Climate-driven change in plant–insect interactions along elevation gradients



  1. Global warming is predicted to dramatically alter communities' composition through differential colonization abilities, such as between sessile plants and their mobile herbivores. Novel interactions between previously non-overlapping species may, however, also be mediated by altered plants' responses to herbivore attack.

  2. Syndromes of plant defences and tolerance are driven by inherited functional traits, biotic and abiotic conditions, and the geographical and historical contingencies affecting the community. Therefore, understanding climate change-driven herbivore responses and evolution towards a particular plant defence syndrome is key to forecasting species interactions in the near future.

  3. In this paper, we first document variations in herbivory, and plant defences along altitudinal gradients that act as ‘natural experiments’. We then use an empirical model to predict how specialist herbivore abundance may shift with respect to elevation in the near future.

  4. Our field surveys and field experiment showed a decrease in herbivory with elevation. However, contrary to expectations, our meta-regression analyses showed that plant defences, particularly leaf toughness and flavonoid compounds, tend to be higher at high elevations, while secondary metabolites showed no clear trend with elevation.

  5. Based on those results, we discuss how plant communities and species-specific plant defence syndromes will change in response to the climate-driven herbivore colonization of higher altitudes. Particularly, plant from high elevation, due to high protection against abiotic stress may be already ecologically fitted to resist the sudden increase in herbivory pressure that they will likely experience during global change.


Assessing the consequences of climate change on ecosystems dynamics and species interactions is becoming increasingly urgent (Schröter et al. 2005; Tylianakis et al. 2008; Wu et al. 2011; de Sassi, Lewis & Tylianakis 2012). Among the well-documented responses to global warming are organisms' range shifts to higher latitudes and altitudes (Grabherr, Gottfried & Pauli 1994; Parmesan 1996; Stange & Ayres 2001; Hoegh-Guldberg et al. 2008; Burrows et al. 2011; Pateman et al. 2012; Pauli et al. 2012). Such displacements of populations have raised several concerns, including loss of biodiversity (Parmesan 1996, 2006), population mismatches between predators and prey (Durant et al. 2007), parasitoids and their host insect herbivores (Harrington, Woiwod & Sparks 1999) or herbivorous insects and their host plants (Singer & Parmesan 1993), and increased pestilence due to lessened population-regulating factors (Logan, Regniere & Powell 2003; Bentz et al. 2010). Indeed, more important than the effect on any single species are the effects of global warming on complex community-level interactions (van der Putten et al. 2004; Tylianakis et al. 2008; Jamieson et al. 2012).

Insects are among the groups of organisms that exhibit the strongest responses to climate change because they are ectothermic, so thermal changes have strong direct influences on their development, reproduction and survival (Bale et al. 2002). Additionally, shorter generation times and high reproductive rates, coupled with a better ability to disperse, lead insects to respond more quickly to climate change through range shift than longer-lived and less mobile organisms, such as plants (Menéndez 2007). Different migration rates between plant, insect herbivores and their predators may lead to reshuffled communities. Insect herbivores and plants together comprise more than half of the terrestrial macro-biodiversity (Strong, Lawton & Southwood 1984) and display trait-mediated complex networks of interactions that are expected to be affected by climate change (Harrington, Woiwod & Sparks 1999; Visser & Both 2005; Lurgi, López & Montoya 2012). For instance, plant resistance strategies against insect herbivores, including direct physical and chemical defences, the recruitment of natural enemies of the herbivore (i.e. indirect defences), and tolerance (the ability to compensate for tissue loss after herbivory) (Schoonhoven, van Loon & Dicke 2005), have been shown to vary along ecological gradients (Moles et al. 2011; Rasmann & Agrawal 2011; Pellissier et al. 2012) and to be affected by climate change (DeLucia et al. 2012). Therefore, predicting the community-level dynamics of plants and herbivores requires adopting an integrative approach that includes measuring community composition and herbivory dynamics as well as the bottom-up effects of plants on herbivores (i.e. plant defences) and top-down regulation of herbivores by predators and parasitoids.

