William E. Snyder, Department of Entomology, Washington State University, Pullman, WA 99163, USA (fax +509 335 1009; e-mail email@example.com).
1While evidence suggests that undisturbed refuges within agricultural fields conserve natural enemies, few studies have examined whether pest control does actually improve following predator conservation. When the targets of conservation are generalists, polyphagy may complicate the impact of the conserved predators on agricultural pests.
2We examined the use of beetle banks to conserve predatory beetles for the control of pest Diptera. Locally, the community of predatory beetles included several species of small carabid and staphylinid beetles (< 1 cm in length), which eat fly eggs, and one common larger carabid beetle (> 1·5 cm), Pterostichus melanarius, which rarely eats fly eggs but does eat smaller beetles.
3Predator beetle activity densities, but not rates of fly egg predation, increased in fields including beetle banks. A series of field experiments was conducted to examine whether two types of polyphagy, intraguild predation and feeding on non-target prey, could be preventing increased egg predation following successful predator conservation.
4The putative intraguild predator Pterostichus melanarius reduced activity densities of smaller beetles, and thus weakened fly egg predation. The strength of fly suppression increased with increasing densities of small beetles in the absence of Pterostichus melanarius, but not when aphid alternative prey were readily available. In the presence of abundant aphids, egg predation rates did not increase at higher small beetle densities.
5Synthesis and applications. Overall, our results suggest that both intraguild predation and the presence of alternative prey could limit conservation biological control that targets generalist predators. Thus, higher predator densities will not necessarily lead to improved pest control. Ecologists must consider the impact of predator manipulations at multiple trophic levels when assessing the success or failure of conservation biological control.
This present study combined field observations of beetle bank efficacy with a more mechanistic understanding of how interactions involving conserved predators contributed to beetle bank performance. We first established beetle banks in several production fields, and monitored their impact on ground-active predatory beetles over the following two growing seasons. From previous work we knew that the guild of conserved predators included several smaller (< 1 cm adult length) beetle species that feed on fly eggs, and a common larger beetle species (Pterostichus melanarius Illiger, c. 1·5 cm adult length) that rarely eats fly eggs but eats the smaller beetles (Prasad & Snyder 2004). Thus, we conducted a series of experiments to examine whether intraguild predation might be limiting egg predation. Unexpectedly, an aphid outbreak during one of our experiments suggested that these non-target prey might also impact egg predation, a possibility that we examined in an additional field experiment. Thus, our mechanistic work examined two types of trophic interactions that were potentially disruptive to biological control, intraguild predation and predation upon non-target (alternative) prey.
Materials and methods
Several species of anthomyiid flies are major pests of vegetable crops in north-western Washington, USA, and south-western British Columbia, Canada (Finch 1989; Howard, Garland & Seaman 1994). Female flies lay eggs at the base of host-plants, and root feeding by the larvae results in cosmetic injury or death of seedlings. Carabid and staphylinid beetles are predators of the eggs of these pest flies (Finch 1989, 1996). In local vegetable fields, the most common species of epigeal, predatory beetles are several carabids (Coleoptera: Carabidae; Bradycellus congener LeConte, Bembidion tetracolum Say, Bembidion lampros Herbst, Amara littoralis Mannerheim, Amara apricaria Paykull and Pterostichus melanarius) and staphylinids (Coleoptera: Staphylinidae; Aleochara bilineata Gravenhorst, Philonthus politus L. and an Aleocharine morphospecies) (Prasad & Snyder 2004).
on-farm beetle bank performance
A beetle bank was established on each of three organic mixed-vegetable farms, located in Ladner, British Columbia, Canada, and in Mt Vernon and Carnation, Washington, USA. A fourth beetle bank was established in a radish Rhaphnus sativus L. field at Washington State University's research station in Mt Vernon, which was managed conventionally for nutrients and weeds but received no insecticide applications. All four beetle banks were planted between April and June 2002 by broadcasting orchard grass Dactylis glomerata L. seed in strips 1·5 m wide and 30–60 m long. Two banks were raised approximately 50 cm, while the remaining two were level with the surrounding field; co-operating growers independently determined the dimensions of their beetle banks using previous designs as a guideline (Thomas, Wratten & Sotherton 1991; Frampton et al. 1995; Lee, Menalled & Landis 2001). In 2003, we surveyed the beetle fauna in the three organically managed beetle bank fields. In 2004, the beetle bank on one of the organic farms was tilled under, so we conducted our survey with the remaining two organic fields and the beetle bank field at the research station.
