Ecological Modelling 182 (2005) 199–216

Insect behavioural ecology and other factors affecting the control efficacy of agro-ecosystem diversification strategies R.P.J. Pottinga,∗ , J.N. Perryb , W. Powellb a

Laboratory of Entomology, Wageningen University, P.O. Box 8031, 6700 EH Wageningen, The Netherlands b Plant and Invertebrate Ecology Division, Rothamsted Research, Harpenden, UK Received 30 June 2003; received in revised form 8 June 2004; accepted 4 July 2004

Abstract In the last decade there is an increased interest in the design and use of diversified pest-suppressive agro-ecosystems. A diversification approach aims to manipulate the spatial dynamics of herbivores by adding a trap crop that attracts and retains herbivores in the non-crop area or by adding a disruptive crop that induces an emigration response. Empirical studies have shown that there is a wide variation in insect herbivore response to vegetation diversification. To increase the predictability and reliability of this approach it is necessary to understand the mechanisms underlying herbivore population response to diversified agro-ecosystems. We use a spatially explicit, individual-based, simulation framework, with a strong emphasis on the behavioural ecology of insects, to explore the factors that influence the population regulatory effect of agro-ecosystem diversification. The reported wide variation in population response of herbivores to diverse agro-ecosystems is replicated in this study. In our simulations we found that the population regulation effect of diversification can be positive, negative or negligible. Behavioural factors that influenced the spatial dynamics of herbivore populations were the colonisation pattern, movement speed and sensory mode of finding host plants. Simulations show that the strength of inhibition of flight by the trap crop (i.e. arrestment) is the most important parameter to manipulate the spatial dynamics of insects with post-alighting host recognition behaviour. For herbivore species that use olfactory or visual cues to find host plants, the mechanism of aggregation in the trap crop is a combination of attraction and arrestment and hence the population regulatory effect of the trap crop is higher compared to herbivores with post-alighting host recognition behaviour. An important factor that influences the efficacy of the disruptive cropping strategy is the strength of the emigration-inducing effect of the vegetation. The simulation framework is a valuable tool to test hypotheses on insect behaviour and dynamics in heterogeneous environments and can be used to determine optimal diversification systems and hence generate guidance for establishing environmentally benign pest control strategies. © 2004 Elsevier B.V. All rights reserved. Keywords: Individual-based model; Spatial dynamics; Agro-ecosystem diversification; Insect behaviour; Movement; Trap crop; Intercrop

1. Introduction ∗

Corresponding author. E-mail address: [email protected] (R.P.J. Potting).

0304-3800/$ – see front matter © 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.ecolmodel.2004.07.017

In the last decade concerns for sustainability have replaced the maximisation of productivity as the tar-

200

R.P.J. Potting et al. / Ecological Modelling 182 (2005) 199–216

get for agricultural development. This has generated increased interest in integrating agro-ecosystem diversification with integrated crop management (Wood and Lenn´e, 1999). As a pest management strategy it relies on adding specific pest-disrupting or natural enemy enhancing vegetation to the agro-ecosystem, with the aim of decreasing pest density and damage in the main crop. However, field studies of herbivore population response to diversified environments show an enormous variation in the level of population regulation. In a review of studies on herbivore population response to diversified agro-ecosystems, Andow (1991) reported that the population density of insect herbivores in polycultures compared to monocultures was in 52% of the studies lower, in 15% higher, in 13% equal and in 20% of the studies variable. This variation in pest control potential may be one of the main reasons why farmers have not yet accepted agro-ecosystem diversification in integrated crop management (Banks and Ekbom, 1999). Thus, there is an urgent need to develop a mechanistic framework to understand and predict the response of herbivores and natural enemies to spatial arrangements of vegetation in agricultural systems. In this paper, we use a modelling approach to get more insight in the mechanisms underlying the efficacy of diversification strategies and to give an indication of the reliability of agro-ecosystem diversification as a pest management strategy. The source of variation of the pest regulatory effect of diversification is thought to be a combination of differences in ecology of the herbivore and ecological aspects of the vegetation in the agro-ecosystem (Andow, 1991). Insect herbivores differ widely in their behavioural ecology, such as their colonisation pattern of agricultural fields, within-field movement behaviour and hostplant finding behaviour (Bernays and Chapman, 1994; Schoonhoven et al., 1998). Empirical studies have demonstrated that different herbivore species respond differently to diversification (e.g. Bach, 1988b; Banks, 1998). Differences in the ecological properties of the additional vegetation in diversified agro-ecosystems include the strength and mode of behavioural modifying effect on the herbivore, as well as the scale of diversification (i.e. the fraction and spatial configuration of the additional vegetation). In general two types of crops are used to manipulate herbivore response: disruptive (repellent) crops or attractive trap crops (Vandermeer, 1989). Banks and Ekbom (1999)

modelled herbivore colonisation and movement in diversified agro-ecosystems and showed that both percent crop cover and herbivore movement rates were critical in determining herbivore densities. The objective of our study was to extend this modelling approach by varying both the behavioural ecological parameters of the animal and the composition and spatial arrangement of the vegetation in the ecosystem. The behavioural and chemical ecology of insects, and in particular the searching behaviour for resources such as host plants by herbivores or prey by carnivores, has been well investigated (Bell, 1991a,b; Vet and Dicke, 1992; Papaj and Lewis, 1993). Most of these studies were conducted using laboratory bioassays and small-scale field trials and for several species there is a detailed knowledge of individual foraging behaviour and interactions with both plants and antagonists. However, there is still a huge, unrealised potential in the use of this insect-specific ecological information to elucidate insect community and population dynamics in open heterogeneous field systems. Although there are some exciting novel techniques for tracking insects in natural environments (Riley and Drake, 2002), understanding and predicting insect dynamics in open field systems largely depends on modelling techniques (Holt and Cheke, 1997). We present an individual-based, spatially explicit, simulation framework, with a strong emphasis on sensory abilities and movement patterns of insects. Object-oriented, individual-based modelling is a powerful and flexible way to bridge the gap between individual behavioural ecology and population dynamics (DeAngelis and Gross, 1992; Judson, 1994). Such models can handle biological detail at a low level of organisational complexity (e.g. behavioural of individuals) and calculate the consequent higher scale trophic dynamics (e.g. Uchmanski and Grimm, 1996; van Lenteren and van Roermund, 1997; Berec, 2002; Mamedov and Udalov, 2002). Thus, these models are perfectly suited to use information from small-scale experiments to identify relevant mechanisms of population dynamics at a field-scale and allow extrapolation to novel conditions, for example to predict the response of insect herbivores to agro-ecosystem diversification. In this paper, we first explain how we have generalised and incorporated the main aspects of the behavioural and dispersal ecology of insects in the simulation environment. We explore the influence of

R.P.J. Potting et al. / Ecological Modelling 182 (2005) 199–216

aspects of the ecology of the herbivore and the cropping environment on the efficacy of trap cropping and disruptive cropping as a pest management strategy. The simulation model will allow us to explain which behavioural and ecological characteristics are important and to predict which cropping practices can be used as a reliable sustainable pest management strategy.

