Fifty-four articles fitted the selection criteria and were included in the study. In the overall analysis, the mean total score for a research paper was 32% of the maximum score possible; 39 (72%) of studies scored only 21–40% and only two studies scored more than 50% (Figure 1A). The mean score for five of the parameters analysed ranged from 28% to 44% (Figure 1B). Only one of the analysed studies reported a change in management practice to incorporate predation into pest-management programmes as a result of reported research findings, resulting in a mean overall score of only 1.9% for this parameter (Figure 1B). Studies which comprehensively investigate the impact of predators on the population dynamics of pest insects are difficult to do. They require careful design and planning, necessitate frequent field data collection over extended periods of time, and are extremely labour intensive. The fact that few research projects have the resources to painstakingly study all aspects of a given system and that even those that do are likely to publish results in a series of papers which separately examine a single aspect of the system probably contributes to these low scores. However, the analysis clearly identifies deficiencies in the approaches currently adopted to assess the impact of arthropod predators on lepidopteran pest populations. Consequently, in many systems appropriate ecological evidence to support the inclusion of predators in IPM programmes has simply not been collected.
Scale of experimental studies
The issue of scale is fundamental to any ecological study (Englund & Cooper, 2003). Meaningful estimates of the impact of predators on the population dynamics of lepidopteran pests are difficult to determine, as the scale over which insects can move is often far greater than that of the study site (usually a small plot) or management unit (a farm field). Consequently, measurements of generational mortality, an essential metric for valid decision-making in any IPM programme, are frequently confounded by immigration and emigration events of both the pest and potentially important predators. To appositely assess interactions between pest insects and their predators for pest management, studies need to be designed at the scale which determines pest and predator population dynamics or, at the very least, at the scale at which pest-management decisions are made, usually a field. Landscape effects can have a greater influence on the diversity of predatory insects than farm-management practices (Weibull et al., 2000, 2003). However, in our literature survey only one study (Bianchi et al., 2005) investigated landscape effects on predation (Figure 2B) but it did not identify the key predator species and only investigated predation events on sentinel egg cards.
Figure 2. (A) Scale of studies: proportion of studies conducted at landscape, farm, field, and plot scales. (B) Measurements of prey abundance: proportion of studies which did not measure prey abundance, which measured it in isolation of predator abundance, and which measured it following predator population manipulation. (C) Impact of predation: proportion of studies in which specific stages of prey were utilized. (D) Measurements of (putative) predator abundance: proportion of studies which did not measure predator abundance, which measured it in isolation of prey abundance, and which measured it and related it to impact on the prey population.
Download figure to PowerPoint
Several studies were conducted at the landscape scale and either investigated treatment effects between paired large fields (>1 ha) within a given landscape (Head et al., 2005; Furlong et al., 2008) or compared farms adopting different pest-management practices within a given agricultural region (Furlong et al., 2004b; Delate et al., 2008). Such studies can be extremely informative as they allow direct comparison of the effect of given management practices (e.g., organic, IPM, conventional, cultivation or transgenic crops) at the scale at which they are actually deployed. The vast majority of studies (42 of 54) were conducted at the plot scale, usually much <1 ha. Issues of cost, labour, and, in the case of transgenic crops, licensing agreements often limit the scale at which studies can be conducted to a size that is considerably smaller than the scale at which such crops are grown in agriculture. Although this can be justified, the size and spatial distribution of plots can significantly affect arthropod abundance in experimental crops (Prasifka et al., 2005).
Consequently, extrapolation of results across scales is not always possible and results of small-scale experiments should be interpreted with caution when they are projected to a larger ecological scale. In fact, there is a real danger that such extrapolations will lead to spurious conclusions regarding predator efficacy; research questions should be formulated and experiments designed at a scale which is relevant to the organisms under investigation. In practice, this should be an area circumscribed by the average distance moved by a predator from when it emerges and starts feeding to when it ceases feeding (ignoring long-distance migration).
Predator population manipulation and assessment of prey mortality
To assess the impact of natural enemies on a target prey population, prey survivorship should be compared in the presence and absence of predators over a minimum of a single generation. Experimental methods to assess natural enemy impact have been extensively reviewed (Bellows & Van Driesche, 1999; Luck et al., 1999) and favoured approaches include the physical exclusion of natural enemies by barriers, prey augmentation or enrichment and the insecticidal check method advanced by De Bach (1946). Of the literature surveyed, only 23 of the 54 studies investigated used one or more of these approaches to manipulate predator populations and compared the effects on the target prey population with an appropriate control (Figure 2B). The remaining studies either did not measure prey abundance or they did so without manipulating predator density (Figure 2B). Such approaches are meaningless for the evaluation of the pest control potential of putative predators, because without an appropriate control, no conclusion is possible.