Concurrently, in both years, we surveyed beetle densities and egg predation rates in three control fields that lacked a beetle bank, selected to minimize differences in field management between fields with and without beetle banks. In 2003 the control fields consisted of one mixed vegetable field, in Mt Vernon, and two broccoli Brassica oleracea L. monocultures in Ladner; all fields were managed organically. In 2004, control fields consisted of two organic mixed-vegetable fields, one each in Mt Vernon and Ladner. The third control field was a cauliflower Brassica oleracea L. monoculture, located in Mt Vernon, under conventional weed and nutrient management but no post-transplant insecticide applications.
Ground beetle activity was assessed using pitfall traps (design as in Prasad & Snyder 2004) three times each year between 20 and 25 May, 8 and 12 August and 25 and 29 September. In fields containing beetle banks, traps were placed within 5 m of banks (and no less than 20 m from field margins), while in control fields traps were placed 20 m from field margins.
To determine whether beetle banks indirectly impacted predation of pest-fly eggs, our focal prey stage, we measured predation of sentinel Diptera eggs during the 2004 field season. Previous results indicated no preference among the commonly occurring carabid and staphylinid species for eggs of the housefly Musca domestica L. vs. eggs of the economically important vegetable pest Delia radicum L. (Prasad & Snyder 2004); thus we used the easily propagated Musca domestica eggs as a surrogate (Prasad & Snyder 2004). A group of five eggs (< 24 h old) was placed on 1-cm2 pieces of peat cut from transplant pots, placed at the base of a plant and then covered with a 0·5-cm layer of soil (Finch & Elliot 1994). We placed five egg cards per field, spaced 1 m apart and positioned either 5 m from the beetle bank or, in control fields, 20 m from the nearest field margin. Egg cards were placed in fields for 24 h concurrent with each pitfall survey (May, August and September).
field cage experiments
Field cage experiments were conducted in a 0·81-ha radish field at the Washington State University farm in Mt Vernon. Experimental units were 2 × 2 × 2-m field cages, covered on all sides but the bottom with a fine mesh screen and positioned to contain three parallel rows of plants (a full description of cages is provided in Prasad & Snyder 2004). Between experiments cages were moved to a new section of the field and reassembled.
Experiment 1 was designed to isolate the impact of the intraguild predator Pterostichus melanarius on fly egg consumption by smaller beetles. The experiment had an additive design (sensuGoldberg & Scheiner 2001); in addition to 28 small beetles, cages received either no Pterostichus melanarius (0X), seven Pterostichus melanarius (1X) or 28 Pterostichus melanarius (4X). Ratios of Pterostichus melanarius to small beetles, and overall beetle densities, approximated the range observed at different times of the season in our beetle bank demonstration fields (see the Results). We also included two control treatments, all species trapped-out but not replaced (Removal) and uncaged and unmanipulated 4-m2 reference plots (Open). The experiment was conducted in 2003, as two replicated blocks separated in time, an appropriate approach when a limited number of replicates can be conducted simultaneously (Gotelli & Ellison 2004). The first block began on 25 July, and the second on 13 August, with five replicates of each treatment per block (total n= 50).
In the second experiment our objective was to examine whether, in the absence of the putative intraguild predator Pterostichus melanarius, fly egg consumption would increase with increasing densities of smaller beetles. This replicated the early season rarity of Pterostichus melanarius and higher small beetle densities in the presence of beetle banks, observed in our open-field data (see the Results). Cages received either seven (1X) or 28 small beetles (4X); Removal and Open treatments were again included as controls. This experiment was also conducted in two temporal blocks, during 2003, beginning on 10 June in the first block and on 1 July in the second block, with five replicates of each treatment within each block (total n= 40).
