Host Plant Phenology, Insect Outbreaks and Herbivore Communities – the Importance of Timing

Received: 28 March 2019 | Accepted: 27 October 2019 DOI: 10.1111/1365-2656.13151

RESEARCH ARTICLE

Host plant phenology, insect outbreaks and herbivore communities – The importance of timing

Adam Ekholm1 | Ayco J. M. Tack2 | Pertti Pulkkinen3 | Tomas Roslin1

1Department of Ecology, Swedish University of Agricultural Sciences, Uppsala, Sweden Abstract 2Department of Ecology, Environment 1. Climate change may alter the dynamics of outbreak species by changing the and Plant Sciences, Stockholm University, phenological synchrony between herbivores and their host plants. As host plant Stockholm, Sweden 3Natural Resources Institute Finland, phenology has a genotypic component that may interact with climate, infestation Läyliäinen, Finland levels among genotypes might change accordingly. When the outbreaking herbi-

Correspondence vore is active early in the season, its infestation levels may also leave a detectable Adam Ekholm imprint on herbivores colonizing the plant later in the season. Email: [email protected] 2. In this study, we first investigated how the spring phenology and genotype of Funding information Quercus robur influenced the density of the spring-active, outbreaking leaf miner Swedish University of Agricultural Sciences Acrocercops brongniardellus. We then assessed how intraspecific density affected Handling Editor: Sharon Zytynska the performance of A. brongniardellus and how oak genotype and density of A. brongniardellus affected the insect herbivore community. 3. We found that Q. robur individuals of late spring phenology were more strongly infested by A. brongniardellus. Conspecific pupae on heavily infested oaks tended to be lighter, and fewer heterospecific insect herbivores colonized the oak later in the season. Beyond its effects through phenology, plant genotype left an imprint on herbivore species richness and on two insect herbivores. 4. Our results suggest a chain of knock-on effects from plant phenology, through the outbreaking species to the insect herbivore community. Given the finding of how phenological synchrony between the outbreak species and its host plant influences infestation levels, a shift in synchrony may then change outbreak dynamics and cause cascading effects on the insect community.

KEYWORDS climate change, community ecology, pest outbreak, phenology, temporal dynamics

1 | INTRODUCTION environmental change (Björkman, Bylund, Klapwijk, Kollberg, & Schroeder, 2011). Studies have so far identified several factors driv- Insect outbreaks can have vast ecological impacts by reducing ing insect outbreaks, including climate (Marini et al., 2017), induced plant growth (Chen et al., 2018; Cooke & Roland, 2007) and alter- responses in plants (Abbott & Dwyer, 2007) and densities of multiple ing the richness, abundance and composition of the insect commu- predators (Dwyer, Dushoff, & Yee, 2004). nity (Roslin & Roland, 2005). Exactly why insect herbivore species One factor causing variation in plant susceptibility to insect her- sometimes rapidly increase in numbers and defoliate plant resources bivores is the spring phenology of plants. Phenological variation has is a question that is not yet completely understood. This hampers proven important in determining larval growth rate (Schwartzberg our understanding of how insect outbreaks may change with global et al., 2014), insect fitness (Tikkanen & Julkunen-Tiitto, 2003) and

J Anim Ecol. 2020;89:829–841. wileyonlinelibrary.com/journal/jane © 2019 British Ecological Society | 829 830 | Journal of Animal Ecology EKHOLM et al. the structure of insect communities (Crawley & Akhteruzzaman, the environment can both individually and interactively contribute 1988; Sinclair et al., 2015). Since plant and insect phenology are to explaining outbreak dynamics and the structure of insect commu- sensitive to temperature (Parmesan, 2007; Thackeray et al., 2016), nities. Distinguishing between the two is thus an important objec- an asymmetrical response to temperature at different trophic levels tive in order to understand how communities respond to a changing (Thackeray et al., 2016) could lead to a shift in interaction strength, environment. resulting in a weakened (Burkle, Marlin, & Knight, 2013; Kharouba In this study, we investigated how spring phenology and geno- et al., 2018; Stalhandske et al., 2016) or a strengthened interaction type of the host plant affect the density of a single outbreak insect (Kharouba et al., 2018; Schwartzberg et al., 2014). herbivore, and how variation in the density of the outbreak herbi- As the chemical properties of young leaves tend to change rap- vore and plant genotype affects the insect herbivore community. idly early in the season (Falk, Lindroth, Keefover-Ring, & Raffa, 2018; In particular, we focused on fitness-related proxies of conspecific Feeny, 1970; Salminen et al., 2004; Tikkanen & Julkunen-Tiitto, individuals of an early-season outbreak herbivore, on the commu- 2003), certain herbivores might benefit from feeding on the leaves nity structure of other herbivore species colonizing the plant later in during a narrow time window. A shift in climate may therefore allow the season, and on aphid population dynamics. By manipulating the some outbreak species to become better synchronized with their spring phenology of 159 grafted oaks (Quercus robur) and then ex- host plants. For instance, Pureswaran et al. (2015) hypothesized that posing them to a natural setting with high densities of the outbreak with a warming climate, the eastern spruce budworm (Choristoneura herbivore species, we ask: fumiferana) can become more synchronized with its secondary host plant, the black spruce (Picea mariana). This may then cause a geo- 1. What is the impact of tree spring phenology and genotype graphical shift in spruce budworm distribution and thus increase on the density of an outbreaking insect herbivore? To what the vulnerability of the black spruce forest to spruce budworm extent does the susceptibility of trees to the outbreaking in- outbreaks. Similarly, Jepsen et al. (2011) attributed the northward sect herbivore change with climatic conditions, i.e. is there a expansion of the moth Agriopis aurantiaria to an increasing pheno- genotype × environment interaction? logical synchrony with its host plant at higher latitudes. 2. What effect does the outbreaking insect herbivore have on the How variation in the density of the outbreak species affects the performance of conspecifics? rest of the community is a crucial question. Outbreak species could 3. How does the density of the outbreak species and tree genotype, interact directly with other insect herbivores by for example, preda- respectively, affect the structure of the insect herbivore commu- tion, or indirectly by depleting the plant resource or by inducing a nity and the population dynamics of an herbivore active through- response in the plant which alters plant quality. In terms of plant-in- out the season? duced responses, previous studies have demonstrated that insect 4. To what extent does tree genotype influence the performance of herbivores feeding on different parts of the same plant individual conspecifics and shape the insect community beyond its effects can have a negative affect each other's fitness (Tack, Ovaskainen, through phenology? Harrison, & Roslin, 2009), and that insect herbivores colonizing a plant in one year could have both a negative and positive effect on colonization by herbivores in the following year (Tack et al., 2009). 2 | MATERIALS AND METHODS Similarly, insect herbivores colonizing a plant early in the season could affect the performance of species colonizing the plant later in 2.1 | Study system the season (Van Zandt & Agrawal, 2004) and thereby alter the structure of the insect community (Roslin & Roland, 2005; Stam, Dicke, & The pedunculate oak Quercus robur (Fagaceae) sustains one of the Poelman, 2018; Timms & Smith, 2011; Van Zandt & Agrawal, 2004). most species-rich insect communities in Europe (Southwood, 1961; Hence, the appearance of an early-season outbreak species could Southwood, Wint, Kennedy, & Greenwood, 2004), and has been have a major ecological impact on the plant-associated food web. used as a model system for plant–herbivore interactions in numer- Plant phenology and other phenotypic traits can both be influ- ous studies (e.g. Crawley & Akhteruzzaman, 1988; Ekholm, Tack, enced by the environment and the plant-specific genotype. Plant Bolmgren, & Roslin, 2019; Feeny, 1970; Pohjanmies et al., 2015; genotypes can differ in traits such as carbon-nitrogen ratio, leaf sec- Tack et al., 2009). Apart from a distinct peak of herbivore chewers ondary metabolites and plant size (Barbour et al., 2015; Falk et al., in the early season, the Q. robur herbivore community is dominated 2018). Such a genotypic variation can influence herbivore abun- by sucking, galling or mining species for the remainder of the season dance (Barbour et al., 2015), community composition (Barbour et al., (Southwood et al., 2004). In Sweden, Q. robur is common through- 2015) and food web complexity (Barbour et al., 2016). In fact, her- out the southern part of the country and reaches its northern dis- itable variation in the plant-specific insect community has even been tribution limit at approximately 60°N (Stenberg & Mossberg, 2003). framed as parts of an ‘extended phenotype’ (Whitham et al., 2003). Phenology of Q. robur varies with latitude (Ekholm, Tack, Bolmgren, However, if the environment changes, then the relative impact of et al., 2019), but is also highly variable within oak populations, where plant genotype on the insect community could also change (G × E in- individuals can differ in their spring phenology by up to a month teraction). Hence, heritable traits related to specific genotypes and (Crawley & Akhteruzzaman, 1988; Wesolowski & Rowinski, 2008). EKHOLM et al. Journal of Animal Ecolog y | 831

