plutella xylostella, (c) parasitism of aphids, and (d) hyperparasitism of aphids. The full model was the same for all three systems except that no link between floral resources and host density was present in the aphid parasitism model, because aphids do not feed on flowers. Arrows represent causal paths from predictor to response variables. The two dashed arrows represent paths that were added after initial inspection of the residual covariance matrix and overall model fit. The number on each path in the parsimonious models is the value of the unstandardized partial regression coefficient, indicating whether the relationship is positive or negative. The statistical significance of individual regression coefficients is indicated by the colour of the line (black,  0·05; dark grey, 0·05<  0·10; light grey > 0·10). The thickness of the line indicates the magnitude of the standardized path coefficients (Tables S4a–c, Supporting Information). For the four endogenous variables, squared multiple correlations (R2) are given to represent the variance explained by all the associated pathways linking that variable

Download figure to PowerPoint

Proportion parasitism of P. xylostella pupae, hyperparasitism of aphid mummies, alternative crucifer hosts and floral resources were arcsine-square-root-transformed, whereas the proportional parasitism of aphids was 4th-root-transformed, and P. xylostella, aphid and mummy density were ln (x + 1)-transformed prior to analysis to meet the assumptions of multivariate normality.

The full SEM model was fit and tested using a maximum likelihood approach (Kline 2005), and the most parsimonious final SEM models were identified using an information theoretical approach (see Appendix S4, Supporting Information). Furthermore, we conducted a sensitivity test of the validity of the causal hypotheses underlying our SEMs by testing a ‘maximal model’ containing all potential direct and indirect links between ultimate and proximate variables (Appendix S4, Supporting Information). We also tested for possible spatial autocorrelation in the residuals of the final models using SAM 3·0 (Rangel, Diniz-Filho & Bini 2006; Appendix S4, Supporting Information). Finally, we ran confirmatory generalized linear models to validate the relative effect sizes of proximate variables identified as important drivers of parasitism in the SEM (Appendix S5, Supporting Information).

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

Herbivore densities and parasitism rates

In the assessment of herbivore density, an average of 38 P. xylostella pupae were counted on the 25 plants at each site. During laboratory rearing to estimate parasitism rate, an average of 66 of the collected P. xylostella emerged per site (median, 73; range, 17–92). Parasitism was determined as the proportion of individuals that emerged as the larval parasitoid Diadegma semiclausum Hellen (82·1% of parasitoids) or the pupal parasitoid Diadromus collaris Gravenhorst (17·9% of parasitoids). Diadegma semiclausum was hyperparasitized by Trichomalopsis sp., but only at low frequencies at five of the sites and with no hyperparasitism detected at the other sites. Therefore, landscape effects on hyperparasitism rate of P. xylostella were not tested.

On average, 2500 live aphids were counted on the 25 plants at each site. The aphids counted were dominated by B. brassicae (90·6% of aphids), with the remaining aphids being M. persicae (9·4% of aphids). On average, 167 aphid mummies (parasitized aphids) were counted on the 25 plants at each site. Primary parasitism of aphids was estimated by dividing the number of mummies counted by the number of live aphids plus mummies at each site. During laboratory rearing to estimate hyperparasitism rate, an average of 45 of the collected aphid mummies survived and emerged per site (median, 43; range, 10–88), and hyperparasitism of aphids was determined as the proportion of individuals that emerged as Alloxysta sp. (83·9% of hyperparasitoids), Asaphes sp. (15·9% of hyperparasitoids) and Dendrocerus sp. (0·2% of hyperparasitoids). Diaeretiella rapae McIntosh was the only primary parasitoid that emerged during rearing from aphid mummies, so all primary parasitism was attributed to this species.

Discriminating the direct and indirect effects of habitat type diversity and annual crop cover on host–parasitoid interactions

Across our study sites, habitat type diversity and annual crop cover in the landscape were uncorrelated (sampling period 1: r = 0·06, = 0·75; sampling period 2: r = 0·09, = 0·65). Examples of variation in habitat type diversity and annual crop cover among the study landscapes are shown in Figs S2, S3 in Supporting Information.

