1 Responses of a Scandinavian sub-Arctic dwarf shrub heath community to 9 years of nutrient and temperature treatments were examined. Our objective was to assess the responses of plant and insect herbivore communities to these treatments, and to determine how vegetation responses scale-up to those of a second trophic group.
2 There were strong effects of nutrient addition on the above-ground biomass of both dominant (dwarf shrubs) and subordinate (grasses and mosses) plant functional groups, with responses by the latter being of greater magnitude. Responses to temperature were less frequent and of a smaller magnitude.
3 There were marked changes in the abundance of insect herbivores in response to the treatments. Changes in the above-ground biomass of subordinate plant species had a greater impact on the composition of the insect herbivore community than the smaller responses of dominant dwarf shrubs. For example, the abundance of a moss-feeding Heteropteran in fertilized plots was as little as 6% that of controls, while Homoptera specializing on grasses were over 400% more abundant. In addition, gramnivorous taxa (the Delphacidae) were present only in those plots that received nutrients.
4 Despite some species-specific effects of the perturbations on the quality of dwarf shrub annual shoots (defined as the concentrations of nitrogen and phenolic compounds), little change in leaf herbivory was observed. Insect herbivores removed less than 1% of annual biomass from dominant dwarf shrub species.
5 It is proposed that insect community change was driven by subordinate plant groups and not by the dominant dwarf shrub species, suggesting a wider importance of subordinate species for community structure.
Surface air temperatures are increasing across large areas of the world, with some of the greatest changes occurring in Arctic regions (Rind 1999). Specifically, a regional analysis within Europe has suggested that mean annual warming in Arctic Norway will be 0.8 °C per decade (Hanseen-Bauer & Forland 2001). Warming is likely to have significant ecological impacts in Arctic regions where growth is limited by low temperatures and nutrient availability (Callaghan & Jonasson 1995). In addition to the direct influence of warming on the growth and development of organisms, stimulated soil nutrient mineralization (Rustad et al. 2001) is likely to have an important influence on ecosystem function in environments where water in non-constraining. Warming may accelerate decomposition of standing soil organic matter and enhance nutrient availability, although there have been indications, both from a meta-analysis (Rustad et al. 2001) and a short-term field study (Luo et al. 2001), that rates of respiration may acclimatize. There have additionally been large regional increases in nitrogen deposition over the past 100 years in the Arctic (Woodin 1997), and these are predicted to continue through the 21st century (Lee 1998).
Boreal and tundra environments are frequently dominated by dwarf shrub species such as Vaccinium (Shevtsova et al. 1995). Short-term experiments have frequently recorded increased growth of dwarf shrubs and other vascular plants with nutrient addition (e.g. Parsons et al. 1994) and, to a lesser extent, warming (e.g. Michelsen et al. 1996). However, analyses of mid- to long-term experiments have shown that species interactions within the community finally determine responses to treatments, leading to reduced growth of some species or functional groups such as bryophytes and lichens (Cornelissen et al. 2001; Shaver et al. 2001).
The mass ratio theory, discussed by Grime (1998), proposes that the extent to which a plant species influences ecosystem function is proportional to its abundance, that is, its degree of dominance. Therefore it might be anticipated that in sub-Arctic heath communities the responses of the dominant dwarf shrubs to perturbations would contribute significantly to the functional response of the heath community. In this study we tested this ‘community response’ by examining a second trophic group, the insect herbivores. Evidence for the mass ratio theory has largely been acquired from observational studies, but here we use experimental perturbation to examine the extent to which the dominant plants influence the insect herbivore community.
Insect herbivores that feed above ground are considered to be the most responsive element of the Arctic invertebrate fauna to the direct effects of environmental change, because those guilds that live either in the soil or in water are probably more buffered against changes in climate (Hodkinson et al. 1998). Responses of insect herbivores are likely to be complex and individualistic, both because of the diversity of herbivore form and life cycle, and because of interactions with host plants and predators. While vertebrate herbivores are likely to have larger impacts on their host plants, the short generation time of invertebrate herbivores may produce more rapid changes in abundance, especially relative to their longer-lived host plants (Masters et al. 1998). Furthermore, as the geographical range of herbivores is frequently encompassed within the range of their host plant, the capacity to respond to warming by extending their range into existing host plant areas is significant (Hodkinson & Bird 1998).