A subset of the oak insect community, consisting of special- conditions in the greenhouse (including light conditions) resembled ized leaf mining and gall-inducing species, is particularly amenable those outside of the greenhouse. The greenhouse covering was to study due to the sedentary life stage of the larvae (Coulianos & made of clear glass and the control oaks were placed in an open field Holmåsen, 1991; Pschorn-Walcher, 1982; Svensson, 1993). Gall in- (c. 10 m from a forest edge), to minimize variation in photoperiod ducers and leaf miners are common on Quercus spp. and as guilds between the two treatments. As soon as the buds started to elon- they have a global distribution. The gall-inducing wasps are bivoltine gate and the tip of the leaves were visible on the earliest oaks in the with one sexual and one asexual generation per year (Coulianos & greenhouse treatment, we transported oaks from both treatments Holmåsen, 1991), while the leaf mining species are either uni- or bi- (between 2nd and 4th of May) to the field site in Länna, Sweden voltine (Pschorn-Walcher, 1982; Svensson, 1993). Beyond gall-induc- (59.87°N, 17.96°E). ing and leaf-mining species, a common aphid on oaks is Tuberculatus Our experiment was based on a design where tree phenology annulatus (Hemiptera). This aphid produce several generations per was advanced though brief exposure to elevated temperatures in a year and can be found on the underside of oak leaves (Heie, 1980, greenhouse environment. For practical reasons, the control group 1982). In Sweden, T. annulatus is active during the full vegetation was maintained under ambient, outdoor conditions. Although we season and is mainly found on Q. robur (Heie, 1982). made every effort to match all conditions beyond temperature be- Within the oak insect community, several species are known for tween treatments, it is difficult to exclude the existence of other, large temporal fluctuations in density among years (Blanchet et al., hidden differences. Our current inference is thereby conditional on 2018; Hails & Crawley, 1991). One species in particular has recently the assumption that the greenhouse environment introduced no caused outbreaks in Sweden: the leaf mining species Acrocercops other effects beyond advancing phenology—an assumption which brongniardellus (Lepidoptera: Gracillariidae). Adults of this species we find reasonable to make but impossible to test conclusively. are active early in the spring and oviposit on oak trees. After oviposi- Once at the field site, we divided the oaks into five blocks, tion, larvae enter the leaf, start feeding and—after some weeks—exit with each block containing at least one replicate oak from each the leaf by descending on a silk thread, to then pupate in the soil. Genotype × Temperature combination (28–35 oaks per block; a few Usually, there are several larvae feeding on the same leaf (Bengtsson of the oaks died during the experiment). We randomly assigned the & Johansson, 2011; Svensson, 1993) and within outbreak areas, position of each oak within each block. In order to detect differences larvae often consume a high fraction of the leaf tissue, thus drasti- in phenology between genotypes and temperature treatments, we cally reducing the total photosynthetically active leaf area per oak scored spring phenology on the tree level based on a modified BBCH- (Appendix S3; Figure S1). scale (using a subset of the original phenological stages) and defined the final stage (stage 19) as “day of leaf maturation” (Appendix S1; Table S1; Meier, 2001). Tree level phenology was recorded every 2.2 | Study design third or fourth day from the 26th of April until the 23rd of May, with a final survey at the 30th of May. 2.2.1 | Oak phenology

In this study, we utilized a set of 159 grafted oaks planted in 50 L 2.2.2 | Impact of oak phenology and genotype on pots. The material was established in years 2010–2013 and con- incidence and abundance of A. brongniardellus tains seven different clones (referred to as genotypes in this paper), propagated by grafting twigs from seven individual source oaks to To investigate the impact of oak genotype and phenology on the in- random root stocks (one twig per root stock). The source oaks rep- cidence and abundance of A. brongniardellus, we surveyed each oak resent the variation present within natural populations, as they were on two dates: 1st and 10th–11th of June. On the first date, we sur- all growing within a 5 km2 island in the southwestern archipelago of veyed the incidence of A. brongniardellus, where we defined “abun- Finland (Ekholm, Roslin, Pulkkinen, & Tack, 2017). dant” as more than five infested leaves per oak and “absent” as five To establish a difference in spring phenology among the grafted or less infested leaves per oak. At the second date, we surveyed the oaks, we advanced the phenology for half of the oak trees of each abundance (i.e. the proportion of infested leaves per shoot) by ran- genotype (n = 76) by placing them in a warm greenhouse from the domly selecting c. 10 shoots per oak and then counting the number 28th of March until the 2nd of May, 2018. Due to some (limited) of infested leaves per shoot. variation in size among oaks, we used constrained randomization to achieve a similar average oak size in the two treatment groups. To avoid positional effects, we randomized each oak in the greenhouse 2.2.3 | Effects of A. brongniardellus on conspecific treatment to a new position within the group on a weekly basis. As performance the greenhouse could not be set to a temperature matching ambient temperature, we kept the control oaks (n = 83) outdoors, allowing To estimate the performance of A. brongniardellus on oaks with dif- them to track the local temperature and thereby express a phenol- ferent genotypes and infestation levels of A. brongniardellus, we fo- ogy similar to local oak trees. For all aspects beyond temperature, cused on pupal weight. This metric is a proxy of individual fecundity, 832 | Journal of Animal Ecology EKHOLM et al. since heavier individuals within a species tend to be more fecund 28th of August; 155–158 oaks surveyed per occasion). To exclude (Haukioja & Neuvonen, 1985; Honĕk, 1993; Tammaru, Kaitaniemi, predators from feeding on aphids, we covered each branch with a & Ruohomaki, 1996). For this purpose, we used two different meth- mesh bag, starting at the first day of counting (4th of July). Leaves ods, both drawing on the fact that full-grown larvae of A. brongn- lost during the experimental period were not replaced. We then re- iardellus will readily pupate on almost any substrate, spinning a small lated the proportion of infested leaves and population growth rate silken cocoon. For the first method (pupation ex situ), we collected of T. annulatus on each tree to the infestation levels of A. brongn- one shoot per available oak (n = 102), with a minimum of three in- iardellus and oak genotype. Population growth rate was defined as fested leaves (for lightly infested oaks, we collected the shoot with the natural logarithm of the number of aphids found on a leaf at time the highest infestation), on the 30th of May. The shoots were then t divided by the number found at the previous sampling date t − 1, stored in closed plastic vials (400 ml) together with a humid paper i.e. loge (Nt + 1)/(Nt−1 + 1). For each date, the leaf-specific population tissue and placed in a climate chamber (20°C, 18L: 6D). After larvae growth rate was averaged for each oak and then loge-transformed. had exited the leaf and pupated in the plastic vial, we collected all pupae (mean number of pupae per oak 5.19 ± 0.36, range 1–19) and determined their sex and weight. Sex was determined based on the 2.3 | Statistical analyses shape of the segment adjacent to the terminal abdominal segment.

