Introduction

Changes in land-use systems and increasing land-use intensity have been identified as main drivers of biodiversity loss in all types of terrestrial habitats, from agricultural landscapes to forests (Dormann 2007; Lambin et al. 2001; Sala et al. 2000). The consequences of management regimes for interaction-based ecosystem processes and services are fairly well studied in agricultural cropping systems (Chaplin-Kramer et al. 2011; Letourneau et al. 2012; Rand et al. 2012). Thus, we know that anthropogenic intensification in agriculture can strongly impact species interactions and ecosystem services such as biological control (reviewed by Tscharntke et al. 2005). Herbivore parasitoids and predators have important functions in controlling insect herbivore outbreaks, but with increasing land-use intensity, the abundance and diversity of the herbivores’ natural enemies decreases (reviewed by Bianchi et al. 2006; Jonsson et al. 2012).

In forest ecosystems, biological control mechanisms are not well studied overall and thus our knowledge of the role of forest management regimes on this important ecosystem service is weak (Didham et al. 1996). Forest management might influence arthropod communities either directly by mechanically reducing population size during harvest activities or indirectly by, e.g., affecting habitat availability (shelter, overwintering structures, etc.), habitat complexity and heterogeneity, tree species composition or prey availability (e.g., Brunet et al. 2010; Lange et al. 2011; Paillet et al. 2010). Structural simplification as well as the substantial reduction of plant species richness might be the consequences of converting structurally highly complex natural forests into one-layered plantations of a single tree species. As structural complexity and plant architecture of habitats has been described as one main factor influencing movement and host finding of herbivores as well as their antagonists (Goodwin and Fahrig 2002; Hannunen 2002; Obermaier et al. 2008), forest management intensification might highly affect interactions between plants, herbivores and insect predators and parasitoids. Additionally vegetation structure might be important as shelter, etc. for herbivore antagonists. Only a few studies, however, have investigated the role of forest management regime on higher trophic level insects such as parasitoids and predators and even less on interactions between trophic levels (e.g., Dodd et al. 2012). In boreal forests of Sweden, Hilszczanski et al. (2005) showed that at the species level the assemblages of saproxylic ichneumonid parasitoids are significantly affected by the forest management regime. For temperate beech forests, Sobek et al. (2009) concluded from their study on herbivory and relative abundance of predators across a tree diversity gradient in the Hainich National Park in central Germany that beech herbivory is mediated bottom-up by resource concentration as well as regulated top-down by natural enemies.

In the past, numerous laboratory and greenhouse experiments showed that parasitoids and predators use volatiles released from herbivore-damaged plants as major cues in finding their hosts/prey (reviewed by Dicke 2009; Heil 2008; Unsicker et al. 2009). However, experimental evidence for the role of herbivore-induced volatiles in indirect defense under field conditions is scant (Hare 2011). Very few studies document the effect of specific volatiles or volatile blends for insect attraction under field conditions, but almost all of these were performed in agricultural systems (i.e., Bernasconi Ockroy et al. 2001; Degen et al. 2012; Simpson et al. 2011). Klemola et al. (2012) investigated rates of caterpillar parasitism in herbivore-infested versus almost non-infested birch stands. In this study caterpillars on herbivore-infested birch trees were significantly more frequently parasitized than caterpillars on control trees. The authors argue that this difference is due to an attraction of parasitoids to herbivore-induced volatiles released from birch (Klemola et al. 2012).

The aim of our study was to investigate how the abundance of insect herbivore enemies is influenced by management practices in forests dominated by the European beech (Fagus sylvatica). We assessed enemy abundance by determining their attraction to herbivore-induced volatiles. Therefore we first investigated the role of tree volatile emission in mediating natural enemy attraction in beech forests. We hypothesize that young beech trees infested with Lymantria dispar caterpillars will release a quantitatively and qualitatively different odor blend compared to non-infested control trees and thus attract more parasitoids and herbivore enemies than the controls. Furthermore, we hypothesize that the composition of species attracted to the herbivore-induced beech odor will differ with respect to forest management regime due to differences in the population densities of insects and varying host-finding success. Studies in agricultural systems and grasslands have already shown that higher trophic-level predators and parasitoids can be strongly affected by land-use intensification (e.g., Tylianakis et al. 2007; Herbst et al. 2013) and concurrent changes in plant species richness and structural vegetation complexity (Randlkofer et al. 2010; Meiners and Obermaier 2004). For forest ecosystems, these relationships are poorly studied.