The potential implications of climate change for host plant–insect herbivore interactions have been investigated through observational studies (e.g. Tenow et al. 1999), but relatively long-term field manipulation studies in the Arctic that examine both vegetation and herbivore responses are lacking, and are necessary if we are to address important issues of scaling-up in climate-change predictions (Lawton 2001).
The overall objectives of our study were (i) to determine the long-term (9 years) responses of vegetation of a sub-Arctic dwarf shrub heath to nutrient addition and warming, and (ii) to measure the impact on insect herbivores. Specifically, we sought to quantify impacts on (i) insect herbivory of three dominant dwarf shrub species, (ii) the abundance of the dominant insect herbivore families, and (iii) the interaction between insect herbivores and plants that may arise as a result of treatment-induced changes in plant quantity and (for three dominant dwarf shrub species) plant quality.
We tested the following hypotheses: (i) that the dominant plant group, the dwarf shrubs, would show the greatest responses and that the largest impact on insect herbivores would be on species feeding on this group, and (ii) that for three of the dwarf shrub species, treatment-induced increases in plant quality (reductions in phenolics and increases in nitrogen concentrations) and quantity would be matched by increases in shrub herbivore abundance and leaf herbivory.
field site and study species
The field site was located within an open birch forest, Betula pubescens ssp. czerepanovii[Orl.] Hämet-Athi, approximately 2 km south-east of the Abisko Scientific Research Station, Swedish Lapland (c. 68°N, 19°E), where the mean July temperature is 11.0 °C and mean summer precipitation is c. 120 mm (Andersson et al. 1996; Abisko Scientific Research Station Meteorological Station). The understorey vegetation was composed largely of ericaceous dwarf shrubs, described by Sonesson & Lundberg (1974) as the Empetrum type, co-dominated by Empetrum nigrum ssp. hermaphroditum (Hagerup) and species of Vaccinium. A complete description of the understorey vegetation at this site is given by Wookey et al. (1993). This study was an examination of the understorey vegetation, with particular focus on three dwarf shrub species within the dominant genus, Vaccinium, and the insect herbivore community present on that vegetation. Biomass, annual shoot (leaf plus stem) quality and herbivory were measured in two deciduous species, Vaccinium myrtillus (L.) and Vaccinium uliginosum (L.), and an evergreen, Vaccinium vitis-idaea (L.). Biomass of the remaining components of the community – grasses, lichens and bryophytes – was also measured. Details of the species present within these functional groups are given by Press et al. (1998). Responses by the insect herbivore community were determined by two independent assessments that reflected the most appropriate techniques available. First, herbivory of the three Vaccinium species by leaf-chewing larvae of Lepidoptera and Symphyta (Hymenoptera) was quantified. Neither visual estimations of insect abundance nor vacuum sampling of this group yielded reliable data (Richardson 2000), and impacts were therefore assessed from the resulting damage. The density of the second group of insect herbivores present at this site, the Hemiptera, was assessed from vacuum samples, as herbivory by these groups could not be quantified in the field. Host plant preferences for Hemiptera were ascertained from Southwood & Leston (1959) and Dolling (1991).
Experimental plots measuring 150 × 150 cm were established in 1991 and maintained until 1999. Initially, six replicate plots were randomly assigned (across all plots not in blocks) to each treatment combination of nutrient addition and warming in a factorial design, but due to plot damage, only four replicates of each treatment were utilized in this analysis. Nutrient addition was intended to simulate both aerial deposition and enhanced rates of mineralization and decomposition in warmer soils. Inorganic fertilizer solution was applied with a watering can as 3 L solutions of NH4NO3 and KH2PO4 (10 : 10 : 12.6 g N : P : K m−2 year−1) on six occasions between snowmelt (late May) and early August, by which time annual growth of the dominant dwarf shrub species was completed. To avoid any risk of scorching, a rinse of c. 2 L water was applied after each nutrient addition in order to remove nutrients from contact with leaf surfaces. A supplementary precipitation manipulation, previously included in this experiment (e.g. Potter et al. 1995), was not included in this study. The ecological consequences of the additional water added with the fertilizer were controlled for in the first 5 years of the experiment during which no effects were detected (Parsons et al. 1994, 1995; Potter et al. 1995; Press et al. 1998).