Elevation gradients are optimal ecological surrogates for inferring global change-driven dynamics (Pickett 1989; Körner 2007; Garibaldi, Kitzberger & Chaneton 2011; Beier et al. 2012; Rasmann, Alvarez & Pellissier 2013) acting as ‘natural experiments’ by providing variation in abiotic factors under which biotic interactions can be evaluated (Preszler & Boecklen 1996; Darrow & Bowers 1997; Salmore & Hunter 2001; Yarnes & Boecklen 2005; Zehnder et al. 2009). Compared to much larger-scale latitudinal gradients, elevation gradients minimize the confounding effects of historical and biogeographical differences in, for instance, plants and herbivore species pools (Hodkinson 2005).

In this contribution, we first aim to describe variation in herbivory levels and plant resistance strategies along elevation gradients. We purposely excluded an in-depth analysis of parasitism pressure along elevation gradients, subject of a quantitative review discussed below (Péré, Jactel & Kenis 2013). Second, we forecast levels of plant–herbivore interactions along elevation gradients in warmer climates and discuss the results in light of current elevation patterns in plant defence traits. Our working hypothesis is that if we can identify consistent trends in biotic changes along elevation gradients, we should be able to extrapolate our results to future global change dynamics such as increases in temperature.

Herbivory along elevation gradients

Because insects are ectothermic, it is generally expected that insect herbivory rates decrease with increasing elevation. Most studies on the variation of herbivory rates with elevation (or even with latitude) have been framed by (i) focusing on the ambient herbivory rate resulting from all insects present at a given altitude on a single plant species, or (ii) analysing the ambient herbivory from all insects on all plants in a local community at a given altitude. These studies have generally shown that herbivory rates decrease with increasing elevation, but the pattern is quite variable, with some potentially increasing or nonlinear responses (Hodkinson 2005). Even within the same study site, opposing patterns (increasing or decreasing with latitude) have been reported, depending on the herbivory measurement method (Adams & Zhang 2009; Zhang, Adams & Zhao 2011). A recent study on herbivory of forest insects along 24 elevation gradients in Europe confirms this lack of consistent effect of elevation, reporting great variability among four sampled tree species (L. Marini et al. unpublished). However, overall, the responses of different insect feeding groups (e.g. leaf chewers, miners, sap feeders, gall makers) were found to be either neutral or positive (Marini et al. submitted). Some studies have reported that the pattern of decrease in herbivory rate with elevation is clearer for generalist herbivores than for specialists (Scheidel, Rohl & Bruelheide 2003; Hodkinson 2005). However, the relative importance of specialists and generalists may also vary with elevation, although the evidence is mixed [higher diversity and specialization with increasing altitude in the tropics (Rodríguez-Castañeda et al. 2010) but the opposite patterns for diversity and specialization in temperate zones (Pellissier et al. 2012)].

A third approach for measuring herbivore variation along elevation gradients is to analyse the average herbivory rate at the community level for all of the different species that are present in a community. This approach allows understanding how plant species succession, with contrasting ecological and antiherbivory strategies along an elevation gradient, may drive the observed herbivory rate, and reciprocally, how herbivory affects the plant community structure along the gradient. We here provide results for herbivory rates on natural and experimentally established community of plants along elevation gradients.


Briefly, leaf herbivory rates were analysed for six dominant tree species (i.e. Quercus petraea, Fagus sylvatica, Abies alba, Picea abies, Pinus uncinata and Pinus cembra) commonly found along elevation in the French Alps from 500 to 2000 m (see Methods S1 in Supporting information for a detailed description of the methods). Quercus petraea, F. sylvatica and A. alba are associated with warmer habitats, whereas P. abies, P. uncinata and P. cembra are associated with colder climate. Six plots (4 × 4 m) per altitude were evenly distributed across seven altitudes. First, randomly selected natural seedlings (<50 cm height) were scored for leaf damage by visually assessing the percentage of leaf area eaten for every leaf on every individual seedling. For practicality, we decided to focus only on leaf chewer damage, omitting piercing-sucking damage as well as pathogen attack. Second, at the same elevations, we established common gardens of the same six species of trees. After germination, we monitored leaf herbivory on sown seedlings as above. Although we acknowledge that herbivory rates are likely to vary with ontogeny (Boege & Marquis 2005), monitoring herbivory on seedlings is valuable because this lifestage is key for community assembly and therefore herbivore impact is predicted to be stronger.

