Global Change Biology Global Change Biology (2015) 21, 106–116, doi: 10.1111/gcb.12682

Sap-feeding insects on forest trees along latitudinal gradients in northern Europe: a climate-driven patterns € ERMAN3, EUGENIA S. M I K H A I L V . K O Z L O V 1 , A N D R E Y V . S T E K O L S H C H I K O V 2 , G U Y S OD 2 1 1 L A B I N A , V I T A L I Z V E R E V and E L E N A L . Z V E R E V A 1 Section of Ecology, Department of Biology, University of Turku, Turku 20014, Finland, 2Zoological Institute, Russian Academy of Sciences, Universitetskaya nab. 1, St. Petersburg, 199034 Russia, 3V€alitalontie 43, Helsinki 00660, Finland

Abstract Knowledge of the latitudinal patterns in biotic interactions, and especially in herbivory, is crucial for understanding the mechanisms that govern ecosystem functioning and for predicting their responses to climate change. We used sap-feeding insects as a model group to test the hypotheses that the strength of plant–herbivore interactions in boreal forests decreases with latitude and that this latitudinal pattern is driven primarily by midsummer temperatures. We used a replicated sampling design and quantitatively collected and identified all sap-feeding insects from four species of forest trees along five latitudinal gradients (750–1300 km in length, ten sites in each gradient) in northern Europe (59 to 70°N and 10 to 60°E) during 2008–2011. Similar decreases in diversity of sap-feeding insects with latitude were observed in all gradients during all study years. The sap-feeder load (i.e. insect biomass per unit of foliar biomass) decreased with latitude in typical summers, but increased in an exceptionally hot summer and was independent of latitude during a warm summer. Analysis of combined data from all sites and years revealed dome-shaped relationships between the loads of sap-feeders and midsummer temperatures, peaking at 17 °C in Picea abies, at 19.5 °C in Pinus sylvestris and Betula pubescens and at 22 °C in B. pendula. From these relationships, we predict that the losses of forest trees to sap-feeders will increase by 0–45% of the current level in southern boreal forests and by 65–210% in subarctic forests with a 1 °C increase in summer temperatures. The observed relationships between temperatures and the loads of sap-feeders differ between the coniferous and deciduous tree species. We conclude that climate warming will not only increase plant losses to sap-feeding insects, especially in subarctic forests, but can also alter plant-plant interactions, thereby affecting both the productivity and the structure of future forest ecosystems. Keywords: aphids, Betula pendula, Betula pubescens, biotic interactions, climate change, hemiptera, herbivory, phloem feeders, Picea abies, Pinus sylvestris, temperature Received 8 March 2014 and accepted 18 June 2014

Introduction The impact of climate change on biotic interactions is one of the hot topics in modern ecology. Although research is expanding in this area, multiple gaps remain in our knowledge regarding the consequences of climate change for species interactions across trophic levels (Jamieson et al., 2012). In particular, the consequences of possible changes in background herbivory for plant fitness are essentially unknown, because studies that address plant responses to herbivory are clearly dominated by those that explore the effects of severe defoliation (Zvereva et al., 2012). As a result, the potential effects of climate change on insect feeding on forest trees are generally discussed in relation to eruptive species, which are mostly defoliators (Ayres & Lombardero, 2000; Dukes et al., 2009). Insect responses to climate change are context-dependent, but all aspects of insect Correspondence: Mikhail V. Kozlov, tel. +358 2 333 5716, fax +358 2 333 6598, e-mail: [email protected]

106

outbreak behaviour, and therefore herbivory pressure on plants, are expected to intensify as the climate warms (Logan et al., 2003). The validity of the latter prediction for background levels of herbivory remains to be established. Sap-feeding insects significantly reduce growth, reproduction and photosynthesis of woody plants (Zvereva et al., 2010). Even at the background levels of population density, sap-feeders (mostly aphids) affect the biogeochemistry of forest ecosystems by modifying carbon, nitrogen and potassium fluxes, both directly and through interactions with ants and other arthropods (M€ uhlenberg & Stadler, 2005; Domisch et al., 2009). However, despite the long history of research on sap-feeding pest species (including aphids, scales and true bugs), ecosystem-level studies generally quantify insect herbivory based on the amount of foliage eaten away by defoliators (Cebrian & Lartigue, 2004; Wolf et al., 2008; but see Andrew & Hughes, 2005a). This approach obviously underestimates the overall losses of net primary production (NPP) to insects because the © 2014 John Wiley & Sons Ltd

S A P - F E E D E R S I N L A T I T U D I N A L G R A D I E N T S 107 loss of assimilates due to sap-feeders in some situations can greatly exceed the loss of foliage due to defoliators (Schowalter et al., 1981), approaching 19% of annual NPP (Llewellyn, 1972). The shortage of information on the spatial and temporal patterns of the background levels of plant losses to sap-feeding insects hampers our understanding of the impacts of this group of herbivores on the structure and functions of current and future ecosystems. For example, only one (Andrew & Hughes, 2005a) of thirty-eight studies considered in a meta-analysis of latitudinal gradients in herbivory by Moles et al. (2011) included data on sap-feeding insects. We are aware of only two studies that searched for largescale geographical patterns in the abundance of sapfeeders (Hysell et al., 1996; Andrew & Hughes, 2005b), and even these studies did not report herbivore load (i.e. the biomass of herbivores per host plant biomass), which can be used as a proxy of plant losses to sap-feeders (Petermann et al., 2010) and, consequently, as an index of the strength of plant-herbivore interactions. Latitudinal gradients have been recently advertized as natural laboratories to study potential impacts of changing temperatures on terrestrial organisms (De Frenne et al., 2013). Importantly, the temperature gradient between the equator and poles is nonlinear: within the latitudes of 20° North and South mean annual temperature shows little dependence on latitude, but then the slope of the temperature gradient increases towards the poles (Terborgh, 1973; Greenwood & Wing, 1995). In addition, insects living between the 25th and 50th parallels in both hemispheres have broad thermal tolerances (Deutsch et al., 2008) while insects living at higher latitudes are generally controlled by temperature (Bale et al., 2002; Deutsch et al., 2008) and are highly responsive to climate changes (Hodkinson & Bird, 1998). As a result, latitudinal gradients in biotic interactions at high latitudes are likely to be stronger than at low latitudes (Kozlov et al., 2013), as confirmed by a study of the geographical pattern of birch losses to defoliating insects and leafminers (Kozlov, 2008). These losses strongly decreased with latitude in boreal forests of northern Europe between 60 and 70°N, but showed no latitudinal changes in central Europe between 48 and 60°N (Kozlov, 2008). Most of the studies that explored latitudinal patterns in herbivory were conducted between the 25th and 50th parallels, where the insects can be expected to show relatively minor latitudinal variation in performance, and this can explain why no directional trend in herbivory was discovered in the meta-analysis by Moles et al. (2011). On the other hand, herbivory was recently found to decrease with latitude in Nothofagus © 2014 John Wiley & Sons Ltd, Global Change Biology, 21, 106–116

