Growth and community responses of alpine dwarf shrubs to in situ CO2 enrichment and soil warming


Author for correspondence:
Melissa A. Dawes
Tel: +41 81 417 0271


  • Rising CO2 concentrations and the associated global warming are expected to have large impacts on high-elevation ecosystems, yet long-term multifactor experiments in these environments are rare.
  • We investigated how growth of dominant dwarf shrub species (Vaccinium myrtillus, Vaccinium gaultherioides and Empetrum hermaphroditum) and community composition in the understorey of larch and pine trees responded to 9 yr of CO2 enrichment and 3 yr of soil warming at the treeline in the Swiss Alps.
  • Vaccinium myrtillus was the only species that showed a clear positive effect of CO2 on growth, with no decline over time in the annual shoot growth response. Soil warming stimulated V. myrtillus growth even more than elevated CO2 and was accompanied by increased plant-available soil nitrogen (N) and leaf N concentrations. Growth of Vaccinium gaultherioides and E. hermaphroditum was not influenced by warming. Vascular plant species richness declined in elevated CO2 plots with larch, while the number of moss and lichen species decreased under warming.
  • Ongoing environmental change could lead to less diverse plant communities and increased dominance of the particularly responsive V. myrtillus in the studied alpine treeline. These changes are the consequence of independent CO2 and soil warming effects, a result that should facilitate predictive modelling approaches.


Atmospheric CO2 concentrations are predicted to reach 730–1020 ppm by the year 2100 and, as one consequence of this change, the global mean surface air temperature is expected to increase by 1.8–4.0°C during the same period (Meehl et al., 2007). Elevated CO2 concentrations might enhance carbon uptake by plants through a direct stimulation of photosynthesis, and responses might be particularly strong at high elevation where atmospheric pressure, and therefore CO2 partial pressure, are lower (Körner, 2003). Similarly, ongoing global warming is already documented to have particularly large ecological impacts on high-latitude and high-elevation regions where plants grow close to their low temperature limit (Walther, 2003; Dorrepaal et al., 2009). For these reasons, research on environmental change in alpine and arctic ecosystems has increased substantially in recent years, and has included both manipulation experiments (see review by Dormann & Woodin, 2002) and observations of natural vegetation change (e.g. Wilson & Nilsson, 2009; Hill & Henry, 2011).

Elevated CO2 concentrations have been experimentally applied to high-latitude and high-elevation vegetation in only a few studies. Field experiments in a late successional alpine grassland in the Swiss Central Alps (Körner et al., 1997), in a forest heath ecosystem in subarctic Sweden (Gwynn-Jones et al., 1997) and in the wet tussock tundra in Alaska (Tissue & Oechel, 1987) indicate that responses of plant growth and biomass to rising CO2 concentrations are generally small or nonexistent but that co-occurring species often vary in their responses (Dormann & Woodin, 2002). More broadly, CO2 enrichment studies of various plant types conducted in relatively natural growth conditions have revealed that other variables, such as climate and availability of nutrients or water, can influence CO2 effects (e.g. Niklaus & Körner, 2004; Dawes et al., 2011) and that any initial growth stimulation often declines after a few years (Körner, 2006). In ecosystems where nitrogen (N) availability is low, this temporal response pattern can be caused by increasing N limitation over time as a result of increased N immobilization in plant biomass and long-lived soil organic matter and of enhanced microbial activity (Luo et al., 2004). Studies lasting several years are clearly important for determining temporal dynamics of plant responses to CO2 enrichment and interactions with other environmental factors.

In contrast to the small number of CO2 manipulation experiments, many relatively long-term warming experiments have been completed in (sub)arctic (Chapin & Shaver, 1985; Parsons et al., 1994; Shevtsova et al., 1997; Aerts et al., 2009) and alpine (Harte & Shaw, 1995; Kudo & Suzuki, 2003; Kudernatsch et al., 2008) ecosystems. Increased plant growth and biomass production with warming were observed in several experiments, although responses were often small or transient and varied across species and study sites (Rustad et al., 2001; Walker et al., 2006). Many high-latitude and high-elevation environments are characterized by low availability of soil nutrients, particularly N, as a result of low-temperature constraints on decomposition and mineralization (Nadelhoffer et al., 1992; Körner, 2003), and higher soil temperatures tend to accelerate these processes in systems that are not water limited (Cornelissen et al., 2007; Kammer et al., 2009). Therefore, plant growth responses to warming might be caused by enhanced nutrient availability in addition to a direct effect of increased rates of photosynthesis at higher temperatures. Similar to experimental CO2 enrichment, warming experiments lasting several years are needed to elucidate responses at the individual species, community and ecosystem scales, especially as shifts in competitive interactions and plant–soil feedback processes can alter responses over time (Wookey et al., 2009).