Herbivory levels were quantile-regressed against degree-days sum above 5·5 °C (DDS) for each elevation (Pearson correlation between DDS and elevation, = −0·98 for natural sites and = −0·85 for experimental sites). Because our data were strongly skewed towards zero, we used quantile regressions, which do not require any assumption concerning the shape of the error distribution, to capture the maximum potential herbivory rate by the 95% quantile. For both the natural population observation and the experimental data, we pooled all tree species to analyse community-level responses, as well as we scored individual species responses.


In naturally established seedlings, we found a nonlinear response of herbivory to DDS, with a clear trend of lower herbivory at higher elevations (Fig. 1a, and see AIC table in Methods S1, Supporting information). However, individual tree species showed idiosyncratic responses to DDS (Fig. 1b). In contrast, experimentally sown seedlings exhibited an overall decrease in insect herbivory with increasing elevation (i.e. with decreasing DDS), both at the community level (Fig. 1c), and for all species analysed individually (Fig. 1d). The data for experimentally sown seedlings also showed that for similar elevations, species from warmer climates (Q. petraea, F. sylvatica and A. alba) experienced greater herbivory than species from colder climates (P. abies, P. uncinata and P. cembra) (Fig. 1d).

Figure 1.

Variation in leaf damage on tree seedlings along an elevation gradient in the French Alps for both naturally established seedlings and experimentally sown seedlings outside of their natural distribution. Panels (a) and (b) show results from the survey of natural populations, whereas panels (c) and (d) show results of the common garden experiment. Additionally, panels (a) and (c) the community response (i.e. overall average across six species), and panels (b) and (d) show individual tree species responses. The elevation gradient is represented by the degree-day sum (>5·5°C). A higher DDS represents a warmer climate at a lower elevation. Lines representing the 95% quantile of the leaf herbivory rate at the community level (all species together) and the species level, estimated using quantile regression, are presented. Grey points are individual herbivory rate measurements. Species and their abbreviations from low to high altitude are Quercus petraea (QUPE), Fagus sylvatica (FASY), Abies alba (ABAL), Picea abies (PIAB), Pinus uncinata (PIUN) and Pinus cembra (PICE).

Overall, these results suggest that herbivore pressure on plants decreases with increasing elevation, which is consistent with recent work (Pellissier et al. 2012). The lack of a clear pattern at the species-specific level for the naturally established seedlings, as opposed to the pattern observed for the experimental planted seedlings, may be explained by the very restricted climatic conditions under which natural seedlings are found. Hence, the pattern of variation in herbivory rate at the community level may be strongly affected by the succession of species along the elevation gradient.

Plant defences along elevation gradients

According to theory, high-elevation living plants, because they experience lower levels of herbivory, are expected to have lower levels of defences against herbivores, compared to their relatives at lower altitudes (Coley & Barone 1996). When measuring resistance against a generalist herbivore across 16 pairs of high- vs. low-elevation species, we recently found that caterpillars grew larger on high-elevation species overall (Pellissier et al. 2012). The exact reasons for this pattern, such as a decline in defences or the higher nutritional quality of the leaves of alpine plants, have not yet been determined. A number of plant quality traits, such as foliar nitrogen (Erelli, Ayres & Eaton 1998; Hengxiao et al. 1999; Richardson 2004), defensive chemistry (Erelli, Ayres & Eaton 1998; Hengxiao et al. 1999; Salmore & Hunter 2001; Alonso et al. 2005; Zehnder et al. 2009), structural compounds (Richardson 2004) and leaf morphology traits (Hengxiao et al. 1999), change with elevation, but not in a predictable manner. Very likely, both abiotic and biotic stresses contribute shaping secondary chemistry phenotypes. Leaves that have high concentrations fibre are stronger, more resistant to damage and potentially longer-lived than leaves with low fibre concentrations (Abrahamson et al. 2003; Richardson 2004). Additionally, phenolic compounds can simultaneously protect leaves from photodamage by acting as antioxidants and reduce insect performance (Close & McArthur 2002; Forkner, Marquis & Lill 2004).