In the SEM analyses, all three models for parasitism rates (Fig. 1) had acceptable goodness-of-fit indices, with all residual covariance values <2 for both the full and final models, chi-square values non-significant for all final models (P. xylostella parasitism: Chi-square = 17·5, df = 22, = 0·73; aphid parasitism: Chi-square = 15·6, df = 18, = 0·69; aphid hyperparasitism: Chi-square = 22·2, df = 23, = 0·59) and all RMSEA values below 0·001. However, the relative strength and significance of direct and indirect paths varied significantly between models (Fig. 1). Primary parasitism rates of P. xylostella and aphids, as well as hyperparasitism rates of aphid mummies, all decreased with an increasing proportion of annual crop cover in the landscape. By contrast, habitat type diversity in the landscapes only had a weak (positive) effect on aphid parasitism, and no effect on parasitism rates of P. xylostella or hyperparasitism rates of aphid mummies (Figs 1,2).

image

Figure 2.  The relationship between landscape composition and parasitism rates of Plutella xylostella (a, b), parasitism rates of aphids (c, d), and hyperparasitism rates of aphids (e, f). Parasitism of diamondback moth and hyperparasitism of aphids are presented as arcsine-square-root-transformed proportions (in radians), and parasitism of aphids is presented as 4th-root-transformed proportions. Untransformed proportion parasitism of diamondback moth was, on average, 0·68 (range, 0·30–1·00), for aphid parasitism 0·07 (range, 0·00–0·31) and for hyperparasitism of aphids 0·79 (range, 0·17–1·00). The slopes of the fitted lines are calculated from the implied covariances around the sample means, derived from the SEM analyses

Download figure to PowerPoint

In the SEM, we found that the negative effects of annual crop cover on parasitism rates were mediated by different mechanisms for different species, with increased frequency of insecticide application most important for P. xylostella parasitism, habitat disturbance most important for aphid hyperparasitism, and both these variables being important for aphid parasitism (Fig. 1b–d, Table S4a–c, Supporting Information). Although the total effect of habitat disturbance on aphid parasitism rates was negative, the final model also retained a small positive, albeit non-significant, indirect effect of this variable on aphid parasitism via increased abundance of crucifer hosts (Fig. 1c, Table S4b, Supporting Information). There was also a non-significant direct negative effect of annual crop cover in the landscape retained in the most parsimonious aphid parasitism model. The positive total effect of habitat diversity on aphid parasitism rate was mediated predominantly by a negative association between habitat diversity and habitat disturbance. Finally, there was a negative direct effect of habitat diversity on aphid density and a negative direct effect of annual crop cover on mummy density, but apart from that no paths including host density were retained in any of the models.

No spatial autocorrelation was detected in any of the SEM models, and no alternative approaches to constructing the habitat disturbance index had a significant effect on SEM results (Appendix S3, Supporting Information). The alternative SEMs testing a ‘maximal model’ with all potential direct and indirect links between ultimate and proximate variables also had no significant effect on the SEM results (Table S5 Supporting Information). Finally, the results from the confirmatory generalized linear models did not change interpretation of the most important factors determining variation in parasitism (Tables S6, S7, Supporting Information).

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

Drivers of landscape effects

We found that annual crop cover in the landscape had a strong negative effect on parasitism rates of two important crop pests (P. xylostella and aphids), whereas habitat diversity only had a minor positive impact on parasitism rates of one of the pests (aphids). These landscape-scale effects were driven primarily by the frequency and magnitude of different processes associated with the intensity of annual cropping practices in our study system (insecticide application and habitat disturbance, such as ploughing and harvesting), rather than by any direct effects of greater resource availability for parasitoids. To our knowledge, this is the first study to identify the relative landscape-level influences of habitat diversity and amount of intensive land use on host–parasitoid interactions in agroecosystems.