Despite clear predictions that rising CO2 concentrations will be accompanied by increased temperatures, relatively few studies combining CO2 enrichment and experimental warming have been conducted within intact systems (Beier, 2004). In a glasshouse study conducted in tussock tundra vegetation (Toolik Lake, AK, USA), CO2-induced stimulation of net primary productivity (NPP) lasted only one season when applied alone but was sustained over three growing seasons when combined with a 4°C temperature increase and unaltered water availability (Oechel et al., 1994). However, most field studies of natural vegetation combining elevated CO2 and warming have been conducted in grassland systems and have yielded conflicting results regarding how the combined changes influence plant productivity, community composition and N cycling (Shaw et al., 2002; Hovenden et al., 2008; Engel et al., 2009; Dijkstra et al., 2010). While experiments combining CO2 enrichment and nutrient addition have been conducted in alpine environments (Körner et al., 1997; N. Inauen, pers. comm.), to our knowledge no previous in situ studies of high-elevation systems have simultaneously manipulated CO2 concentration and temperature.

Dwarf shrubs are a major component of arctic and alpine plant communities, and shifts in the growth and reproductive output, abundance and/or distribution of these species under environmental change are likely to have important ecological consequences. In this study, we determined the effects of 9 yr of free air CO2 enrichment (FACE) and 3 yr of soil warming on three dominant ericaceous dwarf shrub species, deciduous Vaccinium myrtillus (bilberry) and Vaccinium gaultherioides (northern bilberry) and evergreen Empetrum hermaphroditum (crowberry), growing in the understorey of treeline trees in the Swiss Alps. Studying three co-occurring species allowed us to explore whether individual dwarf shrub species respond differently to the manipulations. We hypothesized that: (1) any initial stimulation of above-ground growth of these dwarf shrub species in response to CO2 enrichment would decline over time; (2) soil warming would lead to enhanced dwarf shrub growth, reflecting a direct stimulation and/or an indirect benefit from increased soil N availability; (3) soil warming would alleviate temperature or nutrient constraints on the growth response to elevated CO2, yielding a positive interactive effect of the two experimental treatments on dwarf shrub growth; and (4) species-specific responses of understorey vegetation to the experimental treatments would lead to shifts in species composition in the experimental plots.

Materials and Methods

Site and experimental design

The study site is located at Stillberg, Davos in the Central Alps, Switzerland (9°52′E, 46°46′N). The CO2 enrichment and soil warming experiment covers an area of 2500 m2 and is situated on a northeast-exposed 25–30° slope at 2180 m asl, corresponding to or slightly above the natural climatic treeline (Hättenschwiler et al., 2002). The site is located within a 5-ha long-term afforestation research area where tree seedlings were planted into the intact dwarf shrub community in 1975 by the Swiss Federal Institute for Forest, Snow and Landscape Research (WSL). During the experimental period from 2001 to 2009, the mean annual precipitation was 1175 mm and the mean annual air temperature was 2.5°C, with February (mean −5.2°C) being the coldest month and July (mean 10.5°C) the warmest month (WSL climate station at 2090 m asl).

The experiment consists of 40 hexagonal 1.1-m2 plots, 20 with a Pinus mugo ssp. uncinata (DC.) Domin (mountain pine) individual in the centre and 20 with a Larix decidua Mill. (European larch) individual in the centre. These trees are now 39 (pine) and 37 (larch) yr old but are < 3.5 m tall and have a stem basal diameter of < 10 cm. The trees are sparsely distributed and do not form a closed canopy; thus, each plot contains a single tree surrounded by a dense cover of understorey vegetation including the dominant dwarf shrub species Vaccinium myrtillus L., Vaccinium gaultherioides Bigelow (group V. uliginosum agg.) and Empetrum nigrum ssp. hermaphroditum (Hagerup) Böcher, plus several herbaceous (e.g. Avenella flexuosa, Gentiana punctata, Homogyne alpina, Leontodon helveticus and Melampyrum pratense) and nonvascular species. For the dwarf shrub species targeted in this study, regeneration within the existing heath zone at the treeline occurs mainly by vegetative spread (Körner, 2003).