Despite a relatively long history, a comprehensive analysis of plant defence traits along elevation gradients is still lacking. We thus resolved to explore the literature for putative defence traits of plants that were measured along elevation gradients and summarize this with a formal meta-regression analysis as in Castagneyrol & Jactel (2012).


We scanned the literature in the ISI Web of Science data base, using combinations of relevant terms (plant*, defen*, elevation*, altitude*). The survey was completed using the Google Scholar search engine and a reference list of individual papers. We only considered studies in which plant traits were measured in relation to resistance against herbivore attack. Because of statistical constraints [the variance of the effect size = 1/(n−3)], we removed from consideration studies in which trait variation was measured with less than four elevation points. We also included our own data from the Baccara project (, in which we measured leaf toughness in six woody plant species along elevation gradients of European mountain regions. We identified a total of 40 studies that met our conditions, providing a total of 99 case studies (Table S1, Supporting information). The relationships between plant defensive traits and elevation were measured by extracting correlation coefficient (r) values from the text and tables. If correlation coefficients were not directly available, the numerical values were extracted by digitising the figure and plotting simple linear regressions. A positive value implied that plant defensive trait values increased with increasing elevation (Table S2, Supporting information). We tested the effects of four covariates on the relationship between plant chemical defences and altitude. First, we divided plant species into herbaceous and woody plants (including shrubs). Second, we divided plant chemicals into flavonoid and nonflavonoid secondary metabolites (SM). This division was based on the assumption that flavonoid-based compounds can be involved in pigmentation, UV protection and freezing tolerance in plants (Close & McArthur 2002). Therefore, it is more likely to expect a positive correlation between flavonoids and elevation than between nonflavonoid SMs, which should only be involved in defence against herbivores and elevation. Third, we separated whether measures were taken on reproductive (flowers, seeds, fruits) or vegetative organs of the plants (leaves, branches, trunks). We carried out different meta-analyses, using a hierarchical approach to avoid the need to use individual data points more than once in any given analysis (Whittaker 2010). First, we tested the effect of the type of plant (herbaceous vs. woody plants) on a subset of data comprising flavonoids and nonflavonoids defences in reproductive and vegetative organs and found no significance differences. We therefore pooled data from both types of plants in further analyses. We then tested the effect of the type of chemical defence within each category of organs.


Overall, we detected a significant increase in plant physical defence traits and flavonoids with elevation (Fig. 2). This was driven by leaf toughness and flavonoids in the reproductive organs (see positive regression coefficients in Fig. 2). We found no effect of elevation on other types of defences (nonflavonoids in general and flavonoids in vegetative organs).

Figure 2.

Coefficients of correlation (shown with confidence intervals) between elevation and direct chemical and physical plant defences for total defence (black circle), and the vegetative organs (black squares) and the reproductive organs (open squares). Chemical defences were separated into flavonoid-based compounds and nonflavonoid-based compounds. Physical defences were only analysed in woody species. Asterisks indicate significant deviations from a null relationship.

Forecasting shift in herbivore abundance along elevation under warmer climate

As an illustration of how herbivore abundance may shift in response to climate change along the elevation gradient, we modelled the maximum abundance of butterflies with respect to elevation in response to temperature and projected the model under three different climate change scenarios.


Total butterfly abundance data were collected in 192 plots 50 × 50 m in size in a region of the Western Swiss Alps (700 km2 Pellissier, Alvarez & Guisan 2012). All butterfly species belonging to the Papilionoidea superfamily (sensu Heikkilä et al. 2012) were monitored. We counted the number of individual butterflies as a measure of the abundance of herbivores at each site. For each site, we then determined the annual mean temperature, calculated as described by Zimmermann & Kienast (1999). We calculated the maximum abundance of butterflies in sampled communities in 20 equal classes of increasing temperature. Maximum abundance permits calculating an extreme scenario of herbivore pressure for a given set of environmental conditions. We related the maximum abundance to the mean annual temperature of the corresponding class using a linear model. We projected this model considering temperature anomalies for three climate change scenarios, that is, A2, A1B1 and RCP3PD. Those scenarios were developed as part of the Swiss Climate Change Scenario CH2011 project conducted by the Center for Climate Systems Modelling (, with new generations of climate models with higher resolution combining global climate models (e.g. ECHAM) and regional climate models (COSMO).