We selected European beech (F. sylvatica) for our experimental approach as it is the most dominant deciduous tree species in Central European forests (Brunet et al. 2010) and occurs across various forest management regimes. European beech is a late successional species and has significant economic value for timber production. Apart from the numerous studies on the role of European beech volatiles in atmospheric chemistry (i.e., Dindorf et al. 2006; Holzke et al. 2006; Tollsten and Müller 1996), there is virtually only one study documenting the release of herbivore-induced volatiles by beech (Joó et al. 2010).

We carried out laboratory and field experiments to address the following questions:

  1. 1.

    What is the herbivore-induced volatile blend released by F. sylvatica?

  2. 2.

    Is there differential attraction of forest insects, in particular natural enemies, to herbivore-infested versus control trees in the field?

  3. 3.

    Is there an effect of forest management on the attraction of forest insects to herbivore-infested beech trees?

Materials and methods

Study region and sites

The study was conducted in the Hainich-Dün region within the German Biodiversity Exploratories Project (Fischer et al. 2010). This region comprises a range of hills in north-west Thuringia, Germany (10°10′24″–10°46′45″E, 50°56′15″–51°22′43″N). Altitude ranges between 285 and 550 m a.s.l. (Fischer et al. 2010), whilst the mean annual temperature range is 6.5–8 °C, and the mean annual precipitation is between 500 and 800 mm. The region consists of a large closed forest area of 16,000 ha dominated by broad-leaved trees. The forests are managed with varying intensity ranging from unmanaged forests within the National Park Hainich to intensively managed age-class forests.

We performed our study at 18 forest sites of different forest management regimes that were dominated by European beech (F. sylvatica L.) (Table 1). Unmanaged forests (no-management sites, n = 5 sites) exclusively occur within the national park. They are characterized by uneven-aged (up to 250 years) mixed deciduous forest (Fagus sylvatica, with Fraxinus excelsior L., Acer pseudoplatanus L. and others). The core area of the national park, where there was no forest management for at least 40 years, comprises 261 ha and is unique for Germany in terms of its age, structure and extent. In the selection-cutting beech forests (low-management sites, n = 4) only selected trees are harvested and trees in the gaps recruit from seed fall. This type of management creates a permanent forest of uneven-aged stands of different tree species of different age. Age-class beech forests (high-management sites, n = 9) are characterized by homogeneous even-aged stands of different developmental stages (age classes); we studied thickets (n = 3) characterized by trees of a diameter at breast height (DBH) <7 cm which corresponds to a tree age of <30 years, pole wood (n = 3; 7–15 cm DBH; 30–50 years), and old timber (n = 3; >30 cm DBH; >90 years) (Table 1). The rotation for high-management beech forests is around 160 (± 30) years. Because of a change in management practices, the younger high-management forests did not originate from clear-cuts with subsequent planting, but included both natural seedling recruitment and the preservation of some mature trees (120–180 years old) that were cut only once the young cohort reached an age of somewhere between 10 and 20 years. Both the no-management forests and the high-management forests are characterized by a closed canopy without big gaps. These shady conditions result in a low cover of the shrub layer, in particular in the studied young age-class beech stands (4 ± 3 SE %). The cover of the shrub layer increased from old age-class timber stands (22 ± 6 %) to unmanaged forests (26 ± 5 %), to the uneven-aged (selective cutting) stands (38 ± 10 %). Age-class thicket stands had by definition (woody species <5 m) a relatively high shrub layer (47 ± 14 %) [data from vegetation relevés, see Boch et al. (2013); Boch, personal communication].

Table 1 Number of experimental plots in forest types of different management intensity and the corresponding number of beech phytometers [beech phytometers infested with Lymantria dispar caterpillars (Herbivory), non-infested phytometers (Control)] and flight-interception traps installed in these plots, respectively
Full size table

Plant and insect material

We decided to use transplants instead of naturally occurring beech trees in order to avoid confounding factors that we could not control for in the field (e.g., different soil properties at the sites, different damage levels by herbivores and plant pathogens). The F. sylvatica trees were all around 1.50 m high and were purchased from a tree nursery, grown from regional seeds (Müller-Münchehof, Seesen-Münchehof, Germany). Twelve trees were used for the laboratory and 54 for the field experiment.