The CO2 enrichment experiment started in early June 2001, at which point the 40 plots were assigned to 10 groups of four neighbouring plots (two larch and two pine trees per group) in order to facilitate the logistics of CO2 distribution and regulation. Half of these groups were randomly assigned to an elevated CO2 treatment (mean concentration ± 1 SD 575 ± 52 ppm from 2001 to 2009) while the remaining groups served as controls and received no additional CO2 (c. 380 ppm). During daytime hours only, the system released pure CO2 gas through laser-punched drip irrigation tubes hung vertically around a hexagonal frame surrounding each plot. Interruptions in CO2 release as a result of adverse weather conditions or technical failure meant that plants received supplementary CO2 for 73–87% of the seasonal daytime-only treatment periods. The set-up and performance of the CO2 enrichment facility have been described in detail previously (Hättenschwiler et al., 2002; Handa et al., 2006; Dawes et al., 2011).

In spring 2007, one plot of each tree species identity was randomly selected from each of the 10 CO2 treatment groups and assigned a soil warming treatment, yielding a balanced design with a replication of five individual plots for each combination of CO2 concentration, warming treatment and tree species. Soil warming was applied during the 2007–09 growing seasons using heating cables laid on the ground surface underneath the dwarf shrub layer (Hagedorn et al., 2010). As the plots contained trees over 2 m tall and were located on steep, rocky terrain, this method was more feasible and required less disturbance than the use of open top chambers (OTCs) or infrared radiators and additionally allowed greater soil warming than OTCs. Heating was turned on immediately after snowmelt and turned off just before the site was covered in snow for the winter, thereby avoiding an interaction between soil temperature and snow cover duration. Warming increased the growing season mean soil temperatures at 5 cm depth by 3.9 ± 0.3°C in 2007, 4.4 ± 0.5°C in 2008 and 3.1 ± 0.4°C in 2009 (mean ± 1 SE; = 10). Air temperature was increased at up to 20 cm above the ground (0.9 ± 0.1 K) but no temperature difference was detected at 50 cm height (Hagedorn et al., 2010). Warming also had a slight drying effect which was most pronounced in late summer of each year. Averaged over the growing season, volumetric soil water content (0–10 cm depth) was reduced by c. 15% and air humidity (10 cm height) was reduced by c. 10%, irrespective of CO2 treatment (no effect) and tree species identity (Hagedorn et al., 2010). Nonetheless, soil water potential in both warmed and unwarmed plots was always higher (less negative) than −300 hPa, indicating high water availability. See Fig. S1 in the Supporting Information for details of how the warming treatment influenced soil temperature and moisture.

Plant growth and plot species composition

During the autumns of 2002 (second season of CO2 enrichment) to 2009 (final season of study), after seasonal shoot elongation of the dwarf shrubs was complete, we measured the length of the new shoot increment on the longest branch of five to seven randomly selected ramets in each plot. These shoots experience less self-shading than shoots closer to the ground and are therefore less likely to have reduced growth as a result of competition for light among shoots within an individual ramet. Further, a study of dwarf shrub biomass allocation conducted outside our experimental plots indicated that this measurement is highly correlated with gross annual above-ground biomass production on a given ramet for our three study species (S. Wipf et al., unpublished). Shoots produced in 2002 were only measured in half of the plots, with a replication of five for each combination of CO2 treatment and tree species. In all years, ramets located within 10 cm of the plot border were not measured in order to avoid potential edge effects. We additionally measured the overall understorey vegetation height (mean of three random point measurements) and the maximum ramet height of V. myrtillus, V. gaultherioides and E. hermaphroditum in each plot during late summer in 2008 and 2009. These measurements are positively correlated with total above-ground biomass per unit area (understorey height) and per ramet (individual species height) at our research site (S. Wipf et al., unpublished). During the peak of the growing season in 2005 and 2009, we recorded all vascular and nonvascular (moss and lichen) species present in each plot to document changes in species composition and richness.