The model for maximum butterfly abundance vs. the mean annual temperature had a good fit to the data (R2 = 0·61). We found a decrease in maximum butterfly abundance in communities at higher elevations (Fig. 3). Subsequently, our model provides a way to forecast the relative increase in herbivore pressure with elevation. The ratio of ‘future to current butterfly herbivore abundance' indicates that the impact of herbivores is likely to increase dramatically above a threshold of 2200 m, where the two worst climate change scenarios A2 and A1B indicate more than a doubling in abundance of lepidopteran herbivores (Fig. 4). Below this elevation, the increase in this ratio is less marked because of currently higher abundance of butterflies in those conditions.

Figure 3.

Relationship between butterfly maximum abundance for the current climate (dark) as well as for three climate change scenarios (A2, A1B, RCP3PD, light) for two time periods (2035 and 2085) and elevation.

Figure 4.

Relationship between ‘future to current butterfly abundance’ ratio for three climate change scenarios (A2, A1B, RCP3PD) for the 2085 period. The ratio increases with elevation.


Using our own empirical data and information available in the literature, we found evidence for a decrease in herbivory pressure along elevation gradients. This should lead to an overall reduction in plant defences in altitude; however, our meta-analysis showed that leaf toughness in trees and flavonoids in reproductive organs of herbs increase with elevation.

Several high-elevation plants might be thus selected by the abiotic environments to be ecologically fitted, and promptly respond to the sudden increase in herbivore abundance during global warming.

Herbivory and plant defences along elevation gradients

Our results suggest that, contrary to classical predictions, a reduction in herbivory pressure in harsher environments does not necessarily lead to a decrease in plant defences (Schemske 2009). Accordingly, manipulative experiments to test plant palatability at different altitudes lead to contrasting results. Across three species of slugs, a decrease in herbivory with altitude was only detected in three out of six plant species (Scheidel & Bruelheide 2001). Across sixteen pairs of plant species from high and low altitudes, we found an overall reduction in resistance against a generalist insect herbivore, despite considerable variability of responses across pairs of species (Pellissier et al. 2012). These contrasting results demonstrate that how plant defences vary with elevation are complex and is not solely the result of herbivore pressure. Indeed, several abiotic factors may contribute to selection on plant traits that in turn also protect plants against herbivores. First, a severe climate might cause selection for increased leaf toughness and decreased specific leaf area (Wright et al. 2004). Second, drought has been shown to inhibit plant secondary metabolism (Gutbrodt, Mody & Dorn 2011), thus increase susceptibility to herbivores and pathogens (Jactel et al. 2012). And third, delayed snowmelt has been shown to increase leaf nitrogen and herbivory rates (Torp et al. 2010a,b).

Moles et al. (2011) proposed that climate affects the relative cost of losing leaf tissue, with the cost being higher in harsh, unproductive environments (Coley, Bryant & Chapin 1985), resulting in different selective regimes at different latitudes (Johnson & Rasmann 2011). Therefore, selection pressure due to herbivory may be greater in harsher climate even if the herbivory rate is lower than in a warm productive climate. Thus, along with climate influencing physical and chemical protective traits, herbivores should also be expected to select for higher defence at high elevations. At low elevations, low defence may also be selected because species have fast growth rates and a high capacity to recover after damage (Herms & Mattson 1992; Bee, Kunstler & Coomes 2007).

The results of the meta-analysis are consistent with the results of our herbivory survey on tree seedlings. At a given elevation, species from higher elevations (mainly conifers) experience lower herbivory rates (Fig. 1d) and are associated with a lesser diverse community of herbivores (Brandle & Roland 2001), than deciduous species from lower elevations.

Hence, plant species from higher elevations may have developed tougher leaves as an adaptation to severe climatic conditions, but this may indirectly confer increased resistance to herbivores. This suggests that the selective forces of abiotic conditions (e.g. cold hardiness) might be stronger than biotic ones (e.g. resistance to herbivores) along elevation gradients, although this might be organ-specific, depending on the trait analysed. Because it is generally harder to reproduce at high elevation, elevation itself would select for increased protection of the highest-value organs (i.e. more flavonoids in reproductive organs), according to the optimal defence theory (Zangerl & Rutledge 1996).