As results from one block of experiment 2 suggested that aphid alternate prey could disrupt egg predation by small beetles (see the Results), a third experiment was conducted to examine the effect of aphids on egg predation by small beetles. In this experiment 28 small beetles were added to all cages, along with no aphids (No), 20 aphids added to a single radish plant per cage (Low) or 20 aphids added to 10 plants per cage (High). Removal and Open treatments were again included, and this experiment was conducted once starting on 19 May 2004. All treatments were replicated five times (total n= 25).
Experiment 1 was conducted as a pulse experiment (Gotelli & Ellison 2004), with beetles added to cages only on day 1 (Table 1; for a similar experiment run as a press see Prasad & Snyder 2004). In contrast, a press design (Gotelli & Ellison 2004) was appropriate for experiments 2 and 3 because, for these experiments, our goal was to mimic predation rates in a field adjacent to a beetle bank with immigrant predators arriving regularly (Table 1; Snyder & Wise 1999). Ambient beetle densities (1X treatments) reflected typical densities in conventionally managed vegetable fields (Prasad & Snyder 2004). Beetles were added to cages as an assortment of the seven species or morphospecies of common smaller beetles: Bradycellus congener, Bembidion tetracolum, Bembidion lampros, Amara spp., Aleochara bilineata, Philonthus politus and an Aleocharine morphospecies (Prasad & Snyder 2004), with the restriction that at least one individual from each taxon was added to each cage on each release date. Beetles were field-collected < 24 h prior to being released into cages. We used pitfall traps to lower predator densities in the Removal treatments, and to assess predator activity–density in all other treatments (Table 1).
Table 1. Summary of the experimental protocol for beetle addition, activity–density monitoring and sentinel egg card handling, for each of the three field cage experiments
Beetles added to cages (addition cages only)
Egg cards added to cages (removed 48 h later)
Pitfall traps opened to monitor activity–density* (opened for 48 h)
Traps in Removal treatments were open throughout each experiment.
The delay in experiment 3 was because of heavy rain showers.
Predation rates in field cages were measured using 20 sentinel Musca domestica eggs, as described previously, but with two cards of 10 eggs placed at the base of two haphazardly selected plants per plot (Table 1). Full pitfall trapping and predation measurement protocols are described in detail in Prasad & Snyder (2004). In experiment 2, aphids were counted on five randomly selected plants per plot, at the conclusion of each block. Aphids were not manipulated experimentally in experiment 2. For experiment 3 green peach aphids Myzus persicae Sulzer were collected from the field and reared on radish plants in a greenhouse, under natural photoperiod at 21 ± 2 °C. At the end of the experiment, one randomly selected leaf from each plant per replicate was removed and the number of aphids was counted.
For our on-farm data, beetle activity–densities (activity–densities because pitfall trapping measures a mix of both the number of beetles present and their movement) were analysed using repeated-measures manova (von Ende 1993), following square root transformation to homogenize variances. Relationships between beetle activity–densities and sentinel egg predation rates were examined using multiple regression, with the critical P-value adjusted to reflect multiple comparisons (Jones 1984).
For the field cage experiments, predation of fly eggs in all experiments, and predator activity–densities in experiments 2 and 3, were first analysed using repeated-measures manova. However, treatment–time interactions were never significant in a way that was clearly ecologically significant, and so for the sake of brevity only statistical outputs for the main treatment, block and treatment × block effects are presented and discussed herein; the full time series data, and associated repeated-measures analysis, are presented in Prasad (2005). We first tested for cage effects, by comparing Open plots with the appropriate caged treatment (the treatment designed to most closely match open-field conditions). Next, we tested for the efficacy of beetle manipulations by comparing Removal with the appropriate addition treatment(s). Finally, for analysis of main treatment effects, only the beetle addition treatments were compared to one another. We analysed initial and final beetle activity–densities in experiment 1 using two-way anova, and final aphid densities in experiment 3 using one-way anova. When appropriate, post-hoc tests were performed using a Tukey-Kramer honestly significant difference (HSD) test (α = 0·05). All analyses were conducted in systat version 9·0 (SPSS, Chicago, IL).