Leaf morphology

During early August 2006–09, c. 50 leaves of V. myrtillus and of V. gaultherioides were harvested from each plot to measure leaf traits. Empetrum hermaphroditum leaves were not sampled because individual leaf area and mass could not be measured accurately for the small, folded leaves. Leaves were collected from the whole canopy, with samples evenly distributed across the distinct ramets present throughout a given plot but again avoiding ramets near the plot edges. Leaves were scanned within 12 h of harvesting and mean leaf area was calculated using ImageJ version 1.43 k (measurements not available from 2008; Rasband, 1997–2008). Leaves were subsequently dried at 60°C for at least 24 h and weighed to obtain the average dry leaf mass and specific leaf area (SLA) per plot for each species.

Soil inorganic N and leaf N concentration

Treatment effects on plant-available soil N were assessed by measuring the soil inorganic N pool size at 0–5 cm depth in August 2004, 2007 and 2009. In each plot, six soil cores with a diameter of 2 cm were taken and the bulked sample was extracted with 0.5 M K2SO4. We additionally sampled soil water at 3–7 cm soil depth once per month during each vegetation period, using two ceramic suction cups (SoilMoisture Equipment Corp., Santa Barbara, CA, USA) per plot and applying a suction of 400 hPa overnight (Hagedorn et al., 2008). Soil extracts and soil solution were analysed for NH4+ by colorimetry using an automated flow injection analysis (FIAS-300; PerkinElmer Inc., Waltham, MA, USA) and for NO3 by ion chromatography (DX-120; Dionex, Sunnyvale, CA, USA). We additionally measured leaf N concentration for the two Vaccinium species to estimate treatment effects on plant N status (Marschner, 1995). Dried Vaccinium leaves were ground, as one bulk sample per plot for each species, and N concentration (mg N per g dry mass) was measured using a C-N analyser (EA-1110; Carlo Erba, Milan, Italy).

Plot-level environmental conditions

Light and soil moisture conditions were measured at the plot level to investigate whether understorey plant responses to CO2 enrichment and soil warming were driven by changes in these environmental variables associated with the treatments (indirect effects). Hemispherical photographs taken in each plot during the seasonal peak of leaf area in 2008 were used to quantify the amount of shade experienced by understorey plants near the end of the experiment when tree canopy size varied most among plots. We defined ‘canopy shading’ as the percentage of sky obstructed by the tree trunk and canopy, trees in the area surrounding the plot, and topographic elements (see Fig. S2 for details). On average, canopy shading was slightly greater for plots with larch than for those with pine (for a summary of plot conditions, see Table S1). Additionally, stimulation of larch growth under elevated CO2 meant that tree canopy cover tended to be greatest for elevated CO2 plots containing a larch tree (Dawes et al., 2011). Seasonal mean soil volumetric water content was calculated for each plot during 2007–2009 to quantify inter-plot variation in soil moisture, including drying associated with the warming treatment (see Fig. S1 for methods). Snowmelt date in spring and plot topography (slope or ridge) were also estimated for each plot and considered in statistical tests. However, these two parameters did not vary significantly between treatment groups (Table S1) or show a correlation with response variables and are not presented in the results below.

Statistical analysis

Treatment effects on the growth of each individual dwarf shrub species, on understorey vegetation height, and on soil inorganic N were tested with Type I analysis of (co)variance, using repeated measures linear mixed effects models fitted with REML. To test the CO2 enrichment effect, we used a model including CO2 concentration, plot tree species identity and their interaction as between-subject fixed factors, and treatment year (categorical variable) and all two- and three-way interactions with year as within-subject fixed factors. For the 3 yr when both CO2 enrichment and soil warming were applied (2007–09), likelihood ratio tests indicated that none of the interactions between the two treatments contributed significantly to model fits for any of the parameters measured (Table S2). We therefore pooled warmed and unwarmed plots for tests of the CO2 effect. We used a similar statistical approach to test soil warming effects, with temperature treatment replacing CO2 treatment in the models. Soil warming effects on understorey vegetation height and total ramet height were not analysed. We included values averaged over 2005 and 2006 as a pretreatment covariable in tests of shoot growth responses to warming, 2006 values in models involving leaf morphology and N concentration, and 2004 values in soil inorganic N analysis. We used a single model to test effects of CO2 enrichment and soil warming on species richness in 2009, using values from 2005 as a covariable even though some CO2 effects might have occurred before that point.