An increase in air and soil temperatures was achieved using geodesic dome-shaped polythene open-top chambers (OTCs), which were approximately 90 cm from the apex to soil surface. These were constructed in the field at the end of spring each year, when snow cover was c.≤5% in all plots, and removed in early September when leaves of deciduous dwarf shrub species were dehiscing. A gap of c. 12 cm between the base of the polythene and the soil was incorporated into the design to facilitate ventilation and free movement of ground-dwelling invertebrates. OTCs elevated air temperature by 2–4 °C, with greatest warming at midday (Havström et al. 1993), and soils (50 mm below soil surface) by c. 1 °C (Wookey et al. 1993; Richardson 2000).
In addition to the air- and soil-warming effects of OTCs, other less desirable effects can arise. Those of relevance to this study are briefly dealt with here, while more comprehensive reviews can be found in Kennedy (1995) and Marion et al. (1997). The most significant artefact of the chambers is soil drying, and both this study and that of Robinson et al. (1995) estimated OTC soil moisture to be 27–29% less than controls. Additionally, the relative humidity at shrub canopy height was approximately 7% less in OTCs relative to controls (Richardson 2000), an effect also reported by Havström et al. (1993) and Wookey et al. (1993). Despite these microclimatic covariables, Hollister & Webber (2000) reported that plant responses to warming within OTCs corresponded well with responses to natural warming events.
The barrier effect of OTCs to wind may be of significance for insect herbivores, although the gap at the base and large opening at the top was some remedy to this, compared with closed-chamber designs (e.g. Strathdee et al. 1995). Similarly, the gap around the base will have allowed invertebrate predators to enter the plots, although the barrier effect of the chamber walls against parasitoids, indeed against the herbivores themselves, almost certainly reduced densities within OTCs (Richardson et al. 2000). This effect in warming experiments is unavoidable, unless costly infrared lamp warming devices are employed (e.g. de Valpine & Harte 2001), and these are currently rare in climate-change studies. Of the vertebrate herbivores that might have fed on the study vegetation, only hares were likely to have been excluded. Lemmings were observed within similar OTCs elsewhere in the region, and reindeer were rarely present in the lowland area around the field site during the summer months when OTCs were in place (S.J.R., personal observations).
Mean soil temperatures at 50 mm beneath the surface were +2 °C higher in plots receiving nutrients, with the greatest warming in the afternoon (maximum 2.9 °C) and the smallest effect early in the morning (minimum 1.6 °C). This occurred because this treatment stimulated above-ground biomass production and consequently increased the insulation of the soil relative to plots not receiving nutrients (Richardson 2000). Plots with OTCs that also received nutrients thus received an additional warming effect. Additionally, relative humidity at dwarf shrub canopy height was c. 10% greater in fertilized plots, again as a consequence of stimulated above-ground biomass production. Changes in microclimate associated with the increase in grass biomass represent the conditions this system will experience with global change, and thus are viewed here not as artefacts, but as components of the manipulation.
Total above-ground biomass of the dwarf heath vegetation was assessed in August 1991, after one growing season of treatments, and in August 1999 after nine growing seasons. Within the central 100 × 100 cm of each experimental plot, two or three 10 × 10 cm subsamples in 1991 and 1999, respectively, were taken at random and all above-ground biomass was removed. Material was sorted to dwarf shrub species or functional group, and oven-dried at 60 °C for 48 h before the mass was determined.
Annual shoot (leaf plus stem) quality, herbivory and herbivore abundance data were collected in 1998 after eight growing seasons of treatments. Where possible annual shoot chemistry and leaf herbivory were sampled from all four replicates of each treatment, but where inadequate numbers of annual shoots existed as a consequence of perturbation effects, replication was sometimes reduced (see d.f. in Table 3). For each of the three Vaccinium species, 10 randomly selected annual shoots were collected from the central 100 × 100 cm of each experimental plot into waxed envelopes on 5th July 1998, stored in a cool bag while being transported between the field and the laboratory, and frozen within 4 h at −80 °C. Samples were removed, subsequently freeze-dried, and milled (Glen Creston Mills, Stanmore, UK) to produce a homogenized sample with a particle size of c. 1 mm. The extraction and determination of phenolics followed the Folin–Ciocalteu method described by Waterman & Mole (1994), using the procedure in Kerslake & Hartley (1997). Total nitrogen was determined colorimetrically from acid digests using the sodium hydroxide continuous-flow method (Allen 1989). A known weight of material was digested at 370 °C with 3 mL of an acid mixture (33 g salicylic acid per L concentrated sulphuric acid) and approximately 1 g of 10 : 1 lithium sulphate : copper sulphate catalyst. Samples were digested for between 5 and 8 h until they cleared, during which time all nitrogen was converted to ammonium. The concentration of the latter was determined in a subsample of the digest using a colorimetric assay and a Tecator flow-injection analysis system (Tecator FIA, Foss-Tecator AB, Höganäs, Sweden).