201

2. Methods 2.1. General model structure In the simulations presented here, a univoltine population of adult herbivorous insects colonises, and subsequently disperses within, a diversified agricul-

Fig. 1. Flow diagram of the behavioural events for each individual for each timestep in the simulation model.

202

R.P.J. Potting et al. / Ecological Modelling 182 (2005) 199–216

Table 1 Overview of parameters in the simulation model Parameter

Default

Alternative states, values

Environment Grid size Fraction non-crop vegetation Spatial configuration non-crop vegetation

100 × 100 cells 0.25 Intercropped rows

Not used Not used Border, 25 patches

Animals Colonisation pattern Natural daily mortality rate Plant specific flight tendency (f) Maximum movement length (s) Emigration tendency (e) Sensory mode In-flight perception probability (p) Memory vector size (m) Damage threshold (D)

At random (airborne) 0.001 0.5 10 0.05 Contact 0.7 5 100

Directional (sideways) Not used 0.05 (trap crop) to 1 (repel crop) 1–20 Not used Olfactory, visual Not used Not used 1–100

Simulations Population size Timesteps Runs per scenario

500 50 40

Not used Not used Not used

tural landscape during a fixed time. The simulation model includes behaviour-based stochasticity and spatially explicit structures based on vegetation density and composition. A flowchart for the order of events for each individual (for each timestep) is given in Fig. 1 and an overview of the parameters in the model is given in Table 1. Basically, the model keeps track of the age, movement and position of each individual. Individual movement decisions are based on the plant type and quality encountered at the current position. These generally result in arrestment responses on preferred habitat types and displacement responses on less or non-preferred habitat types. By keeping track of the number of herbivore visits in each cell, predictions can be made on the expected crop damage – resulting from particular agro-ecosystem diversification strategies.

sion to the herbivore. The preference or non-preference for particular plant/habitat types is simulated by setting habitat-specific movement tendencies and detection probabilities for the herbivore (see below). To investigate the potential effect of agro-ecosystem diversification on herbivore damage levels in the main crop, the environment was initialised to consist of main crop vegetation (75% of total area) and herbivoreattractive or repellent non-crop vegetation (25% of total area). To investigate the potential effect of the spatial arrangement of the non-crop vegetation on damage levels, the non-crop vegetation was placed in three main spatial arrangements (Fig. 2). These consisted of: (1) a non-crop border of 7 rows around the crop, (2) noncrop strips in an intercrop (five strips each with width of five rows) and (3) non-crop in patches of varying size (default: 25 patches of 10 × 10 cells).

2.2. Environment The spatial environment is represented by a square lattice of 100 × 100 individual cells. The environment is heterogeneous in space and time: each cell represents a plant of a specified type (i.e. plant species) and each cell can be empty or occupied by one or more animals. The model is initialised with a specified fraction and spatial arrangement of up to 10 plant types. The plant types are typified by their level of attraction or repul-

Fig. 2. Examples of spatial structures used in simulations (each based on a coverage of 25% trap crop (black area) as a proportion of the total area of 10,000 cells). (A) Intercrop 5 blocks each 5 rows wide and 100 long (B) Border over 7 outermost rows (C) 25 Patches each 10 × 10 cells.

R.P.J. Potting et al. / Ecological Modelling 182 (2005) 199–216

2.3. Colonisation pattern In annual agro-ecosystems crop damage is typically caused by herbivorous insects that move at low altitude into fields from surrounding areas or immigrate after moving greater distances at higher altitudes in the air. The latter are usually small insects, such as whiteflies, aphids and thrips that have a limited ability to control their flight direction and have sometimes been described as aerial plankton (Johnson, 1969; Price, 1976). Thus, the colonisation pattern of these small herbivores is largely due to passive, random, high-altitude aerial dispersal. In contrast, larger insects such as beetles, moths and butterflies can be strong fliers with an ability for directional flight, often upwind. Besides aerial colonisation, these insects may also actively colonise a crop system from wild hosts in surrounding areas. For example, the Colorado potato beetle (Leptinotarsa decemlineata,) locates and invades fields through directional flight and by walking, resulting in an aggregation of colonising beetles along field-edges (Blom et al., 2002). The effect of colonisation pattern of the herbivore population was investigated by varying the colonisation source. The initial population was either located randomly, simulating airborne colonisation (Fig. 3B) or, to simulate an insect species with an active directional colonisation pattern, the colonisation was initialised from a source population at one side of the grid. Here, the x position of the colonising animals was randomly drawn from an exponential distribution with a mean of 0.5 and the y position randomly drawn be-

203

tween 1 and 100 or vice versa, resulting in colonisation from the left side or top side of the field (see Fig. 3A). 2.4. Natural mortality & migration Survival was modelled at the individual level. Agespecific mortality was modelled by specifying that an individual will survive to the next day with probability 1−(0.001 × age of individual). Dead or emigrated animals were removed from the simulation environment. Thus, those n individuals that remained in the simulation environment were carried over to the next timestep (to form the population at time t + 1) with agen = agen + 1 and new n(x, y) position. 2.5. Movement behaviour At each timestep each animal is allowed to move. The movement tendency and movement length depends on the type and state of the currently occupied cell. Each plant type has a specified effect on the flight tendency of the visiting animals (see below). The state of the plant (i.e. cell) is determined by the herbivore damage level, which is used as an indirect indicator of animal density on that plant. First it is established if the animal will make a short movement, no further than the neighbouring plant, or initiate a flight away from the currently occupied cell. A flight is initiated when a randomly generated number between 0 and 1 is below the flight probability f of the current position, which is initiated according to the type and state of the plant. In general, low quality plants induce flights

Fig. 3. Typical examples of airborne colonisation patterns, either (A) with active directional component from source at left side or (B) at random.

204

R.P.J. Potting et al. / Ecological Modelling 182 (2005) 199–216

and high quality plants induce ‘area restricted search’ (arrestment) responses (Morris and Kareiva, 1991). If no flight is initiated, the animal has an equal probability (0.5) of either staying in its current cell or hopping to a neighbouring cell. The direction of a hop or flight is random. The movement length of a flight is generated by randomly selecting a number between 1 and a specified maximum movement length s. For each generated flight there is a background probability e (by default e = 0.05) for emigration from the field. The effect of displacement speed of the herbivore on trap crop efficacy was investigated by varying s. We compared maximum displacement speeds of 1, 3, 5, 10 and 20 cells per timestep, where slow moving (i.e. walking, hopping) animals are represented by s = 1 (i.e. nearest neighbour dispersal), and the fastest movers by s = 20 (i.e. flight dispersal). 2.6. Migration Animals could leave the simulation environment by emigration to the air column or at the grid edge. For each generated flight there was a fixed probability (e = 0.05) that the animal would leave the field by airborne emigration. If the generated flight path crossed one of the field-edges there was a fixed probability (e = 0.05) that the animal would leave the field. For each returning animal a random position was generated in a square window of size s, with as base the (x, y) position of the exit point of the flight path (Fig. 4). 2.7. Host-plant preference ranking Most herbivorous insects have a specific host-plant range and within this range they usually exhibit a hierarchy in their preference for particular host-plant species (Bernays and Chapman, 1994). An insect seeking a suitable oviposition or feeding site will preferentially select and spend more time on the most preferred plant species compared to less preferred plant species and will avoid or move away from non-host plants (Thompson and Pellmyr, 1991). Thus, insects adjust their rate of movement in response to the quality of their current environment. For example, host-plant induced movement patterns in Lygus rugulipennis affected the spatial distribution of individuals in heterogeneous arenas (Hannunen and Ekbom, 2001).