pumilio forests in Argentina, between 38 and 55°S (Garibaldi et al., 2011). More studies are needed on latitudinal patterns in insect herbivory in boreal and subarctic forests to test the generality of the conclusions drawn by Moles et al. (2011) and Kozlov et al. (2013). The studies on latitudinal patterns, in addition to furthering the understanding of the effects of geography and climate on ecological processes, may also provide empirical parameterization to allow incorporation of plant losses due to insects into climate-driven biogeochemical models (Wolf et al., 2008). The aim of our study was to test, using sap-feeding insects as the model group, the hypotheses that (i) the strength of plant-herbivore interactions in boreal forests decreases with latitude and that (ii) this latitudinal pattern is driven primarily by midsummer temperatures. During 2008–2011, we explored the relationships between the latitude, climate and several characteristics of sap-feeding insect communities on four species of main forest trees of the boreal zone in northern Europe, from 59 to 70°N and from 10 to 60°E. In addition to testing the hypotheses listed above, we asked whether latitudinal patterns in plant losses to sap-feeding insects differ among geographical regions, study years and tree species.

Materials and methods

Study objects The sap-feeding insects considered in this study belong to the order Hemiptera. They include leafhoppers (suborder Cicadomorpha), true bugs (suborder Heteroptera), aphids and psyllids (suborder Sternorrhyncha). These insects are generally small and they feed on phloem of their host plants. In contrast to defoliators, the feeding marks of sap-feeders cannot be used to estimate the plant biomass that was actually consumed by insects.

Study area and sampling sites Sampling was conducted along five latitudinal gradients 750– 1300 km in length (Fig. 1: N, between Olso and Andselv, Norway; F, between Turku and Nuorgam, Finland; R, between St Petersburg and Murmansk, Russia; A, between Vologda and Arkhangelsk, Russia; K, between Vologda and Inta, Russia). The northernmost study sites in all gradients were selected in subarctic forests close to the distribution limits of the investigated tree species, while the southernmost sites were selected at the approximate borderline between the southern taiga and temperate deciduous forests. All gradients were located in Scandinavian and Russian taiga. The forests of this ecoregion are usually dominated by Scots pine (Pinus sylvestris) or Norway spruce (Picea abies) but also have significant numbers of downy and silver birches (Betula pubescens and B. pendula, respectively).

108 M . V . K O Z L O V et al.

Fig. 1 Study area and study sites. The name of each site contains the code for the gradient and the approximate latitude (for the coordinates of study sites and the list of tree species sampled at each site, see Table S1).

We sampled insects from canopies of four above-mentioned tree species. The sampling sites (ten in each gradient) were located in forests typical for each locality; care was taken to select a representative site where all four study species grow naturally. This was impossible in some areas, and in thirteen of the fifty sampling sites we collected samples from three species of trees, and in two localities from two species of trees (Table S1). In most study sites, the sampled area did not exceed 2000 m2. Each site was sampled twice a year. The R gradient was sampled each of 4 years (23–29 July and 21–25 August 2008, 24–29 June and 24–28 August 2009, 22–27 June and 29 July–2 August 2010, 12–16 June and 10–14 August 2011), while all other gradients were each sampled during 1 year (N gradient: 29 June–2 July and 27–30 August 2011; F gradient: 25–26 June and 2–4 September 2008; A gradient: 16–18 June and 7–9 August 2010; K gradient: 18–20 June and 1–3 September 2009). Mean temperatures in July were previously identified as the best predictor of plant losses to defoliators and leafminers in our study region (Kozlov, 2008; Kozlov et al., 2013). These temperatures (averaged from all records made at the weather stations nearest to our study sites from 1 to 31 July) were obtained from the web-based archive (www.rp5.ru) for each study year. Multiyear averages were calculated using New_LocClim (FAO, 2006).

Sampling and processing We sampled haphazardly selected mature trees with lower branches accessible from the ground, avoiding trees that were intensively visited by wood ants (i.e. when five or more ants were found on a branch selected for sampling); less than 5% of trees were rejected on this basis. We excluded trees frequented by ants to minimize impacts of small-scale heterogeneity in the distribution of wood ant hills on site-specific values of aphid abundance. The early and late summer samples were generally collected from different trees. One branch about 50 cm in length (with approximately 80 leaves on birches, or 500 needle pairs on Scots pine or 4000 needles on

Norway spruce) was collected from five individual trees of each of the four species at each site on each sampling date. One of two collectors placed a mesh bag attached to a ring (60 cm in diameter) under the selected branch, while the second collector cut the branch in such a way that it fell into the bag, together with the insects that dropped from the branch when disturbed. The average temperature at the time of sampling was 16.2 °C and the likelihood was low that many herbivores escaped by flying. The bag was immediately closed, labelled and transported to the laboratory, where all invertebrates were collected and preserved in alcohol. All aphids were identified by A. V. Stekolshchikov, psyllids by E. S. Labina, and leafhoppers and true bugs by G. S€ oderman. Voucher specimens were deposited in the Zoological Institute of the Russian Academy of Sciences in St Petersburg and in the Zoological Museum of the University in Helsinki. The sap-feeding insects greatly differ in their body size (Table S2); therefore, herbivore load (insect biomass per unit of plant biomass) is a better descriptor of the intensity of insect-plant relationships than is the relative abundance of insects (the number of individuals per unit of plant biomass). However, we used both indices to allow comparison with published data, which mostly refer to the abundance of sapfeeders. The need to preserve collected specimens for identification did not allow direct measurement of their biomass, which was therefore calculated from their body length as follows: weight (mg) = 0.0396 9 length (mm)2.62 (Rogers et al., 1976). Nymphs were visually attributed to one of three (aphids) or two (all other insects) size classes on the basis of their body lengths relative to adults, and the number of nymphs of each size class was recorded separately. The average body lengths of large, intermediate and small aphid nymphs were 0.80, 0.62 and 0.45 of the adult body lengths, respectively. The body lengths of large and small nymphs of other insects were 0.67 and 0.33 of the adult body lengths, respectively. The foliage from the collected branches was dried at +80 °C for 48 h and weighed. The abundance and load of sap-feeding insects were expressed in the number of © 2014 John Wiley & Sons Ltd, Global Change Biology, 21, 106–116

S A P - F E E D E R S I N L A T I T U D I N A L G R A D I E N T S 109 specimens (abbreviated as ‘exx’, from exemplars) and in the insect biomass (g) per unit of dry weight of foliage (exx kg 1 and g kg 1, respectively).

