Gone with the wind: low availability of volatile information limits foraging efficiency in downwind-flying parasitoids

Parasitoids need to find their plant-feeding hosts in complex environments that contain multiple other plant and insect species. They usually rely on herbivore-induced plant volatiles to locate herbivore-infested plants from a distance and their foraging efficiency may be reduced when volatile information is not available. Downwind foraging during times when high wind speeds prevent odour-guided upwind flights may create foraging situations with limited accessibiliy to volatile information. We hypothesized that parasitoids forage less efficiently by landing on nonhost-damaged or undamaged plants when they are forced to fly downwind and tested this in a wind tunnel experiment. We released the parasitoid Cotesia glomerata (Hymenoptera, Braconidae) either upwind or downwind of a plant stand and observed their foraging behaviour. During downwind foraging, parasitoids were less successful in host finding and needed more time until they managed to oviposit in a host caterpillar compared to the upwind foraging situation. The observed increase in foraging time was caused by prolonged foraging on nonhost-infested and undamaged plants in the downwind situation, indicating that parasitoids leave an unprofitable patch that does not contain host caterpillars earlier, when they perceive volatiles from other herbivore-infested plants located upwind. Volatile information on the availability of herbivore-infested plants within a plant stand seems to be crucial for efficient foraging in plant stands that contain a mixture of host-infested, nonhost-infested and undamaged plants. Parasitoid foraging efficiency may thus be strongly reduced when high wind speed prevents odour-guided upwind flight.

Parasitoids need to find their plant-feeding hosts in complex environments that contain multiple other plant and insect species. They usually rely on herbivore-induced plant volatiles to locate herbivoreinfested plants from a distance and their foraging efficiency may be reduced when volatile information is not available. Downwind foraging during times when high wind speeds prevent odour-guided upwind flights may create foraging situations with limited accessibiliy to volatile information. We hypothesized that parasitoids forage less efficiently by landing on nonhost-damaged or undamaged plants when they are forced to fly downwind and tested this in a wind tunnel experiment. We released the parasitoid Cotesia glomerata (Hymenoptera, Braconidae) either upwind or downwind of a plant stand and observed their foraging behaviour. During downwind foraging, parasitoids were less successful in host finding and needed more time until they managed to oviposit in a host caterpillar compared to the upwind foraging situation. The observed increase in foraging time was caused by prolonged foraging on nonhost-infested and undamaged plants in the downwind situation, indicating that parasitoids leave an unprofitable patch that does not contain host caterpillars earlier, when they perceive volatiles from other herbivore-infested plants located upwind. Volatile information on the availability of herbivore-infested plants within a plant stand seems to be crucial for efficient foraging in plant stands that contain a mixture of host-infested, nonhost-infested and undamaged plants. Parasitoid foraging efficiency may thus be strongly reduced when high wind speed prevents odour-guided upwind flight.
© 2020 The Author(s). Published by Elsevier Ltd on behalf of The Association for the Study of Animal Behaviour. This is an open access article under the CC BY license (http://creativecommons.org/licenses/ by/4.0/).
To forage efficiently in complex, patchy environments, animals need to acquire information about the location and profitability of a resource patch. Long-range cues such as resource-associated volatiles are used by many invertebrate and vertebrate animals to navigate towards a resource patch (Mumm & Dicke, 2010;Nevitt, Veit, & Kareiva, 1995;Stutz, Banks, Proschogo, & McArthur, 2016). After a patch has been reached, animals need to decide how long they should forage within the current patch before leaving to search for another patch. According to the marginal value theorem, optimal residence time in a patch should increase with increasing patch profitability and decreasing profitability of the overall habitat (Charnov, 1976). Patch profitability is often assessed based on the within-patch encounter rate with resource items or resourceassciated cues, while habitat quality can be evaluated based on information acquired during previous patch visits (Krebs, Ryan, & Charnov, 1974;Persons & Uetz, 1996;Wajnberg, 2006) and through long-range cues about the presence of other resources close by (Mayland, Margolies, & Charlton, 2000;Tentelier & Fauvergue, 2007).
Predatory insects and parasitoids that attack insect herbivores need to find plants infested with suitable prey or host herbivores in environments that contain multiple other plant and insect species which often do not provide suitable resources (Aartsma et al., 2017(Aartsma et al., , 2019de Rijk, Dicke, & Poelman, 2013). It is assumed that natural enemies deal with the enormous complexity of their foraging environment by following hierarchical foraging steps, which include habitat location, location of the herbivore-infested plant (refrered to as a patch), prey or host location (on the plant) as well as prey or host acceptance, and by using different cues and strategies at the different foraging steps (Aartsma et al., 2019;Hassell & Southwood, 1978;Hatano, Kunert, Michaud, & Weisser, 2008; van Alphen, Bernstein, & Driessen, 2003;Vinson, 1976). The effect of patch size (e.g. number of host herbivores on a plant), patch distribution and habitat quality (e.g. total number of hosts in the habitat) on parasitoid foraging behaviour and patch allocation times has been intensively studied in simple foraging arenas that only contain host patches and in some cases empty patches (Wajnberg, 2006 and cited literature). During foraging in their natural environments, female parasitoids need to find their hosts among multiple nonhost species which can occur together with the host on the same plant (mixed patch) or on a different plant (nonhost patch;de Rijk et al., 2013). Even though it has been repeatedly found that parasitoids attack fewer hosts when they are foraging in an environment that also contains nonhosts (Bukovinszky et al., 2012;De Rijk, Wang, Papagiannaki, Dicke, & Poelman, 2016;Meisner, Harmon, & Ives, 2007;Vos, Berrocal, Karamaouna, Hemerik, & Vet, 2001), we have little information on how the presence of nonhost patches influences parasitoid foraging behaviour, such as approach rates to nonhost-infested plants and the tendency to leave nonhost patches (but see Bukovinszky et al., 2012;De Rijk, Yang, Engel, Dicke, & Poelman, 2016;Vosteen, van den Meiracker, & Poelman, 2019).
We hypothesized that, owing to the reduced potential to use volatile information, the foraging efficiency of the parasitoid C. glomerata would be lower during downwind than during upwind foraging. We predicted that parasitoids would spend longer on nonhost-infested plants in downwind foraging situations, since they are not able to perceive HIPVs from the host-infested plant located further downwind and are thus less motivated to leave the nonhost-infested plant. To test this hypothesis, we performed wind tunnel experiments in which we released C. glomerata either upwind or downwind of a plant stand containing host-infested, nonhost-infested and undamaged plants. To test how foraging direction influences foraging behaviour and foraging efficiency of inexperienced and oviposition-experienced C. glomerata, we recorded foraging time until host oviposition, duration and frequency of visits to nonhost-infested and undamaged plants as well as the occurrence of nonhost oviposition.

