Ethics Statement

No specific permits were required for the described field studies. Land-holders granted us permission to access their property and conduct the arthropod sampling.

Hebivore-parasitoid Interactions

We collected and reared insect herbivores (Lepidopteran larvae) from a range of crop and non-crop habitat-patches to construct a quantitative food web. In 2009 we sampled fields and native perennial vegetation (NPV) patches within a circular area of 10 km diameter near Cootamundra NSW (lat. 34.72°S, long. 147.73°E). Transects (1 m width) were laid down monthly in arable fields (wheat, barley and canola), pasture fields and NPV. In total 502 m of cereal transect was searched, 440 m of canola, 600 m of fallow arable field, 260 m of pasture, and 360 m of NPV. In 2010 additional sampling was conducted in a second landscape 76 km to the North-East near Young (lat.34.42°S, long. 148.46°E). At both sites 6 canola fields, 6 wheat fields, 12 pasture fields and 9–12 NPV patches were sampled on each monthly visit. The vegetation was sampled by beating into a box (37 cm×30 cm×12 cm) or using a suction sampler (converted leaf blower Stihl^® Sh863, 11 cm diameter pipe placed over the extraction fan). One suction sample consisted of a 120 cm×30 cm area. A sock was placed over the intake tube to collect the insects and these were immediately emptied into a box and examined in the field. In total 720 boxes were sampled in wheat fields, 630 boxes in canola, 2650 suction samples in pasture, and 2797 boxes and suction samples in NPV patches. Herbivores were placed in individual 30 oz cups (with their host plant) and transported back to the laboratory in a chilled cooler-box. Extra plant material was added as the larvae matured in a temperature-controlled glasshouse. Each individual was maintained until it died, an adult host emerged, or a parasitoid emerged. The adult host or parasitoid was pin- (or card-) mounted and sorted into families or morphospecies and further identified by taxonomists if possible (see acknowledgements). For the hosts that produced parasitoids the identification of similar adult lepidopterans collected at the same time were assumed to be the host for that parasitoid (this was cross-checked using the subsequent parasitoid identification). Males of certain species can be difficult to assign and in some cases we assigned a genus based on the female parasitoids that had been collected at the same time in the same location. If no females were reared then the sub-family identification or morphospecies code was used.

To supplement the herbivore-parasitoid interactions identified during the field work we collated information from the published literature. All herbivore species collected in the field were listed and the genus and species names used in an ISI Web of Knowledge^SM search. The search results were restricted to parasitoid species that are known to attack larval stages of the host, and are likely to be present in temperate climates in south-eastern Australia. An honours thesis [18] and hard copy taxonomic publications (e.g. [19]) were searched. We did not cite all references for each herbivore-parasitoid interaction, but collated as many unique interactions as possible.

Malaise Trapping

We quantified the degree of insect movement across edges in the Cootamundra landscape to characterise the relative permeability of these edges. We chose three edge-types: canola adjacent to canola (the control edge), canola adjacent to cereal, and canola adjacent to NPV (Table S1 in File S1). Twelve bi-directional malaise traps (4 replicates per edge-type) were placed within the 10 km diameter circle sample area (nearest traps were 382 m apart, furthest 5.13 km apart). Each edge-type replicate was spatially independent (i.e. no field edge shared traps). Each trap resembled a small white tent (dimensions 170 cm height at front, 170 cm length, 96 cm height at back) made out of fine mesh material. One end was held up with a tent pole and had two collection bottles filled with 70% ethanol (∼250 ml) and ∼5 ml of detergent. The black mid-vein of the tent functioned as an interception trap positioned parallel to the interface. Insects flying from one direction hit the mesh and climbed upwards entering the collection bottle, while insects flying from the other direction were captured in the bottle on the opposite side. The trap catch represents insects moving from one habitat patch to another, and is not a reflection of the abundance in each habitat-patch. The traps were inspected and bottles changed weekly throughout the canola growing season. The catch from the 24 bottles for six sample weeks was used in analyses (early season: 21/05/2009, 11/06/2009, 18/06/2009, 25/06/2009; late season: 22/09/2009, 6/10/2009). Vegetation around the base of the trap was mown during the sampling period. The samples were sieved through a fine mesh strainer (0.5 mm) and stored in vials of 80% ethanol. Insects were identified to ordinal or family level and classified into functional groups (Herbivores, Predators, and Parasitoids). Some taxa were identified to species if sufficient numbers were found in the traps. On the odd occasion (11 bottles out of 144 throughout the season) when a trap was ripped or pulled down by livestock during the sampling week the catch was excluded from the data set.