Fig. 4. Graphical representation of implementation of emigration events and possible return paths at field edge. Cell a indicates start position, cell b position of border crossing and cell c randomly generated end position. Dotted line arrows indicate possible return paths within return window. Possible emigration events are indicated by e1 (probability of airborne emigration for each generated flight) and e2 (probability of emigration at field edge).

In the simulation environment, habitat types are characterised by their effect on the flight tendency f of the animal: a preferred habitat is typified by a low flight probability (fp = 0.05) compared to the main crop (fn = 0.5) or repellent plant (fr = 1). There is considerable interest in using highly preferred vegetation as a trap crop strategy in pest management (Hokkanen, 1991). To investigate the behavioural and environmental factors affecting the control mechanism of trap cropping, population dynamics of typified herbivore species were simulated in agro-ecosystems containing trap crop vegetation in specified spatial arrangements. 2.8. Sensory abilities in host selection behaviour The spatial distribution of the damaging life stages of herbivores is largely determined by the oviposition behaviour of adult females, and different species differ in their host selection behaviour. Some species, usually active fliers, have good sensory abilities, allowing them to use visual or olfactory cues to recognise a preferred host-plant whilst in-flight (Prokopy and Owens 1983;

R.P.J. Potting et al. / Ecological Modelling 182 (2005) 199–216

Fig. 5. Visual representation of implementation of different searching/detection mechanisms. (A) Contact; (B) olfactory; (C) visual. Grey boxes indicate preferred habitat. Dotted line indicates random generated direction and length of next movement. Solid arrow indicates actual path.

Visser, 1986; Finch and Collier, 2000). In species with a limited ability for directional dispersal, plant recognition usually occurs after a female has landed, based on physical or chemical properties of the plant. For example, whiteflies (Homoptera: Aleyrodidae) move short distances from plant to plant until they find acceptable hosts (van Lenteren and Noldus, 1990). We compared three different types of habitat detection mechanisms: contact search, olfactory search and visual search. For contact search we assumed that all animals have the basic sensory capacity to recognise the suitability of a host-plant after alighting, by encountering either stimulatory or deterrent contact semiochemicals on the plant surface or during probing of plant tissues. However, within the model, a contact searcher is assumed unable to detect preferred habitat types along its randomly generated flight path (Fig. 5). For olfactory search, in contrast to post-alighting host recognition, we consider animals that practice prealighting host recognition to have the sensory abilities to recognise and respond to a specific habitat while in-flight. Here, we make a distinction between an olfactory searcher that is assumed to respond to attractant or repellent volatile semiochemicals encountered along its flight path and a visual searcher that can detect and respond to specific visual cues within the model environment (see Fig. 5). The in-flight perception ability (p) of an olfactory searcher was implemented by setting a fixed probability (default p = 0.7) that each cell of the preferred habitat type within the generated movement path can induce a landing response (i.e. interrupted flight). The search path was set at the maximum movement length in the randomly generated direction with a

205

path width of 1 cell. This type of host searching can be seen as olfaction-guided upwind flight, where the wind direction is random for each generated flight and an animal can recognise (i.e. smell) preferred sites within its search path in the forward direction of movement. We assumed that animals can always recognise their preferred habitat type at their current position. Thus, contact searchers are typified by individuals with p = 0 and thus cannot recognise a preferred habitat in-flight, but do recognise and respond to a preferred habitat type after landing. A visual searcher scans the environment around its current position in concentric circles (Fig. 5) up to the maximum movement length. Thus, the area searched for preferred cells is much larger and less unidirectional than that of the olfactory searcher. 2.9. Informational state of animal Awareness has grown of the importance of learning in the searching behaviour of insects. Optimal foraging theory predicts that animals should focus their searching effort towards the most profitable resources (Stephens and Krebs, 1986). Many insect species can learn visual or olfactory cues associated with successful host location and use these learned stimuli in subsequent foraging decisions (Papaj and Lewis, 1993). In heterogeneous environments, this may lead to a preference shift towards rewarding habitat types, also termed preference learning (Turlings et al., 1993). In the simulation environment, movement decisions and perception abilities were dependent on the informational state of the animal, and individuals were able to use information from previous foraging experiences. This was implemented by giving each individual in the model a memory vector that holds the sequence of habitat types of the previous encountered sites. We used the‘sliding memory window’ approach (Mangel and Roitberg, 1989), where the memory vector (i.e. array of fixed size) is constantly updated so that when new information is entered (habitat type of currently visited site) the most distant memory is lost. Foraging decisions of each individual animal are based upon the composition of their memory vector. Thus, an animal with a memory vector of size 5 will remember and base its foraging decision on the habitat types encountered at the last five visited sites. Animal species with a memory vector of 0 will not use information from pre-

206

R.P.J. Potting et al. / Ecological Modelling 182 (2005) 199–216

viously visited sites. This cognitive ability allows us to investigate the effects of storing (by varying the size of memory window) and processing (by varying decision rules) the stored information. The effect of cognitive abilities on the foraging behaviour of animals will be dealt with in a separate paper, but in the simulation results presented here we used the memory function to determine the response to unrewarding (i.e. repellent) habitat types (see below). 2.10. Density dependent behaviour Many insect species alter their behaviour in the presence of conspecifics, often avoiding sites with conspecific competitors. This avoidance behaviour can be induced by visual or physical contact with conspecifics or by encounters with indirect cues such as epidiectic pheromones left behind by ovipositing females (Roitberg and Prokopy, 1987). A common indirect density dependent effect in herbivore populations is mediated through herbivore damage to host plants. Many patterns attributed to density effects in herbivore populations may actually be mediated through herbivore damage to host plants: a decline in plant quality caused by accumulated herbivore damage. For several herbivore species it has been demonstrated that they can detect/respond to plant volatiles induced by insect feeding damage (Bernasconi et al., 1998). The cumulative number of previous visits by herbivores in a particular cell (cumulative herbivore days, CHD) was used as an indicator of herbivore feeding damage in that cell. We assumed that herbivores could access the level of damage in a cell and use this information in their decision to leave low quality cells. Thus, density dependent avoidance behaviour was manifest as indirect feeding competition. The decision not to use or to leave an already damaged cell was dependent on the damage threshold of the habitat type of the cell. If CHD was above the damage threshold of this cell type, it was treated as a repellent habitat type, resulting in an avoidance response (see below). The effect of a density dependent response was investigated by setting a cumulative damage threshold D (1, 3, 5 and 100).