Gypsy moth (L. dispar) caterpillars originated from a culture at the Max Planck Institute (MPI) for Chemical Ecology and were grown on artificial wheat germ diet (gypsy moth diet; MP Biomedicals, Illkirch, France).

Attraction of insects to beech trees

We performed our field experiment for a period of 36 days between 29 May and 3 July 2008 on all plots in parallel. In each of the 18 plots two beech phytometers were installed (total 36 trees, plus 18 trees that were exchanged during the experiment) where one was infested with caterpillars (herbivory treatment, 15 third-instar caterpillars per tree), and the other functioned as a ‘control’ tree. In each plot the distance between the two phytometers was 100 m. To prevent caterpillars in the herbivory treatment from escaping and to avoid predation by birds and ground-living predators, all phytometers were caged (for a detailed description of the experimental set up, see Fig. S1).

Composite flight-interception traps (Fig. S1) were used to catch insects that might be attracted to the beech phytometers. This trap type has been shown to representatively assess activity densities of flying insects of different trophic levels including parasitoid communities in forests (Springate and Basset 1996; Jäkel and Roth 2004). At the top of each beech phytometer (treatment and control) one trap was installed (n total = 36 traps). The traps consisted of a crossed pair of gauze (50 cm × 50 cm) with a funnel of tough material attached to the bottom and to the top; at the end of both funnels sampling jars were mounted filled with killing and preserving agent (3 % copper sulphate solution) and a drop of detergent to reduce surface tension. The traps were emptied three times at 12-day intervals and all arthropods were transferred to 70 % ethanol in the field. Insects were sorted into taxonomic orders in the laboratory. Determination of Coleoptera, Diptera (only parasitoids), Hemiptera, Hymenoptera, Neuroptera and Thysanoptera was performed to a level at which feeding guild (herbivores, suckers–chewers; predators, parasitoids) could be classified. Hemiptera and Thysanoptera were classified as suckers and Coleoptera, Hymenoptera, and Neuroptera as chewers. Among Hemiptera all Heteroptera were identified to species level (by M. M. G.) and classified as herbivores and predators according to Wachmann et al. (2004–2012). All Homoptera and Thysanoptera were classified as herbivores. Coleoptera were identified to a level at which classification into herbivores and predators according to Böhme (2005) was possible (e.g., all specimens of the family Curculionidae were classified as herbivores, whereas in Staphylinidae genus-level determination was necessary). All Hymenoptera were sent to specialists for further identification and classification. Only parasitic Diptera (Tachinidae and Pipunculidae) were identified and integrated in statistical analysis as the other, more than 2,000, trapped Diptera could not be assigned to a specific trophic level (the diversity of trophic levels within families is high and ecological knowledge is not sufficient for a reliable classification). Only flight-capable individuals (no larvae or nymphs) were used for further analysis of abundances as with flight-interception traps only activity densities of these insects can be estimated. Thus all non-winged insects and mites (mostly phoretic species) were excluded. Furthermore, more than 1,300 holometabolic larvae and all Collembola were ignored, as well as around 300 spiders that might have tried to use the traps to build webs.

Herbivory of all phytometers (gypsy moth treatments and control) was measured at the end of the experiment (in the herbivory treatment additionally mid experiment) by visually estimating the percent leaf area loss (caused by L. dispar caterpillars and others). Phytometers of the herbivory treatment were exchanged mid experiment, to avoid complete defoliation, which would have resulted in a drastically reduced emission of herbivore-induced volatiles.