Furlong et al. (2008) used natural enemy exclusion cages and paired cabbage fields that were managed either conventionally (calendar application of broad-spectrum insecticide) or by an IPM strategy to show that the impact of predators and parasitoids on diamondback moth, Plutella xylostella L. (Plutellidae), populations was significantly increased by adoption of the IPM strategy. In a similar approach, Head et al. (2005) studied the predation of sentinel Helicoverpa zea (Boddie) (Noctuidae) eggs in paired fields of conventional and Bt-cotton to demonstrate that typical insecticide use in the conventional crop significantly reduced the impact of predatory arthropods. Grundy (2007) released nymphs of the assassin bug, Pristhesancus plagipennis Walker (Hemiptera: Reduviidae), into cotton crops and showed that the predator, which was compatible with several selective insecticides, significantly reduced the Helicoverpa spp. target pest population when compared with sites where the predator was not released. Although no studies adopted the insecticidal check method to specifically investigate the impact of predation on a target pest species per se, several studies adopted its principles to demonstrate that commonly used conventional insecticides negatively impact on predatory arthropod populations. The more useful of these studies estimated pest mortality rates under the different insecticidal and control regimes (Atanassov et al., 2003; Musser & Shelton, 2003a; Ehler, 2004) but many studies simply compared putative predator diversity and abundance under the different regimes (Figure 2B); such studies are of limited use and provide no information about the impact of predators on prey populations.
Twenty of the 54 papers examined did not measure prey mortality at all (Figure 2C) and of the studies which did, none strictly measured mortality over a complete generation. Even the most comprehensive studies measured mortality across egg and pre-pupal stages but only incompletely in the pupal stage (Furlong et al., 2004a, 2008), or from neonate to adult emergence but only incompletely in the egg stage (Xiao et al., 2007). Of the 34 studies which did attempt to measure prey mortality, 12 (22% of all studies, 35% of studies which measured predation) used experimental cohorts of eggs that were grouped on artificial substrates and then transferred to the field. Although such studies have some merit – for example, they are useful to demonstrate whether or not egg predation occurs in a given system – they are clearly limited in that they only estimate mortality of one life stage of the pest and, in isolation of the generational mortality suffered by the target population, they offer little information about the impact of predation on prey population dynamics. Although Mansfield et al. (2003) used this approach thoughtfully to deploy single eggs of Helicoverpa armigera (Hübner) (Noctuidae) within an experimental crop, these methods are often employed to study sentinel groups of eggs laid by insects which naturally oviposit singly, resulting in unrealistic densities of eggs in pseudo-oviposition sites. Such limitations are likely to affect predator foraging and behaviour and lead to spurious conclusions regarding predator–prey relationships. Ideally, estimates of egg predation should be conducted on eggs laid in natural oviposition sites by field populations of the target pest. Although such studies can be difficult to design and execute, they are possible and can yield important information regarding the biotic mortality factors affecting eggs of lepidopteran pests (Schmaedick & Shelton, 1999; Midega et al., 2006).
A further eight studies investigated the predation of sentinel larvae that were placed in crops. Such studies suffer from similar limitations to those which deploy sentinel laboratory-derived eggs and problems with the approach are accentuated when, as is most often the case, only larger stages of larvae are offered as potential prey items. The latter approach ignores the substantial mortality that occurs in early instars, particularly around hatching (Zalucki et al., 2002). Although the mobile nature of lepidopteran larvae is a real constraint to these studies, realism is diminished further when, in order to aid the location and recovery of prey items not attacked for the estimation of predation rates, prey items are pinned or tethered to plants. In such studies, the fact that interactions between predators and immobilized prey affect both predator hunting and prey escape behaviours, and that consumption of dead prey actually measures scavenging and not predation, is often completely overlooked. Despite the limitations of this approach, Frank et al. (2007) successfully used video recordings of predators attacking immobilized prey items to identify important predators within a vineyard agro-ecosystem. Further development of such techniques will be extremely useful as the approach allows the predator species which attack and consume a target pest in a given system to be identified relatively cheaply and quickly without the need for tedious observations, which may themselves disrupt predator behaviour (Wade et al., 2005). However, such studies should form only a first step in the evaluation of arthropod predators as they are of little significance unless complimented by studies to measure the impact of the predatory species on the target prey population.