1). We compared the control efficacy of deterrent or repellent non-crop vegetation. A deterrent plant differs from a repellent plant in the mechanism of recognition and avoidance. A deterrent cell is only recognised as an unsuitable habitat after landing and settling on the plant (e.g. contact response of sucking insects) resulting in a delayed avoidance flight in the next timestep, whereas a repellent habitat is recognised from a distance when approached in-flight (e.g. response to visual or olfactory cues), inducing an extra prolonged flight away from the plant within the same timestep (Schoonhoven et al., 1998). If an animal encountered five repellent sites in one sequence, this was assumed to induce emigration from the field. The effect of repellency strength of the non-preferred habitat type was simulated by setting an elevated emigration rate e after encountering a repellent site. 2.12. Model output parameters Each computer run simulated the displacement of 500 insects. All runs were terminated after 50 timesteps or after the population became extinct. The number of replicate simulation runs was 40, which resulted in a sufficiently low standard error of the mean. For each timestep (i.e. day) and for each cell, the model keeps track of the animal density and cumulative number of herbivore days (CHD). Herbivore induced crop damage was estimated as the mean number of herbivore days in a crop cell. This was calculated as total CHD divided by the number of crop cells in the system (e.g. 10,000 for monocrop, 7500 for 25% trap crop coverage). The mean CHD in a crop cell, representing average crop damage, was used as the main output parameter to compare simulation runs. To investigate the effect of different management scenarios and herbivore behavioural ecologies the population control impact was assessed by analysing the estimated herbivore damage (mean CHD in a crop cell) resulting from variations in the model inputs. Because the system keeps track of all events in all the cells it was possible to visualise the dynamic state of the simulation, such as the spatially explicit density of animals and CHD.

2.11. Disruptive non-crop vegetation

2.13. Sensitivity to parameter values

A non-preferred habitat type was typified by setting a high probability of an induced flight response (fr =

The effect of behavioural and agro-ecosystem traits on the population dynamics of the herbivore were in-

R.P.J. Potting et al. / Ecological Modelling 182 (2005) 199–216

207

Table 2 Overview of scenarios and factorial designs that were modelled to explore agro-ecosystem diversification efficacy Trap crop efficacy Sensory modality of herbivore (contact, olfactory, visual) and movement speed (s = 1, 5, 10, 20 cells per timestep). Colonisation process (airborne, sideways) and spatial configuration trap crop (border, intercropped rows, 25 patches). Density dependent response (high to low: D = 1, 3, 5, 100) and spatial configuration trap crop (border, intercropped rows). Disruption crop efficacy Disruption mechanism (deterrent, repellent) and strength of disruptive effect (edisruptive crop = 0.05, 0.1, 0.20, 0.3, 0.4).

vestigated using several factorial designs (summarised in Table 2). A two-way ANOVA was used to identify the importance of the different traits on the response of the herbivore population. All the tests revealed significant effects (P < 0.001). Therefore, statistical details are not presented. A sensitivity analysis was carried out by varying the default value of individual input parameters by −20%, −10%, +10% and +20%. The mean expected damage level in the crop was used to compare the effect of individual parameters on the simulation model. The sensitivity analysis was carried out twice, the first time for the contact search strategy and the second time for the olfactory search strategy. The effect of the arrestment strength of the trap crop was studied by varying the plant specific flight tendency of the trap crop. In all cases the spatial configuration was an intercrop, except for the effect of trap crop coverage, which was determined using a different density of patches of 10 × 10 cells.

3. Results 3.1. Trap cropping efficacy in relation to search mechanism To investigate the effect of the host-plant recognition mechanism on the control efficacy of a trap crop, three different search mechanisms were compared for animals with varying displacement speeds foraging in an intercrop environment. The results are summarised in Fig. 6. There was a significant effect of host-plant location mechanism and displacement speed on the mean expected damage level in the crop. The trap crop had the greatest control efficacy (i.e. significantly lower damage in main crop) for insects that detect and respond to a trap crop by vision, with a strong relationship be-

tween movement distance and expected damage. For insect species that use olfactory cues to find preferred plant species, a similar but less strong relationship was found. This is due to the fact that the scanned area is smaller for the olfactory strategy (Fig. 5), resulting in a relatively lower probability of locating trap crop plants for each movement, compared to a visual searcher. Interestingly, for insect species that move randomly through the landscape and can only recognise a preferred host-plant after landing, aggregation in the trap crop still occurred, although at a slower rate and lower level compared to species with more sophisticated sensory abilities. For all three host location strategies there was a strong relationship between displacement speed and expected damage in the main crop. Damage levels decreased sharply with increasing displacement speed. The faster the insects move through the system the higher the probability that they will detect and aggregate in the preferred trap crop (Fig. 7). The fraction of movements towards and within the trap crop vegetation increased with displacement speed (Fig. 8). For slow moving species with nearest neighbour displacement (ML = 1), the majority of transition events occurred between crop cells (Fig. 8). Insects perceive a trap crop environment as favourable and stay longer in these environments, due to a reduced fraction of emigration events. For example, in the monocrop a typical population (500 animals for 50 timesteps, displacement speed 10, contact response) spent 10.667 ± 571 herbivore days (HD) in the crop, whereas in an intercrop system the total mean number of HD was 13.516 ± 454. The fact that this increased activity in a trap crop system generally does not lead to increased damage levels in the crop is due to the fact that the majority of the population is retained within the trap crop area.

208

R.P.J. Potting et al. / Ecological Modelling 182 (2005) 199–216

Fig. 6. The effect of herbivore search strategy and displacement speed on the efficacy of a trap cropping strategy. Data points indicate expected herbivore damage in the crop (mean number of herbivore days in a crop cell) for species with different search strategies (contact, olfaction, vision) as a function of displacement speed (n = 40 replicates). For an explanation of search strategy see Fig. 5. Trap crop (25% of total environment) was arranged as an intercrop in rows.