Potential influences of canopy shading (measured in 2008) and soil moisture (measured in 2007–09) on response variables (measurements from 2007 to 2009 only) were tested using linear regression, including pretreatment predictors in tests relevant to the warming treatment. For species richness, the influence of understorey vegetation height and individual dwarf shrub species ramet height (averaged over 2008–09) were also tested. To determine if dwarf shrub shoot growth responses to soil warming were associated with a change in N availability, we regressed shoot increment length against soil inorganic N pool size and against leaf N concentration for each individual treatment year (2007–09), standardizing both variables to pre-warming values.

For all analyses, random effects associated with specific intercepts for the 10 CO2 treatment groups and 20 warming treatment groups were excluded because they were very small and nonsignificant (Zuur et al., 2009). Therefore, repeated measures tests only included random effects for each individual plot, the unit of measurement, and species richness was analysed without random effects using generalized least squares. Models assuming a normal error distribution were used for all parameters except nonvascular plant species richness, where a Poisson distribution provided a better fit (Venables & Ripley, 2002). Response variables were log-transformed in some cases to improve assumptions of normality and homoscedasticity. For all statistical tests, effects were considered significant at < 0.05. Because of relatively low replication and therefore statistical power, we designated P-values ≥ 0.05 but < 0.10 as marginally significant. All analyses were performed using R version 2.11.1, and mixed models were fitted using the nlme package (Pinheiro et al., 2008; R Development Core Team, 2010).


Annual shoot growth and total ramet height

Repeated measures tests of the CO2 effect on annual shoot growth during 8 yr of treatment yielded a significant stimulation of 12% in V. myrtillus (F1,36 = 7.9, < 0.01), averaged across tree species identities, temperature treatments and years (Fig. 1a). This stimulation was relatively consistent, with no indication of a decline in the signal over time (> 0.9 for CO2 × year interaction). Plot tree species identity also significantly influenced V. myrtillus shoot growth, with longer annual shoot increments in plots with larch than in those with pine, irrespective of CO2 treatment (F1,36 = 5.2, = 0.03). For V. gaultherioides, the interaction between CO2 treatment and tree species identity was significant (F1,36 = 5.9, = 0.02), with a slight negative response to elevated CO2 in plots with larch and a slight positive response in plots with pine (Fig. 1b). Empetrum hermaphroditum shoot growth did not show an effect of CO2 enrichment or tree species, and no interactions were significant (> 0.16; Fig. 1c). Annual shoot growth of all three species varied significantly among individual treatment years (< 0.01) but did not show a clear trend over time (see Table S3 for ANOVA results). Canopy shading did not significantly influence shoot growth of any species in 2007–09 (Table S4). Further, a harvested subsample of annual shoot increments from three age classes demonstrated that, for all three species, longer shoot increments had greater biomass and the length to mass ratio was not significantly influenced by the degree of canopy shading (see Fig. S3 for details). These results indicate that enhanced V. myrtillus shoot increment length under elevated CO2 was caused by direct stimulation rather than by a change in light conditions.

Figure 1.

CO2 effect on the length of annual shoot increments during 8 yr of enrichment. The CO2 effect was calculated as the ratio of the mean shoot increment length for all elevated CO2 plots to the mean shoot increment length for all ambient CO2 plots (= 5 for 2002 and = 8–10 for all other years). Error bars represent ± 1 SE of the ratio, estimated according to Gelman & Hill (2007). Plots are separated by tree species identity (light grey circles, under pine; dark grey circles, under larch), and means are pooled across warmed and unwarmed plots for 2007–09. A dotted line is drawn where the elevated to ambient CO2 ratio is 1.