Organisms
The parasitoid C. glomerata and caterpillars of P. brassicae and M. brassicae were obtained from cultures maintained at the Laboratory of Entomology, Wageningen University under the same conditions as described in Vosteen et al. (2019). Caterpillars were reared on Brussels sprout plants, Brassica oleracea L. var. gemmifera cultivar Cyrus, under climate-controlled conditions (16:8 light:dark photoperiod, at 21 ± 1 C and 50e70% relative humidity). Every summer, the C. glomerata culture is replaced with individuals collected at the field site at Wageningen University, the Netherlands. Parasitoids were reared by placing a Brussels sprout leaf with approximately 200 first-instar P. brassicae in the parasitoid rearing cage for 5e10 min to allow oviposition. Parasitized caterpillars were reared on Brussels sprout plants under the same conditions as the nonparasitized caterpillars. After pupation, parasitoid cocoons were transferred to screen cages (30 Â 30 Â 30 cm, Bugdorm, Taiwan) in a climate cabinet (24 ± 1 C, 12:12 light:dark) and were kept in the absence of plant odours. Emerging parasitoids were supplied with water and honey. Females were separated from males 1 day before the experiment and were used in the experiments at the age of 2e5 days. Since C. glomerata males usually emerge several hours earlier than females and wait on the cocoons to mate with them (I. Vosteen, N. van den Meiracker & E.H. Poelman, personal observation), it can be assumed that all females used in the experiments had mated. Under natural conditions, most females leave the natal patch without mating with their siblings (Gu & Dorn, 2003) and are attracted by the combined odours of males and host-infested plants. After mating, females stop responding to males odours and respond to volatiles from undamaged and herbivore-damaged plants (Xu et al., 2016).
For the experiment we used 4e5-week-old Brussels sprout plants, replacing them every day with a fresh set of plants. Plants were either undamaged or infested by 10 1e2-day-old P. brassicae (hosts) or 10 1e3-day-old M. brassicae (nonhosts) 1 day before the experiment. Caterpillars were placed on one of the newly emerging leaves, from where they could disperse freely on the plant. While P. brassicae usually fed gregariously close to the release site, M. brassicae moved all over the plant, causing highly dispersed feeding damage.