on-farm beetle bank performance
Total beetle activity–densities were significantly higher in fields with beetle banks than in those without in 2003 (treatment, F1,4= 11·2, P = 0·03; time, Wilks’λ= 0·30, F2,3= 3·52, P = 0·16; treatment × time, Wilks’λ= 0·31, F2,3= 3·34, P = 0·17), with a similar trend in 2004 (treatment, F1,4= 4·65, P = 0·097; time, Wilks’λ= 0·24, F2,3= 4·69, P = 0·12; treatment × time, Wilks’λ= 0·47, F2,3= 1·67, P = 0·33). In 2003, small beetle activity–densities were higher in fields with beetle banks at the beginning of the season, but declined and eventually converged to levels seen in fields lacking beetle banks (Fig. 1a; treatment × time, Wilks’λ= 0·001, F2,3= 7·34, P = 0·03). In 2004, fields with beetle banks supported significantly higher small beetle populations throughout the season than fields lacking this refuge (Fig. 1b; treatment, F1,4= 13·53, P= 0·02; time, Wilks’λ= 0·17, F2,3= 7·19, P = 0·07; treatment × time, Wilks’λ= 0·33, F2,3= 3·11, P = 0·19). The most abundant types of small beetles captured over the course of both years from all fields were the two Bembidion species, Bembidion tetracolum and Bembidion lampros (20·6% of total catch), and an Aleocharine morphospecies (23·0%). In 2003, Pterostichus melanarius activity–density was significantly higher in fields with beetle banks than those without beetle banks, across the season (Fig. 1c; treatment, F1,4= 15·31, P = 0·02; time, Wilks’λ= 0·35, F2,3= 2·73, P = 0·21; treatment × time, Wilks’λ= 0·34, F2,3= 2·90, P = 0·20). However, these differences were not distinct in 2004 (Fig. 1d; treatment, F1,4= 1·15, P = 0·35; time, Wilks’λ= 0·22, F2,3= 5·35, P = 0·10; treatment × time, Wilks’λ= 0·63, F2,3= 0·87, P = 0·50).
Predation of sentinel eggs was not correlated with small beetle activity–densities (Fig. 2a–c; May correlation coefficient, −0·03, P= 0·77; August correlation coefficient, −0·01, P= 0·26); the September correlation (correlation coefficient −0·04, P= 0·05) was not significant at the Bonferonni-corrected critical P-value (α = 0·017). Pterostichus melanarius activity–densities were also not correlated with sentinel egg removal (Fig. 2d–f; May correlation coefficient, −0·10, P= 0·24; August correlation coefficient, −0·01, P= 0·13; September correlation coefficient, 0·001, P= 0·80).
field cage experiments
Each experiment was designed as three sets of comparisons to test for cage effects, efficacy of addition and removal techniques, and differences among the main experimental treatments. Cage effects on beetle activity–densities were significant in experiments 1 and 3 (Table 2) and are discussed more fully in Prasad (2005). Small beetle activity–densities and egg predation rates were significantly lower in the Removal than in the corresponding addition treatment(s) for all three experiments (Table 2).
Table 2. Statistical output for analysis of predator beetle activity–density, sentinel egg predation and aphid density for field cage experiments
Treatment × block
Full repeated-measures analysis, of time series data are presented in Prasad (2005).
Efficacy of manipulation: Removal vs. No + Low + High (pooled)
F1,18 = 22·62
P < 0·001
F1,18 = 12·19
P = 0·003
Main treatment effects
F2,12 = 2·19
P = 0·15
F2,12 = 34·82
P < 0·001
F2,12 = 5·49
P = 0·02
Our Pterostichus melanarius manipulation in experiment 1 (treatments 0X, 1X and 4X) successfully altered activity–densities of this predator, but the pattern of these treatment differences varied slightly between the two blocks (treatment–block interaction; Table 2). In block 1, Pterostichus melanarius activity–densities were higher in 4X compared with the other two treatments (Fig. 3a; F2,12= 47·34, P < 0·01, Tukey-Kramer post-hoc comparison for both 0X vs. 4X, and 1X vs. 4X, P < 0·01), but in block 2 significant differences occurred only between 0X and 4X (Fig. 3b; F2,12= 4·88; P = 0·03, post-hoc 0X vs. 4X comparison, P= 0·03). Activity–densities of small beetles were significantly reduced by the addition of Pterostichus melanarius (Table 2 and Fig. 3c,d). Consistent with a negative effect of high Pterostichus melanarius activity–densities on small beetles, egg predation was significantly reduced with increasing Pterostichus melanarius activity–densities (Table 2 and Fig. 3e,f).