To test if density and herbivory of surrounding trees (up to 10 m height) of the natural vegetation affected the trap catches, number of trees were counted and herbivory was measured in a circle of 2-m radius around each phytometer. These variables were considered as covariates in the analyses. Herbivory was assessed along four 2-m radial transects (north-east, south-east, north-west, south-west direction) at three locations (next to the cage, 1- and 2-m distance). At each location one leaf at three heights (lowest, middle, uppermost part of the regeneration up to 2 m height) was collected haphazardly by blindly pointing with a stick to the respective foliage in the tree. Herbivory was measured as average percentage of leaves damaged and as average leaf area consumed. For the calculation of leaf area consumed all leaves were spread on a white board containing a reference square and covered with non-reflecting Plexiglas before they were photographed with a digital camera. Digital photographs were analyzed with the graphics software Adobe Photoshop 8.0. Total remaining leaf area was determined by using the pixel number of the reference square as a template. Herbivore-related missing leaf area was then reconstructed. Leaf area was then determined again for the reconstructed leaves and leaf damage was quantified using the differences in pixel number between the reconstructed and the damaged leaves.

Release of herbivore-induced volatiles by beech trees

To investigate the emission of volatiles from young F. sylvatica trees after L. dispar feeding, 12 trees (~150 cm in height) from the same nursery as those for the field experiment were selected. Six of these entered the caterpillar-herbivory treatment, and six functioned as non-infested control trees. For experimental herbivory and following volatile collections, all trees were transferred to a shaded area outside the MPI-ICE greenhouse in Jena. The trees were completely enclosed with mesh bags and 15 caterpillars in fourth and fifth instar were released on the herbivore-treated individuals. Twenty-two hours after the onset of gypsy moth caterpillar feeding, the volatile collections started. The mesh bags were exchanged by PET foil bags (Toppits Bratschlauch, Minden, Germany). Volatiles were collected for 4 h around noon with 20 mg Super-Q filters using a push–pull system supported by a compressor. Air was purified through a charcoal filter before it entered the headspace of the respective tree at a flow rate of 1 l min−1. Simultaneously air was sucked out of the bags through the Super-Q filter at a rate of 0.5 l min−1. Slight overpressure in the bags was deliberate during volatile collection to avoid contamination from the outside. After the collection the volatile compounds were desorbed by eluting the filter twice with 100 µl dichloromethane containing nonyl acetate as an internal standard (concentration 10 ng µl−1). Samples were stored at −20 °C until further analysis.

To relate volatile emission to the actual leaf biomass on the tree, all leaves were harvested right after volatile collections, dried at 80 °C in a drying oven and then weighed.

Qualitative and quantitative determination of the volatiles emitted from F. sylvatica was conducted using a Agilent 6890 series gas chromatograph coupled to a Agilent 5973 quadrupole mass selective detector (interface temperature 270 °C, quadrupole temperature 150 °C, source temperature 230 °C, electron energy 70 eV) and a flame ionization detector operated at 300 °C. The constituents of the volatile bouquet were separated with a DB-5MS column (Agilent, Santa Clara, CA; 30 m × 0.25 mm × 0.25 µm) and He (mass spectrometer) or H2 (flame ionization detector; FID) as carrier gas. One microliter of the sample was injected splitless at an initial oven temperature of 40 °C. The temperature was held for 2 min and then increased to 155 °C with a gradient of 7 °C min−1, followed by a further increase to 300 °C with 60 °C min−1 and a hold for 3 min.

Compounds were identified by comparison of retention times and mass spectra to those of authentic standards obtained from Fluka (Seelze, Germany), Roth (Karlsruhe, Germany), Sigma (St. Louis, MO) or Bedoukian (Danbury, CT), or by reference spectra in the Wiley and National Institute of Standards and Technology libraries and in the literature (Joulain and König 1998). The absolute amount of all compounds was determined based on their FID peak area in relation to the area of the internal standard using the effective carbon number concept (Scanion and Willis 1985).

Data analysis

We used generalized linear mixed-effects models (GLMM) to test for effects of forest type and treatment (herbivory/control) on: (1) the herbivory of beech trees in the surroundings (proportion of leaves damaged/proportion of leaf area consumed) of the cages; and (2) the abundance of herbivores (chewers/suckers), predators and parasitoids sampled by composite flight-interception traps next to the cages. In both GLMMs, forest site was used as random factor and forest type and treatment as fixed factors. Additionally an interaction term between main factors (treatment, forest type) was considered. In (1) a binomial error distribution with logit link-function was specified. In (2) density of regeneration and herbivory in the surrounding (mean percentage of leaf area consumed) were used as covariates. Here a Poisson error distribution with log link-function was specified. To test for overdispersion we calculated the ratio of sum of squared Pearson residuals and the residual df. In all cases, the ratio was between 0.98 and 1, except for the model regarding suckers where we specified a quasi-Poisson error distribution. Experimental caterpillar herbivory is shown as  % leaf area loss. Because of the unbalanced design regarding forest types we additionally tested if results are consistent when using the recently published quantitative silvicultural management intensity index (SMI) (Schall and Ammer 2013). Analyses were performed with the package lme4 within R 3.0.2 (R Development Team 2013).