Ethical Note
No licence is required for work with insects. In total, 517 female C. glomerata parasitoids, 200 P. brassicae caterpillars and 220 M. brassicae caterpillars were used in the two experiments presented in this study. Additionally, approximately 550 caterpillars were used to condition the oviposition-experienced parasitoids. All insects received an excess of food. Caterpillars were handled gently with a fine brush when they had to be transferred to a new plant. Parasitoids were collected from the cages by covering them with a glass vial and waiting until they climbed into the vial. To avoid stress from overcrowding, the number of parasitoids per emergence cage was kept below 150 individuals. Shortly after they had been used in the experiments, insects were killed by freezing.

Wind Tunnel Set-Up
All experiments were conducted in a wind tunnel (200 x 60 cm and 60 cm high; for a detailed description see Geervliet, Vet & Dicke, 1994), with the same abiotic conditions (0.1 m/s wind speed, 24e25 C, 53e63% relative humidty) and parasitoid release procedure as used by Vosteen et al. (2019). To prevent them from immediately flying towards the ceiling, individual parasitoids were released inside a horizontal glass cylinder (30 cm long, 15 cm diameter) with both ends open (Geervliet et al., 1994) from an upright glass vial. Parasitoids that had not taken flight within 5 min (downwind foraging experiment) or 6 min (undamaged plants experiment) after they had climbed out of the glass vial and those that flew directly to the ceiling of the wind tunnel were considered nonresponsive.

Foraging Direction Experiment
The aim of this experiment was to gain a better understanding of parasitoid foraging behaviour in an environment that contains host-infested, nonhost-infested and undamaged plants and to test whether downwind-flying parasitoids respond differently to undamaged and nonhost-damaged plants than parasitoids in upwind flight.
Parasitoid foraging behaviour was recorded in a stand of eight plants placed in the wind tunnel ( Fig. 1). Two of the plants were infested with 10 nonhost caterpillars (NH1 and NH2) and one plant was infested with 10 host caterpillars (H), while the rest of the plants were undamaged. Plants were positioned either in groups of two (one nonhost-infested and one undamaged plant or two undamaged plants) next to the sides of the wind tunnel or alone (hostinfested or undamaged plant) in the middle of the wind tunnel. Individual plants and plant pairs were alternated and separated by 20 cm (Fig. 1). The distance from the first plant pair to the release site was 25 cm. Female C. glomerata were released either downwind of the plant stand, so that they would enter it in upwind flight (upwind foraging situation, Fig. 1a), or upwind, so that they would enter it in downwind flight (downwind foraging situation, Fig. 1b). The positions of the host and nonhost-infested plants relative to the release site were the same in both foraging situations.
Tested females had either had an oviposition experience or were inexperienced. On the day before they were tested, ovipositionexperienced parasitoids had been allowed one oviposition in a petri dish containing a piece of leaf with 40 to 60 1e2-day-old P. brassicae, which had been taken from a from a heavily infested Brussels sprout plant just before the conditioning was started (same conditioning procedure as in Vosteen et al., 2019).
Parasitoids that started foraging after they were released in the glass cylinder were considered as responsive and observed until they oviposited into a host caterpillar or for a maximum of 45 min. Parasitoid foraging behaviour was observed and several behavioural parameters (see Table 1) were live recorded with the software Observer XT10 (Noldus Information Technology, Wageningen, Netherlands) on a hand-held computer (Pocket Observer 3.1, Noldus). If a parasitoid rested for more than 5 min on a plant or flew to the walls or ceiling of the wind tunnel and stayed there for more than 1 min, observations were terminated and the parasitoid was scored as an unsuccessful forager.
The Observer XT10 was used to calculate the total number, total duration and latency of each behavioural parameter, from which the most informative parameters were selected for further analysis by evaluating whether the respective parameter would help to understand how parasitoid foraging efficiency was influenced by foraging direction and experience. To calculate the total time a parasitoid spent foraging on a specific plant (henceforth 'foraging time'), time spent in foraging flight and time spent foraging on a plant (Table 1) were summed. Each host-or nonhost-infested plant is referred to as a host or nonhost patch, respectively. To compare the patch residence time during reoccurring plant visits, a foraging event was defined as a plant visit with at least one landing on the plant from the beginning of the first foraging flight until the parasitoid left the canopy of that plant to forage elsewhere. Patch residence time describes the duration of a foraging event, while patch foraging time describes the time a parasitoid foraged within a patch until a nonhost attack occurred. Periods longer than 10 s in which the parasitoid sat still on the plant were subtracted from the total patch residence and patch foraging time. Responsiveness, foraging success and the occurrence of a specific behaviour (e.g. foraging on undamaged plants, foraging on nonhost-infested plants, nonhost attacks) were scored as binary data.