Small beetle activity–densities in experiment 2 increased in the high addition rate treatment (4X), while activity–densities remained lower with fewer beetles added (1X) (Table 2 and Fig. 4a,b). Differences in egg predation between 1X and 4X were not consistent between the two blocks (treatment–block interaction; Table 2): there were no differences in egg predation between 1X and 4X for block 1 (F1,8= 0·32, P = 0·58; Fig. 4c) but egg predation was significantly higher in the 4X than 1X treatment in block 2 (F1,8= 6·09, P = 0·04; Fig. 4d).
Although not manipulated experimentally, we observed that aphids were present within cages in both blocks of experiment 2, and we recorded their densities. Mean densities of resident aphids in block 1 (35·60 ± 9·28 aphids leaf−1) were significantly higher than those in block 2 (14·00 ± 2·21; Table 2).
Small beetle activity–densities did not differ between No, Low and High aphid addition treatments in experiment 3 (Table 2 and Fig. 5a). Final mean aphid densities were significantly higher in High (501·77 ± 83·88 aphids leaf−1) than in No (2·83 ± 1·11) or Low (10·17 ± 3·09) treatments (P < 0·01 for both post-hoc comparisons with High; P > 0·05 for No vs. Low comparison). Aphid addition significantly decreased egg predation (Table 2 and Fig. 5b). The significant aphid effect appeared to be driven by reduced egg predation in High compared with No (P = 0·02 for post-hoc High vs. No comparison); egg predation in Low was similar to both High and No (P > 0·05 for all remaining comparisons).
Our study consisted of two components, replicated on-farm measurements of beetle bank performance, and a series of manipulative field cage experiments designed to understand better the interactions among key species. Our on-farm results demonstrated that densities of predatory ground and rove beetles were higher during the growing season in fields with beetle banks compared with fields without beetle banks, consistent with many earlier studies (Lee, Menalled & Landis 2001). However, despite a dramatic increase in beetle activity–densities, there was no relationship between increasing beetle densities and rates of fly egg predation. Evaluation of conservation programmes aimed at promoting natural pest control must include measures of pest suppression, in order to be certain that higher predator densities do indeed lead to improved pest control (Gurr, Wratten & Barbosa 2000; Snyder, Chang & Prasad 2005).
We found that the ratio of small beetles to Pterostichus melanarius varied dramatically over the course of the growing season, particularly in fields including beetle banks. At the beginning of the season the predator guild was composed primarily of small beetles, but by the end of the season numbers of small beetles and Pterostichus melanarius were roughly equal. This pattern was consistent with previous work suggesting that Pterostichus melanarius acts as an intraguild predator of the smaller beetles (Finch & Elliot 1992; Prasad & Snyder 2004); it is only the smaller beetles that are themselves effective egg predators (Prasad & Snyder 2004).
In our field cage experiments we also observed a decline in both small beetle activity–density and a concomitant drop in fly egg predation, in the treatment including the highest Pterostichus melanarius density. This is consistent with the classic intraguild predation scenario, with higher densities of the large, intraguild predator leading to fewer intermediate predators and lower predation rates on herbivores (Polis, Myers & Holt 1989; Rosenheim, Wilhoit & Armer 1993, Rosenheim et al. 1995; Finke & Denno 2004). However, because we measured activity–densities of small predators, rather than absolute densities, we cannot exclude the possibility that small beetles reduced their foraging behaviour to avoid Pterostichus melanarius, and thus were trapped less frequently rather than actually falling victim to intraguild predation (Moran & Hurd 1994; for a similar conclusion regarding impact of carabids on lycosids see Lang 2003). Pterostichus melanarius’ role as an intraguild predator has been documented in several other systems (Dinter 1998; Snyder & Ives 2001). Because Pterostichus melanarius often responds positively to agri-environment schemes (Symondson et al. 1996; Shah et al. 2003; Raworth, Robertson & Bittman 2004), in systems where Pterostichus melanarius acts primarily as an intraguild predator this species may be a particular impediment to the success of conservation biological control programmes.