Table 3. Annual shoot (leaf plus stem) nitrogen (mg g−1) and phenolic (% d.w.) concentrations and leaf herbivory (% leaf mass) for three Vaccinium species (mean ±1 SE). Data were tested for main and interaction effects of warming (OTC) and nutrient addition (F) by anova. F statistics are given with P in parentheses; treatment effects where P < 0.05 are indicated in bold
Species variable (d.f.)
Treatment mean ±1 SE
–OTC − F
–OTC + F
+OTC − F
+OTC + F
OTC × F
V. myrtillus (deciduous)
10.60 ± 2.13
8.51 ± 4.00
13.56 ± 1.31
19.57 ± 1.09
17.94 ± 3.66
18.28 ± 4.86
18.14 ± 1.62
14.65 ± 1.78
0.098 ± 0.088
0.639 ± 0.380
0.829 ± 0.488
3.742 ± 1.532
V. uliginosum (deciduous)
15.75 ± 0.43
16.52 ± 1.21
13.12 ± 0.45
13.20 ± 1.09
13.31 ± 1.98
13.19 ± 1.77
11.59 ± 1.63
13.52 ± 2.71
0.543 ± 0.237
1.090 ± 0.624
0.148 ± 0.133
1.659 ± 0.346
V. vitis-idaea (evergreen)
10.26 ± 0.72
12.43 ± 1.03
7.78 ± 0.19
10.16 ± 1.21
22.32 ± 1.51
21.51 ± 1.20
19.98 ± 0.41
17.16 ± 1.35
0.050 ± 0.050
0.053 ± 0.053
0.000 ± 0.000
0.177 ± 0.177
In order to determine herbivory levels, 10 annual shoots of the three Vaccinium species were randomly selected within the central 100 × 100 cm of each plot, and tagged at the start of the growing season. These annual shoots were harvested at peak biomass in August, and from each plot the proportion of leaves damaged was recorded. All leaves were photocopied onto acetate sheets, and the total leaf area was measured on a Delta-T Devices leaf area meter (Delta-T Devices, Cambridge, UK), with and without the areas consumed coloured in using permanent ink. Regression equations between leaf area and mass were employed to convert this estimate of leaf area loss to leaf mass consumed.
Insect herbivores were collected at peak biomass in early August, and the total number of Hemiptera from each plot was used for analysis. Nomenclature followed Chinery (1995) and Southwood & Leston (1959), in which the Hemiptera are subdivided into two suborders, the Heteroptera and the Homoptera. The assemblage of Hemiptera was sampled using a D-VAC vacuum sampler (D-VAC Co., Ventura, CA). The volume of air sampled by this device can be standardized according to the area and period sampled. The area of the D-VAC sampling nozzle was 0.03 m2, and each plot was sampled for 150 s in 30 five-second subsamples organized in a grid-like fashion within the central 100 × 100 cm of the plot. Herbivore density can thus be reported as individuals m−2 min−1. Although the D-VAC sampled each taxon with varying efficiency (Richardson 2000), comparisons in this study are made between treatments within groups, and no comment on the relative abundance of a group is made. D-VAC samples were transferred to heavy-duty polythene bags in the field, returned to the field station at Abisko within 4 h of collection, and frozen at −30 °C for at least 48 h. Samples were sorted at 40× magnification, and all Hemiptera were removed and stored in ethanol.