Undamaged Plants Experiment
To test whether the presence of undamaged plants influences the foraging efficiency of upwind-flying parasitoids, we observed host-seeking wasps that were released either in the presence of a host-infested plant alone or together with three undamaged plants.
Plants were arranged in a row following the direction of the airflow and all parasitoids were released on the downwind site of this row, allowing them to approach the plants in upwind flight (Fig. 2). The first plant was located 50 cm away from the release site and the distance between the plants was 25 cm, measured from stem to stem, comparable to the set-up used by Vosteen et al. (2019).

Statistical Analysis
All data were analysed with R version 3.3.1 R (The R Foundation for Statistical Computing, Vienna, Austria, http://www.r-project. org) using two-tailed tests.
To account for variation between the different experimental days, data were analysed with either generalized linear mixedeffect models (GLMMs) for binary data (glmer function of the lme4 package; Bates, Maechler, Bolker, & Walker, 2014) or linear mixed-effect models (LMMs) for continuous data on foraging duration (lme function of the nlme package; Pinheiro, Bates, DebRoy, Sarkar, & Team, 2013) with day as a random factor (random intercept; see Table 2 for an overview of the statistical tests). If variance of the random effect was close to zero, a general linear model (GLM) was used for binary data (occurrence of foraging on nonhost-infested plants, nonhost attacks and first encounter with nonhost-infested plant 1). Only a few parasitoids attacked nonhosts, resulting in some experimental days with no or just one nonhost attack. Thus, for analysing the effect of the occurrence of nonhost attacks on patch residence time and patch foraging time, day could not be included as a random factor and a KruskaleWallis test was used instead.
Foraging time until oviposition was log transformed to achieve homogeneity of variances. Owing to zeros in the data sets on foraging duration on nonhost-infested plants in the downwind foraging experiment, a log transformation was not possible and a square root transformation was used instead.
Model simplification was done by backward selection and the model with the lowest Akaike information criterion was chosen as the final model. P values are reported for all explanatory variables that were retained in the final model. In those cases when none of the variables were retained in the final model, P values of explanatory variables of interest are reported to stress that they do not have an effect.

Foraging direction experiment
To test whether foraging direction and oviposition experience influence parasitoid foraging behaviour in a plant patch, several behavioural parameters (responsiveness, foraging success, foraging durations, approaches/min on herbivore-infested plants as well as occurrence of foraging on undamaged plants, foraging on nonhostinfested plants and nonhost oviposition, first encounter with nonhost plant 1) were analysed with GLMMs or LMMs with experience and foraging direction as explanatory variables. Identity of the herbivore-infested plant (NH1, NH2 or H), foraging direction and parasitoid experience were used as explanatory variables during the analysis of the proportion of parasitoids that landed on the first herbivore-infested plant that they approached.