Soil warming had a positive effect on V. myrtillus shoot growth (F1,35 = 12.4, < 0.01; Fig. 2a). Relative to the 2 yr preceding warming, mean annual shoot increment length was 31% greater in warmed plots than in unwarmed plots (averaged across tree species identities, CO2 concentrations and years). Neither V. gaultherioides nor E. hermaphroditum shoot growth responded significantly to warming, although the latter species showed a trend of stimulation in 2009 (Fig. 2b,c). Shoot increment length measured in 2005 and 2006 influenced growth of each species during the three seasons of warming (< 0.04), and the strength of this relationship did not vary significantly with year or temperature treatment. No interactions between temperature treatment, tree species and year were significant (see Table S3 for ANOVA results). Soil moisture did not significantly influence shoot growth of any species, implying that slightly reduced soil moisture under warming was not responsible for responses to the treatment (Table S4).

Figure 2.

Soil warming effect on the length of annual shoot increments in 2007–09. The warming effect was calculated as the ratio of the mean shoot increment length for all warmed plots to the mean shoot increment length for all unwarmed plots (= 20 for Vaccinium myrtillus and Vaccinium gaultherioides; = 15–19 for Empetrum hermaphroditum). Error bars represent ± 1 SE of the ratio, estimated according to Gelman & Hill (2007). Means are pooled across plot tree species identities and CO2 treatment groups. Pre-warming ratios (2005 and 2006) are shown in the shaded region. A dotted line is drawn through the average of these two points, indicating the mean warmed to unwarmed ratio before treatment began.

Average understorey vegetation height was positively correlated with the maximum ramet height of V. myrtillus, generally the most abundant species in the plots. Both parameters were significantly enhanced in plots exposed to several years of CO2 enrichment ( 0.01; Fig. 3a,b), irrespective of tree species identity. There was a marginally significant CO2 × tree species interactive effect on V. gaultherioides ramet height (F1,36 = 4.1, = 0.05), with a positive response in plots with pine but a slight negative response in plots with larch, and ramet height was lower in plots with larch than in those with pine (F1,36 = 4.3, = 0.05; Fig. 3c). Empetrum hermaphroditum ramets were taller in plots with pine than in those with larch (F1,32 = 5.4, = 0.03) but showed no effect of CO2 enrichment (Fig. 3d). Ramet height of E. hermaphroditum showed a slight negative relationship with canopy shading (= 0.09), which might have explained the tree species effect, whereas no other height variables were significantly influenced by shading (Table S4). Overall, these results are consistent with patterns observed for shoot increment length during years 2 to 9 of CO2 enrichment, and they demonstrate that sustained responses in shoot growth yielded cumulative effects at the plot level (see Table S3 for dwarf shrub ANOVA results).

Figure 3.

CO2 effect on the total height of understorey vegetation overall and on the total ramet height of individual dwarf shrub species. Measurements for each plot from 2008 to 2009 were averaged to calculate the values shown here. Left and right panels show averages for plots shared with larch and pine, respectively, and values were pooled across warmed and unwarmed plots. (a) Mean plot understorey vegetation height for each of the two CO2 treatment groups, ± 1 SE (A, ambient CO2; E, elevated CO2; = 10). (b–d) Maximum ramet height of each dwarf shrub species, averaged over each of the two CO2 treatment groups, ± 1 SE (= 10 for Vaccinium myrtillus and Vaccinium gaultherioides; = 8–10 for Empetrum hermaphroditum).

Leaf morphology

Vaccinium myrtillus dry mass and area per individual leaf showed a marginally significant positive response to elevated CO2 (Table 1a; mean values for individual years shown in Table S5). Vaccinium myrtillus leaf area but not mass increased with increasing canopy shading, so it is possible that enhanced leaf area was at least partially a response to slightly reduced light conditions in elevated CO2 plots (Table S4). By contrast, neither leaf mass nor area of V. gaultherioides responded significantly to CO2 enrichment or to the amount of shading (Tables 1b, S4). SLA of both species was unaffected by CO2 enrichment (> 0.19), meaning that V. myrtillus leaf size increased but the area to mass ratio did not change significantly. Greater canopy shading was associated with higher SLA for both species (Table S4), which probably contributed to higher SLA in plots with larch than in those with pine, and V. myrtillus consistently had a higher SLA than V. gaultherioides (Tables 1, S5). Under soil warming, V. myrtillus showed only very limited evidence of a change in leaf morphology (Table 2a). There was a significant temperature × year interaction for V. gaultherioides mass and area per individual leaf, with both traits slightly reduced under warming in 2007 only (Tables 2b, S5). Soil moisture did not show a significant relationship with any measured leaf parameters (Table S4).