Data analysis The distribution of sap-feeders among branches and among study sites was greatly skewed: no insects were found on 1874 of 3000 collected branches, and the maximum number of insects recorded on one branch was 2191. However, this value was excluded from the analysis as an outlier (Grubbs test, Z = 15.0, P < 0.0001); the next maximum numbers (631 and 334) were not identified as outliers. We summed the data per study site 9 study year 9 tree species, and used these pooled samples to calculate the load of sap-feeders and Shannon’s index (H = –Σ [pi 9 ln pi], where pi is the proportion of ith species of the total abundance). Insects identified at the genus level were accounted for as single species when calculating the Shannon’s index if the sample contained no congeneric individuals identified at the species level. Samples with no sap-feeding insects had the H value set to zero. We also calculated the number of species recorded in each study site; note that these values may underestimate actual species richness, because nymphs were not always identified to the species level. Transformed (as ln[1 + √x]) load values met the assumption of normality within individual species of study trees, but due to the heterogeneity of dispersions (Levene’s test: F3,256 = 11.54, P < 0.0001), the characteristics of sap-feeder communities were compared between tree species, study years and gradients using the Kruskal–Wallis test. Pairwise differences were calculated by a Mann–Whitney test, followed by a Bonferroni–Holm correction for multiple comparisons. The effects of study year and latitude on the load of sap-feeding insects were explored by an analysis of covariance, and the effects of the mean temperatures in July were assessed by a regression analysis performed separately for each tree species. The fit of the first-order and second-order regression models was compared using Akaike Information Criterion (AIC). Sources of variation in the relationships of abundance and diversity of sap-feeding insects with latitude were further explored using meta-analysis, which allows data from different tree species to be combined in spite of the heterogeneity of dispersions (see above) and to develop a single conclusion that has greater statistical power than individual results. Latitudinal patterns in community-wide load and diversity of sap-feeders were quantified by calculating Spearman rank correlation coefficients with latitudes of sampling sites. Individual correlations (Table S5) refer to gradient 9 study year 9 tree species combination. To calculate effect sizes (ES), these coefficients were z-transformed and weighted by their sample size (i.e. the number of study sites) using the standard procedure in the MetaWin programme (Rosenberg et al., 2000). In our study, the negative ES values indicate decreases in load or diversity with latitude. If the number of ES in an individual group was nine or less, a bootstrap estimate of the 95% confidence interval (CI95) was used. The effect was

© 2014 John Wiley & Sons Ltd, Global Change Biology, 21, 106–116

considered statistically significant if its CI95 did not include zero. Meta-analysis was performed using random effects categorical models. The variation in the ES among the classes of categorical variables was explored by calculating the heterogeneity index (QB) and testing it against the v2 distribution.

Results

Overall abundance and diversity We collected 13 240 sap-feeding insects: 12 085 aphids of at least 33 species, 481 leafhoppers of at least 24 species, 432 psyllids of at least 19 species and 242 true bugs of at least 26 species (Tables S2 and S3). Aphids contributed 77.6% to the total biomass of sap-feeding insects, leafhoppers 14.4%, psyllids 4.6% and true bugs 3.4%. We identified 10 231 specimens to the species level and the remaining 3009 specimens (among which 2937 aphids) to the genus or family level. A total of 270 sapfeeders of at least 75 species were classified as ‘tourists’ not feeding on plants from which they were collected (Table S3) and therefore excluded from subsequent analyses.

Differences between tree species, gradients and study years The studied tree species differed in all three characteristics of their sap-feeding insect communities (Table S4): abundance (Fig. 2a; Kruskal–Wallis test: v2 = 93.3, df = 3, P < 0.0001), load (Fig. 2b; v2 = 129.4, df = 3, P < 0.0001) and diversity (Fig. 2c; v2 = 107.2, df = 3, P < 0.0001). We collected 12 species feeding on Norway spruce, 13 species on Scots pine, 27 species on silver birch and 29 species on downy birch. The highest average load of sap-feeders was found on silver birch, and it was more than 20-fold higher than the load on Norway spruce, which had the lowest average load (Fig. 2b). The load of sap-feeders (all tree species combined) did not differ among the geographical gradients (v2 = 5.44, df = 4, P = 0.25) but did vary among the study years (v2 = 12.9, df = 3, P = 0.0049). The diversity of sap-feeders was similar among the geographical gradients and among study years (v2 = 1.71, df = 4, P = 0.79 and v2 = 3.54, df = 3, P = 0.32, respectively). When data from individual tree species were analysed separately, both the load and diversity of sap-feeding insects demonstrated idiosyncratic relationships with study year and latitude (Table 1). Restriction of the analysis to the R gradient (which was sampled during all 4 years) maintained a statistically significant year 9 latitude interaction for Norway spruce and downy birch (data not shown), indicating that the

110 M . V . K O Z L O V et al. (a)

Table 1 Sources of variation in the load (biomass of herbivores per unit of foliar biomass; values transformed as ln[1 + √x]) and diversity (Shannon’s index) of sap-feeding insects (ANCOVA, type III sum of squares) Tree species Picea abies

(b)

(c)

Fig. 2 Quartiles, minimum and maximum values of the site-byyear-specific abundance (a), load (b) and diversity (c) of sapfeeding insects on four tree species. Sample sizes: Picea abies, n = 79; Betula pendula, n = 62; B. pubescens, n = 79; Pinus sylvestris, n = 78. Different letters indicate significant (P < 0.05) differences between the values (Mann–Whitney test followed by Bonferroni–Holm correction for multiple comparisons).

variable relationships between study year and latitude could not be explained by our sampling design.