Undamaged plant experiment
To test whether the presence of neighbouring undamaged plants and the position of the host-infested plant relative to the parasitoid release point influenced the time parasitoids needed to find the host-infested plant, an LMM with presence of neighbouring plants and position of the host-infested plant as explanatory variables was performed. To analyse the effect of plant position on the proportion of parasitoids that landed first on one of the undamaged plants before they flew to the host-infested plant, a GLMM with position of the host-infested plant as explanatory variable was performed.

Foraging Direction Experiment
Parasitoid responsiveness was lower in the downwind than in the upwind foraging situation (GLMM: c 2 ¼ 20.483, P < 0.001) and increased after oviposition experience (GLMM: c 2 ¼ 9.825, P ¼ 0.002; Fig. 3a). Fewer parasitoids found the host-infested plant in the downwind foraging situation (GLMM: c 2 ¼ 36.822, P < 0.001; Fig. 3b) and inexperienced parasitoids were less successful at finding the host-infested plant than ovipositionexperienced parasitoids (GLMM: c 2 ¼ 4.054, P ¼ 0.044). Those parasitoids that managed to find the host-infested plant needed in total three times longer until oviposition in the downwind than in the upwind foraging situation (LMM: F 1,85 ¼ 46.377, P < 0.001; Fig. 3c). After they had landed on the host-infested plant, 80% of the parasitoids found their first host caterpillar within 1.5 min, while 92% found the first host within 2.5 min, independent of foraging direction and experience.
In the downwind foraging situation, more parasitoids foraged on the nonhost-infested plants (GLM: LRT ¼ 8.999, P ¼ 0.003; Fig. 4a) and they spent in total three times longer inspecting the nonhost-infested plants compared to parasitoids in the upwind  Fig. 4b). While they were foraging on the nonhost-infested plants, parasitoids occasionally attacked nonhost caterpillars. While this behaviour was observed for fewer than 10% of the parasitoids in the upwind foraging situation, it occurred significantly more often in the downwind situation (GLM: LRT ¼ 13.574, P < 0.001; Fig. 4c), where 22% of the inexperienced and 34% of the experienced parasitoids attacked nonhost caterpillars. Parasitoids in the upwind foraging situation approached significantly more herbivore-infested and undamaged plants per min (1.04 ± 0.08 plants/min) than parasitoids in the downwind foraging situation (0.65 ± 0.07 plants/min; LMM: F 1,113 ¼ 13.700, P < 0.001). Foraging situation influenced which herbivore-infested plant was approached first. Two-thirds of parasitoids made their first approach to the nonhost-infested plant 1 (NH1) in the downwind foraging situation, while only one-third of the parasitoids approached this plant first in the upwind foraging situation (GLM: LRT ¼ 15.78, P < 0.001; Fig. 5a). In the upwind foraging situation, all herbivore-infested plants were approached first at a similar rate. Eighty two per cent of the parasitoids landed on the first herbivore-infested plant they approached, independent of plant position and herbivore species (GLMM: c 2 ¼ 1.644, P ¼ 0.440) and of foraging direction (GLMM: c 2 ¼ 1.114, P ¼ 0.291) and experience (GLMM: c 2 ¼ 0.018, P ¼ 0.895).
Landing on a herbivore-infested plant usually resulted in intensive foraging on this plant, regardless of whether herbivores were hosts or nonhosts. Patch residence times (foraging flight þ on-plant foraging) during the first three approaches of nonhost-infested plants in which the parasitoids landed at least once per approach were longer in the downwind foraging situation (LMM: F 1,62 ¼ 9.884, P < 0.003; Fig. 5b) and also if parasitoids oviposited in nonhosts during this time (LMM: F 1,75 ¼ 28.114, P < 0.001). Parasitoids in the downwind situation tended to forage longer on a nonhost-infested plant before they attacked a nonhost caterpillar compared to parasitoids in the upwind foraging situation (KruskaleWallis test: c 2 ¼ 3.209, P ¼ 0.073; Fig. 5c). Patch residence time during plant visits without nonhost attacks was similar to the time parasitoids spent foraging in a nonhost patch before they attacked a nonhost caterpillar (KruskaleWallis test: c 2 ¼ 1.122, P ¼ 0.289). In the downwind foraging situation, parasitoids spent on average 2.5 min on a nonhost patch before they left or started to attack nonhost caterpillars (leaving: mean ± -SE ¼ 153 ± 16 s; start nonhost attacks: mean ± SE ¼ 159 ± 24 s), while in the upwind foraging situation, parasitoids attacked a nonhost on average after 1 min (mean ± SE ¼ 60 ± 8 s) or left the plant after 1.5 min (mean ± SE ¼ 96 ± 13 s).
In the downwind situation, more parasitoids visited an undamaged plant at least once than in the upwind situation (GLMM: c 2 ¼ 10.993, P < 0.001). In the upwind situation, more inexperienced than oviposition-experienced parasitoids foraged on undamaged plants, while in the downwind situation, experience did not influence foraging on undamaged plants (foraging direction) experience: GLMM: c 2 ¼ 7.384, P ¼ 0.007; Fig. 6a). Parasitoids foraged longer on undamaged plants in the downwind than in the upwind situation (LMM: F 1,119 ¼ 25.690, P < 0.001; Fig. 6b).