Non-metric multidimensional scaling (NMDS) was used to test whether communities of parasitoids differ between treatment and forest type. The analysis was based on Bray–Curtis dissimilarities using the package vegan in R. As a measure of variance, we provide SEs throughout the manuscript.

Results

Natural herbivory in young beech regrowth

Leaf herbivory in naturally occurring young beech regrowth surrounding the beech phytometers showed high variability among forest sites (Fig. 1). However, natural herbivory was neither significantly different between forest types nor between the beech regrowth surrounding the phytometers experimentally infested with gypsy moth caterpillars (30 ± 2 % of all leaves were damaged, 3 ± 0.3 % leaf area loss) and the control phytometers (37 ± 3, 4 ± 0.8 %) plots (Fig. 1; GLMM; p > 0.10 in all cases).

Fig. 1
figure 1

Herbivory (mean and SEs) of beech natural regeneration in the surroundings of the experiment separated by forest type (top panel) and beech phytometers (bottom panel) of herbivory treatment (n = 18) and control (n = 18). Left Average percentage of leaves damaged, right average percentage of leaf area consumed. Note that herbivory of control phytometers was measured only once, at the end of the experiment. Phytometers of the herbivory treatment were exchanged mid experiment

Full size image

Experimental herbivory in beech phytometers

Aphid infestation by Phyllaphis fagi was already detected in all phytometers before the beginning of the experiment. Experimentally exposed caterpillars significantly increased the level of foliar herbivory of beech phytometers compared to the control trees, where herbivory was negligible (Fig. 1). Two weeks after introducing L. dispar caterpillars, 10–90 % of all leaves of the trees in the herbivory treatment were damaged (mean 85 ± 3 %). Average herbivory was 37 ± 6 % leaf area loss (in 30 % of all leaves >50 % of leaf area was consumed). Experimental herbivory in the second cohort of beech phytometers installed half-way through the experiment was also high (leaves damaged 66 ± 7 %; leaf area loss 26 ± 5 %), with often only the leaf veins left by the end of the experiment. In the control trees herbivory was much lower, showing 23 ± 7 % leaves damaged with an average of 4 ± 1 % leaf area loss observed at the end of the experiment, but there was also a high variation in herbivory among forest sites (Fig. 1).

Attraction of insects to beech phytometers

We sampled a total of 7,684 arthropods during the 36 days of the experiments. More individuals were caught next to the herbivory treatment (4,271; mean 237 ± 43) compared to the control plots (3,413, mean 190 ± 23; GLMM, z = 9.77, p < 0.001). Of these we assigned the 1,497 individuals with wings to trophic guilds (see Materials and methods). Chewing herbivores showed highest numbers (585), followed by sucking herbivores (452), parasitoids (269) and predators (191). Among parasitoids we observed 83 (morpho-) species of 21 families (see Table S1).

Experimental caterpillar herbivory in the beech phytometers significantly affected insects caught with the flight-interception traps. The most conspicuous effect was observed in parasitoids. The abundance of parasitoids was significantly positively affected by experimental herbivory (Fig. 2; Table 2) and this was independent of the management regime in the different forest sites. Results were consistent when using SMI as a quantitative measure of land-use intensity measure (GLMM, SMI, z = 0.54, p > 0.5; treatment, z = 3.0, p < 0.01). No significant interaction was observed between treatment and forest type (p > 0.05). When only using the data of the first 2 weeks (before exchanging phytometers in the herbivory treatment) the treatment effect on the abundance of parasitoids was still significant (GLMM, z = 2.0, p < 0.05). Parasitoid abundance was neither influenced by the density of natural beech regrowth nor herbivory in these trees. Additionally, parasitoid species richness was positively affected by experimental herbivory (z-value 2.0, p = 0.042). Most of the caught parasitoids (>90 %) are known to parasitize herbivorous insects (see Table S1). By focusing on known parasitoids of herbivores only, the effect of treatment remained consistent, but was marginally non-significant (z-value 1.7, p = 0.098). When the number of parasitoid individuals was additionally fitted to the model, no additional effect of treatment on parasitoid species richness was observed (GLMM, p > 0.10). Moreover, parasitoid community composition did not differ between herbivory treatment and control plots (NMDS based on Bray–Curtis dissimilarities; Fig. S2).