3.2. Colonisation pattern and trap cropping efficacy The simulation results obtained by varying the colonisation pattern are summarised in Fig. 9. There was a significant effect of the colonisation pattern and of the spatial arrangement of the trap crop and an interaction between these two. The damage level in the monocrop was independent of colonisation pattern, although there is more variation for directional colonisation from one side. The driving factor for the differences between trap cropping treatments and colonisation patterns is the fraction of the population that is captured in the trap crop upon entering the field. For the random airborne colonisation process, this directly related to the proportion of the area that was occupied by the trap crop (data not shown). For side colonisation this was dependent on the position of the trap crop. In this respect, a trap crop employed as a border was far more efficient, compared to random airborne colonisation, if the herbivore population

colonised the field from the side. This is due to the direct interception of a large proportion of the side colonisers in the border arrangement. The intercrop system was relatively insensitive to the colonisation pattern if the side colonising population entered the field perpendicular to the direction of the rows. The control efficacy of a trap crop in patches was completely lost for side colonisers, which is due to the fact that in this situation a large proportion of the population colonises the field directly into main crop cells. 3.3. Density dependent effect The effect of density dependent behaviour on the efficacy of trap cropping was investigated by examining the effect of cumulative herbivore damage on herbivore population response. Insects encountering a trap crop cell where the cumulative number of herbivore days exceeded the threshold reacted with a flight away from that cell. A trap crop with a low damage threshold increased the expected main crop damage in both

R.P.J. Potting et al. / Ecological Modelling 182 (2005) 199–216

209

Fig. 7. Proportion of the herbivore population situated in the trap crop throughout one simulated season in relation to the displacement speed of individual insects. Indicated are results from one typical model run for an olfactory searcher. Displacement speed was varied by setting the maximum movement length (ML) per timestep.

border and intercrop configuration to a damage level expected in a monocrop (Fig. 10). An increase in the damage threshold of the trap crop resulted in a decrease in expected damage to the main crop if the trap crop was deployed as an intercrop, whereas for the border arrangement a decrease in expected damage only occurred at higher damage thresholds (threshold >5). The

increase in damage to the main crop in trap crop systems with a low damage threshold is due to the lower ability of the trap crop to retain the herbivore population as a result of individuals responding to previous visits of conspecifics, resulting in a higher flow of insects from trap crop to main crop. This is reflected in the number of damage-threshold induced flight events

Fig. 8. Overview of proportion of transition events for an olfactory searcher with different displacement speeds (max. 1, 3 or 10 cells per timestep). White area indicates proportion of individual movements from crop cell to crop cell, black area indicates trap to trap, grey area trap to crop and blocked area crop to trap.

210

R.P.J. Potting et al. / Ecological Modelling 182 (2005) 199–216

Fig. 9. The effect of herbivore colonisation process on the efficacy of a trap cropping strategy. Data points indicate expected herbivore damage in the crop (mean number of herbivore days in a crop cell) for species with different colonisation behaviours (airborne, sideways from source at East or North side of field) in three spatial arrangements of the trap crop (Border, Intercrop, Patches as 25% of total environment). In all environmental setups 25% of the cells were trap crop (see Fig. 2 for examples). Simulated herbivore was olfactory searcher with displacement speed 10.

Fig. 10. Influence of indirect density dependence on trap crop efficacy. Indicated are expected damage levels (± S.D.) in main crop in agroecosystems with trap crop vegetations that differ in their herbivore damage thresholds (see text for explanation). Dotted line indicates expected damage level for the main crop when the trap crop is applied as intercrop, solid line indicates damage level when trap crop forms a border.

R.P.J. Potting et al. / Ecological Modelling 182 (2005) 199–216

in trap crop systems with a low threshold (data not shown). This effect is particularly evident in the border system, where the relatively smaller number of trap crop cells neighbouring a main crop cell (border: 2604, intercrop: 5000 out of 10.000 cells) results in a higher probability of animals encountering a trap crop cell which has exceeded the damage threshold. 3.4. Disruptive cropping To investigate the effect of intercropping with a repellent or deterrent plant on herbivore population response, the expected damage levels in simulated agro-ecosystems were estimated for deterrent or repellent plants with varying levels of induced emigration (e). The results are summarised in Fig. 11. The population control efficacy of both repellent and deterrent plants is dependent on the level of induced emigration to the air column. Intercropping with a non–crop vegetation having a neutral (erepel = ecrop ) or low (erepel = 2 × ecrop ) effect on emigration rates actually increased the herbivore damage level in the crop. This is because the same immigrant population is concentrated on a smaller crop area (75% main crop area, 25% non-used repellent area). A repellent/deterrent-based

211

intercropping strategy only works when the non-crop vegetation has a strong emigration-inducing effect on insect behaviour (erepel–crop = 4–8) In the model, the control efficacy of deterrent plants was higher than that of repellent plants. This is caused by the delay in recognising an unsuitable deterrent host-plant. A deterrent plant is recognised after landing and contacting the plant surface, whereas a repellent plant is recognised in-flight. 3.5. Sensitivity to input parameters The results of the sensitivity analysis are summarised in Fig. 12. For the contact sensory mode the parameter that has the largest effect of expected damage is the strength of inhibition of flight by the trap crop crop (i.e. arrestment). For the tested parameter range there is no effect of the damage threshold and a moderate effects of maximum movement length, mortality, emigration and percentage of trap crop vegetation. For the olfactory sensory mode the parameter that has the largest effect of expected damage is maximum movement length, indicating the role of the perceptual range of the animal. The parameters that did not significantly effect the mean expected damage were damage thresh-

Fig. 11. Expected damage levels (± S.D.) in crop in agro-ecosystems with repellent (solid line) or deterrent (dotted line) vegetation with varying emigration-inducing effects on herbivore. Asterisk indicates expected damage without repellent vegetation (monocrop).

212

R.P.J. Potting et al. / Ecological Modelling 182 (2005) 199–216

Fig. 12. Sensitivity of the simulation model to standardised changes of individual input parameters for: (A) contact sensory mode and (B) olfactory sensory mode. Bars indicate the range (min–max) of the normalised values of the mean expected damage level in the crop. Model parameters: (a) arrestment strength of trap crop; (b) damage threshold; (c) maximum movement length per timestep; (d) natural mortality rate; (e) emigration tendency for each flight; (f) percentage of trap crop in environment; (g) in-flight perception probability. Bars indicate minimum and maximum values of normalised mean expected damage for each parameter.

old and detection probability and, in contrast to the contact sensory mode, the arrestment strength of the trap crop. As was found for the contact sensory mode, the parameters mortality, emigration and percentage of trap crop vegetation did have an effect on the mean expected damage.

4. Discussion There are very few studies of population dynamics that systematically vary both behavioural ecological parameters of the animal and the composition and spatial arrangement of vegetation as we have done here. The results of the present study show that it is important to have a good understanding of the behavioural ecology of the target pest before one can determine whether the addition of highly attractive or repellent plants in a particular layout and density is likely to have a significant effect on population density and hence potential crop damage. Among the ecological attributes of the herbivore that may influence the efficacy of agro-ecosystem diversification are colonisation pro-

cess, within field movement behaviour and host-plant searching mechanisms. 4.1. Colonisation process In general, modern agro-ecosystems consist of patches of annual crops, which form suitable uninhabited patches within the landscape that are invaded by agricultural pests. The spatial distribution of the colonising population, in combination with the mobility and colonisation pattern of the herbivore, can greatly influence the efficacy of a diversification strategy. Simulation results show that for herbivores that actively immigrate from a nearby source via the field edge (e.g. Ferguson et al., 2000), a surrounding border trap crop is the optimal strategy, because it intercepts the population and reduces movement to the crop area (Fig. 9). Several Tephritid fruit flies have an active colonisation process by entering a crop area from the border. Recently, it has been demonstrated that these tephritid species can be successfully controlled by a border trap crop (references in Boucher et al., 2003). A trap crop employed as an intercrop also can be efficient in