Undamaged Plants Experiment
The position of the host-infested plant influenced parasitoid responsiveness (GLMM: c 2 ¼ 7.974, P ¼ 0.005; Fig. 7a): 72% of the parasitoids responded if the host-infested plant was placed close to the release site, while 62% responded if it was further away. If undamaged plants were placed in between the release site and the host-infested plant, parasitoids needed more than twice as long to reach the host-infested plant compared to a situation without any neighbouring plants. If the undamaged plants were placed behind the host-infested plant, the effect on foraging time was less pronounced (LMM: position of the nonhost-infested plant: F 1,160 ¼ 17.111, P ¼ 0.001; presence of undamaged neighbouring plants: F 1,160 ¼ 14.158, P ¼ 0.002; position) neighbouring plants: F 1,160 ¼ 3.501, P ¼ 0.063; Fig. 7b). Ninety per cent of the parasitoids made their first landing on the host-infested plant if it was placed close to the release point (position 1), while 40% of the parasitoids landed first on an undamaged plant if these were placed in between the release point and the host-infested plant (binomial GLM: c 2 ¼ 10.057, P ¼ 0.002; Fig. 7c).

DISCUSSION
Foraging direction (upwind or downwind) strongly influenced foraging efficiency and foraging behaviour of C. glomerata. In downwind foraging situations, parasitoids were less motivated to start foraging, less successful at host finding and needed more time until they managed to oviposit in a host caterpillar compared to the upwind foraging situation. The observed increase in foraging time was caused by prolonged foraging on nonhost-infested and undamaged plants in the downwind situation, indicating that perception of volatiles from plants located further away in a plant stand (which is possible only in an upwind situation) enhances effective foraging. In downwind foraging situations, volatile information only becomes available after parasitoids pass a plant and the foraging insects lack information about the presence of  herbivores on plants located further downwind. Our results show that patch residence time during individual visits of nonhost patches was longer in the downwind foraging situation and that parasitoids returned more often to these unprofitable patches in the downwind foraging situation. Even though downwind foraging is not usually considered in studies that address odour-guided foraging in insects, field studies indicate that parasitoids often move downwind or crosswind and may thus have to deal with a limited availability of volatile information (Avila et al., 2013;Canto-Silva et al., 2006;Corbett & Rosenheim, 1996;Desouhant et al., 2003;Fournier & Boivin, 2000;Greatti & Zandigiacomo, 1995;Hendricks, 1967;Langhof et al., 2005;Pomari-Fernandes et al., 2018;Saavedra et al., 1997;Sallam et al., 2001;Williams et al., 2007;Yu et al., 1984). Wind speeds in the field often exceed the flight capacity of parasitoids and impede odour-guided upwind foraging (Vosteen et al., 2020 in press). Under high wind conditions, parasitoids ceased their foraging activity and waited for more favourable conditions, but this reduction in the time available for foraging decreases parasitoid fitness (Weisser, Volkl, & Hassell, 1997). If strong winds persist, parasitoids may thus engage in downwind or crosswind foraging despite the   observed reduction in foraging efficiency during downwind movement. Below, we discuss how foraging direction in relation to wind direction alters the way in which parasitoids perceive host and nonhost-infested plants, move within a plant stand and how this mediates parasitoid behaviour on nonhost-infested plants.