For the second method (pupation in situ), we covered one branch All statistical analyses were implemented in r (R Core Team, 2017) per oak (pupae obtained from 68 oaks) with a mesh bag, just before using the framework of (generalized) linear mixed effects models. the first larvae started to exit the leaf (29th of May). When we were Models were fitted with the functions glmer and lmer for GLMMs confident that all of the larvae had exited the leaf, we collected all and normally-distributed response variables (LMMs), respectively, of the mesh bags (10th–11th of June) and stored them in a climate as available in the package lme4 (Bates, Maechler, Bolker, & Walker, chamber (20°C, 18L: 6D), until they were weighed (mean number of 2015). For generalized linear mixed models, we checked for under- pupae per oak 3.38 ± 0.37, range 1–14) and sex determined. For the and overdispersion with the functions testUniformity and testDis- rationale behind combining two methods, see Appendix S5. persion from the dharma package (Hartig, 2018), and if under- or overdispersion was detected, the model was fitted with a quasi-dis-

tribution using function glmmPQL from the package mass (Venables 2.2.4 | Impact of A. brongniardellus on the insect & Ripley, 2002). Model simplification was applied to all models, with herbivore community interaction terms with a p > .10 (based on ANOVA-tests) excluded from the final model. Non-significant interaction terms shown in To assess the impact of oak genotype and infestation levels of A. the results are derived from the original, non-simplified models. For brongniardellus on the insect community, we surveyed each oak for models with a quasi-distribution or a temporal correlation struc- all herbivores present. Only leaf-mining and gall-inducing species ture, significance was assessed with a marginal ANOVA using func- were found in informative densities, whereas free feeders were rare tion anova.lme from the nlme package (Pinheiro, Bates, DebRoy, & or lacking. This caused us to focus on gallers and leaf miners alone, Sarkar, 2017), while a type III ANOVA from the car package (Fox & scoring these taxa at 4–5 occasions throughout the vegetation sea- Weisberg, 2011) was used for all other models. To account for vari- son (14th of June, 27th of June, 19th of July, 24th of August and ation among blocks, we included block as a random effect in each of 28th of September). During the four first occasions, all branches on the models described below. A detailed overview of the structure of each oak were visually inspected for leaf mines and galls, while for all models can be found in Appendix S1, Table S2. Trend lines in fig- the final occasion, each oak was surveyed for 90 s (a time sufficient ures are least squares means extracted with the function emmeans to establish the presence of species on the small target oaks). As using the package emmeans (Lenth, 2018). To analyse the separate metrics of community structure, we summed observations from the effect of genotype after controlling for any variation shared with whole season, and established species-specific occurrence and the oak phenology, we added phenology as a second fixed effect to the overall species richness of leaf mining and gall-inducing taxa at the models described below (except for those models that already in- level of each individual oak. cluded phenology as a fixed effect). These additional models effec- tively answer questions (iv) of to what extent tree genotype shapes the insect community beyond its effects through phenology. For lack 2.2.5 | Effects of A. brongniardellus on aphid of space, they are reported in Appendix S2 but discussed below. population dynamics

To assess the effects of A. brongniardellus densities and oak geno- 2.3.1 | Impact of oak genotype and phenology on type on the population dynamics of the aphid Tuberculatus annulatus, incidence and abundance of A. brongniardellus we counted aphids on ten marked leaves per tree, where all of the leaves belonged to a single branch. This survey was done at five oc- To investigate how spring phenology and oak genotype affected the casions (4th of July, 23rd of July, 1st of August, 17th of August and incidence and proportion of leaves infested by A. brongniardellus, we EKHOLM et al. Journal of Animal Ecolog y | 833 modelled the (a) oak-level incidence and (b) proportion of infested over all five survey dates, was affected by oak infestation of A. brongn- leaves by A. brongniardellus as a function of spring phenology and iardellus and oak genotype. Since the number of leaves surveyed per oak genotype (Appendix S1; Table S2; models 1 and 2). To detect oak differed slightly among dates, we included the total number of whether oaks of a certain phenology were more heavily infested by surveyed leaves from all dates (oak level mean = 48.79 ± 0.12, range A. brongniardellus than other oaks, we included spring phenology as 26–50) as a covariate (Appendix S1; Table S2; model 6). a linear and a squared term, thereby testing for both linear and non- To study aphid performance, we modelled aphid growth rate linear responses. As A. brongniardellus was absent from a few oaks, and proportion of aphid-infested leaves over time, respectively, we only tested for the interaction between spring phenology and as a function of oak infestation levels by A. brongniardellus, oak genotype when modelling oak-level infestation levels. In the analysis genotype and date. For this particular analysis, we used the subset described below, we use oak infestation levels of A. brongniardellus as of oaks (n = 111) on which aphids had been observed. As the num- an explanatory variable, which we define as the fraction of infested ber of aphids on a leaf at time t is likely to be correlated with the leaves on an oak (i.e. the number of infested leaves divided by the number of aphids at time t − 1, we assumed a temporal correlation total number of leaves). We did this by first calculating the fraction structure with a time step of one (function corCAR1 from pack- of infested leaves per shoot and then averaging among all shoots age nlme; Pinheiro et al., 2017). As the response variables were within an oak. This specific term is referred to as ‘infestation levels of taken at the oak level, we included the random effect oak tree A. brongniardellus’. nested under block in the models. To fit the respective models with a correlation structure, we used the function glmmPQL for proportion of aphid-infested leaves and the function lme from the 2.3.2 | Impact of A. brongniardellus on conspecific package nlme (Pinheiro et al., 2017) for growth (Appendix S1; Table performance S2; models 7 and 8).

To examine whether oak infestation levels of A. brongniardellus, oak genotype and the sex of the individual pupa affected the pupal weight 3 | RESULTS of A. brongniardellus, we modelled the pupal weight of A. brongniardellus as a function of infestation levels of A. brongniardellus, oak genotype 3.1 | Impact of oak phenology and genotype on and sex. To account for variation among oaks, we included oak tree as incidence and abundance of A. brongniardellus a random effect nested under block (Appendix S1; Table S2; model 3). The probability of an oak tree being infested by A. brongniardellus de- creased with an advancement of the day of leaf maturation. Nearly all 2 2.3.3 | Impact of A. brongniardellus on the insect late-phenology oaks were infested by A. brongniardellus ( 2 = 28.81, herbivore community p < .01; Figure 1a), whereas the incidence of A. brongniardellus did 2 not differ among oak genotypes ( 6 = 2.98, p = .81; Figure 1a). The To test for an effect of the infestation levels of A. brongniardellus on the relationship between day of leaf maturation and proportion of in- incidence of other species and on species richness, we modelled spe- fested leaves was hump-shaped, with early- and late-phenology oaks cies-specific incidence (for species Profenusa pygmaea (Hymenoptera: characterized by a lower proportion of infested leaves than oaks of

Tenthredinidae), Macrodiplosis dryobia (Diptera: Cecidomyiidae), intermediate phenology (F2, 134 = 25.43, p < .01; Figure 1b), while the

Phyllonorycter spp. (Lepidoptera: Gracillariidae), Stigmella spp. genotypic effect (F6, 134 = 2.05, p = .06) varied depending on day of