Several previous studies have compared the effects of local land-use intensification and landscape-level complexity on natural enemy attack rates (e.g. Roschewitz et al. 2005; Geiger et al. 2010; Holzschuh, Steffan-Dewenter & Tscharntke 2010). These studies have mostly used organic vs. conventional farming as a proxy for local land-use intensity and have therefore not been able to distinguish which aspects of land-use intensity are most important. Two exceptions are the recent studies by Geiger et al. (2010) and Krauss, Gallenberger & Steffan-Dewenter (2011), which showed that local pesticide application had persistent negative effects on biodiversity, predator–prey ratios and natural enemy attack rates. Our work now shows that insecticide application and habitat disturbance are critical drivers of landscape-level effects on natural enemy attack rates.

Our study contradicts the common assumption that the negative effect of landscape simplification on parasitism rates is caused primarily by a lack of resources for parasitoids (Bianchi, Booij & Tscharntke 2006; Tscharntke et al. 2008). Although empirical evidence supporting this assumption at the landscape scale is surprisingly scant, a few studies have found parasitism rates to be positively related to the availability of some habitats in the landscape, such as forest edges or grasslands, which are known to provide resources for certain parasitoids (Bianchi et al. 2005; Bianchi, Goedhart & Baveco 2008; and see meta-analysis by Chaplin-Kramer et al. 2011). It is likely that the relative importance of resource availability and land-use intensity at the landscape scale depends on the biological characteristics of the species and landscapes studied. For example, the availability of a high diversity of habitat types that provide complementary resources to different parasitoid species might be more important when high overall parasitism rates are caused by the combined effects of a high diversity of parasitoids (Tscharntke et al. 2008). In our study, only a few species of parasitoids contributed to parasitism, which might help explain the low importance of habitat diversity and resource availability for parasitism. Exotic parasitoid species may also be more strongly associated with crop land than native species, and the fact that all species included in this study are exotic to New Zealand (Berry, Cameron & Walker 2000) could also have contributed to the importance of annual crop cover and land-use intensity. Furthermore, annual crop cover might have a particularly strong impact on host–parasitoid interactions when it is positively related to the frequency of pesticide application and other types of disturbance. Naturally, different types of annual crops are grown under differing intensities of agrochemical inputs and disturbance frequencies in different agricultural systems. In our study, insecticides were more frequently applied to crops located in landscapes with high annual crop cover. This is in agreement with a recent study from the mid-western USA (Meehan et al. 2011), but contrasts with European studies that found no such relationship (Roschewitz, Thies & Tscharntke 2005; Herzog et al. 2006).

We were able to discriminate between the mechanistic drivers of landscape effects (i.e. components of land-use intensity and resource availability) on host–parasitoid interactions because our comparative analysis was conducted in a system where (i) habitat diversity and annual crop cover were uncorrelated and (ii) land-use intensity (habitat disturbance and insecticide application) and resource availability (crucifer hosts and floral resources) were largely independent. However, it is important to acknowledge that most agricultural landscapes typically exhibit strong collinearity among the potential drivers of landscape effects. For example, low-intensity organic farms often have higher levels of crop diversity and a higher proportion of fallow or set-aside land compared to conventional farms, and organic farms are often located in complex landscapes with a relatively high proportion of semi-natural habitats (Norton et al. 2009). This could lead to a significant negative relationship both between annual crop cover and habitat type diversity, and between land-use intensity and resource availability in the landscape. The SEM approach we used here would not, in itself, provide the ability to overcome this type of strong collinearity among predictor variables, unless it was also possible to test casual inference with experimental manipulation at the landscape scale (e.g. experimental modification of pesticide application, tillage frequency, floral resource availability). Nevertheless, the magnitude of collinearity could at least be estimated as a means of determining whether apparent correlations between parasitism rates and resource distributions (for example) might actually be confounded by covariance in some other factor such as the distribution of insecticide use. These effects were not evident in our study, perhaps because there were no organic farms present within our study landscapes and no tendency for farmers in structurally complex landscapes to use lower intensity farming practices. Habitat disturbance and insecticide application effects associated with annual cropping do, then, appear to drive variation in host–parasitoid interactions in our study system, rather than resource availability for parasitoids. We acknowledge that causal inference on the relative importance of different mechanistic pathways of landscape effects requires experimental validation, as there is always the possibility that apparent correlative relationships in SEM analyses can be determined by hidden extrinsic drivers that we are not aware of. In our models, any such effects (if they existed) would have to be almost completely collinear with measured variables that were entered into the model, as the only unidentified pathway of landscape effects on parasitism rates was a weak (non-significant) direct effect from annual crop cover to parasitism rate in the aphid parasitism model.