Load Source

Year Latitude Year 9 Latitude Error Pinus Year sylvestris Latitude Year 9 Latitude Error Betula Year pendula Latitude Year 9 Latitude Error Betula Year pubescens Latitude Year 9 Latitude Error

Diversity

d.f F

P

F

P

3 1 3 71 3 1 3 70 3 1 3 54 3 1 3 71

0.02 0.01 0.02 – 0.04 0.0006 0.04 – 0.08 0.39 0.05 – 0.0033 0.01 0.0020 –

5.18 15.55 4.90 – 1.25 2.38 1.14 – 0.75 0.44 0.69 – 1.58 8.70 1.42 –

0.0027 0.0002 0.0038 – 0.30 0.13 0.34 – 0.52 0.51 0.57 – 0.20 0.0043 0.24 –

3.64 6.86 3.55 – 2.88 12.92 2.94 – 2.40 0.74 2.74 – 5.01 6.82 5.43 –

raw data). Combination of all five gradients revealed a decrease in the overall richness of sap-feeder fauna from 44 species between latitudes 59 and 62°N to 40 species between latitudes 62 and 66°N and to 38 species between latitudes 66 and 70°N. The latitudinal patterns in the load of sap-feeders differed between study years, while the latitudinal changes in diversity were similar in all study years (Fig. 3). In typical summers (the average difference between the actual and multiyear temperature of July in our study sites was 0.25  0.07 °C in 2008 and 0.57  0.06 °C in 2009), the total load of sap-feeders on forest trees significantly decreased with latitude. In contrast, during the warm year (+2.42  0.16 °C in 2011), the load was independent of latitude, while during an exceptionally hot summer (+4.87  0.16 °C in 2010), the load increased towards the north (Fig. 3).

Latitudinal patterns in community characteristics

Relationships between community characteristics and midsummer temperature

The meta-analysis demonstrated that although a variation was detected, the load and the diversity of sapfeeders both generally decreased with latitude (ES = 0.28, CI95 = 0.49 to 0.08 and ES = 0.23, CI95 = 0.46 to 0.05, respectively). The average load of sap-feeders in the two northernmost sites of each gradient was 57.7%, and the average number of recorded species was 69.8% of the values observed in the two southernmost sites (consult Table S4 for the

Analysis of combined data from all sites and years revealed dome-shaped relationships between the load of sap-feeders and midsummer temperatures (Fig. 4). All second-order regression models fitted the data better than linear models (Table 2), but in the case of silver birch, the linear model had a lower AIC than the second-order model. The load of sap-feeding insects, as calculated from quadratic regression models, peaked at 17 °C in Norway spruce, at 19.5 °C in Scots pine and © 2014 John Wiley & Sons Ltd, Global Change Biology, 21, 106–116

S A P - F E E D E R S I N L A T I T U D I N A L G R A D I E N T S 111 r = 0.02, n = 78, P = 0.86; silver birch: r = 0.08, n = 62, P = 0.52).

Discussion

Latitudinal patterns in abundance and diversity of sapfeeding insects

Fig. 3 Variation in latitudinal patterns of the load and diversity of sap-feeding insects among study years (results of meta-analysis). The negative ES values indicate decreases in load or diversity with latitude. Horizontal lines denote 95% bootstrap confidence intervals; n = 8 gradient 9 year 9 tree species combinations; QB, between-class heterogeneity.

downy birch and at 22 °C in silver birch. Shannon’s index was positively associated with temperature for insects feeding on downy birch (r = 0.36, n = 79, P = 0.001), but not on the other three tree species (Norway spruce: r = 0.01, n = 79, P = 0.92; Scots pine:

The latitudinal diversity gradient is one of the most striking biogeographical patterns, and its generality is supported by both narrative reviews and meta-analysis (Willig et al., 2003; Hillebrand, 2004). In contrast, the existence of latitudinal patterns in species abundance and biotic interactions remains debatable (Gaston & Blackburn, 2000; Leonard, 2000; Ollerton & Cranmer, 2002). In particular, although insect-plant relationships, intensity of which is measured by plant losses to insects, are often referred to as examples of biotic interactions that weaken with latitude (Coley & Barone, 1996; Schemske et al., 2009), the generality of this pattern had recently been questioned (Moles et al., 2011). Our study provides the first comprehensive assessment of large-scale geographical variation in the communities of sap-feeding insect herbivores in the Northern hemisphere, and it is one of the first studies to show the latitudinal patterns in herbivory based on a sampling design replicated in both space and time. Our results, in line with earlier studies conducted in the same region on defoliators (Kozlov, 2008) and

(a)

(b)

(c)

(d)

Fig. 4 The load of sap-feeding insects on forest trees in relation to midsummer temperature: (a) Picea abies, (b) Pinus sylvestris, (c) Betula pendula, (d) B. pubescens. Each data point corresponds to one study site 9 year combination (i.e. is based on 10 individual samples). For the results of regression analyses, consult Table 2. © 2014 John Wiley & Sons Ltd, Global Change Biology, 21, 106–116

112 M . V . K O Z L O V et al. Table 2 Regression of the load of sap-feeding herbivores (biomass of herbivores per unit of foliar biomass; values transformed as ln[1 + √x]) on mean temperature in July Parameter estimates (mean  SE) Regression model

Tree species

Linear

Picea abies Pinus sylvestris Betula pendula Betula pubescens Picea abies Pinus sylvestris Betula pendula Betula pubescens

Quadratic

Intercept 0.1692 0.0783 0.0724 0.0230 0.7234 0.7296 1.0093 2.0355

       

Temperature 0.0740 0.0831 0.1858 0.1288 0.4253 0.4703 1.2468 0.7313

leafminers (Kozlov et al., 2013), indicate a pronounced poleward decrease in the abundance and diversity of sap-feeding insects on the main forest trees in northern Europe and thus confirm our first hypothesis that the strength of plant-herbivore interactions in boreal forests generally decreases with latitude. Importantly, the lower loads of sap-feeders observed in the northernmost parts of our study region resulted from both a smaller number of herbivore species and their lower abundances. This result adds to the limited evidence (Adams et al., 2010; Kozlov et al., 2013) that the southern populations of a certain tree species are attacked by a larger number of herbivore species than are its northern populations. Consequently, the ongoing expansion of distribution ranges towards the North, reported for many insect species (Parmesan et al., 1999; Warren et al., 2001), can be expected to contribute to an overall increase in herbivory. However, we demonstrated a uniform latitudinal pattern in the diversity of sap-feeders across the study years, indicating that one or two hot summers alone cannot substantially increase the richness of local communities of insect herbivores. Our data demonstrated significant relationships between the load of sap-feeding insects on main forest trees and the midsummer temperatures, thus supporting our second hypothesis. Our findings are in line with an earlier conclusion that the latitudinal variation in folivory rates is partly driven by climate (Garibaldi et al., 2011): mean temperature in July explained 7–20% of the total variation in the load of sap-feeders. The low abundance of sap-feeders in the northernmost study sites is most likely associated with the direct impact of low temperatures on insect performance (Bale et al., 2002). The importance of climate in controlling the populations of sap-feeders is further confirmed by the pronounced variation observed among the study years. Our results agree with those of earlier studies (Zhou et al., 1997; Whittaker & Tribe, 1998) showing that