(Lepidoptera: Nepticulidae), Tischeria ekebladella (Lepidoptera: leaf maturation (Genotype × Environment; F12, 134 = 1.61, p = .10; Tischeriidae), Neuroterus spp. (Hymenoptera: Cynipidae)) and species Figure 1b). richness as a function of infestation levels of A. brongniardellus and oak genotype. Since very rare species will contribute little information, we only included species that were present on >20 oaks in the 3.2 | Effects of A. brongniardellus on conspecific species-specific incidence analysis, a criterion fulfilled by 6 of 14 spe- performance cies. All leaf mining and gall-inducing species except A. brongniardellus were included in metrics of species richness (Appendix S1; Table S2; In both of our rearing methods, we found that the pupal weight models 4 and 5). of A. brongniardellus was lower on oak trees with high infesta- 2 tion by conspecifics (ex situ: 1 = 18.34, p < .01; Figure 2a; in situ: 2 1 = 6.90, p < .01; Figure 2b). We detected no difference in pupal 2 2.3.4 | Effects of A. brongniardellus on aphid weight among oak genotypes (ex situ: 6 = 6.73, p = .35; in situ: 2 2 population dynamics 6 = 6.78, p = .34) or between males and females (ex situ: 1 = 0, 2 p = .95; in situ: 1 = 1.06, p = .30). The same pattern was observed As 47 of the 158 oaks remained free of aphids during the entire growing after controlling for any variation shared with oak phenology season, we tested whether oak-level presence of aphids, summarized (Appendix S2; Table S3). 834 | Journal of Animal Ecology EKHOLM et al.

FIGURE 1 (a) The probability of infestation (>5 leaf mines on an oak) and (b) the proportion of leaves infested by the leaf miner Acrocercops brongniardellus (photo) on oaks of different genotypes and different spring phenology (defined by leaf maturation, i.e. the day when ‘the first leaves are spread out over 90% of the crown’). Leaf maturation was surveyed at seven dates. The x-axis represents the month and the day of the month (dd/mm) and each of the seven oak genotypes is represented by a separate colour. Shown are raw data (circles, jittered vertically) and model predictions (filled dots) with standard errors (error bars) from a generalized linear mixed model

FIGURE 2 Pupal weight of Acrocercops brongniardellus (a) reared from shoots ex situ and (b) from pupae collected in situ (see Section 2 for an explanation of rearing methods) on oaks of different infestation levels of A. brongniardellus. Shown are raw data (circles) and model predictions (line) with standard errors (shaded interval) from a linear mixed model

3.3 | Impact of A. brongniardellus on the insect and Neuroterus spp. were more common on oaks with late phenology herbivore community (Appendix S2; Table S4).

Species richness differed among oak genotypes and was, on all genotypes, negatively affected by infestation levels of A. brongniardellus 3.4 | Effects of A. brongniardellus on aphid (Figure 3a; Table 1). Infestation levels of A. brongniardellus had a sig- population dynamics nificant negative effect on the four leaf miners, but not on the two galling species (Figure 3b–f; Table 1). Only one species, Tischeria eke- The probability of an oak to become infested by aphids was nega- 2 bladella, was detectably affected by oak genotype (Figure 3b; Table 1). tively affected by infestation levels of A. brongniardellus ( 1 = 11.83, After controlling for any variation shared with oak phenology, we p < .01; Figure 4). While genotypes differed in the incidence of aphids 2 found the same pattern as above with the exception that M. dryobia ( 6 = 13.52, p = .04; Figure 4), the negative impact of infestation EKHOLM et al. Journal of Animal Ecolog y | 835

FIGURE 3 The impact of oak tree infestation levels by Acrocercops brongniardellus and oak genotype on (a) species richness and incidence of (b) Tischeria ekebladella, (c) Phyllonorycter spp., (d) Stigmella spp., (e) Profenusa pygmaea and (f) Neuroterus spp. Each of the seven oak genotypes is represented by a separate colour. Asterisks indicate a significant effect of infestation levels of A. brongniardellus, whereas a plus indicates a significant effect of oak genotype. Shown are raw data (circles) and model predictions (lines) from generalized linear mixed models summarized in Table 1 levels by A. brongniardellus on aphid incidence was consistent across p = .72). The same pattern was observed after controlling for any all genotypes (interaction A. brongniardellus infestation × Genotype: variation shared with oak phenology (Appendix S2; Table S5). 2 6 = 6.64, p = .35; Figure 4). The total number of surveyed leaves When focusing on the subset of oaks infested by aphids, we 2 did not have any detectable impact on aphid incidence ( 1 = 0.13, found an effect of infestation by A. brongniardellus on aphid growth 836 | Journal of Animal Ecology EKHOLM et al.

TABLE 1 The effect oak tree infestation levels by Acrocercops brongniardellus and oak genotype on species richness and the incidence of four leaf mining and two gall inducing species, modelled with generalized linear mixed models

A. brongniardellus abundance (A) Genotype (G) A × G

2 2 2 Response df χ /F p df χ /F p df χ /F p Species richness 1, 147 63.06 <.01 6, 147 3.76 <.01 6, 141 0.73 .62 Tischeria ekebladella 1 24.46 <.01 6 18.37 <.01 6 4.01 .68 Phyllonorycter spp. 1 18.39 <.01 6 2.93 .82 6 4.22 .65 Stigmella spp. 1 17.53 <.01 6 3.09 .80 6 5.36 .50 Profenusa pygmaea 1 22.44 <.01 6 10.44 .11 6 6.10 .41 Neuroterus spp. 1 3.27 .07 6 8.08 .23 6 4.47 .61 Macrodiplosis dryobia 1 0.82 .37 6 2.64 .85 6 2.59 .86

Note: For F-ratios, both the numerator and denominator degrees of freedom are presented. Significant p-values (p < .05) are highlighted in bold. Plots of relationships can be found in Figure 3.

and insect outbreaks in turn to modify the structure of herbivore communities (Roslin & Roland, 2005; Timms & Smith, 2011). In our study system consisting of Q. robur and a subset of its herbivore community, we found that spring phenology of the host plant had a large impact on infestation by the outbreak species A. brongniardellus. Higher densities of the outbreaking species affected both its own performance and the structure of the insect herbivore community, by altering species richness and the occurrence of several herbivores. Furthermore, the population dynamics of aphids, which produce high numbers of generations per year, were markedly modified by the density of the outbreaking species. These findings suggest a chain of knock-on effects from phenology through the outbreaking species to the insect herbivore community, the links of which we will inspect below.

FIGURE 4 The probability of finding the aphid Tuberculatus 4.1 | Impact of plant phenology on the presence and annulatus on oaks of different genotypes and with varying levels abundance of an outbreak herbivore of infestation by Acrocercops brongniardellus. Each of the seven oak genotypes is represented by a separate colour. Shown are raw Oaks with early phenology had a lower probability and lower lev- data (circles) and model predictions (lines) from a generalized linear mixed model els of infestation by A. brongniardellus. This study design does not directly expose the causal mechanism underlying this pattern, but (Table 2; Figure 5a) and on the proportion of aphid-infested leaves. we can still speculate. Leaves are often most nutritious, but also The effect on the proportion of aphid-infested leaves was most pro- hold high levels of secondary metabolites just after bud burst (Falk nounced at the end of the season (Table 2; Figure 5b). We found et al., 2018; Salminen et al., 2004). Specialized herbivores can no detectable effect of oak genotype on aphid growth or on the sometimes overcome these defenses (Roslin & Salminen, 2008), proportion of aphid-infested leaves (Table 2). After controlling for allowing them to utilize a highly nutritious resource. By contrast, shared variation with oak phenology, we found the same results herbivores feeding on older leaves can experience lower survival, for proportion of aphid-infested leaves, whereas for growth rate body mass and growth rate (Fuentealba, Pureswaran, Bauce, & the effect of A. brongniardellus infestation turned nonsignificant Despland, 2017; Tikkanen & Julkunen-Tiitto, 2003; Tikkanen & (Appendix S2; Table S6). Lyytikainen-Saarenmaa, 2002), possibly due to lower nitrogen content and/or a tougher leaf surface in older leaves (Falk et al., 2018; Feeny, 1970; Salminen et al., 2004). Hence, many herbivores 4 | DISCUSSION experience a short time window when the leaf quality is high. In the case of the specialist leaf miner A. brongniardellus, this time Variation in host plant phenology has been proposed to affect window could stretch from bud burst until the leaf has developed herbivore performance and the potential for insect outbreaks a tough leaf surface and/or had a reduction in nitrogen content. (Falk et al., 2018; Jepsen et al., 2011; Pureswaran et al., 2015), Therefore, if A. brongniardellus oviposition is delayed relative to EKHOLM et al. Journal of Animal Ecolog y | 837

oak bud burst, the resulting offspring will be subjected to old .99 .11 p leaves with low quality that most likely will influence offspring

p < .05). performance.