Effects on different trophic levels

We found that both primary parasitism rates and hyperparasitism rates of aphids decreased with increasing annual crop cover in the landscape. This suggests that primary parasitoids are more sensitive to agricultural expansion than their herbivore hosts, and in turn hyperparasitoids are more sensitive than primary parasitoids. We are aware of two other studies that investigated landscape effects on fourth trophic-level processes in agroecosystems and both found that hyperparasitism rates decreased with increasing cover of annual crops and decreasing landscape complexity (Gagic et al. 2011; Rand, van Veen & Tscharntke 2012). Thus, previous work suggests that the negative effects of agricultural expansion on primary parasitoids may be somewhat ameliorated by the even larger effect it has on hyperparasitoids. Our results also support the theoretical prediction that higher trophic levels should be increasingly sensitive to the disturbances associated with land-use change (Post 2002). To date, empirical evidence for this theory, in general terms, has been relatively weak (Post 2002), although it has been shown that primary parasitism rates may decrease with increasing intensity of grazing (Kruess & Tscharntke 2002) and parasitoids are sensitive to broad-spectrum insecticides (Croft 1990).

The observed effects of annual crop cover and land-use intensity on host–parasitoid interactions were not an artefact of density-dependent responses to variation in host density (Costamagna, Menalled & Landis 2004). There was no significant relationship between host density and parasitism rate in any of the host-parasitoid systems studied (Fig. 1b–d). At the same time, though, this lack of a relationship provides no evidence that higher parasitism rates in less intensively managed landscapes would lead to increased pest suppression (Thies & Tscharntke 1999), as has also been found in a meta-analysis of natural enemy responses to landscape complexity (Chaplin-Kramer et al. 2011). Parasitism rates of aphids were probably too low (0–31%) to have a significant effect on aphid densities, but for P. xylostella parasitism rates were much higher (30–100%) and this is likely to influence P. xylostella densities in subsequent generations (P. xylostella has multiple generations per year in New Zealand).

The only path including herbivore density that was retained in any of the SEM models comprised a negative direct effect of habitat diversity on aphid density. However, the mechanism driving this relationship is unclear. In this case, there was no direct relationship between parasitism rate and aphid density, so the lower aphid density in more diverse landscapes could not be attributed to increased parasitism. Aphid density was also unrelated to host-plant density, of either cultivated brassicas (there was no effect of annual crop cover, which was comprised on average of ca 80% brassica crops) or alternative crucifer hosts (the effect of habitat diversity was not mediated by cover of alternative crucifer hosts; Fig. 1c).

Conclusions

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

Studies considering the implications of land-use intensification on biodiversity and ecosystem services have mostly focussed on the local effects of this process (e.g. Holzschuh et al. 2007; Krauss, Gallenberger & Steffan-Dewenter 2011). However, it has recently been shown that the proportion of organic farming in the landscape may have significant effects on biodiversity (Gabriel et al. 2010) and that the species richness of certain taxa are negatively affected by variables such as the percentage of intensively fertilized land (Billeter et al. 2008). This suggests that land-use intensity at the landscape level may also be an important determinant of biodiversity loss and reduction in ecosystem services on farmland. In our work, we show that reduction in disturbance and insecticide applications associated with lower levels of intensive land use can explain why complex landscapes often have higher natural enemy abundance and enhanced biological control. While we recognize that the relative importance of different drivers of landscape effects probably depends on multiple factors, including landscape structure and the identity of the species involved, we argue that land-use intensity has been underestimated as a driver of landscape effects on ecosystem services. Until we improve our understanding of the conditions under which different drivers of landscape effects operate, we suggest that farming practices that decrease the extent and frequency of agrochemical inputs and disturbances should be given equivalent weight to practices that increase habitat diversity in the landscape, if we are to maintain natural pest control services in agroecosystems.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