Above-ground biomass data were transformed such that they met the assumptions for parametric analyses. The biomass of lichens and bryophytes, and the biomass of each of the three Vaccinium species, were transformed by [√X + √(X + 1)] (Zar 1996); untransformed dwarf shrub and total biomass data each satisfied parametric assumptions; and grass biomass was rank-transformed. Although rank transformation provides weak estimations of interaction terms and a time estimate is lost in repeated measures (Conover & Iman 1981), it was considered more powerful than nonparametric tests for main effects within each sampling time. Time, nutrient addition, warming, and all interaction terms were tested for using a repeated-measures anova (SPSS Inc. Chicago, IL). Nitrogen and phenolic data were transformed using [√X + √(X + 1)] (Zar 1996), and were tested for treatment effects from two-way anovas (minitab v. 11, SPSS Inc.). The same transformation was employed for regression analyses (Zar 1996). The effects of the treatments on insect density and the level of herbivory in the growing season were tested for using GLM with Poisson distribution assumptions (s-plus, MathSoft, Seattle, WA).
plant community composition
Nine seasons of nutrient addition caused significant changes in species composition of the dwarf shrub vegetation, although the total amount of above-ground biomass remained unaltered (Fig. 1a; Table 1). Warming alone did not elicit a response from any functional group. Up to 95% of the total biomass was dwarf shrub, and after nine seasons of treatments there was significantly less above-ground biomass of this group in fertilized plots (Fig. 1b; Table 1). This effect was not apparent after one season of treatments, and is largely accounted for by the response of evergreen dwarf shrubs after nine seasons (Fig. 1d; Table 1). In 1999, the biomass of evergreen species in plots that received nutrients was 27% that of untreated controls. In contrast, there was no impact of either treatment on the total biomass of deciduous dwarf shrubs (Fig. 1c).
Table 1. Repeated-measures anovaF statistics (with P values in parentheses) for main and interaction effects of warming (OTC), nutrient addition (nutrients) and time (1991 and 1999 samplings) on above-ground biomass of sub-Arctic heath functional groups (d.f. = 1,12); data are indicated in bold where P < 0.05
OTC × nutrients
OTC × time
Nutrients × time
OTC × nutrients × time
Deciduous dwarf shrubs
Evergreen dwarf shrubs
The three dwarf shrub Vaccinium species studied in detail provided further insight into the response of the dwarf shrub functional group. The evergreen species V. vitis-idaea largely accounted for the response of all evergreen shrubs (Fig. 2; Table 2). However, the absence of a significant response by all deciduous dwarf shrubs collectively to either treatment was not supported by analysis of the two deciduous Vaccinium species in this study (Table 2). After just one season of treatments, the biomass of V. myrtillus was 92% greater in plots that were both warmed and fertilized (Fig. 2), although either warming or nutrient addition individually reduced biomass (Fig. 2). However, after nine seasons of treatments, biomass in the interaction plots was only 34% of the control biomass and either warming or nutrient addition individually had no effect on biomass (Table 2). There were no significant effects of either treatment on the biomass of the deciduous species V. uliginosum (Table 2).
Table 2. Repeated-measures anovaF statistics (with P values in parentheses) for main and interaction effects of warming (OTC), nutrient addition (nutrients) and time (1991 and 1999 samplings) on above-ground biomass of three Vaccinium dwarf shrub species (d.f. = 1,12); data are indicated in bold where P < 0.05
OTC × nutrients
OTC × time
Nutrients × time
OTC × nutrients × time
The greatest treatment impacts were on three subordinate groups of species: the grasses, the bryophytes and, to a lesser extent, the lichens. After nine seasons of treatments, there was 45 times more grass biomass, principally Calamagrostis lapponica (Wahlenb.) Hartm. and Deschampsia flexuosa (L.) Trin, in plots that had received nutrients (Fig. 1e; Table 1). Although grass biomass was also greater in warmed plots after nine seasons, this effect was not significant (Table 1). The biomass of lichens was significantly less in fertilized plots after nine seasons (Fig. 1f). The highly variable distribution of lichen biomass between plots that were warmed but did not receive nutrients (Fig. 1f) perhaps obscured a statistically significant effect of warming, as there was an apparent loss of biomass within this group in warmed plots (Fig. 1f). While treatments initially stimulated biomass production by bryophytes, after nine seasons there was less biomass in all treated plots (Fig. 1g), although this effect was not significant (P = 0.054) for warming (Table 1).
annual shoot quality and herbivory of vaccinium
Nutrient addition and warming elicited species-specific responses of nitrogen and phenolics for the three Vaccinium species examined. In contrast to biomass responses (above) there were significant warming effects on these variables. Warming affected the nitrogen concentration of both deciduous species, although both the magnitude and direction of responses were different. Warming increased foliar nitrogen by 73% in V. myrtillus, while for V. uliginosum nitrogen was 19% less in warmed plots relative to controls (Table 3). Warming had no effect on phenolics of either of these two deciduous species (Table 3), whereas warmed annual shoots of the evergreen species V. vitis-idaea contained 15% less phenolics than controls (Table 3). Nutrient addition did not significantly affect foliar nitrogen or phenolic concentrations of either deciduous species, while fertilized annual shoots of the evergreen species contained 25% more nitrogen than controls (Table 3).