0.0038 0.0091 0.0245 0.0238 0.1000 0.1035 0.1472 0.2629

       

0.0044 0.0049 0.0106 0.0076 0.0489 0.0544 0.1402 0.0840

Model characteristics Temperature2

R2

F

P

– – – –

0.01 0.04 0.08 0.11 0.07 0.08 0.09 0.20

0.75 3.42 5.38 9.84 2.67 3.24 3.07 9.45

0.388 0.068 0.024 0.002 0.076 0.043 0.054 0.0002

0.0029 0.0027 0.0034 0.0067

   

0.0014 0.0015 0.0038 0.0024

sap-feeders living in relatively cool climates can undergo rapid increases in population size with increases in summer temperatures. These effects may result from the impacts of increased temperature on both insects and their hosts, because feeding on plants growing at elevated temperatures generally improves herbivore performance (Zvereva & Kozlov, 2006). De Frenne et al. (2013) proposed that latitudinal gradients can serve as natural laboratories that allow a better understanding of the potential impacts of changing temperatures on terrestrial organisms. They argued in particular that latitudinal patterns in biological characteristics should be evaluated against temperature and covarying environmental factors, rather than using latitude as a ‘surrogate’. Our results showing variation in the relationships between the load of sap-feeders and latitude among study years strongly support this opinion. One of the most interesting findings of our study is the discovery of an annual variation in the strength and even in the direction of the latitudinal changes in the load of sap-feeding herbivores on forest trees. This demonstrates that weather conditions during the study year(s) should be taken into account in the analysis of latitudinal patterns in biotic interactions, and that oneyear observations, which formed the basis of the majority of earlier studies (reviewed in the meta-analysis by Moles et al., 2011), can easily generate misleading conclusions. For example, if we were to base our conclusions only on observations obtained for the year 2010, we would have to state that plant losses to sap-feeders increase with latitude (Fig. 3), which is in fact opposite to the general trend. The among-year variation in the strength and direction of latitudinal patterns is an important addition to the list of factors (discussed by De Frenne et al., 2013) that can confound analyses across latitudes. Our data demonstrate that the annual variation in the direction of the latitudinal patterns in the load of sap© 2014 John Wiley & Sons Ltd, Global Change Biology, 21, 106–116

S A P - F E E D E R S I N L A T I T U D I N A L G R A D I E N T S 113 feeding herbivores on forest trees is explained by the dome-shaped relationships of this index of plant-herbivore interactions with midsummer temperatures. To our knowledge, this pattern, presumably reflecting the maximum load of herbivores at an optimal temperature, has not been detected in earlier studies, which approximated climatic or latitudinal patterns in herbivory with a linear function. Interestingly, the load of insects feeding on conifers showed weaker relationships with climate and peaked at lower temperatures than the load of insects feeding on birches. This pattern corresponds to the results of meta-analysis (Zvereva & Kozlov, 2006), which demonstrated that angiosperm plants, but not gymnosperms, respond to temperature elevation by decrease in their secondary defensive compounds, and therefore species feeding on deciduous trees benefit from temperature elevation more than conifer-feeding species.

Current and future losses of forest trees to sap-feeders Data are scarce regarding the background density of sap-feeding insects feeding on forest trees in northern Europe and most have been obtained from a single study site and/or a single study year. For example, the average density of Hemiptera on Norway spruce in Toftaholm, southern Sweden, in the spring of 2010 was 0.3 exx 50 cm 1 branch (Edenius et al., 2012), which is approximately equal to 40 exx kg 1 (d.w.) of needles. In Turku, southwestern Finland, in May of 2002, the densities of aphids on silver birch ranged from 140 to 430 exx kg 1 (Valkama et al., 2005). In S€ avar, northern Sweden, in July of 1987, the average density of aphids on silver birch was 250 exx kg 1 (den Herder et al., 2009). These data indicate that the abundances of sapfeeding insects recorded in our study (Fig. 2a) are within the ‘normal’ range for boreal forests; i.e. they represent the background levels of herbivory. The mean load of sap-feeding insects on the studied tree species ranged 0.036–0.764 g kg 1 (d.w.) of foliage (Fig. 2b), and 77.6% of this load was imposed by aphids. Several aphid species were found to cause a 3.3–3.7 mg reduction in plant growth for each milligram gained in aphid biomass during the development of these insects (Van Hook et al., 1980; Mackay & Lamb, 1996). The number of aphid generations varies between the northern and southern ends of our gradients, from three to ten in conifer-feeding species and from four to thirteen in birch-feeding species (Table S2). Multiplying aphid load, plant losses per unit of aphid biomass and number of generations, we conclude that Norway spruce and Scots pine in our study region annually lose 0.05–0.5% of their foliar biomass to sap-feeders, while birches lose 1.0–3.5%. These estimates are somewhat © 2014 John Wiley & Sons Ltd, Global Change Biology, 21, 106–116