Analysis

The movement patterns measured by the 24 bottles (12 traps×2 bottles/trap) were: canola to canola (Control) 8 bottles, from canola to cereal 4 bottles, from cereal to canola 4 bottles, from canola to NPV 4 bottles, from NPV to canola 4 bottles. An ordination approach was used to explore the differences in community composition between the different edge-types. Each bottle was classified according one of five treatments that reflected edge-type and the crop next to the bottle (Table S1 in File S1). We used non-metric multidimensional scaling (MDS) in PRIMER (v6.1.13) [20] to order the samples such that their interpoint distances reflect the pattern of variation across multiple taxa. A matrix of total abundance of each taxon (for five sample dates, the first sample was removed due to very low numbers) by bottle was calculated. The abundance counts were square-root transformed (to down-weight the highly abundant species) and a Bray Curtis (also known as Sorensen) similarity matrix was created. A randomization test was used to assess the optimal number of dimensions for MDS ordination. A permutational analysis of variance using PERMANOVA+ (v1.0.3) [21] was used to test if there were significant differences in community composition between the treatments. Type III partial sums of squares were used, with unrestricted permutation of raw data, and 9999 permutations. Pair-wise comparisons were then made between the five treatments.

A generalized linear modelling (GLM) approach in GenStat [22] was used to assess if edge-type was important for explaining the variation in the density of trap catches across time. The model being assessed included the factors sampling date (6 sampling dates, Date), edge-type (5 different treatments with an unbalanced design, Edge) and the interaction between Edge and Date. The insect counts in each bottle per week were used as the response variates. We initially used a Poisson error distribution but found unacceptably high overdispersion. The final models used a negative binomial error distribution and a log link function. For each model, the aggregation parameter (k), which specifies the amount of overdispersion, was first estimated by the Newton-Raphson method [23] then used to fit the final model. Six sampling dates were used for all functional groups and most taxa, however for some taxa we could only obtain an adequate fit using the final two end of season sampling dates (in which trap density was higher). The interaction term was dropped from the final model if extra variation it explained was not significant. The differences between Edges across the sampling dates were examined by calculating the predicted values for the five Edge means, and their standard errors, from the appropriate GLM with back-adjustment from the log scale to the original scale.

High parasitoid numbers were obtained for one late-season sample date (22/09/2009) and further analysis was conducted on parasitoid taxa collected during this week. The parasitoids were grouped into Aphidiinae (Braconidae, aphid parasitoids), Microgastrinae (Braconidae), other Braconidae, Diplazon sp. (Ichneumonidae), other Ichneumonidae, and other hymenopteran parasitoids. A multivariate ANOVA in GenStat [22] was used to assess if any parasitoid group showed strong patterns with respect to treatment.

Do Native Vegetation Patches Provide Alternative Non-pest Hosts for Important Parasitoids?

There are many ways in which NPV may potentially improve productivity and environmental sustainability of farms. If NPV provides alternate hosts for parasitoid species we may see greater population abundance of parasitoid species across a landscape, and/or greater number of pests parasitised in nearby crop fields. In this study we found a good diversity of parasitoid species using hosts in NPV patches, however we recorded few instances of cross-over with species using hosts in both NPV patches and crop fields (Table 1). This is not to say such cross-overs could not occur. The literature review highlighted parasitoid species such as Brachymeria phya that have been recorded attacking Plutella xylostella (diamond-back moth) on canola and Merophyas divulsana (lucerne leaf roller on pastures). Other common non-pest herbivores (e.g. Uraba lugens, gum-leaf skeletonizer) have a large number of parasitoid species attacking them, and could potentially lead to cross-overs in certain years (Table S3 in File S1). We can use the malaise trap data to explore if the lack of cross-overs is due to a lack of movement between these patches (i.e. parasitoids in NPV only attack hosts in NPV, and those in crop fields only attack hosts in crop fields, with little movement between patches).

Do Natural Enemies Move From Native Vegetation Patches into Adjacent Crop Fields?