Losses of leaf biomass to herbivores over one growing season were low, and ranged from <0.1% in the evergreen species to c. 0.5% from the two deciduous species (Table 3). While there was no effect of either treatment on the amount of leaf mass consumed from the evergreen species V. vitis-idaea, herbivory of both deciduous species was between four and six times greater in fertilized plots, and there was an additional, highly significant sixfold increase in herbivory with warming for V. myrtillus (Table 3).
Regression plots of leaf herbivory against nitrogen, phenolics and plant abundance suggested that annual shoot chemical composition determined more of the interspecific and between-treatment variation in herbivory than plant abundance. A positive regression between herbivory and annual shoot nitrogen accounted for 59% of the variation in leaf herbivory, while the negative relationship between phenolics and herbivory was not significant, and accounted for only 26% (Fig. 3). The above-ground biomass (abundance) of each of the three Vaccinium species accounted for none of the variation in leaf herbivory. Individual treatments were scattered within these relationships with no obvious treatment-induced shifts (Fig. 3).
the insect herbivore community
The community of herbivorous Hemiptera was affected by the two treatments, and the greatest responses were observed from families that fed on either bryophytes or grasses (Table 4). The addition of nutrients resulted in the establishment of Delphacidae, a gramnivorous family, which was otherwise absent from the heath (Fig. 4). Additionally, there was a positive effect of nutrient addition on the abundance of Cicadellidae (Fig. 4), a family that includes a large proportion of gramnivorous species. However, this increase was marked by a negative interaction effect with the warming and OTC treatment (Fig. 4; Table 4). The Tingidae in this study consisted of a single species, Acalypta nigrina (Fallén), which is bryophagous. The abundance of this species was significantly less in all treated plots (Fig. 4), and in four out of the eight plots that received nutrients it was completely absent. The two groups in this herbivore community that feed largely on dwarf shrubs (Psylloidea and Aphidoidea) did not respond to the two treatments (Fig. 4; Table 4).
Table 4. anovaF statistics (with P values in parentheses) for main and interaction effects of warming (OTC) and nutrient addition (nutrients) on the density of five hemipteran groups, and of total Hemiptera, present on a dwarf shrub heath vegetation in early August (peak biomass) 1998 (d.f. = 1,12)
OTC × nutrients
The direct and indirect impacts of warming on above-ground biomass of a regionally important vegetation type, dwarf shrub heath, were quantified after one and nine seasons of simulated change, and the consequences of plant community change for insect herbivore density and herbivory of three dominant species were assessed. Large-scale responses by grasses and bryophytes had significant implications both for other plant groups and for insect herbivores, while the responses by dominant dwarf shrub species were of relatively smaller magnitude and consequence.
plant community change
This is one of very few long-term factorial nutrient and warming field manipulations in the sub-Arctic that have examined whole-community responses. Two other directly comparable studies have been undertaken at an open heath site in the Abisko area of Scandinavia (Jonasson et al. 1999; Graglia et al. 2001), and a tussock tundra site in Alaska (Chapin et al. 1995; Shaver et al. 2001). All three studies observed marked changes in plant community composition with nutrient addition, with a large reduction in moss biomass common to all. Chapin et al. (1995) and this study both observed that total above-ground biomass remained unaltered despite marked changes in species composition. In all three studies, one species or group of species increased with nutrient addition, but these were not consistent between experiments. In a meta-analysis of high-latitude perturbation experiments, Dormann & Woodin (in press) identified grasses as the only plant functional group to respond significantly to nutrient addition, a result derived from earlier results of our experiment (Press et al. 1998) and from the open heath experiment in Abisko (Graglia et al. 2001). However, it was the deciduous dwarf shrub Betula nana that responded most to fertilization at the tussock site in Alaska (Chapin et al. 1995; Bret-Harte et al. 2001).