: population Infestation by A. brongniardellus was the highest on trees 0.13 1.76 F of intermediate phenology. A similar pattern was observed by Wesolowski and Rowinski (2008), who found lower defoliation and a lower amount of frass fall of folivorous caterpillars from Q. robur with late spring phenology. The authors explained this pattern by a 6, 315 6, 431 df G x D reduction in temporal synchrony between herbivores and oaks with late phenology. In our study, oaks of late phenology could likewise

.15 have escaped high levels of herbivory by flushing when part of the -values are highlighted in bold ( bold in highlighted are -values p <.01 female moth population had ceased ovipositing. Overall, our findings indicate that A. brongniardellus is a phenological specialist, which is highly dependent on matching the phenological stage of the host 2.11 17.85 F plant for successful colonization.

4.2 | Effects of an outbreaking herbivore on 1, 437 1, 315 df A x D conspecific performance

Higher densities of A. brongniardellus came with lower pupal weight— <.01 <.01 p a clear indication of intraspecific, density-dependent competition. Such intraspecific, density-dependent competition in leaf miners has earlier been demonstrated in several studies (Doak & Wagner, 18.84 128.05 F 2015; Faeth, 1992; MacQuarrie, Spence, & Langor, 2010; Stiling, Brodbeck, & Strong, 1984; Tack et al., 2009). In our study, the low pupal weight of A. brongniardellus on heavily infested oaks could be due to (a) direct interference competition resulting from a higher 1, 437 df 1, 322 Date (D) density of larvae per leaf (Doak & Wagner, 2015), (b) exploitative competition over nutrients on heavily infested oaks, or (c) a stronger , oak genotype and date on two metrics of the performance of the annulatus aphid Tuberculatus herbivory-induced response by heavily infested oaks (Underwood, .27 p .63 2000, 2010). We are not able to separate between these three mechanisms, nor are they mutually exclusive. What we can infer is that competition overall results in substantial effects on the size of 1.29 F 0.73 the next generation, with likely effects on the insect's fitness as well as future outbreak dynamics. 6, 99 6, 99 df Genotype (G)

4.3 | The impact of an outbreaking herbivore

.02 on the insect herbivore community and aphid p <.01 population dynamics (A) Interspecific competition is an important process in shaping the 5.60 111). For F -ratios,n = 111). degrees of freedom for the numerator and denominator are shown, separated by a comma. Significant p 14.78 F structure of the insect community (Kaplan & Denno, 2007). In our study, we observed lower species-specific incidence and lower species richness on oaks heavily infested by A. brongniardellus. The short 1, 99 1, 99 df A. brongniardellus overlap in the activity period between A. brongniardellus and the ma- jority of the species included in this study (Coulianos & Holmåsen,

- 1991; Pschorn-Walcher, 1982; Svensson, 1993) suggests that direct interactions between A. brongniardellus and later colonizing herbivores are of secondary importance. Therefore, the negative effect of A. brongniardellus on species’ incidence likely consists of indirect fested leaves fested Proportion in Growth rate Response Note: As our modelling framework, we used linear- and generalized linear mixed models with a temporal correlation structure.observed In the models, during the we only survey used period data from ( oaks on which aphids had been Plots of relationships can be found in Figure 5. TABLE 2 TheTABLE effect of oak tree infestation levels brongniardellus by Acrocercops growth rate and proportion of aphid-infested leaves interactions between temporally segregated species. 838 | Journal of Animal Ecology EKHOLM et al.

FIGURE 5 The effect of oak tree infestation levels by Acrocercops brongniardellus and date on (a) the growth rate of the aphid Tuberculatus annulatus and (b) proportion of aphid- infested leaves, on oak trees of which aphids have been observed during the vegetation season. Date are represented by a separate colour. Shown are model predictions (lines) from a (a) linear and (b) generalized linear mixed model summarized in Table 2