R.P.J. Potting et al. / Ecological Modelling 182 (2005) 199–216

controlling species immigrating from the crop edge, when the rows are perpendicular to the main immigration direction (Fig. 9). In contrast, for passive airborne immigrants it is important to optimise the probability of detection of the trap crop after they have entered the field. Our simulation results show that the intercrop and patch arrangement are the most efficient in controlling species with this colonisation pattern (Fig. 9). A good example of the application of an intercropped trap crop border is the successful control of the green mirid with lucerne in cotton (Mensah and Khan, 1997). In our simulations, it is assumed that immigrants do not distinguish between the different vegetation types whilst entering the system. If insect species can perceive and respond to vegetation in polycultures during colonisation from the air column, this could significantly influence the control efficacy of diversification (Banks and Ekbom, 1999). However, most field studies have indicated that selective immigration is unimportant in determining herbivore densities (Kareiva, 1982; Risch et al., 1983; Bach, 1988a,b) and in general insect herbivores colonise monocultures and polycultures to the same extent (Bernays and Chapman, 1994; R¨amert and Ekbom, 1996). 4.2. Movement behaviour Arthropod herbivores differ widely in their movement behaviour (Morris and Kareiva, 1991; Turchin, 1998). Variability in movement behaviour in response to environmental heterogeneity can express itself in the tendency to initiate a movement, the duration of a movement and the choice of a given location. Our simulation results show that species-specific mobility is an important factor determining the control efficacy of diversified agro-ecosystems. Fast moving species aggregate more rapidly in the trap crop than slow moving species (Fig. 7) and this relation is relatively insensitive to host-plant detection mechanisms (Fig. 6). The success of a trap cropping scheme is dependent on the relative movement speed of the target herbivore in relation to the density and spatial configuration of the trap crop vegetation. These three factors determine the probability that the herbivore will come in to contact with the trap crop. Using a spatially explicit population model, (King and With, 2002) also found that dispersal success was an increasing function of dispersal speed. Fast moving animals located the preferred habi-

213

tat in a fragmented landscape faster than slow moving animals. 4.3. Search mechanism It is interesting to note that insects that are not attracted to the trap crop can also be controlled with a trap crop (Fig. 6, contact strategy). This effect is due to the arrestment response: once a contact searcher has found a preferred host-plant it has a tendency to stay there. The rate at which a trap crop can attract and retain a herbivore is an important factor for population control efficacy. Our simulations show that the strength of inhibition of flight by the trap crop (i.e. arrestment) is the most important parameter to manipulate the spatial dynamics of insects with post-alighting host recognition behaviour (Fig. 12A). However, to achieve successful population regulation it is necessary that these type of insects move relatively fast (Fig. 6). To give an example of this type of insect: whiteflies move short distances from plant to plant until they find acceptable hosts, once they have landed in a habitat (van Lenteren and Noldus, 1990). Smith and McSorley (2000) showed that a trap crop strategy did not achieve the desired control effect for the whitefly Bemisia tabaci in cotton fields. In contrast to contact searchers, the arrestment strength of the trap crop has no effect on expected damage for herbivores that can use olfactory or visual cues for finding preferred host plants (Fig. 12B). This is caused by the fact that the attraction towards the trap crop overrules the arrestment effect. For olfactory and visual searchers the mechanism of aggregation in the trap crop is a combination of attraction and arrestment and hence the population build-up in the trap crop is much faster compared to contact searchers (data not shown). A trap crop strategy has the highest efficiency for insect species that use visual cues to find their preferred host plants (Fig. 6). This increase in control impact is due to the fact that visual stimuli are not restricted to a particular direction, such as in upwind orientation to olfactory cues. Tephritid fly species have good sensory abilities (Prokopy et al., 1973; Prokopy and Owens, 1983). Recently, it has been demonstrated that the Papaya fruit fly, Mediterranean fruit fly and Pepper maggot fly can be succesfully controlled by a border trap crop (Aluja et al., 1997; Boucher et al., 2003; Cohen and Yuval, 2000).

214

R.P.J. Potting et al. / Ecological Modelling 182 (2005) 199–216

4.4. Density dependent avoidance behaviour

4.6. Disruptive crop strategy

Our simulation results show that density dependent avoidance behaviour can reduce the control efficacy of a trap crop, by inducing movements into the main crop if the plant-stress threshold is reached locally in the trap crop (Fig. 10). There are some examples of density dependent dispersal behaviour in field experiments. For example, (Van Den Berg and Rebe, 2002) found that a trap crop of forage sorghum surrounding a maize crop was highly attractive for oviposition by the stemborer Chilo partellus. However, the resulting density of stemborer larvae in the trap crop was too high to maintain and larvae moved from the trap crop to the main crop resulting in higher damage of maize in the trap crop system. (Luther et al., 1996) reported that Indian mustard and a cabbage variety acted as attractive trap crops for Plutella xylostella and Pieris brassicae, harbouring high numbers of insects. However, the damage level in the main (cabbage) crop was generally higher in the trap crop treatment compared to the damage level in the monocrop. The authors suggest this may be due to higher colonisation rates in the trap crop treatment and an overflow of insects from the trap crop to the main crop. One way to circumvent the redistribution of herbivores to the main crop has been to treat the trap crop with an insecticide at the appropriate time (Hokkanen, 1991). Simulation results show that an insecticide treatment of the trap crop enhances the control efficacy of the trap crop (Potting et al., 2002).

It is often assumed that repellent volatiles inhibit orientation towards the repellent plant (Foster and Harris, 1997). However, our simulation results show that this mechanism alone will lead to an increase of herbivore activity in the main crop (Fig. 11). Repellents may induce an avoidance response, but do not necessarily in˚ duce airborne emigration (e.g. Asman et al., 2001). It is interesting to note that Held et al. (2003) found that interplanting repulsive geraniums with roses increased the density of the Japanese Beetle (Popilla japonica) on roses. For a significant reduction in crop damage to be achieved, the repellent has to have a strong emigrationinducing effect on the herbivore. In this case, simulation results show that a seed mixture of crop and repellent is the best strategy, because it optimises contact with emigration-inducing sources (Potting et al., 2002), in particular if the repellent plant inhibits detection of neighbouring crop plants (data not shown). However, there are likely to be practical difficulties with seed mixes if crop and non-crop have to be separated at harvest. The fact that there are few examples of repellent chemicals or plants being effective at a fieldscale may be due to the failure of many repellents to induce sufficient emigration as opposed to local movements (Finch and Collier, 2000; Finch et al., 2003). Another factor, not investigated in this study, that may reduce the control efficacy of repellents and deterrents is that numerous pest insects lose their responsiveness after repeated exposure (Roitberg and Prokopy, 1987; Bernays and Chapman, 1994). Habituation to a plantderived deterrent within 24 h was demonstrated for the Japanese beetles Popillia japonica (Held et al., 2001).