conservative, because we excluded trees frequented by ants (usually having greater densities of aphids) and did not account either for metabolic losses and for honeydew production by adult aphids or for significant reduction in photosynthesis due to sap-feeding (Zvereva et al., 2010). Nevertheless, our estimates of background losses to sap-feeders are of the same order of magnitude as the background losses of forest trees to defoliators in our study region. For example, losses of needle biomass to chewing and mining insects in southern Karelia, Russia, were 0.66% in Norway spruce and 0.47% in Scots pine (Galasjeva & Pisareva, 1991). In southern Sweden, defoliators consumed 0.70% of the needle biomass of Scots pine (Larsson & Tenow, 1980). Losses in silver and downy birches to defoliators in Fennoscandia varied from 1–2% at 70°N to 5–7% at 60°N (Kozlov, 2008). Regression analysis (Table 2) suggests that the average load of sap-feeding insects on birches and Scots pine is likely to increase by 10–30% of its current level in the southern boreal forests (59–62°N, where the average temperature of July recently ranges 15–18 °C) and by 45–90% in the northern boreal forests (68–70°N, where the average temperature of July recently ranges 11–14 °C) with a 1 °C increase in temperature. The average load of sap-feeding insects on Norway spruce will either remain the same or increase by 7–15% of its current level in the southern boreal forests and increase by 60–120% in the northern boreal forests with a 1 °C increase in temperature. These estimates are of the same order of magnitude as results from field experiments with open-top chambers conducted in Spitsbergen, where the densities of aphids feeding on Salix polaris increased by 25–45% with an approximately 1 °C temperature increase (Gillespie et al., 2013). The summer temperatures within our study region are predicted to increase by 1.5–3.5 °C by the end of the XXI century (G€ ottel et al., 2008). Due to the discovered dome-shaped relationships between the abundance of sap-feeders and the midsummer temperature, this increase may have different consequences in the northern and in the southern parts of the study region. In southern boreal forests, the load of sap-feeders on Norway spruce, downy birch and Scots pine is likely to decline after the mean temperature of July increases over 18–20 °C. At the same time, the load of sap-feeders in the northern boreal forests will increase within the entire range of the realistic climate change scenarios. A warming of 1 °C not only changes the average aphid abundance but also increases the number of generations produced per year by two or three (Yamamura & Kiritani, 1998). This estimate fits well with the projections made from the numbers of generations observed at the northern and southern borders of our study

114 M . V . K O Z L O V et al. region (Table S2). One additional generation might be expected for some species of leafhoppers and true bugs (Kiritani, 2006). Taking all these changes into account, we conclude that the losses of Norway spruce to sap-feeders are likely to increase by 20% of the current level in southern boreal forests and by 65% in subarctic forests with a 1 °C increase in temperature. This will occur exclusively due to the increase in the number of generations of sap-feeding insects, even if their abundance will remain unchanged. If, however, the abundance will change as predicted by the marginally significant quadratic model, then the losses of Norway spruce to sap-feeding insects will increase by 20–25% of the current level in southern boreal forests and by 160–210% in subarctic forests with a 1 °C increase in temperature. In other species of forest trees, the effects of increased numbers of generations will be significantly enhanced by an increase in the average abundance of sap-feeders, thereby yielding higher estimates of the relative increase in plant losses on both the southern and northern margins of our study region: by 40% and 150% in Scots pine and by 45% and 150% in birches, respectively. Thus, the expected increase in the losses to sap-feeders is likely to be at least three times as high in subarctic forests than at the southern boreal forest. This conclusion is consistent with the earlier prediction (based on the data on birch-feeding leafminers) that climate warming will result in a stronger relative increase in herbivory at higher latitudes than at lower latitudes (Kozlov et al., 2013). In absolute values, the losses in foliar biomass of Norway spruce due to sap-feeders are likely to increase by 0.02–0.1%, losses of Scots pine by 0.05–0.1%, and losses of birches by 0.3–1.0% with a 1 °C increase in temperature. This estimate approaches the expected increase in losses of birch foliage to defoliators (0.91% by 1 °C: Kozlov, 2008). Our predictions on the future losses of forest trees to sap-feeding herbivores are based on the patterns we observed in latitudinal gradients. These predictions obviously have some limitations, in particular because they do not account for the consequences of steadily increasing ambient CO2 concentrations. The abundance of phloem feeders generally increases with an increase in ambient CO2 (Robinson et al., 2012), and combined impacts of elevated CO2 and temperature are likely to cause further enhancement in aphid population growth. Moreover, the future net losses may be even higher than those predicted from changes in herbivore abundance because the elevation of ambient CO2 may enhance the consumption rates of plant-eating insects due to decreased nitrogen concentrations in plant

tissues (Lincoln et al., 1993). However, it remains unclear if the latter prediction applies to sap-feeders, because only a few studies addressing their performance actually measured CO2-induced changes in phloem composition of the host plants (Robinson et al., 2012). The scarcity of knowledge on mechanisms that drive sap-feeder responses to elevated CO2 hampers predictions on the impacts of climate change on this feeding guild of insects.

Impacts of background herbivory on ecosystem structure and functions Modern concepts of plant responses to herbivory are based primarily on studies of short term, severe damage. The effects of background herbivory therefore remain poorly understood. We recently demonstrated that, after 7 years of treatments, the annual removal of 4–8% of leaf biomass from downy birch – the level corresponding to the average foliar losses due to background herbivory in much of its distribution range (Kozlov, 2008) – resulted in a significant (34–45%) growth reduction that was proportional to the applied damage (Zvereva et al., 2012). The effects of background loads of sap-feeders on growth of our tree species remain unknown, but sap-feeders generally impose a more severe negative impact on the performance of woody plants than do defoliators. Woody plants usually enhance photosynthesis following defoliation, thereby facilitating recovery, whereas sap-feeding damage reduces photosynthesis, thereby slowing recovery (Zvereva et al., 2010). Direct background losses of biomass might be similar in our study trees when attacked by sap-feeders and defoliators, but sucking insects may cause a greater reduction in overall plant fitness. The expected increase in abundance of sap-feeding insects will therefore have pronounced effects on the structure and functions of future forest ecosystems. Existing vegetation and biogeochemical models generally disregard effects of insect damage on NPP and vegetation structure (Cramer et al., 2001; Galbraith et al., 2010), although recent developments in LPJGUESS model include simulation of spruce bark beetle outbreaks (J€ onsson et al., 2012) and coupling with a population model of large (vertebrate) grazers (Pachzelt et al., 2013). Not surprisingly, inclusion of leaf area loss in birches to defoliators (as a function of the temperature) into the LPJ-GUESS model alters predictions of future forest composition in northern Europe (Wolf et al., 2008). In particular, a decrease is predicted in the proportion of forests that are dominated by birches. Interestingly, our results indicate that climate-driven changes in plant losses to sap-feeding insects in the © 2014 John Wiley & Sons Ltd, Global Change Biology, 21, 106–116

S A P - F E E D E R S I N L A T I T U D I N A L G R A D I E N T S 115 northern (subarctic) forests will be larger in Norway spruce than in birches. In contrast, in southern boreal forests climate warming will result in lower load of sap-feeders on Norway spruce, while birch damage will increase. These differential changes in herbivory pressure may alter plant-plant interactions in favour of the conifers in the southern boreal forests and in favour of birches in the northern (subarctic) forests as the climate warms.