An advantage of being an early colonizer is the possibility to uti- pattern of higher incidence of Neuroterus spp. on oaks with a late lize a resource before one's competitors reach them (McClure, 1980). phenology has also been detected in a study on Neuroterus quercus- If the early colonizer reaches outbreak densities, then there will be baccarum (Sinclair et al., 2015). a low amount of resources left for later colonizers to consume. In Late-colonizing herbivores could also affect each other, beyond our study, A. brongniardellus at outbreak levels consumed leaves in any effect of previous densities of A. brongniardellus. Yet, such ef- a way similar to other outbreaking herbivores – by consuming a ma- fects are likely to be weak. In a previous study of the same system, jority of the nutritious tissue and leaving little resources for later we showed that inter- and intraspecific competition among leaf min- colonizers (Appendix S3; Figure S1). In addition, trees with heavier ers (Phyllonorycter spp., Tischeria ekebladella) does occur, but that the infestation by A. brongniardellus showed higher levels of abscised effects are negligible at the densities typical in nature (Tack et al., leaves during the late summer (Appendix S4; Figure S2), which may 2009). Such effects will clearly be secondary as compared to the fact have shifted the oviposition preference of females from late-season that oaks with high infestation levels of A. brongniardellus simply had herbivore species towards non-infested oaks (Gripenberg, Mayhew, little leaf mass left for later species to colonize. Parnell, & Roslin, 2010). The negative effect of A. brongniardellus in- Taken together, high infestation levels of A. brongniardellus had a festation on aphid population dynamics is consistent with an earlier negative effect on four out of six species, most likely due to exploit- study conducted by Johnson, Mayhew, Douglas, and Hartley (2002). ative competition but possibly also a result of induced responses by They demonstrated that damage on the leaf midrib caused by the the plant. However, different feeding guilds appeared to respond leaf miner Eriocrania spp. reduced the survival of the phloem feeding differently to infestation by A. brongniardellus, suggesting that out- aphid Euceraphis betulae. That study found no effect of the resulting, breaking densities of A. brongniardellus play an important role in de- elevated levels of phenolic compounds on aphids. Instead, reduced termining the guild-level structure of the local insect community. survival among aphids was attributed to a disruption of phloem flow caused by the leaf miner. The same mechanisms may then be at play also in our study system—mined leaves will probably have a higher 4.4 | Impact of host plant genotype on the insect proportion of damaged leaf veins, resulting in a food source of low herbivore community and aphid population dynamics quality and translating to higher mortality among aphids. However, the pattern observed could also be a result of a lower amount of leaf The lack of an effect of host plant genotype on our specialized her- mass on highly infested oaks and thereby less available resources to bivore insect community could be a results of specialized herbivores support an abundant aphid population, or induced response by the being less influenced by plant genotype than generalists (Barker, plant. Holeski, & Lindroth, 2019). Previous studies have revealed host Two out of six of our target species (Neuroterus spp. and M. dryo- genotype as a weak predictor of species richness in our study sys- bia) were not negatively affected by infestation of A. brongniardel- tem (Pohjanmies et al., 2015; Tack, Ovaskainen, Pulkkinen, & Roslin, lus. The incidence of these gall inducing species were unaffected by 2010), although the genetic diversity of host plants has been shown infestation levels of A. brongniardellus and higher on oaks with late to affect herbivore abundances (Pohjanmies et al., 2015). Such geno- bud burst (Appendix S2; Table S4). The lack of an effect of infesta- typic effects may materialize both through effects on plant-specific tion levels of A. brongniardellus on Neuroterus spp. and M. dryobia phenology, and through effects on other traits such as plant size or might reflect the capability of plant galls to act as sinks for resources foliage density (Barbour et al., 2015). diverted from other parts of the plant. For instance, previous stud- Our study design allowed us to test whether plant individuals of ies have demonstrated that galls can stimulate photosynthesis (Fay, the same genotype but exposed to different spring climates differed Hartnett, & Knapp, 1993) and possibly also allocate carbon from in their susceptibility to A. brongniardellus (i.e. a genotype x envi- neighboring leaves (Bagatto, Paquette, & Shorthouse, 1996). The ronment interaction). Yet, we detected no statistically significant EKHOLM et al. Journal of Animal Ecolog y | 839 interaction, neither did we detect any main effect of genotype. The AUTHORS’ CONTRIBUTIONS only effect detected was that of plant phenology, suggesting that All authors conceived the ideas and designed methodology; A.E. col- differences between early and late trees will persist into the fu- lected and analysed the data; A.E. led the writing of the manuscript. ture. In a recent meta-analysis, Barker et al. (2019) found that most All authors contributed critically to the drafts and gave final approval phenotypic variation in several plant traits of the Salicaceae family for publication. was explained by genotype and environment individually, whereas little variation was explained by the G x E interaction. In fact, they DATA AVAILABILITY STATEMENT found that plant genotype had a big impact on insect herbivore All data have been uploaded to the Dryad Digital Repository: https:// performance. doi.org/10.5061/dryad.83bk3j9mr (Ekholm, Tack, Pulkkinen, & In terms of the herbivore insect community, we found that host Roslin, 2019). genotype had an effect on species richness and on the incidence of T. ekebladella. Similarly, we also detected an effect of host gen- ORCID otype on aphid incidence, where one particular genotype (“G4” in Adam Ekholm https://orcid.org/0000-0001-6199-0018 Figure 4) tended to show a higher probability of infestation than did other genotypes. This pattern matches the results of a previous REFERENCES study using the same oaks, where the very same genotype (“G4” Abbott, K. C., & Dwyer, G. (2007). Food limitation and insect outbreaks: in Figure 4) tended to show a higher abundance of aphids in an Complex dynamics in plant-herbivore models. Journal of Animal Ecology, environment with elevated temperatures (Faticov, Ekholm, Roslin, 76, 1004–1014. https://doi.org/10.1111/j.1365-2656.2007.01263.x & Tack, 2019). In terms of overall effects, indirect effects of oak Bagatto, G., Paquette, L. C., & Shorthouse, J. D. (1996). Influence of galls genotype (mediated via phenology) were more pronounced than of Phanacis taraxaci on carbon partitioning within common dande- lion, Taraxacum officinale. Entomologia Experimentalis Et Applicata, 79, direct effects of genotype on the insect herbivore community, im- 111–117. https://doi.org/10.1111/j.1570-7458.1996.tb008 15.x plying that direct effects of genotype were of minor importance. Barbour, M. A., Fortuna, M. A., Bascompte, J., Nicholson, J. R., Julkunen- Overall, our study suggests genotype-associated plant traits other Tiitto, R., Jules, E. S., & Crutsinger, G. M. (2016). Genetic specificity than phenology are of relatively minor importance in explaining in- of a plant-insect food web: Implications for linking genetic varia- festation levels of A. brongniardellus and the structure of the insect tion to network complexity. Proceedings of the National Academy of Sciences of the United States of America, 113, 2128–2133. https://doi. herbivore community. org/10.1073/pnas.1513633113 Barbour, M. A., Rodriguez-Cabal, M. A., Wu, E. T., Julkunen-Tiitto, R., Ritland, C. E., Miscampbell, A. E., … Crutsinger, G. M. (2015). 5 | CONCLUSIONS Multiple plant traits shape the genetic basis of herbivore community assembly. Functional Ecology, 29, 995–1006. https://doi. org/10.1111/1365-2435.12409 The current findings from our grafted oaks subjected to experimen- Barker, H. L., Holeski, L. M., & Lindroth, R. L. (2019). Independent tal manipulation of phenology suggest a chain of knock-on effects and interactive effects of plant genotype and environment on from phenology via the outbreaking species to the insect herbivore plant traits and insect herbivore performance: A meta-analysis with Salicaceae. Functional Ecology, 33, 422–435. https://doi. community. The outbreak species was found to compete both with org/10.1111/1365-2435.13249 itself and with other herbivore species feeding on the same plant. Bates, D., Maechler, M., Bolker, B., & Walker, S. (2015). Fitting linear Given the finding of how phenological synchrony between the out- mixed-effects models using lme4. Journal of Statistical Software, break species and its host plant influences infestation levels, a shift 67(1), 1–48. https://doi.org/10.18637/ jss.v067.i01 Bengtsson, B. Å., & Johansson, R. (2011). Nationalnyckeln till Sveriges in synchrony may then change outbreak dynamics and cause cascad- flora och fauna. Fjärilar: Bronsmalar-rullvingemalar. Lepidoptera: ing effects on the insect community. Roeslerstammiidae-Lyonetiidae. Uppsala, Sweden: Ardatabanken, SLU. What is further worth noting is that we, in our case, manipulated Björkman, C., Bylund, H., Klapwijk, M. J., Kollberg, I., & Schroeder, M. oak phenology separate from the insect community, thereby exper- (2011). Insect pests in future forests: More severe problems? Forests, 2, 474–485. https://doi.org/10.3390/f2020474 imentally manipulating the levels of synchrony between the host Blanchet, F. G., Roslin, T., Kimura, M. T., Huotari, T., Kaartinen, R., plant and its herbivorous insects. In future studies, simultaneous Gripenberg, S., & Tack, A. J. M. (2018). Related herbivore species manipulation of both trophic levels will bring further insight – but show similar temporal dynamics. Journal of Animal Ecology, 87, 801– logistically, this is no trivial exercise. 812. https://doi.org/10.1111/1365-2656.12807 Burkle, L. A., Marlin, J. C., & Knight, T. M. (2013). Plant-pollinator interactions over 120 years: Loss of species, co-occurrence, and function. ACKNOWLEDGEMENTS Science, 339, 1611–1615. https://doi.org/10.1126/scien ce.1232728 We thank Rasmus Jansson for helping us with the fieldwork Chen, L., Huang, J. G., Dawson, A., Zhai, L. H., Stadt, K. J., Comeau, P. G., and Holmen Skog for allowing us to conduct the experiment on & Whitehouse, C. (2018). Contributions of insects and droughts to their property. Financial support from the Swedish University of growth decline of trembling aspen mixed boreal forest of western Canada. Global Change Biology, 24, 655–667. https://doi.org/10.1111/ Agricultural Sciences, Faculty of Natural Resources and Agricultural gcb.13855 Sciences (NJ) and Department of Ecology, is gratefully acknowl- Cooke, B. J., & Roland, J. (2007). Trembling aspen responses to drought edged. We have no conflict of interest to declare. and defoliation by forest tent caterpillar and reconstruction of recent 840 | Journal of Animal Ecology EKHOLM et al.