We thank the 30 landowners who gave permission to work on their farms. A. Dumbleton of PGG Wrightson Ltd provided invaluable information on brassica cropping in Canterbury. I.H. Lynn and J. Barringer at Landcare Research assisted with vegetation classification. S. Blyth, N. Jørgensen, J. Martin, M. Mackintosh, R. Neumegen, S. Orre, S. Sam and N. White assisted with field and laboratory work. Financial support was provided by the Tertiary Education Commission through the Bio-Protection Research Centre at Lincoln University. M. Jonsson also acknowledges support from a grant from the Swedish Research Council for Environment, Agricultural Sciences and Spatial Planning (FORMAS) to the project: Multifunctional Agriculture: Harnessing Biodiversity for Sustaining Agricultural Production and Ecosystem Services (SAPES). F. Bianchi, D. Crowder, B. Ekbom, R.M. Ewers, A.-K. Kuusk, O. Lundin, T.A. Rand, W.E. Snyder, J. Stenberg, J.M. Tylianakis, C. Winqvist, two anonymous reviewers and associate editor provided valuable comments on previous versions of this manuscript.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information
  • Berry, N.A., Cameron, P.J. & Walker, G.P. (2000) Integrated Pest Management for Brassicas. Crop & Food Research, Christchurch, New Zealand.
  • Bianchi, F.J.J.A., Booij, C.J.H. & Tscharntke, T. (2006) Sustainable pest regulation in agricultural landscapes: a review on landscape composition, biodiversity and natural pest control. Proceedings of the Royal Society B, 273, 17151727.
  • Bianchi, F.J.J.A., Goedhart, P.W. & Baveco, J.M. (2008) Enhanced pest control in cabbage crops near forest in The Netherlands. Landscape Ecology, 23, 595602.
  • Bianchi, F.J.J.A., van Wingerden, W.K.R.E., Griffioen, A.J., van der Veen, M., van der Straten, M.J.J., Wegman, R.M.A. & Meeuwsen, H.A.M. (2005) Landscape factors affecting the control of Mamestra brassicae by natural enemies in Brussels sprout. Agriculture, Ecosystems and Environment, 107, 145150.
  • Billeter, R., Liira, J., Bailey, D., Bugter, R., Arens, P., Augenstein, I., Aviron, S., Baudry, J., Bukacek, R., Burel, F., Cerny, M., De Blust, G., De Cock, R., Diekötter, T., Dietz, H., Dirksen, J., Dormann, C., Durka, W., Frenzel, M., Hamersky, R., Hendrickx, F., Herzog, F., Klotz, S., Koolstra, B., Lausch, A., Le Coeur, D., Maelfait, J.P., Opdam, P., Roubalova, M., Schermann, A., Schermann, N., Schmidt, T., Schweiger, O., Smulders, M.J.M., Speelmans, M., Simova, P., Verboom, J., van Wingerden, W.K.R.E., Zobel, M. & Edvards, P.J. (2008) Indicators for biodiversity in agricultural landscapes: a pan-European study. Journal of Applied Ecology, 45, 141150.
  • Chaplin-Kramer, R., Rourke, M.E.O., Blitzer, E.J. & Kremen, C. (2011) A meta-analysis of crop pest and natural enemy response to landscape complexity. Ecology Letters, 14, 922932.
  • Costamagna, A.C., Menalled, F.D. & Landis, D.A. (2004) Host density influences parasitism of the armyworm Pseudaletia unipunctata in agricultural landscapes. Basic and Applied Ecology, 5, 347355.
  • Croft, B.A. (1990) Arthropod Biological Control Agents and Pesticides. John Wiley & Sons, New York, USA.
  • Fahrig, L., Baudry, J., Brotons, L., Burel, F.G., Crist, T.O., Fuller, R.J., Sirami, C., Siriwardena, G.M. & Martin, J.-L. (2011) Functional landscape heterogeneity and animal biodiversity in agricultural landscapes. Ecology Letters, 14, 101112.
  • Gabriel, D., Sait, S.M., Hodgson, J.A., Schmutz, U., Kunin, W.E. & Benton, T.G. (2010) Scale matters: the impact of organic farming on biodiversity at different spatial scales. Ecology Letters, 13, 858869.
  • Gagic, V., Tscharntke, T., Dormann, C.F., Gruber, B., Wilstermann, A. & Thies, C. (2011) Food web structure and biocontrol in a four-trophic level system across a landscape complexity gradient. Proceedings of the Royal Society B, 278, 29462953.
  • Geiger, F., Bengtsson, J., Berendse, F., Weisser, W.W., Emmerson, M., Morales, M., Ceryngier, P., Liira, J., Tscharntke, T., Winqvist, C., Eggers, S., Bommarco, R.B., Pärt, T., Bretagnolle, V., Plantegenest, M., Clement, L.W., Dennis, C., Palmer, C., Oñate, J.J.G.I., Hawro, V., Aavik, T., Thies, C., Flohre, A., Haenke, S., Fisher, C., Goedhart, P.W. & Inchausti, P. (2010) Persistent negative effects of pesticides on biodiversity and biological control potential on European farmland. Basic and Applied Ecology, 11, 97105.