Fig. 2
figure 2

Average difference between herbivore treatment and control separated by forest type in abundance of herbivores (a sucking, b chewing), predators (c) and parasitoids (d). Positive values indicate higher abundance in herbivore treatment, negative values indicate higher abundance in control treatment

Full size image
Table 2 Effects of treatment and forest type on the abundance of flight-capable insect herbivores (sucker, chewer), predators and parasitoids assessed by composite flight-interception traps next to the phytometers established in each forest plot
Full size table

In contrast, the abundance of predators was not affected by treatment (Fig. 2; Table 2). This effect was independent of forest type, as including an interaction term treatment × forest type in the model revealed no significant interaction (p > 0.05). Furthermore, management did not significantly affect predator abundance when using forest type (Table 2) nor when using SMI (GLMM, z = −1.8, p = 0.11).

Among the herbivores caught in the traps next to the phytometers, only sucking insects were positively affected by experimental herbivory, independent of forest management type (interaction treatment × forest type, p > 0.05). In contrast, chewing insect abundance was negatively affected by experimental herbivory in old stands, but positively in young high-management forest stands (Fig. 2). When an interaction term treatment × forest type was included in the model, this was significant (Table 2). Moreover, both abundance of sucking and chewing insects was significantly affected by management type (Fig. 2; Table 2) and SMI (GLMM, suckers, z = −2.8, p < 0.01; chewers, z = −2.9, p < 0.01). Sucking insect abundance was higher in old stands, in particular in the low-management sites (no management 12 ± 2.9, low management 21 ± 5.3, high-management old timber 14 ± 3.6), than in young stands (high-management pole wood 9.5 ± 2.3, high-management thicket 3.3 ± 0.67). Chewing insects showed highest abundance in no-management forests (31.90 ± 8.63 SE) followed by low-management (14 ± 3.9) and high-management sites (old timber 9 ± 2.4, pole wood 8.3 ± 2.9, thicket 7.3 ± 2).

All trophic guilds showed high variation in the number of sampled individuals among forest sites in herbivory treatment as well as in control plots (Fig. S3). Forest type significantly affected the abundances of herbivores, but not of predators and parasitoids, resulting in a substantial difference in guild composition among forest types (Table 2): herbivores generally showed low abundances in high-management thickets; sucking insects among them were most abundant in low-management sites and chewing insects in no-management forests.

Constitutive and herbivore-induced volatile emission from beech phytometers

We identified over 20 volatile compounds belonging to five major groups, namely monoterpenoids, sesquiterpenoids, homoterpenoids, green leaf volatiles and aromatics, in the volatile blend of young F. sylvatica trees (Table 3; Fig. S5). Sesquiterpenes and the homoterpene 4,8-dimethyl-1,3,7-nonatriene (DMNT) were significantly induced by experimental herbivory of gypsy moth caterpillars by showing up to 30-fold increased concentrations, whereas herbivory had no significant effect on the emission of other major groups of volatiles (Fig. 3). Among individual compounds, the highest increase in emission due to herbivory was observed in the sesquiterpene germacrene D (160-fold) followed by β-bourbonene (91-fold), another sesquiterpene, and the monoterpene α-terpinolene (55-fold). The most abundant compound in the herbivore-induced and the constitutive volatile blend of the beech trees was the monoterpene sabinene. The monoterpenes, α-pinene, trans-β ocimene and α-terpinolene, were also significantly induced by caterpillar feeding (Table 3). The only detectable green leaf volatile was (Z)-3-hexenyl-acetate, and methyl-salicylate was the only aromatic compound that was measured. Both occurred in similar concentrations in control and herbivory treatment.