outbreaks in Ontario. Canadian Journal of Forest Research, 37, 1586– Heie, O. E. (1980). The Aphidoidea (Hemiptera) of Fennoscandia and 1598. https://doi.org/10.1139/X07-015 Denmark. I. Fauna Entomologica Scandinavica. Klampenborg, Coulianos, C. C., & Holmåsen, I. (1991). GALLER En fälthandbok om gallbild- Denmark: Scandinavian Science Press Ltd. ningar på vilda och odlade växter. Stockholm, Sweden: Interpublishing Heie, O. E. (1982). The Aphidoidea (Hemiptera) of Fennoscandia and AB. Denmark. II. Fauna Entomologica Scandinavica. Klampenborg, Crawley, M. J., & Akhteruzzaman, M. (1988). Individual variation in the Denmark: Scandinavian Science Press Ltd. phenology of oak trees and its consequences for herbivorous insects. Honěk, A., & Honek, A. (1993). Intraspecific variation in body size and fe- Functional Ecology, 2, 409–415. https://doi.org/10.2307/2389414 cundity in insects – A general relationship. Oikos, 66, 483–492. https:// Doak, P., & Wagner, D. (2015). The role of interference competition in doi.org/10.2307/3544943 a sustained population outbreak of the aspen leaf miner in Alaska. Jepsen, J. U., Kapari, L., Hagen, S. B., Schott, T., Vindstad, O. P. L., Basic and Applied Ecology, 16, 434–442. https://doi.org/10.1016/j. Nilssen, A. C., & Ims, R. A. (2011). Rapid northwards expansion of baae.2015.04.001 a forest insect pest attributed to spring phenology matching with Dwyer, G., Dushoff, J., & Yee, S. H. (2004). The combined effects of sub-Arctic birch. Global Change Biology, 17, 2071–2083. https://doi. pathogens and predators on insect outbreaks. Nature, 430, 341–345. org/10.1111/j.1365-2486.2010.02370.x https://doi.org/10.1038/natur e02569 Johnson, S. N., Mayhew, P. J., Douglas, A. E., & Hartley, S. E. (2002). Ekholm, A., Roslin, T., Pulkkinen, P., & Tack, A. J. M. (2017). Dispersal, Insects as leaf engineers: Can leaf-miners alter leaf structure host genotype and environment shape the spatial dynamics of a par- for birch aphids? Functional Ecology, 16, 575–584. https://doi. asite in the wild. Ecology, 98, 2574–2584. https://doi.org/10.1002/ org/10.1046/j.1365-2435.2002.00654.x ecy.1949 Kaplan, I., & Denno, R. F. (2007). Interspecific interactions in phytopha- Ekholm, A., Tack, A. J. M., Bolmgren, K., & Roslin, T. (2019). The forgotten gous insects revisited: A quantitative assessment of competition season: The impact of autumn phenology on a specialist insect herbi- theory. Ecology Letters, 10, 977–994. https://doi.org/10.1111/ vore community on oak. Ecological Entomology, 44, 425–435. https:// j.1461-0248.2007.01093.x doi.org/10.1111/een.12719 Kharouba, H. M., Ehrlen, J., Gelman, A., Bolmgren, K., Allen, J. M., Travers, Ekholm, A., Tack, A. J. M., Pulkkinen, P., & Roslin, T. (2019). Data from: S. E., & Wolkovich, E. M. (2018). Global shifts in the phenological syn- Host plant phenology, insect outbreaks and herbivore communities – chrony of species interactions over recent decades. Proceedings of The importance of timing. Dryad Digital Repository, https://doi.org/ the National Academy of Sciences of the United States of America, 115, 10.5061/dryad.83bk3j9mr 5211–5216. https://doi.org/10.1073/pnas.17145 11115 Faeth, S. H. (1992). Interspecific and intraspecific interactions via Lenth, R. (2018). emmeans: Estimated marginal means, aka least-squares plant-responses to folivory – An experimental field-test. Ecology, 73, means. R package version 1.2.2. Retrieved from https://CRAN.R- 1802–1813. https://doi.org/10.2307/1940031 project.org/package=emmeans Falk, M. A., Lindroth, R. L., Keefover-Ring, K., & Raffa, K. F. (2018). MacQuarrie, C. J. K., Spence, J. R., & Langor, D. W. (2010). Using classi- Genetic variation in aspen phytochemical patterns structures win- fication tree analysis to reveal causes of mortality in an insect pop- dows of opportunity for gypsy moth larvae. Oecologia, 187, 471–482. ulation. Agricultural and Forest Entomology, 12, 143–149. https://doi. https://doi.org/10.1007/s00442-018-4160-0 org/10.1111/j.1461-9563.2009.00460.x Faticov, M., Ekholm, A., Roslin, T., & Tack, A. J. M. (2019). Climate and Marini, L., Okland, B., Jonsson, A. M., Bentz, B., Carroll, A., Forster, B., host genotype jointly shape tree phenology, disease levels and insect … Schroeder, M. (2017). Climate drivers of bark beetle outbreak dy- attacks. Oikos. https://doi.org/10.1111/oik.06707 namics in Norway spruce forests. Ecography, 40, 1426–1435. https:// Fay, P. A., Hartnett, D. C., & Knapp, A. K. (1993). Increased photosynthe- doi.org/10.1111/ecog.02769 sis and water potentials in silphium-integrifolium galled by cynipid McClure, M. S. (1980). Competition between exotic species – Scale insects on wasps. Oecologia, 93, 114–120. https://doi.org/10.1007/BF003 hemlock. Ecology, 61, 1391–1401. https://doi.org/10.2307/1939048 21200 Meier, U. (2001). Growth stages of mono-and dicotyledonous plants, BBCH Feeny, P. (1970). Seasonal changes in oak leaf tannins and nutrients as Monograph. Berlin, Germany: Federal Biological Research Centre for a cause of spring feeding by winter moth caterpillars. Ecology, 51, Agriculture and Forestry. 565–581. https://doi.org/10.2307/1934037 Parmesan, C. (2007). Influences of species, latitudes and methodol- Fox, J., & Weisberg, S. (2011). An {R} companion to applied regression (2nd ogies on estimates of phenological response to global warming. ed.). Thousand Oaks, CA: Sage. Retrieved from http://socserv.socsci. Global Change Biology, 13, 1860–1872. https://doi.org/10.1111/ mcmaster.ca/jfox/Books/Companion j.1365-2486.2007.01404.x Fuentealba, A., Pureswaran, D., Bauce, E., & Despland, E. (2017). How Pinheiro, J., Bates, D., DebRoy, S., Sarkar, D., & R Core Team. (2017). nlme: does synchrony with host plant affect the performance of an out- Linear and nonlinear mixed effects models. R package version 3.1-131. breaking insect defoliator? Oecologia, 184, 847–857. https://doi. Pohjanmies, T., Tack, A. J. M., Pulkkinen, P., Elshibli, S., Vakkari, P., & Roslin, org/10.1007/s00442-017-3914-4 T. (2015). Genetic diversity and connectivity shape herbivore load Gripenberg, S., Mayhew, P. J., Parnell, M., & Roslin, T. (2010). A meta- within an oak population at its range limit. Ecosphere, 6(6). https:// analysis of preference-performance relationships in phytopha doi.org/10.1890/ES14-00549.1 gous insects. Ecology Letters, 13, 383–393. https://doi.org/ Pschorn-Walcher, H. (1982). Unterordnung Symphyta, Pflanzenwespen. 10.1111/j.1461-0248.2009.01433.x In W. Schwenke (Ed.), Die Forstschädlinge Europass (pp. 4–234). Hails, R. S., & Crawley, M. J. (1991). The population-dynamics of an alien Hamburg/Berlin: Paul Parey. insect – Andricus quercuscalicis (Hymenoptera, Cynipidae). Journal of Pureswaran, D. S., De Grandpre, L., Pare, D., Taylor, A., Barrette, M., Animal Ecology, 60, 545–562. https://doi.org/10.2307/5297 Morin, H., … Kneeshaw, D. D. (2015). Climate-induced changes in Hartig, F. (2018). DHARMa: Residual diagnostics for hierarchical (multi- host tree-insect phenology may drive ecological state-shift in boreal level/mixed) regression models. R package version 0.2.0. Retrieved forests. Ecology, 96, 1480–1491. https://doi.org/10.1890/13-2366.1 from https://CRAN.R-project.org/package=DHARMa R Core Team. (2017). R: A language and environment for statistical comput- Haukioja, E., & Neuvonen, S. (1985). The relationship between size ing. Vienna, Austria: R Foundation for Statistical Computing. and reproductive potential in male and female Epirrita autumnata Roslin, T., & Roland, J. (2005). Competitive effects of the forest tent cat- (lep, geometridae). Ecological Entomology, 10, 267–270. https://doi. erpillar on the gallers and leaf-miners of trembling aspen. Ecoscience, org/10.1111/j.1365-2311.1985.tb00723.x 12, 172–182. https://doi.org/10.2980/i1195-6860-12-2-172.1 EKHOLM et al. Journal of Animal Ecolog y | 841