4.5. Other factors influencing trap cropping efficacy The sensitivity analysis showed that the levels of mortality rate and emigration rate determine the population size of the herbivores in the environment and hence have an influence on mean expected damage (Fig. 12). The fact that population density is an important factor that induces variation in trap cropping efficacy has been noted by several field studies (Buntin, 1998) The fraction of trap crop vegetation in the environment influences the expected damage level in the crop (Fig. 12). This corroborates the results of the simulation studies of Banks and Ekbom (1999) who found that herbivore densities declined with increasing trap crop cover.

4.7. Concluding remarks The wide variation in population response of herbivores to diverse agro-ecosystems (Andow, 1991) is replicated in this study using different simulation scenarios by varying the behavioural features of the herbivore and ecological characteristics of the agro-ecosystem. In our simulations we found that the expected damage in diversified systems was similar, lower, or higher. We show that both the behavioural ecology of the target herbivore and the ecological characteristics of the agro-ecosystem affect the control efficacy of the added vegetation.

R.P.J. Potting et al. / Ecological Modelling 182 (2005) 199–216

Although it is risky to generalise, our simulations show that small arthropod herbivores, such as mites, thrips, aphids and whiteflies, that have an airborne colonisation pattern, limited host detection ability and slow displacement speed, are the most difficult to control with a diversification strategy. There are only a few published studies on the behavioural manipulation of this type of insect in the field. A combination of attractants and repellents was not a feasible strategy for population control of the hop aphid (L¨osel et al., 1996), and a trap crop strategy did not achieve the desired control effect for the whitefly Bemisia tabaci in cotton fields (Smith and McSorley, 2000). Both studies conclude that the ecology of these herbivores limits the feasibility of their behavioural manipulation. In contrast to small insects, bigger insects, such as beetles, butterflies and moths are generally highly mobile, have the ability for directed flights and good sensory abilities that enable oriented movements. As demonstrated with our simulations, we expect good control effects of diversification for these types of herbivore, given that plants are chosen with a strong behaviour-altering effect and employed in an optimal spatial arrangement. It is interesting to note that for the trap crop systems that are successfully applied in agricultural practice, there are no small herbivores involved but mainly relatively large beetles (Hokkanen, 1991) and tephritid flies (Boucher et al., 2003). Acknowledgement We thank one anonymous referee for constructive comments on the manuscript. This work was supported by funding from DEFRA. Rothamsted International receives grant-aided support from the BBSRC. References Aluja, M., Jimenez, A., Camino, M., Pinero, J., Aldane, L., Castrejon, V., Valdes, M.E., 1997. Habitat manipulation to reduce papaya fruit fly (Diptera: Tephritidae) damage: Orchard design, use of trap crops and border trapping. J. Econ. Entomol. 90, 1567–1576. Andow, D.A., 1991. Vegetational diversity and arthropod population response. Ann. Rev. Entomol. 36, 561–586. ˚ Asman, K., Ekbom, B., R¨amert, B., 2001. Effect of intercropping on oviposition and emigration behavior of the leek moth (Lepidoptera: Acrolepiidae) and the diamondback moth (Lepidoptera: Plutellidae). Environ. Entomol. 30, 288–294.

215

Bach, C.E., 1988a. Effect of host plant patch size on herbivore density: patterns. Ecology 69, 1090–1102. Bach, C.E., 1988b. Effect of host plant patch size on herbivore density: underlying mechanisms. Ecology 69, 1103–1117. Banks, J.E., 1998. The scale of landscape fragmentation affects herbivore response to vegetation heterogeneity. Oecologia 117, 239–246. Banks, J.E., Ekbom, B., 1999. Modelling herbivore movement and colonization: pest management potential of intercropping and trap cropping. Agric. For. Entomol. 1, 165–170. Bell, W.J., 1991a. Searching behavior patterns in insects. Ann. Rev. Entomol. 35, 447–467. Bell, W.J., 1991b. Searching behaviour. In: The Behavioural Ecology of Finding Resources. Chapman and Hall, London. Berec, L., 2002. Techniques of spatially explicit individual-based models: construction, simulation and mean-field analysis. Ecol. Model. 150, 55–81. Bernasconi, M.L., Turlings, T.C.J., Ambrosetti, L., Bassetti, P., Dorn, S., 1998. Herbivore-induced emissions of maize volatiles repel the corn leaf aphid Rhopalosiphum maidis. Entomol. Exp. Appl. 87, 133–142. Bernays, E.A., Chapman, R.F., 1994. Host-plant Selection by Phytophagous Insects. Chapman and Hall, New York. Blom, P.E., Fleischer, S.J., Smilowitz, Z., 2002. Spatial and temporal dynamics of colorado potato beetle (Coleoptera: Chrysomelidae) in fields with perimeter and spatially targeted insecticides. Environ. Entmol. 31, 149–159. Boucher, T.J., Ashley, R., Durgy, R., Sciabarrasi, M., Calderwood, W., 2003. Managing the pepper maggot (Diptera: Tephritidae) using perimeter trap cropping. J. Econ. Entomol. 96, 420– 432. Buntin, G.D., 1998. Cabbage seedpod weevil (Ceutorhynchus assimilis, Paykull) management by trap cropping and its effect on parasitism by Trichomalus perfectus (Walker) in oilseed rape. Crop Prot. 17, 299–305. Cohen, H., Yuval, B., 2000. Perimeter trapping strategy to reduce Mediterranean fruit fly (Diptera: Tephritidae) damage on different host species in Israel. J. Econ. Entomol. 93, 721–725. DeAngelis, D.L., Gross, L.J., 1992. Individual-Based Models and Approaches in Ecology. Chapman and Hall, New York. Ferguson, A.W., Klukowski, Z., Walczak, B., Perry, J.N., Mugglestone, M.A., Clark, S.J., Williams, I.H., 2000. The spatiotemporal distribution of adult Ceutorhynchus assimilis in a crop of winter oilseed rape in relation to the distribution of their larvae and that of the parasitoid Trichomalus perfectus. Entomol. Exp. Appl. 95, 161–171. Finch, S., Collier, R.H., 2000. Host-plant selection by insects – a theory based on ‘appropriate/inappropriate landings’ by pest insects of cruciferous plants. Entomol. Exp. Appl. 96, 91–102. Finch, S., Billiald, H., Collier, R.H., 2003. Companion planting – do aromatic plants disrupt host-plant finding by the cabbage root fly and the onion fly more effectively than non-aromatic plants? Entomol. Exp. Appl. 109, 183–195. Foster, S.P., Harris, M.O., 1997. Behavioural manipulation methods for insect pest-management. Ann. Rev. Entomol. 42, 123–146. Hannunen, S., Ekbom, B., 2001. Host plant influence on movement patterns and subsequent distribution of the polyphagous herbi-