Table 3 Volatile compounds of Fagus sylvatica released from undamaged leaves (Control) and leaves damaged by gypsy moth caterpillars for 24 h (Herbivory)
Full size table
Fig. 3
figure 3

Emission of major groups of volatiles from young Fagus sylvatica trees infested locally with Lymantria dispar caterpillars as compared to non-infested control trees. Mean ± SE, sample sizes n = 6. Asterisks indicate significant differences between controls and herbivore-infested trees (*p < 0.05; Mann–Whitney U-tests). MT Monoterpenes, ST sesquiterpenes, AROMA aromatic compound (methyl salicylate), HT homoterpene (4,8-dimethyl-1,3,7-nonatriene), GLV green leaf volatile [(Z)-3-hexenyl acetate], OTHERS non-analyzed, ns non-significant

Full size image

Discussion

In this study we investigated the effects of herbivore-induced volatiles released from beech trees on the attraction of herbivores and their predators and parasitoids in forest stands of differing management intensities. Next to caterpillar-infested beech phytometers significantly more parasitoids were caught as compared to non-infested control trees. This pattern was observed irrespective of forest type (from unmanaged to intensively managed) although the composition of trophic guilds in the traps did vary in response to forest management regime. Neither the number of naturally occurring beech saplings in the proximity of our experiment nor the herbivory levels in these explained the abundance and diversity of parasitoids caught. Our results suggest that differences in the structural complexity of forests as a consequence of management regime only play a minor role for parasitoid activity. An alternative explanation could, however, also be that the attraction to herbivore-induced volatiles has overcome any difference that there may have been in terms of forest structure effects on parasitoid abundance. Despite the change in abundance, no change in parasitoid community composition across management regimes could be detected.

When young beech trees in our experiment were attacked by L. dispar caterpillars they emitted significantly more volatiles than the non-infested control trees. The headspace consisted of around 21 compounds from within five major groups of volatiles (monoterpenes, sesquiterpenes, green leaf volatiles, aromatic compounds and the homoterpene DMNT). While individual compounds emitted from European beech such as the green leaf volatile (Z)-3-hexenyl-acetate, the homoterpene DMNT and the sesquiterpenes (E)-β-caryophyllene and germacrene D have already been reported for other, taxonomically unrelated plant species, such as maize, lima bean, poplar, plantain, clover, oak and many others (Turlings and Tumlinson 1992; Heil and Silva Bueno 2007; Fontana et al. 2009; Kigathi et al. 2013; Danner et al. 2011; Ghirardo et al. 2012), the volatile blend of F. sylvatica shows some unique features: the bouquet is overall dominated by the monoterpene sabinene and the aromatic compound methyl-salicylate, both present constitutively. A number of mono- and sesquiterpenes are significantly induced by caterpillar feeding and green leaf volatiles are only present in minor amounts. Qualitatively, the volatile blend in this study was very similar to the aphid-induced bouquet measured by Joó et al. (2010). These similarities are likely due to an infestation of our beech phytometers with the same aphid species P. fagi as in the study by Joó et al. (2010). Although aphid infestation might have slightly leveled volatile emission from control trees and caterpillar-infested beech trees, the relative proportions of volatiles measured in our study are very different. Thus the pronounced differences in volatile emission between caterpillar-infested and control tree were certainly brought about by experimental caterpillar herbivory.

The pronounced difference in volatile emission between caterpillar-infested and control trees that we measured in a subset of beech phytometers under comparable abiotic conditions outside the MPI-ICE greenhouse in Jena shows that our experimentally induced herbivory was efficient and most likely accounts for the differences in total numbers of herbivores, predators, and parasitoids caught in the experiment. As L. dispar caterpillars were continuously present in the herbivore-treated beech phytometers, we cannot exclude specific volatiles emitted from the caterpillars and the frass to attract herbivore natural enemies. Yet a study by Clavijo McCormick et al. (2014) revealed that only very low amounts of volatiles are detected in the headspace of caterpillars and feces when measured alone. In laboratory studies some of the major herbivore-induced volatile compounds released from beech, such as germacrene D, DMNT and ocimene, have already been shown to attract parasitoids [Mumm and Dicke (2010) and references therein]. When autumnal and winter moth caterpillars were exposed in birch trees previously defoliated by caterpillars the rate of parasitization was more than twice as high as in caterpillars exposed on control trees (Klemola et al. 2012). The authors argue that herbivore-induced volatiles released from birch account for this parasitization pattern. Although we did not investigate in situ parasitization in our experiment, we are convinced that the differences in herbivores, parasitoids, and predators trapped in our experiment are also due to the differences in volatile emission between the caterpillar-infested and the control trees.