Roslin, T., & Salminen, J. P. (2008). Specialization pays off: Relation to body size and consequences to population dynamics. Contrasting effects of two types of tannins on oak specialist Oikos, 77, 407–416. https://doi.org/10.2307/3545931 and generalist moth species. Oikos, 117, 1560–1568. https://doi. Thackeray, S. J., Henrys, P. A., Hemming, D., Bell, J. R., Botham, M. S., org/10.1111/j.0030-1299.2008.16725.x Burthe, S., … Wanless, S. (2016). Phenological sensitivity to climate Salminen, J. P., Roslin, T., Karonen, M., Sinkkonen, J., Pihlaja, K., & across taxa and trophic levels. Nature, 535, 241–245. https://doi. Pulkkinen, P. (2004). Seasonal variation in the content of hydro- org/10.1038/nature18608 lyzable tannins, flavonoid glycosides, and proanthocyanidins in Tikkanen, O. P., & Julkunen-Tiitto, R. (2003). Phenological variation as oak leaves. Journal of Chemical Ecology, 30, 1693–1711. https://doi. protection against defoliating insects: The case of Quercus robur org/10.1023/B:JOEC.0000042396.40756.b7 and Operophtera brumata. Oecologia, 136, 244–251. https://doi. Schwartzberg, E. G., Jamieson, M. A., Raffa, K. F., Reich, P. B., org/10.1007/s00442-003-1267-7 Montgomery, R. A., & Lindroth, R. L. (2014). Simulated climate Tikkanen, O. P., & Lyytikainen-Saarenmaa, P. (2002). Adaptation of a gen- warming alters phenological synchrony between an outbreak insect eralist moth, Operophtera brumata, to variable budburst phenology herbivore and host trees. Oecologia, 175, 1041–1049. https://doi. of host plants. Entomologia Experimentalis Et Applicata, 103, 123–133. org/10.1007/s00442-014-2960-4 https://doi.org/10.1046/j.1570-7458.2002.00966.x Sinclair, F. H., Stone, G. N., Nicholls, J. A., Cavers, S., Gibbs, M., Butterill, Timms, L. L., & Smith, S. M. (2011). Effects of gypsy moth establish- P., … Schonrogge, K. (2015). Impacts of local adaptation of forest ment and dominance in native caterpillar communities of northern trees on associations with herbivorous insects: Implications for oak forests. The Canadian Entomologist, 143, 479–503. https://doi. adaptive forest management. Evolutionary Applications, 8, 972–987. org/10.4039/n11-025 https://doi.org/10.1111/eva.12329 Underwood, N. (2000). Density dependence in induced plant resistance Southwood, T. R. E. (1961). The number of species of insect associated to herbivore damage: Threshold, strength and genetic variation. Oikos, with various trees. Journal of Animal Ecology, 30, 1–8. https://doi. 89, 295–300. https://doi.org/10.1034/j.1600-0706.2000.890210.x org/10.2307/2109 Underwood, N. (2010). Density dependence in insect per- Southwood, T. R. E., Wint, G. R. W., Kennedy, C. E. J., & Greenwood, formance within individual plants: Induced resistance to S. R. (2004). Seasonality, abundance, species richness and spec- Spodoptera exigua in tomato. Oikos, 119, 1993–1999. https://doi. ificity of the phytophagous guild of insects on oak (Quercus) can- org/10.1111/j.1600-0706.2010.18578.x opies. European Journal of Entomology, 101, 43–50. https://doi. Van Zandt, P. A., & Agrawal, A. A. (2004). Community-wide impacts of org/10.14411/eje.2004.011 herbivore-induced plant responses in milkweed (Asclepias syriaca). Stalhandske, S., Olofsson, M., Gotthard, K., Ehrlen, J., Wiklund, C., & Ecology, 85, 2616–2629. https://doi.org/10.1890/03-0622 Leimar, O. (2016). Phenological matching rather than genetic vari- Venables, W. N., & Ripley, B. D. (2002). Modern applied statistics with S ation in host preference underlies geographical variation in host (4th ed.). New York, NY: Springer. plants used by orange tip butterflies. Biological Journal of the Linnean Wesolowski, T., & Rowinski, P. (2008). Late leaf development in pedun- Society, 119, 1060–1067. https://doi.org/10.1111/bij.12838 culate oak (Quercus robur): An antiherbivore defence? Scandinavian Stam, J. M., Dicke, M., & Poelman, E. H. (2018). Order of herbivore ar- Journal of Forest Research, 23, 386–394. rival on wild cabbage populations influences subsequent arthro- Whitham, T. G., Young, W. P., Martinsen, G. D., Gehring, C. A., Schweitzer, pod community development. Oikos, 127, 1482–1493. https://doi. J. A., Shuster, S. M., … Kuske, C. R. (2003). Community and ecosys- org/10.1111/oik.05265 tem genetics: A consequence of the extended phenotype. Ecology, Stenberg, L., & Mossberg, B. (2003). Den nya nordiska floran. Stockholm: 84, 559–573. https://doi.org/10.1890/0012-9658(2003)084[0559: Wahlström & Widstrand. CAEGAC]2.0.CO;2 Stiling, P. D., Brodbeck, B. V., & Strong, D. R. (1984). Intraspecific competition in Hydrellia valida (Diptera, Ephydridae), a leaf miner of spartina-alterniflora. Ecology, 65, 660–662. https://doi. SUPPORTING INFORMATION org/10.2307/1941431 Additional supporting information may be found online in the Svensson, I. (1993). Lepidoptera calender. Kristianstad (In Swedish). Supporting Information section. Self-published. Tack, A. J. M., Ovaskainen, O., Harrison, P. J., & Roslin, T. (2009). Competition as a structuring force in leaf miner communities. Oikos, 118, 809–818. https://doi.org/10.1111/j.1600-0706.2008.17397.x How to cite this article: Ekholm A, Tack AJM, Pulkkinen P, Tack, A. J. M., Ovaskainen, O., Pulkkinen, P., & Roslin, T. (2010). Roslin T. Host plant phenology, insect outbreaks and herbivore Spatial location dominates over host plant genotype in structur- communities – The importance of timing. J Anim Ecol. ing an herbivore community. Ecology, 91, 2660–2672. https://doi. 2020;89:829–841. https://doi.org/10.1111/1365-2656.13151 org/10.1890/09-1027.1 Tammaru, T., Kaitaniemi, P., Ruohomäki, K., & Ruohomaki, K. (1996). Realized fecundity in Epirrita autumnata (Lepidoptera: Geometridae):