  • Herzog, F., Steiner, B., Bailey, D., Baudry, J., Billeter, R., Bukacek, R., de Blust, G., de Cock, R., Dirksen, J., Dormann, C.F., de Filippi, R., Frossard, E., Liira, J., Schmidt, T., Stöckli, R., Thenail, C., van Wingerden, W. & Bugter, R. (2006) Assessing the intensity of temperate European agriculture at the landscape scale. European Journal of Agronomy, 24, 165181.
  • Holzschuh, A., Steffan-Dewenter, I. & Tscharntke, T. (2010) How do landscape composition and configuration, organic farming and fallow strips affect the diversity of bees, wasps and their parasitoids? Journal of Applied Ecology, 79, 491500.
  • Holzschuh, A., Steffan-Dewenter, I., Kleijn, D. & Tscharntke, T. (2007) Diversity of flower-visiting bees in cereal fields: effects of farming system, landscape composition and regional context. Journal of Applied Ecology, 44, 4149.
  • Kline, R.B. (2005) Principles and Practice of Structural Equation Modeling. Guilford Press, New York, N.Y., U.S.A.
  • Kovach, J., Petzoldt, C., Degni, J. & Tette, J. 2010. A Method to Measure the Environmental Impact of Pesticides. Available from: http://nysipm.cornell.edu/publications/eiq/ default.asp. IPM Program, Cornell University, Geneva, New York.
  • Krauss, J., Gallenberger, I. & Steffan-Dewenter, I. (2011) Decreased functional diversity and biological pest control in conventional compared to organic crop fields. PLoS One, 6, e19502.
  • Kruess, A. & Tscharntke, T. (2002) Grazing intensity and the diversity of grasshoppers, butterflies, and trap-nesting bees and wasps. Conservation Biology, 16, 15701580.
  • Lavandero, B., Wratten, S.D., Shishehbor, P. & Worner, S. (2005) Enhancing the effectiveness of the parasitoid Diadegma semiclausum (Helen): Movement after use of nectar in the field. Biological Control, 34, 152158.
  • Meehan, T.D., Werling, B.P., Landis, D.A. & Gratton, C. (2011) Agricultural landscape simplification and insecticide use in the Midwestern United States. Proceedings of the National Academy of Sciences USA, 108, 1150011505.
  • Norton, L., Johnson, P., Joys, A., Stuart, R., Chamberlain, D., Feber, R., Firbank, L., Manley, W., Wolfe, M., Hart, B., Mathews, F., Macdonald, D. & Fuller, R.J. (2009) Consequences of organic and non-organic farming practices for field, farm and landscape complexity. Agriculture, Ecosystems and Environment, 129, 221227.
  • Post, D.M. (2002) The long and short of food-chain length. Trends in Ecology and Evolution, 17, 269277.
  • Rand, T.A., Tylianakis, J.M. & Tscharntke, T. (2006) Spillover edge effects: the dispersal of agriculturally subsidized insect natural enemies into adjacent natural habitats. Ecology Letters, 9, 603614.
  • Rand, T.A., van Veen, F.J.F. & Tscharntke, T. (2012) Landscape complexity differentially benefits generalized fourth, over specialized third, trophic level natural enemies. Ecography, 35, 97104.
  • Rangel, T.F.L.V.B., Diniz-Filho, J.A.F. & Bini, L.M. (2006) Towards an integrated computational tool for spatial analysis in macroecology and biogeography. Global Ecology and Biogeography, 15, 321327.
  • Roschewitz, I., Thies, C. & Tscharntke, T. (2005) Are landscape complexity and farm specialisation related to land-use intensity of annual crop fields? Agriculture, Ecosystems and Environment, 105, 8799.
  • Roschewitz, I., Hücker, M., Tscharntke, T. & Thies, C. (2005) The influence of landscape context and farming practices on parasitism of cereal aphids. Agriculture, Ecosystems and Environment, 108, 218227.
  • Steffan-Dewenter, I., Münzenberg, U. & Tscharntke, T. (2001) Pollination, seed set and seed predation on a landscape scale. Proceedings of the Royal Society B, 268, 16851690.
  • Thies, C., Roschewitz, I. & Tscharntke, T. (2005) The landscape context of cereal aphid-parasitoid interactions. Proceedings of the Royal Society B, 272, 203210.
  • Thies, C. & Tscharntke, T. (1999) Landscape structure and biological control in agroecosystems. Science, 285, 893895.
  • Tilman, D., Fargione, J., Wolff, B., D’Antonio, C., Dobson, A., Howarth, R., Schindler, D., Schlesinger, W.H., Simberloff, D. & Swackhamer, D. (2001) Forecasting agriculturally driven global environmental change. Science, 292, 281284.
  • Tscharntke, T., Bommarco, R., Clough, Y., Crist, T.O., Kleijn, D., Rand, T.A., Tylianakis, J.M., van Nouhuys, S. & Vidal, S. (2008) Conservation biological control and enemy diversity on a landscape scale. Biological Control, 45, 238253.
  • Tylianakis, J.M., Didham, R.K., Bascompte, J. & Wardle, D.A. (2008) Global change and species interactions in terrestrial ecosystems. Ecology Letters, 11, 113.