The effects of predators on prey individuals as they developed from the egg through to larval stages were investigated in eight studies, whereas a further three studies investigated the impact of predators on pests as they developed through the larval and pupal stages (Figure 2C). Such studies are more useful than studies of egg, larval or pupal stages alone as they investigate mortality across multiple stages of the life cycle and consequently provide more meaningful information on the likely impact of predation on pest population dynamics. Hooks et al. (2006) conducted manipulative experiments on naturally laid cabbage white butterfly, Pieris (Artogeia) rapae (L.) (Pieridae), eggs and resultant larvae to conclusively show that endemic spiders could suppress the pest population and increase crop yield. In a similar approach, Xiao et al. (2007) used a combination of sticky barriers and fine mesh cages to determine that ant predation of the citrus leaf miner, Phyllocnistis citrella Stainton (Gracillariidae), had the greatest impact of all biotic agents on pest populations in Florida citrus groves. Such studies are extremely informative and provide clear and unambiguous evidence that endemic predators can depress pest populations. This information is vital before any sensible plan can be made to usefully exploit the predatory fauna in a given agro-ecosystem. However, the experiments require an understanding of the system, careful planning and execution, and are time consuming and labour intensive to conduct – all factors which probably explain why studies of this type are so underrepresented in the literature (Figure 2C).
Predator abundance and diversity
Studies that only measured the abundance and/or diversity of putative predators in agricultural or forest landscapes were expressly excluded from the analysis. Nevertheless, only 30 (55%) of analysed studies measured predator abundance and related it to impact on the target pest population in some way (Figure 2D). No attempt was made to measure predator abundance in 14% of studies and in a further 31% of studies predator abundance and/or diversity was measured but no attempt was made to link this metric with pest suppression (Figure 2D). Such studies of biodiversity in agricultural landscapes have become fashionable in recent times, but they are of very little value and add nothing to our understanding of the impact of predators. Although generalist predators can contribute to the effective biological control of pests (Symondson et al., 2002) increasing predator abundance does not necessarily result in greater pest mortality (Thomson & Hoffmann, 2009) or pest population suppression. Information on prey preference and diet breadth of many putative predators in the field is lacking and the tacit assumption in many studies that predators have catholic diets is incorrect (Thompson, 1951). Extending this point, it does not follow that increased predator biodiversity will necessarily increase predation rates and hence pest population suppression. Indeed, a recent review of studies investigating the relationship between predator biodiversity and prey suppression found that such diversity can have positive, negative or neutral effects on prey suppression because of niche complementarity, intra-guild predation and functional redundancy respectively (Straub et al., 2008). Not surprisingly, the impact of predator diversity on pest populations is highly context specific and although predator biodiversity was shown to significantly impact on aphid populations on collards (Snyder et al., 2006) a similar study on aphid populations on potato indicated that the presence of a key predator species, rather than predator biodiversity, mediated the degree of aphid population suppression (Straub & Snyder, 2006). Studies such as these support the contention of Landis et al. (2000) that the ‘right diversity’ rather than biodiversity per se is important, and clearly expose the shortcomings of studies which simply measure biodiversity or pest-management strategies which promote predator abundance and diversity, but do not conduct appropriate experiments to quantify impacts on target pest populations.
There is considerable evidence that organic farming promotes greater arthropod biodiversity than conventional approaches to agriculture (Bengtsson et al., 2005) and the literature is replete with claims that this will necessarily lead to increased ecosystems services and improved pest control (Letourneau & Bothwell, 2008). Although firm evidence that the increased biodiversity promoted by organic agriculture does increase pest suppression through the provision of a higher diversity of natural enemies is currently lacking (Letourneau & Bothwell, 2008) these systems do provide an opportunity to test the mechanisms of pest suppression at a scale appropriate to pest and natural enemy management. Recent advances in molecular approaches to detect specific prey items in the guts of predatory arthropods (Symondson, 2002; Sheppard & Harwood, 2005) provide methods by which predatory species which actually feed on prey items of interest can be identified relatively easily and cheaply. To date, these methods have principally focused on identifying trophic links within predator–prey food webs. Although not without pitfalls of their own (e.g., difficulties in distinguishing between predation and scavenging) if carefully linked with appropriate predator–prey population studies and additional molecular methods that enable the extent of predator and prey movement in agro-ecosystems to be determined (Hagler & Naranjo, 2004), these techniques offer enormous potential to provide insight into predator–prey interactions at the field scale. Such integrated approaches will provide entomologists with the capabilities to identify the predatory species which suppress pest populations, the crucial but often neglected first step in any conservation-based approach to pest management. The challenge now is to make use of this potential and take it out of the laboratory and apply it in the field.