Confusion by Nonhost-Infested Plants
Parasitoids in the downwind foraging situation spent in total more than twice as long exploring nonhost-infested plants as those in the upwind foraging situation. Confusion of C. glomerata by nonhost-infested plants has already been observed in upwind foraging situations in a wind tunnel study and probably occurs because these parasitoids do not distinguish between HIPVs from plants infested with host or nonhost caterpillars .
Irrespective of the foraging situation, the majority (81%) of C. glomerata started to forage on the first herbivore-infested plant that they approached. In the upwind foraging situation, all three infested plants could already be perceived at the release point and neither herbivore identity nor plant position influenced which plant was approached first. In the downwind foraging situation, the majority of parasitoids first approached the nonhost-infested plant 1, which was closest to the release point, and foraged intensively on this plant. In the upwind foraging situation, the parasitoids quickly explored the entire plant stand by approaching different infested and undamaged plants at a fast rate, while in the downwind foraging situation, switches between the different plants in the stand occurred less frequently.

Foraging Behaviour on Nonhost-Infested Plants
When parasitoids landed on any of the herbivore-infested plants, they usually foraged intensively on the plant by flying between the leaves and walking on the plant surface. Patch exploration behaviour and residence time on nonhost-infested plants differed between the two foraging situations. During upwind foraging, parasitoids stayed on average about 1.5 min on a nonhost-infested plant. This time appeared to be sufficient to find a host caterpillar if hosts were present on a plant, since from those parasitoids that managed to oviposit in a host caterpillar, 80% found the first host within 1.5 min after arrival at the host-infested plant. According to the studies of Bukovinszky et al. (2012) and Steven et al. (2019), it is not the contact with nonhost caterpillars that triggers patch leaving in C. glomerata and Cotesia kariyai, but the lack of contact with hosts or host-associated cues such as frass. This foraging strategy, in which the parasitoid leaves a patch after a certain amount of time, unless contact with hosts or highly specific host-associated cues stimulates further patch exploration, ensures that early contact with nonhosts does not prevent the detection of hosts on dual infested plants.
HIPVs emitted from neighbouring plants can be used to estimate habitat profitability; for example, a parasitoid and a predatory mite remained longer in a patch when HIPVs or their absence indicated a low habitat profitability (Maeda, Takabayashi, Yano, & Takafuji, 1998;Mayland et al., 2000;Tentelier & Fauvergue, 2007). Similarly, in the downwind foraging situation in our experiment, C. glomerata foraged longer on nonhost-infested plants than in the upwind foraging situation before they either left the plant or started to attack nonhosts. This result suggests that C. glomerata perceives the habitat in the downwind foraging situation as less profitable due to the unavailability of volatile information from neighbouring plants that indicate herbivore presence and adjusts its giving-up time accordingly, thereby increasing its chances of detecting a host to 92%, if the preferred gregarious host is present on the plant.
Nonhost caterpillars were more readily attacked by C. glomerata in the downwind than the upwind foraging situation, indicating   that the perceived habitat profitability alters the parasitoid's responses towards nonhost caterpillars. Parasitoids use contact chemicals such as cuticular hydrocarbons to recognize suitable hosts (e.g. Muratori, Le Ralec, Lognay, & Hance, 2006;Roux et al., 2007) and individuals that are highly motivated from previous host contact are more likely to attack nonhosts (Bukovinszky et al., 2012;de Bruijn, Vet, & Smid, 2018;Vosteen et al., 2019). Attacks on nonhosts thus seem to be merely the result of host selection mistakes by highly motivated individuals, but our results suggest that they may be adaptive if hosts are rare. Suitability of different insect species as hosts for a specific parasitoid species can be seen as a continuum between highly suitable hosts and completely unsuitable nonhosts (Brodeur & Vet, 1995;Desneux, Barta, Hoelmer, Hopper, & Heimpel, 2009) and higher acceptance of low-quality hosts in low-quality habitats may increase parasitoid fitness (Tentelier & Fauvergue, 2007). Time-limited parasitoids may thus benefit from laying eggs in insect species outside their normal host range in situations when hosts are not available, since survival rates in some of the nonhost species may be higher than zero.