216

R.P.J. Potting et al. / Ecological Modelling 182 (2005) 199–216

vore Lygus rugulipennis (Heteroptera: Miridae). Environ. Entmol. 30, 517–523. Held, D.W., Eaton, T., Potter, D.A., 2001. Potenial for habituation to a neem-based feeding deterrent in Japanese beetles Popillia japonica. Entomol. Exp. Appl. 101, 25–32. Held, D.W., Gonsiska, P., Potter, D.A., 2003. Evaluating companion planting and non-host masking odors for protecting roses from the Japanese beetle (Coleoptera: Scarabaeidae). J. Econ. Entomol. 96, 81–87. Hokkanen, H.T., 1991. Trap cropping in pest management. Ann. Rev. Entomol. 36, 119–138. Holt, J., Cheke, R.A., 1997. Modelling. In: Walton, M.P. (Ed.), Methods in Ecological and Agricultural Entomology. CABI, pp. 351–378. Johnson, C.G., 1969. Migration and Dispersal of Insects by Flight. Methuen, London, p. 763. Judson, O.P., 1994. The rise of the individual-based model in ecology. Trends Ecol. Evol. 9, 9–14. Kareiva, P., 1982. Experimental and mathematical analyses of herbivore movement: quantifying the influence of plant spacing and quality on foraging discrimination. Ecol. Monogr. 52, 261–282. King, A.W., With, K.A., 2002. Dispersal success on spatially structured landscapes: when do spatial pattern and dispersal behavior really matter? Ecol. Model. 147, 23–39. L¨osel, P.M., Lindemann, M., Scherkenbeck, J., Maier, J., Engelhard, B., Campbell, C.A.M., Hardie, J., Pickett, J.A., Wadhams, L.J., Elbert, A., Thielking, G., 1996. The potential of semiochemicals for control of Phorodon humuli (Homoptera: Aphididae). Pestic. Sci. 48, 293–303. Luther, G.C., Valenzuela, H.R., Defrank, J., 1996. Impact of cruciferous trap crops on lepidopteran pests of cabbage in Hawaii. Environ. Entomol. 25, 39–47. Mamedov, A., Udalov, S., 2002. A computer tool to develop individual-based models for simulation of population interactions. Ecol. Model. 147, 53–68. Mangel, M., Roitberg, B.D., 1989. Dynamic information and oviposition decisions in a fruit fly. Ecol. Entomol. 14, 181–189. Mensah, R.K., Khan, M., 1997. Use of Medicago sativa (L.) interplantings trap crops in the management of the green mired Creontiades dilutus (Stal) in commercial cotton in Australia. Int. J. Pest Manage. 43, 197–202. Morris, W.F., Kareiva, P.M., 1991. How insect herbivores find suitable host plants: the interplay between random and nonrandom movement. CRC Press, Boston. Papaj, D.R., Lewis, A.C. (Eds.), 1993. Ecological and Evolutionary Perspectives. Chapman and Hall, New York. Potting, R.P.J., Perry, J.N., Powell, W., 2002. Possibilities and constraints of agroecosystem management as a pest management strategy: a simulation approach. In: Proceedings of the BCPC Conference, Pests and Diseases, Brighton, pp. 91–96. Price, P.W., 1976. Colonization of crops by arthropods: nonequilibrium communities in soybean fields. Environ. Entmol. 5, 605–611. Prokopy, R.J., Moericke, V., Bush, G.L., 1973. Attraction of apple maggot files to oder of apples. Environ. Entomol. 2, 743–749. Prokopy, R.J., Owens, E.D., 1983. Visual detection of plants by herbivorous insects. Ann. Rev. Entomol. 28, 337–364.

R¨amert, B., Ekbom, B., 1996. Intercropping as a management strategy against carrot rust fly (Diptera: Psilidae): a test of enemies and resource concentration hypotheses. Environ. Entomol. 25, 1092–1100. Riley, J.R., Drake, V.A., 2002. Technologies for insect movement and migration research. Comput. Electron. Agric. 35, 57–62. Risch, S., Andow, D.A., Altieri, M.A., 1983. Agroecosystem diversity and insect pest control: data tentative conclusions and new research directions. Environ. Entmol. 12, 625–629. Roitberg, B.D., Prokopy, R.J., 1987. Insects that mark host plants. An ecological, evolutionary perspective on host-marking chemicals. BioScience 37, 400–406. Schoonhoven, L.M., Jermy, T., van Loon, J.J.A., 1998. Insect–plant biology: from physiology to evolution. Chapman and Hall, London, p. 409. Smith, H.A., McSorley, R., 2000. Potential of field corn as a barrier crop and eggplant as a trap crop for management of Bemisia argentifolii (Homoptera: Aleyrodidae) on common bean in North Florida. Florida Entomol. 83, 145–158. Stephens, D.W., Krebs, J.R., 1986. Foraging Theory Monographs in Behavior and Ecology. Princeton University Press, Princeton, p. 247. Thompson, J.N., Pellmyr, O., 1991. Evolution of oviposition behavior and host preference in Lepidoptera. Ann. Rev. Entomol. 36, 65–89. Turchin, P., 1998. Quantitative Analysis of Movement. Sinauer Associates, Sunderland, USA. Turlings, T.C.J., Wackers, F., Vet, L.E.M., Lewis, W.J., Tumlinson, J.H., 1993. Learning of host-finding cues by hymenopterous parasitoids. In: Papaj, D.R., Lewis, A.C. (Eds.), Insect Learning: Ecological and Evolutionary Perspectives. Chapman and Hall, New York, pp. 51–78. Uchmanski, J., Grimm, V., 1996. Individual-based modelling in ecology: what makes the difference. Trends Ecol. Evol. 11, 437– 441. Van Den Berg, J., Rebe, M., 2002. Infestation patterns of stem borers in maize surrounded by graminaceous trap crops. In: Van Den Berg, J., Uys, V.M. (Eds.), Proceedings of the Southern African Stem Borer Management Project, ARC, Pietermaritzburg, pp. 41–44. Vandermeer, J., 1989. The Ecology of Intercropping. Cambridge University Press, Cambridge. Vet, L.E.M., Dicke, M., 1992. Ecology of infochemical use by natural enemies in a tritrophic context. Ann. Rev. Entomol. 37, 141– 172. Visser, J.H., 1986. Host odor perception in phytophagous insects. Ann. Rev. Entomol. 31, 121–144. van Lenteren, J.C., van Roermund, H.J.W., 1997. Better biological control by a combination of experimentation and modelling. In: Kropff, E.A. (Ed.), Application of Systems Approaches at the Field Level. Kluwer, Dordrecht, pp. 349–363. van Lenteren, J.C., Noldus, L.P.J.J., 1990. Behavioural and ecological aspects of whitefly-plant relationships. In: Gerling, D. (Ed.), Whiteflies: Bionomics, Pest Status and Management. Intercept, Andover, pp. 47–89. Wood, D., Lenn´e, J.M., 1999. Agrobiodiversity: characterization, utilization and management. CABI, Wallingford.