Supporting Information

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

Appendix S1. Selection of field sites and quantification of land-use patterns

Appendix S2. Resources for parasitoids

Appendix S3. Sensitivity analysis for the index of habitat disturbance

Appendix S4. Sensitivity analysis for structural equation models

Appendix S5. Confirmatory linear model analyses

Table S1. Summary statistics for landscape composition in circular landscapes with a 500-m radius around the centre of each study area.

Table S2. Summary statistics averaged across the 30 sites for proximate variables potentially explaining landscape effects.

Table S3. The estimated level of disturbance (k-value) used for different land-cover classes when calculating the habitat disturbance index.

Table S4. Standardized path coefficients from the structural equation models in Fig. 1, showing the direct effects, indirect effects and total effects of factors influencing parasitism.

Table S5. Standardized path coefficients of ‘maximal’ structural equation models for each response variable (carried out as a sensitivity analysis to test the robustness of causal hypotheses in the primary SEM models), showing the direct effects, indirect effects and total effects of factors influencing parasitism

Table S6. Confirmatory generalized linear mixed model (GLMM) analyses testing the relative influence of ultimate and proximate factors affecting parasitism.

Table S7. Coefficients for the best-fit ultimate-driver and proximate-driver GLMM models for parasitism rates.

Fig. S1. Map of New Zealand and the Canterbury region with the locations of the 30 study fields indicated with black circles.

Fig. S2. Examples of land-use maps of landscape sectors (500-m radius around study transect) with high and low habitat type diversity.

Fig. S3. Examples of land-use maps of landscape sectors (500-m radius around study transect) with high and low annual crop cover.

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials may be re-organized for online delivery, but are not copy-edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

FilenameFormatSizeDescription
JPE_2130_sm_Updated-Supporting-Info.doc354KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.