These studies suggest that competition for light in tundra vegetation is intensified by the amelioration of nutrient limitation brought about by fertilization. Fast-growing species with adequate plasticity to respond to the nutrients provided compete more effectively for light at the expense of slower-growing species of lower stature (Bret-Harte et al. 2001). This shading effect has the greatest impact on bryophytes and lichens. For these groups, indirect effects of climate change, mediated through increasing vascular plant biomass, may be of greater significance than the short-term positive responses to nutrient addition we observed after one growing season. Indeed, a sophisticated analysis by Cornelissen et al. (2001) demonstrated that the mid- to long-term decline of lichen biomass across a suite of high-latitude perturbation studies was strongly correlated with increased vascular plant biomass. Our conclusion regarding the impacts of long-term nutrient addition includes the suggestion that the shading effect, currently pronounced for cryptogams, will also affect evergreen dwarf shrub species. Consequently, the loss of species richness reported from this site after five seasons (Press et al. 1998) will be exacerbated.
The conservative and often transient responses by plant biomass to warming corresponded well with other studies (e.g. Hartley et al. 1999; Graglia et al. 2001). At a similar experimental site in Abisko, Hartley et al. (1999) reported that warming by either OTCs, or OTCs in combination with buried heating cables, had no effect on cover of the three Vaccinium species common to both studies. We can also compare our results with three published meta-analyses of warming experiments. Rustad et al. (2001) examined warming experiments throughout temperate and Arctic regions, and concluded that plant productivity would increase by 15–23%, with greater responses at high latitudes. Second, Arft et al. (1999) examined 4 years of warming responses from a series of polar experiments in which herbaceous species such as grasses increased with warming, while increased growth of woody species was only transient. Lastly, there was a significant positive biomass response to warming by both deciduous and evergreen dwarf shrubs and grasses in the meta-analysis of Dormann & Woodin (in press). The predictions of these meta-analyses are that warming would increase total above-ground biomass, and that grasses and evergreen dwarf shrubs would account for this increase. In this experiment, although there was a non-significant 24% increase in above-ground biomass in warmed plots that corresponds well with the expectation of Rustad et al. (2001), there was no clearly positive response to warming by grasses, deciduous or evergreen dwarf shrubs.
The impact of herbivory on plant performance per se has not been quantified in this study; rather, assessments of leaf damage, annual shoot quality and herbivore density are used to provide an indication of how the interaction between plants and herbivores will respond to environmental change.
Changes in annual shoot chemical composition and herbivory were frequently species-specific, yet two simple patterns of response did emerge. First, herbivory of both deciduous species increased with nutrient addition, although this occurred without the expected changes in nitrogen and phenolics. Second, warming affected the nitrogen concentration in annual shoots of all three species, but herbivory coresponded in only one instance. It was anticipated that fertilization would increase the concentration of nitrogen in leaf tissues, reduce allocation to phenolic compounds (Bryant et al. 1983), and consequently increase losses to herbivory (e.g. Nordin et al. 1998). The expectation with warming was less clear, as warming can dilute leaf nitrogen through stimulated growth (e.g. Michelsen et al. 1996) or, theoretically, produce responses similar to fertilization (through stimulated soil nutrient mineralization). However, we found only a single example of a treatment-induced shift in plant quality generating a corresponding change in herbivory, despite the fact that our regression analysis (Fig. 3) suggests that plant quality is a determinant of herbivory in this system. Furthermore, the warming response by V. myrtillus, in which both tissue quality and herbivory increased, contrasts with the study of Laine & Henttonen (1987) who reported that tissue quality in this species decreased with warming and was lower in warm years, when the direct benefit to insect herbivores would be greatest. Hartley et al. (1999) similarly detected a leaf nitrogen dilution effect with warming in V. myrtillus, although this only occurred in 1 of 5 years examined. In cold systems where insect mobility is strongly limited by temperature, Niemeläet al. (1982) suggested that small changes in plant quantity would have significant consequences for insect herbivores. However, this was not corroborated by our regression analysis of dwarf shrub abundance and herbivory (Fig. 3). Notably, the c. 90% decrease in V. vitis-idaea biomass in warmed plots that received nutrients had no repercussions for herbivory. The unexpected and species-specific results from this study underline the difficulty of predicting how herbivory rates may alter under scenarios of global change.
the abundance of hemiptera
The impact of nutrient addition and warming on the abundance of insect herbivores will be a consequence of both direct impacts of higher temperatures on rates of development (Speight et al. 1999), and indirect impacts manifested through plants, in this study via changes in plant abundance.