Undamaged Plants Distract Foraging Parasitoids
Foraging on undamaged plants was observed in both foraging situations, but a significantly higher percentage of parasitoids showed this behaviour in the downwind situation. Parasitoids spent more than twice as long exploring undamaged plants in the downwind than the upwind situation. Cotesia glomerata usually uses HIPVs to distinguish infested from undamaged plants (e.g. Vosteen et al., 2019), but these volatile cues are not available when a plant is approached downwind. Consequently, parasitoids in the downwind foraging situation were only able to perceive whether the plant was infested or not after they had come close to it or passed it. Our second experiment revealed that undamaged plants increased foraging time before the host-infested plant was located in upwind foraging situations when they were placed between the parasitoid release site and the host-infested plant. Forty per cent of the parasitoids landed on one of these undamaged plants, indicating that it is not only the physical structure that slows down parasitoid flight, but that overlapping odour plumes from the damaged and undamaged plants also make it difficult for the approaching parasitoid to distinguish between them. Large numbers of undamaged plants, for example in monoculture fields, may thus delay host finding in parasitoids.

Oviposition Experience Increased Foraging Motivation
It is generally assumed that learning increases parasitoid foraging efficiency, but recent studies suggest that in the presence of nonhost-infested plants the benefit of learning is context dependent. If C. glomerata had gained host oviposition experience on the same plant species as that harbouring the nonhosts in the foraging experiment, their foraging efficiency decreased compared to inexperienced parasitoids, because they were more strongly attracted towards the nonhost-infested plants than the latter . In de Bruijn et al.'s (2018) study, foraging efficiency of upwind-flying parasitoids was reduced when nonhosts were present on the plant species on which the parasitoid had gained oviposition experience, while host caterpillars were present on a different plant species, indicating that learning of information that does not match the current foraging environment decreases parasitoid foraging efficiency.
When hosts and nonhosts were on the same plant species, a decrease in foraging efficiency of oviposition-experienced C. glomerata was observed in a two-choice experiment and in one of two experiments in which a row of three nonhost-infested plants was placed in front of the host-infested plant ). In the current study, foraging efficiency in a plant stand with undamaged, nonhost-infested and host-infested plants was higher for oviposition-experienced parasitoids than for inexperienced individuals and oviposition experience did not influence the time spent foraging on nonhost-infested plants. These results suggest that the increased attraction of oviposition-experienced parasitoids to nonhost-infested plants may be most pronounced in simple foraging environments, while in more complex environments, other factors such as foraging motivation may determine foraging success and efficiency.
Those parasitoids that did not manage to oviposit usually gave up before they found the host-infested plant and differences in foraging motivation between oviposition-experienced and inexperienced parasitoids probably explain the observed differences in foraging success in our study. Oviposition experience has been reported to increase the motivation of parasitoids to start foraging (Geervliet et al., 1998;Peñaflor et al., 2017) and more ovipositionexperienced than inexperienced parasitoids started foraging both in the upwind foraging situation where volatile cues were available and in the downwind foraging situation where no volatile cues were present at the release site. Additionally, in the upwind foraging situation, fewer oviposition-experienced parasitoids foraged on undamaged plants, indicating that the sensitization towards HIPVs that followed an oviposition experience on a host- infested leaf  triggered avoidance of undamaged plants. The effect of oviposition experience on parasitoid foraging efficiency in complex environments thus seems to depend on various factors and associative learning and sensitization may increase foraging success in some environments, while in others, the effect is negative or neutral.

Conclusion
Availability of volatile information influenced the way in which parasitoids moved within a plant stand containing host-infested, nonhost-infested and undamaged plants, as well as their patch residence time on nonhost-infested plants. In downwind foraging situations, in which volatile information was limited compared to upwind foraging situations, parasitoids foraged less efficiently, suggesting that high wind speeds, which force parasitoids to fly downwind, may reduce their host-finding success under field conditions. Future studies on parasitoid foraging behaviour should thus not be done on isolated plants or in simple two-choice assays, but in complex foraging environments with several host and nonhost-infested plants and further take the effects of wind direction on movement of odour plumes into account.