Within the group of smaller beetles, where intraguild predation is unlikely (Prasad & Snyder 2004), one would expect increased egg predation with increasing predator densities. Yet, even during May in beetle bank fields, there was no relationship between small beetle density and sentinel egg predation. While results from one block of experiment 2 demonstrated that higher small beetle activity–densities resulted in greater egg predation, an unexpected aphid outbreak in the other block suggested an additional disruptive force wherein particularly high aphid densities, combined with plants large enough for foliage to be in contact with the ground, further inflating predator–aphid encounter rates, prevented increasing egg predation with increasing small beetle density. In experiment 3, we directly manipulated aphid densities, and confirmed that fly egg predation rates declined at high aphid densities. A number of studies have reported that predation of a target herbivore is limited if generalist predators prefer or are distracted by alternative prey (Abrams & Matsuda 1996; van Baalen et al. 2001; Harmon & Andow 2004). Aphids, pollen and fungi have all been shown to disrupt predation of target pests by generalist predators (Dennis, Wratten & Sotherton 1991; Hazzard & Ferro 1991; Eubanks & Denno 2000; Musser & Shelton 2003). Providing generalist predators with alternate sources of prey, in order to maintain or build up their populations in fields when pests are absent, can be an effective conservation biological control tactic (Settle et al. 1996; Landis, Wratten & Gurr 2000). But, when conservation schemes promote higher densities of alternate prey at the same time that pests are present, predators can be distracted by or become satiated on the alternate prey, disrupting biological control (Harmon & Andow 2004). Generalists that feed on multiple prey species complicate conservation biological control. However, this does not render such programmes futile; rather, practitioners need to understand how key natural enemies are responding to multiple-prey communities (Eubanks & Denno 2000; Harmon et al. 2000).
Beetle banks, like some other agri-environmental schemes, target generalist predators. However, the broad diet breadths of generalists makes their community-wide impact difficult to predict (Rosenheim et al. 1995). Previous work in our study system has been characterized by high carabid and staphylinid feeding rates on fly eggs in the laboratory (Finch & Elliot 1994; Finch 1996) followed by disappointing performances in the field (Humphreys & Mowat 1993; Finch & Elliot 1994; Kromp 1999). The manipulative field experiments reported here allowed us to identify two disruptive forces: intraguild predation and predation of aphid alternative prey. These are likely to be common in the field and thus will limit the effectiveness of conservation biological control. The identity of the target prey (pest) species is also likely to be important. For example, in our system Pterostichus melanarius is disruptive because this predator is too large to feed on fly eggs (Finch 1996) and instead feeds on the smaller beetles that are the most common egg predators. However, if our target pest was another, larger herbivore, a mollusc for example, Pterostichus melanarius would be likely to make a positive contribution to control (Symondson et al. 1996). Our results suggest that the indistinct and context-dependent trophic role of these generalists will often limit egg predation under field conditions. However, the polyphagous nature of generalist predators will probably limit the maximum contribution of generalists to pest suppression, rather than eliminate the benefits of including generalists in the community. For example, throughout our study more eggs were eaten in treatments containing predators than in the no-predator controls.
synthesis and applications
Recent efforts to incorporate conservation biological control within agriculture have been characterized by a non-mechanistic approach, resulting in limited insight, after the fact, into why particular conservation schemes succeed or fail (Gurr, van Emden & Wratten 1998; Kean et al. 2003; Snyder, Chang & Prasad 2005). Perhaps more troubling, the data collected are often insufficient to allow performance to be assessed (Kleijn & Sutherland 2003). We advocate a more mechanistic understanding of how predator conservation impacts agricultural communities, either empirically (this study), using models (Kean et al. 2003) or through a combination of both empiricism and theoretical work (Snyder & Ives 2003). Mechanism-centred approaches facilitate a more directed (Kean et al. 2003) and predictive (Gurr, van Emden & Wratten 1998) approach to management for conservation biological control of species-diverse food webs.
Christiansen Seed Company donated radish seeds. N. O’Neal, E. Haakenson, G. Johnson, B. Harris, M. Yonquist, M. Ballon and the Snow Family allowed field access. J. LeBonte (Oregon Department of Agriculture) provided beetle identification. This research was supported by grants from the Organic Farming Research Foundation and USDA's Western Sustainable Agriculture Research and Education program.