The parasitoid species attracted to the herbivore-infested beech trees in our experiment are neither known specialists of L. dispar caterpillars nor have they been documented to be restricted to the parasitization of beech herbivores. Thus we speculate that long-distance attraction of parasitoids towards herbivore-infested beech trees is rather species unspecific. It is conceivable that long-distance signaling is mediated by the most abundant herbivore-induced volatile compounds, which are in fact the ones that generally occur in the blends of most herbivore-induced plant species investigated so far (Clavijo McCormick et al. 2012).

Plant-induced attraction of herbivore antagonists is a well-known phenomenon (Dicke 2009; Heil 2008; Unsicker et al. 2009), but has rarely been shown under field conditions. A few studies performed in agricultural systems and in trees could show that herbivore-induced volatiles can attract parasitoids under field conditions (Braasch et al. 2012; Büchel et al. 2011, 2013; Clavijo McCormick et al. 2014; Degen et al. 2012; Simpson et al. 2011). Our study, with the study of Klemola et al. (2012), is one of the first to show the possible existence of this mechanistic link in forests. Surprisingly, predaceous insects were not significantly attracted by herbivore-induced beech volatiles. Flight-interception traps are not adequate to capture mostly flightless herbivore enemies such as ground beetles or ants.

We sampled a high number of insects of all trophic guilds and we were able to detect differences among forest types and between herbivory treatment and the control. Adult herbivores showed the most conspicuous response to forest type. While chewing insects showed highest abundance in no-management forests, sucking insects were most abundant in low-management forests. This could not be explained by differences in plant diversity (Boch et al., unpublished data; Fig. S2), but might by due to structural differences between forest types, i.e., high sucking insect diversity in low-management forests might be caused by the more open structure of these forests (Gossner 2009).

It is conceivable that more intensive management might disconnect the tri-trophic interaction between herbivores, plants and parasitoids resulting in less effective pest control in these forests (e.g., Klein et al. 2002). We, however, could not find a strong indication for this as the main effect of herbivory-induced attraction of parasitoids occurred independently of forest type. The difference in attraction of parasitoids between experimental and control trees was, however, stronger in unmanaged forests and forests with low (i.e., selection-cutting) compared to forests with high management intensity (i.e., age-class forests) (error bars crossed the zero line only in high-management forests; Fig. 2). One might speculate that this suggests a more effective pest control of herbivores in less intensively managed forests, which might have higher structural diversity and thus enable higher population densities of natural enemies. Surprisingly, sucking herbivores were also attracted to herbivore-induced phytometers, although this effect was much less pronounced than in parasitoids. Certainly insect herbivores can deploy plant volatiles to find adequate host plants. There are examples for this phenomenon in the recent literature (e.g., Carroll et al. 2006; Halitschke et al. 2008; Robert et al. 2013).

Several studies have shown that land use might affect trophic interactions which are important for ecosystem processes (Gladbach et al. 2011; Mulder et al. 2011; Tylianakis et al. 2008). For example it has been shown that plant diversity can stabilize food webs (Haddad et al. 2011) and in heterogeneous landscapes an increase in natural enemy diversity benefit pest control (Tylianakis and Romo 2010). Our mechanistic understanding of these links is, however, still rudimentary.

Although we know from numerous laboratory and field studies with short-lived herbaceous plant species that herbivore-induced plant volatiles attract natural enemies and parasitoids of herbivores, our study is among the first to demonstrate that this phenomenon also occurs in long-lived deciduous tree species such as F. sylvatica. Our data also suggest that indirect beech defense is independent of forest management type but further experiments with, for example in situ parasitization as a response tree volatile emission have to be conducted to confirm our findings.