High-elevation plant species may thus be ecologically fitted (but not adapted) to herbivore pressure, so rapid shifts of herbivore range under climate change would result in limited damage to high-elevation plants, and ultimately, high plant defences in alpine environments could play an important role modulating the range expansion of mobile insects following increasing temperatures.

Effects of climate changes on insect herbivory along elevation gradients

Under the scenario of increasing temperature, along with increasing winter precipitations and summer drought events forecasted by climate models (Rowell 2009; Raisanen & Eklund 2012), how would we expect insect herbivory to change? In this paper, we provide an example of a predictive model based on lepidopteran abundance along elevation gradients suggesting that climate change will lead to increased herbivore abundance at higher altitudes and that herbivory impact will be stronger, particularly above the tree line. Undoubtedly, incorporating other herbivore groups besides butterflies is needed to generalize this model. Nevertheless, our forecasted shift is also a function of the potential evolution of plant species' defence abilities, and their natural insect enemies.

First, quantities of defensive secondary metabolites may be more easily modulated under changing herbivore pressure in some plant species groups (e.g. through phenotypic plasticity or epigenetic or genetic changes) while other may have reached an evolutionary dead end, having totally lost their ability to metabolise chemical defences as in Agrawal et al. (2009). Hence, plants may show adaptive responses to variation in the herbivory rate; however, the lability vs. the conservatism of plants defence traits is difficult to estimate for the purpose of climate change forecasts.

Second, global warming might also affect insect herbivore abundance through increased top-down regulation by herbivores' natural enemies (Hance et al. 2006; Thomson, Macfadyen & Hoffmann 2010; Björkman, Berggren & Bylund 2011; Klapwijk et al. 2011; de Sassi, Staniczenko & Tylianakis 2012). Previous reviews have reported that predation and parasitism generally appear to decline with altitude (Hodkinson 2005). A recent quantitative review of the scientific literature reported similar patterns of insect parasitoid richness and parasitism rate decrease along 140 elevation gradients in a wide range of natural habitats (Péré, Jactel & Kenis 2013). On average, the relative decrease in the parasitism rate was approximately 15% per 100 m for individual parasitoid species. The decrease was greater for more exposed parasitoids, that is, ectoparasitoids and parasitoids of ectophagous insects (Péré, Jactel & Kenis 2013). With increasing temperature, we should expect natural enemies to increase their frequencies at higher elevations, thus affecting community dynamics and herbivore pressure on plants. For instance, it has been shown that size-dependent predator shifts towards larger herbivores are mediated by the interactive effect of warming and nitrogen enrichment (de Sassi, Staniczenko & Tylianakis 2012). However, in some cases, climate change might favour host–parasitoid synchrony. For example, warmer temperatures early in the year were found to favour parasitoids being more in synchrony with their host, the butterfly Melitaea cinxia (Van Nouhuys & Lei 2004). Finally, predators might not be adapted to recognize volatiles emitted from the novel, high-elevation host plants, thus further limiting the efficacy of predators during climate-driven rapid range exaptation (Raffa, Powell & Townsend 2013).


If we are to reliably predict the effects of future global change on ecosystems dynamics, then the greatest challenge resides in interpreting how biotic and abiotic ecological factors and evolutionary process act in concert. Our finding that plant defences increase with elevation, in contrast to the classical expectation, clearly show that the interplay of these factors may lead to nontrivial effects. Phylogenetic comparative studies and common garden experiments along elevation gradients for the survey of plant defence, herbivory rate and predation pressure should help to advance our understanding of the potential effects of climate change on ecosystem functioning.


Some of the research reported here was conducted as part of the BACCARA project, which received funding from the European Commission's Seventh Framework Program (FP7/ 2007–2013), under grant agreement no. 226299. GK was supported by a Marie Curie International Outgoing Fellowship within the 7th European Community Framework Program (Demo- traits project, no. 299340), SR was supported by Swiss National Science Foundation Ambizione grant PZ00P3_131956/1, and LP was supported by Danish Council for Independent Research grant no 12-126430. We are grateful to Lorenzo Marini for providing data on physical tree defence along elevation gradients.