Acknowledgements We thank N. Zvereva, A. Popova, M. Inozemtseva and L. Krasheninin for fieldwork and laboratory assistance, and two anonymous reviewers for useful comments to an earlier draft of the manuscript. The study was supported by the Academy of Finland (project 122133 and researcher exchange grants), by a strategic research grant from the University of Turku and by the Ministry of Education and Science of the Russian Federation (project 16.518.11.7070).

References Adams JM, Brusa A, Soyeong A, Ainuddin AN (2010) Present-day testing of a paleoecological pattern: is there really a latitudinal difference in leaf-feeding insectdamage diversity? Review of Palaeobotany and Palynology, 162, 63–70. Andrew NR, Hughes L (2005a) Herbivore damage along a latitudinal gradient: relative impacts of different feeding guilds. Oikos, 108, 176–182. Andrew NR, Hughes L (2005b) Diversity and assemblage structure of phytophagous Hemiptera along a latitudinal gradient: predicting the potential impacts of climate change. Global Ecology and Biogeography, 14, 249–262. Ayres MP, Lombardero MJ (2000) Assessing the consequences of global change for forest disturbance from herbivores and pathogens. Science of the Total Environment, 262, 263–286. Bale JS, Masters GJ, Hodkinson ID et al. (2002) Herbivory in global climate change research: direct effects of rising temperature on insect herbivores. Global Change Biology, 8, 1–16. Cebrian J, Lartigue J (2004) Patterns of herbivory and decomposition in aquatic and terrestrial ecosystems. Ecological Monographs, 74, 237–259. Coley PD, Barone JA (1996) Herbivory and plant defenses in tropical forests. Annual Review in Ecology and Systematics, 27, 305–335. Cramer W, Bondeau A, Woodward FI et al. (2001) Global response of terrestrial ecosystem structure and function to CO2 and climate change: results from six dynamic global vegetation models. Global Change Biology, 7, 357–373. De Frenne P, Graae BJ, Rodrıguez-Sanchez F et al. (2013) Latitudinal gradients as natural laboratories to infer species’ responses to temperature. Journal of Ecology, 101, 784–795. Deutsch CA, Tewksbury JJ, Huey RB, Sheldon KS, Ghalambor CK, Haak DC, Martin PR (2008) Impacts of climate warming on terrestrial ectotherms across latitude. Proceedings of the National Academy of Sciences of the United States of America, 105, 6668–6672. Domisch T, Finer L, Neuvonen S et al. (2009) Foraging activity and dietary spectrum of wood ants (Formica rufa group) and their role in nutrient fluxes in boreal forests. Ecological Entomology, 34, 369–377. Dukes JS, Pontius J, Orwig D et al. (2009) Responses of insect pests, pathogens, and invasive plant species to climate change in the forests of northeastern North America: what can we predict? Canadian Journal of Forest Research, 39, 231–248. Edenius L, Mikusinski G, Witzell J, Bergh J (2012) Effects of repeated fertilization of young Norway spruce on foliar phenolics and arthropods: implications for insectivorous birds’ food resources. Forest Ecology and Management, 277, 38–45. FAO (2006) New_LocClim, Local Climate Estimator version 1.10. Environment and Natural Resources Service Agrometeorology Group FAO/SDRN, Rome, Italy. Available at: ftp://ext-ftp.fao.org/SD/Reserved/Agromet/New_LocClim/ (accessed 18 July 2014).

© 2014 John Wiley & Sons Ltd, Global Change Biology, 21, 106–116

Galasjeva TV, Pisareva SD (1991) The level of background damage of needles and leaves by arthropods and fungi in the ‘Kivatch’ nature reserve. In: Entomological Studies in the ‘Kivatch’ Nature Reserve (eds Yakovlev EB, Mozolevskaya ET), pp. 99– 112. Forest Institute, Petrozavodsk (in Russian). Galbraith D, Levy PE, Sitch S, Huntingford C, Cox P, Williams M, Meir P (2010) Multiple mechanisms of Amazonian forest biomass losses in three dynamic global vegetation models under climate change. New Phytologist, 187, 647–665. Garibaldi LA, Kitzberger T, Ruggiero A (2011) Latitudinal decrease in folivory within Nothofagus pumilio forests: dual effect of climate on insect density and leaf traits? Global Ecology and Biogeography, 20, 609–619. Gaston KJ, Blackburn TM (2000) Pattern and Processes in Macroecology. Blackwell Publishing, Malden, USA. Gillespie MAK, Jonsdottir IS, Hodkinson ID, Cooper EJ (2013) Aphid–willow interactions in a high Arctic ecosystem: responses to raised temperature and goose disturbance. Global Change Biology, 19, 3698–3708. G€ ottel H, Alexander J, Keup-Thiel E et al. (2008) Influence of changed vegetation fields on regional climate simulations in the Barents Sea Region. Climate Change, 87, 35–50. Greenwood DR, Wing SL (1995) Eocene continental climates and latitudinal gradients. Geology, 23, 1040–1048. den Herder M, Bergstr€ om R, Niemel€a P, Danell K, Lindgren M (2009) Effects of natural winter browsing and simulated summer browsing by moose on growth and shoot biomass of birch and its associated invertebrate fauna. Annales Zoologici Fennici, 46, 63–74. Hillebrand H (2004) On the generality of the latitudinal diversity gradient. American Naturalist, 163, 192–211. Hodkinson ID, Bird J (1998) Host-specific herbivores as sensors of climate change in Arctic and alpine environments. Arctic and Alpine Research, 30, 78–83. Hysell MT, Wagner MR, Grier CC (1996) Pattern of herbivore abundance along a climatic gradient: guild shift of leaf-feeding insects and changes in leaf characteristics. Environmental Entomology, 25, 977–982. Jamieson MA, Trowbridge AM, Raffa KF, Lindroth RL (2012) Consequences of climate warming and altered precipitation patterns for plant–insect and multitrophic interactions. Plant Physiology, 160, 1719–1727. J€ onsson AM, Schroeder LM, Lagergren F, Anderbrant O, Smith B (2012) Guess the impact of Ips typographus – an ecosystem modelling approach for simulating spruce bark beetle outbreaks. Agricultural and Forest Meteorology, 166, 188– 200. Kiritani K (2006) Predicting impacts of global warming on population dynamics and distribution of arthropods in Japan. Population Ecology, 48, 5–12. Kozlov MV (2008) Losses of birch foliage due to insect herbivory along geographical gradients in Europe: a climate-driven pattern? Climate Change, 87, 107–117. Kozlov MV, van Nieukerken EJ, Zverev V, Zvereva EL (2013) Abundance and diversity of birch-feeding leafminers along latitudinal gradients in northern Europe. Ecography, 36, 1138–1149. Larsson S, Tenow O (1980) Needle-eating insects and grazing dynamics in a mature Scots pine forest in central Sweden. Ecological Bulletin, 32, 269–306. Leonard GH (2000) Latitudinal variation in species interactions: a test in the New England rocky intertidal zone. Ecology, 81, 1015–1030. Lincoln DE, Fajer ED, Johnson RH (1993) Plant–insect herbivore interactions in elevated CO2 environments. Trends in Ecology and Evolution, 8, 64–68. Llewellyn M (1972) Effects of Lime aphid Eucallipterus tiliae (Aphididae) on growth of lime Tilia 9 vulgaris Hayne. 1. Energy requirements of aphid population. Journal of Applied Ecology, 9, 261–282. Logan JA, Regniere J, Powell JA (2003) Assessing the impacts of global climate change on forest pests. Frontiers in Ecology and the Environment, 1, 130–137. Mackay PA, Lamb RJ (1996) Dispersal of five aphids (Homoptera: Aphididae) in relation to their impact on Hordeum vulgare. Environmental Entomology, 25, 1032–1044. Moles AT, Bonser SP, Poore AGB, Wallis JR, Foley WJ (2011) Assessing the evidence for latitudinal gradients in plant defence and herbivory. Functional Ecology, 25, 380–388. M€ uhlenberg E, Stadler B (2005) Effects of altitude on aphid-mediated processes in the canopy of Norway spruce. Agricultural and Forest Entomology, 7, 133–143. Ollerton J, Cranmer L (2002) Latitudinal trends in plant-pollinator interactions: are tropical plants more specialised? Oikos, 98, 340–350. Pachzelt A, Rammig A, Higgins S, Hickler T (2013) Coupling a physiological grazer population model with a generalized model for vegetation dynamics. Ecological Modelling, 263, 92–102. Parmesan C, Ryrholm N, Stefanescu C et al. (1999) Poleward shifts in geographical ranges of butterfly species associated with regional warming. Nature, 399, 579–583.