The 40-fold increase in grass biomass and almost total loss (93% decrease) of bryophytes after nine seasons of nutrient addition had major impacts on the insect herbivore community, notably the appearance in fertilized plots of two grass-feeding groups, the Delphacidae and the Cicadellidae (Dolling 1991), and the loss of moss-feeding individuals of A. nigrina (Tingidae) (Southwood & Leston 1959). Although bryophyte nitrogen concentrations were higher in fertilized plots at this site (for example a 74% increase in Polytrichum commune; Burch 2001), the impact of host plant decline was evidently of greater importance. Neither of the shrub-feeding groups (Aphidoidea and Psylloidea) responded to nutrient addition, despite a 52% decrease in host plant abundance. The apparent insignificance of shrub biomass for shrub-feeding herbivore density and levels of herbivory (see above) may be due to the smaller percentage change in abundance of shrubs compared with the changes in grasses and bryophytes. Shrubs are the dominant component of the plant community, and so remain a relatively abundant resource under all treatments.
The two-thirds decrease in A. nigrina density with warming may be a consequence of reduced bryophyte host plant abundance, and possibly a direct warming impact that may have enhanced water loss from an insect typical of damp bryophyte vegetation. As A. nigrina moves along the ground and is frequently flightless (Southwood & Leston 1959), the negative warming effect is unlikely to be a barrier artefact of the OTC, as gaps at the bottom would have facilitated its passage between treated and untreated vegetation. However, the negative interaction effect of warming and nutrient addition on the abundance of Cicadellidae is suggested to be a result of an OTC barrier effect. Grass biomass increased within fertilized OTCs, and the low occupancy of this host plant ‘space’ by Cicadellidae is not thought to be ecological, but rather a negative barrier effect of the OTC per se (Richardson et al. 2000).
The large increase in abundance of an Arctic aphid species following warming reported by Strathdee et al. (1995) was greatest at the altitudinal and latitudinal limits of the species, possibly where top-down regulation was least and climatic constraints were greatest. A similar phenomenon was discussed by Whittaker & Tribe (1998) for a British homopteran. Although we did not record any warming responses at this sub-Arctic site, where species may be well within their distributional limits, we are cautious to invoke top-down regulation within OTCs where, with the exception of ground-dwelling predators, predation is likely to have been excluded. It is difficult to account for the paucity of warming responses, and further manipulative field experiments are needed in order to understand the impacts of warming on mobile organisms.
subordinate species, functional diversity and community change
In this study we proposed that responses to simulated climate change at the second trophic level, the insect herbivores, would be ‘driven’ by those of the dominant plant species, in accordance with the mass ratio theory (Grime 1998). Although dwarf shrubs accounted for between 60 and 95% of understorey community biomass, the greater responses to our climatic manipulations by two subordinate groups – grasses and cryptogams – had the greatest impacts on the responses of insect herbivores to the treatments. The suggestion that subordinate species may contribute to ecosystem function, either directly (George & Bazzaz 1999a; 1999b; Walker et al. 1999) or via enhanced functional diversity (Tilman et al. 1997; Hodgson et al. 1998; Diaz & Cabido 2001), has recently attracted considerable attention. Field manipulation studies that relate these aspects of diversity to ecosystem processes are a priority in ecological research (Grime 1997; Wardle et al. 1997), especially in the application of ecological theory to our understanding of ecosystem responses to disturbances and environmental change.
We thank the Director and staff of the Abisko Scientific Research Station for allowing the use of facilities. We gratefully acknowledge the support of NERC (Arctic Terrestrial Ecology Programme) and ESRC (Global Environmental Change Programme) for their support of the earlier phases of this study. We are grateful to Professors Terry Callaghan and John Lee for allowing us to use the experimental plots as well as for helpful discussions and advice. The manuscript was significantly improved by comments from René van der Wal, David Gibson and an anonymous referee. Wendy Ruscoe and Richard Duncan provided statistical advice. Patrick Ericsson, John Jackson, Helen Temple and Jo Wilding are gratefully acknowledged for their assistance in the field. S.J.R. was supported by a NERC studentship and a Swedish Royal Academy of Sciences scholarship.