The malaise trap results showed that there were significant differences in the numbers of insects moving across the different edge-types and that changed across time, with numbers increasing as the season progressed (Table 2). We expected to see greater movement of natural enemies from NPV into canola than in comparison to the control edge, if they were using alternate hosts and prey in the NPV patch. However, this pattern was not consistently found across all natural enemy functional groups or taxa studied (Table 2, Fig. 2). Lower densities of predator taxa were found moving from NPV into canola than across the control edge. For parasitoid taxa greater densities were found moving from NPV into canola, however there were similar (but slightly lower) numbers moving from cereal fields into canola. When we looked in more detail at two parasitoid taxa (Fig. 3), it was clear that these two groups differ in their use of NPV patches late in the season. Herbivores also moved frequently out of NPV (Fig. 2) and this pattern was driven by the movement of aphids late in the season. This result may be due to a number of factors, including the proportion of habitat specialists versus generalists [24] in each functional group, and the dispersal abilities of the taxon involved.

Pest control services rely on the ability of natural enemies to move frequently into crop fields and once there attack and kill pest species. In this study we examined both movement patterns and species interactions to investigate the role native vegetation in pest control services. The malaise trap data suggest that Microgastrinae parasitoids readily move from cereal fields into canola (Fig. 2), yet relatively few taxa were reared from these habitats. There are two possible explanations for the lack of concordance between the species interactions and movement data. Firstly, parasitoids are using hosts in multiple habitat patches across the landscape but our sampling strategy failed to record these interactions. For example, we did not sample herbivores in the tree canopy and this could be a large source of potential alternate hosts. Important interactions can be easily missed if host density is low, and many interactions are likely to be sporadic across space and time. Rearing specimens for more than two field seasons may be required to record some of these interactions. Secondly, the parasitoids collected in the malaise traps are moving through these habitat patches but are not using alternate host resources. They may be using other patches for floral resources or places of refuge, but only using hosts in the crop fields [25], [26].

Rearing of herbivore hosts provided some interesting species interactions that had not previously been recorded in the literature. For example, we reared a specimen that is likely to be Taxoneuron nigriceps (Braconidae, Cardiochilinae) from Helicoverpa punctigera. This species has been recorded parasitising Heliothis virescens and H. zea in the United States, and has been introduced and recorded attacking H. assulta and H. armigera in Asia (see [27] and references within), yet has not been documented as attacking Helicoverpa species in Australian grain crops. Conversely, there are many species of parasitoids that were documented in the literature that were not recorded throughout this study (Table S3 in File S1). The lack of overlap between this study and the literature suggests that parasitoids of pest species in grains cropping landscapes are generally not well characterised. Greater taxonomic effort is needed to clarify the spatial and temporal distribution of important species interactions.

The use of directional traps to understand movement between habitat-patches has been relatively under-utilised (the few exceptions are [3], [28]). Whilst malaise traps are useful for examining movement patterns, translating these into abundance estimates in the crop is difficult. Some species have shown a positive response to edges such that their abundance or activity is increased at the interface [24], [29] and may not reflect the levels observed further into the crop field. The traps give us a good indication of which taxa are moving between different components of the landscape and the functional connectivity of these patches [2]. However, our insights are limited only to insects that are strong flyers and exclude some taxa that may be able to avoid the traps (e.g. some bee species). The wind strength and direction may contribute to variability in trap catch, however given the traps are close to the ground and capturing insects for 7 days, the effect would be difficult to quantify.

Quantifying how species perceive habitat patches and matrix in agricultural landscapes is a challenge [2], [30]. Movement patterns can help us to identify patches that act as a source for species at particular times during the season. The taxa studied here displayed different response to edge-type and some were less likely to move across edges. How this relates to the role of NPV in the provision of pest control services is difficult to judge. At the landscape-scale there is evidence that the configuration and quantity of non-crop vegetation can impact pest control services [13], [31]. In this study we found that alternate hosts in NPV represent a potential resource for parasitoids, however we also saw large numbers of parasitoids moving from cereal fields into canola fields. In contrast, some predatory species such as Neuropterans, appear to be true habitat generalists and are able to move and utilise resources across a range of habitats. Examining how populations of such species respond if all NPV was removed from a cropping landscape would be the next step in assessing the true value of these patches.