116 M . V . K O Z L O V et al. Petermann JS, M€ uller CB, Weigelt A, Weisser WW, Schmid B (2010) Effect of plant species loss on aphid–parasitoid communities. Journal of Animal Ecology, 79, 709–

Yamamura K, Kiritani K (1998) A simple method to estimate the potential increase in the number of generations under global warming in temperate zones. Applied

720. Robinson EA, Ryan GD, Newman JA (2012) A meta-analytical review of the effects of elevated CO2 on plant–arthropod interactions highlights the importance of interacting environmental and biological variables. New Phytologist, 194, 321–336. Rogers LE, Hinds WT, Buschbom RL (1976) General weight vs length relationship for insects. Annals of the Entomological Society of America, 69, 387–389.

Entomology and Zoology, 33, 289–298. Zhou X, Perry JN, Woiwod IP, Harrington R, Bale JS, Clark SJ (1997) Temperature change and complex dynamics. Oecologia, 112, 543–550. Zvereva EL, Kozlov MV (2006) Consequences of simultaneous elevation of carbon dioxide and temperature for plant–herbivore interactions: a meta-analysis. Global Change Biology, 12, 27–41. Zvereva EL, Lanta V, Kozlov MV (2010) Effects of sap-feeding insect herbivores on

Rosenberg MS, Adams DC, Gurevitch J (2000) MetaWin: Statistical Software for MetaAnalysis Version 2.0. Sinauer, Sunderland. Schemske DW, Mittelbach GG, Cornell HV, Sobel JM, Roy K (2009) Is there a latitudinal gradient in the importance of biotic interactions? Annual Review in Ecology and Systematics, 40, 245–269. Schowalter TD, Webb JW, Crossley DA (1981) Community structure and nutrient

fitness of woody plants: a meta-analysis of experimental studies. Oecologia, 163, 949–960. Zvereva EL, Zverev VE, Kozlov MV (2012) Little strokes fell great oaks: minor but chronic herbivory substantially reduces birch growth. Oikos, 121, 2036–2043.

content of canopy arthropods in clear-cut and uncut forest ecosystems. Ecology, 62, 1010–1019. Terborgh J (1973) On the notion of favorableness in plant ecology. American Naturalist, 107, 481–501. Valkama E, Koricheva J, Ossipov V, Ossipova S, Haukioja E, Pihlaja K (2005) Delayed induced responses of birch glandular trichomes and leaf surface lipophilic compounds to mechanical defoliation and simulated winter browsing. Oecologia, 146, 385–393. Van Hook RI, Nielsen MG, Shugart HH (1980) Energy and nitrogen relations for a Macrosiphum liriodendri (Homoptera: Aphididae) population in an East Tennessee Liriodendron tulipifera stand. Ecology, 61, 960–975. Warren MS, Hill JK, Thomas JA et al. (2001) Rapid responses of British butterflies to opposing forces of climate and habitat change. Nature, 414, 65–69. Whittaker JB, Tribe NP (1998) Predicting numbers of an insect (Neophilaenus lineatus: Homoptera) in a changing climate. Journal of Animal Ecology, 67, 987–991. Willig MR, Kaufman DM, Stevens RD (2003) Latitudinal gradients of biodiversity: pattern, process, scale and synthesis. Annual Review in Ecology and Systematics, 34, 273–309. Wolf A, Kozlov MV, Callaghan TV (2008) Impact of non-outbreak insect damage on vegetation in northern Europe will be greater than expected during a changing cli-

Supporting Information Additional Supporting Information may be found in the online version of this article: Table S1. The coordinates of study sites and the list of tree species sampled at each site. Table S2. Overall abundance and life history data of species feeding on the studied tree species. Table S3. Overall abundance and life history data of species that, most likely, did not feed on the studied tree species. Table S4. Site-by-year-specific values of abundance, load and diversity of sap-feeding insects. Table S5. Spearman rank correlation coefficients between the load or diversity of sap-feeding insects and latitudes of study sites.

mate. Climate Change, 87, 91–106.

© 2014 John Wiley & Sons Ltd, Global Change Biology, 21, 106–116