Table 1.   Results of an analysis of (co)variance repeated measures test of the effect of CO2 enrichment and tree species on Vaccinium myrtillus and Vaccinium gaultherioides leaf morphology from 2006 to 2009
 SLA (cm2 g−1)aMass per leaf (mg)Area per leaf (cm2)a,b
  1. aData not available from 2008.

  2. bAny missing leaf parts were filled to correct for losses caused by herbivory (< 4% of potential area per individual leaf).

  3. SLA, specific leaf area.

(a) Vaccinium myrtillus
 CO2 concentration1, 361.780.191, 363.900.061, 368.790.01
 Tree species1, 369.01< 0.011, 365.150.031, 3618.90< 0.01
 Year2, 705.340.013, 10815.64< 0.012, 7115.51< 0.01
 CO2 × tree species1, 360.530.471, 360.190.671, 360.120.73
 CO2 × year2, 700.030.973, 1080.280.842, 710.810.45
 Tree species × year2, 702.990.063, 1082.090.112, 718.19< 0.01
 CO2 × tree × year2, 700.650.533, 1080.090.972, 711.200.31
(b) Vaccinium gaultherioides
 CO2 concentration1, 360.950.341, 360.300.591, 360.010.92
 Tree species1, 3610.61< 0.011, 361.640.211, 360.220.64
 Year2, 6913.20< 0.013, 1055.88< 0.012, 6924.73< 0.01
 CO2 × tree species1, 361.560.221, 362.630.111, 361.600.21
 CO2 × year2, 691.080.353, 1051.750.162, 691.540.22
 Tree species × year2, 691.880.163, 1050.390.762, 691.690.19
 CO2 × tree × year2, 691.370.263, 1050.780.512, 691.410.25
Table 2.   Results of an analysis of covariance repeated measures test of the effect of soil warming on Vaccinium myrtillus and Vaccinium gaultherioides leaf morphology from 2007 to 2009
 SLA (cm2 g−1)aMass per leaf (mg)Area per leaf (cm2)a,b
  1. aData not available from 2008.

  2. bAny missing leaf parts were filled to correct for losses caused by herbivory (< 4% of potential area per individual leaf).

  3. SLA, specific leaf area.

(a) Vaccinium myrtillus
 2006 covariable1, 3519.11< 0.011, 3536.39< 0.011, 3534.25< 0.01
 Temperature1, 350.140.711, 35< 0.01> 0.991, 350.200.66
 Tree species1, 350.720.401, 350.020.891, 35< 0.010.99
 Year1, 356.740.012, 7217.16< 0.011, 355.900.02
 Temperature × tree species1, 350.430.521, 350.020.891, 350.270.61
 Temperature × year1, 353.670.062, 721.080.351, 352.200.15
 Tree species × year1, 354.700.042, 721.120.331, 350.080.78
 Temperature × tree × year1, 351.450.242, 720.330.721, 350.030.86
(b) Vaccinium gaultherioides
 2006 covariable1, 3470.22< 0.011, 3458.09< 0.011, 3447.11< 0.01
 Temperature1, 342.110.161, 340.140.711, 340.060.81
 Tree species1, 344.610.041, 341.960.171, 340.170.68
 Year1, 3314.15< 0.012, 683.520.041, 3339.43< 0.01
 Temperature × tree species1, 340.080.781, 340.410.521, 340.700.41
 Temperature × year1, 330.220.652, 683.960.021, 335.440.03
 Tree species × year1, 332.540.122, 680.210.811, 335.620.02
 Temperature × tree × year1, 330.460.502, 680.690.511, 330.070.79

Soil inorganic N and leaf N concentration

In the soil solution, NH4+ and NO3 concentrations were always below the detection limit (0.1 mg N l−1). Concentrations of K2SO4-extractable NO3 were also not detectable, indicating that there was no net nitrification. Soil inorganic N pool size was not significantly influenced by CO2 treatment, tree species identity, year or any interactive term (> 0.19). Leaf N concentration of the two Vaccinium species increased with increasing canopy shading, and concentrations were generally higher in V. gaultherioides than in V. myrtillus (Tables S4, S5). There was a trend of lower leaf N concentration under elevated CO2 in both Vaccinium species, particularly in 2006 and 2007, but this effect was not statistically significant (marginally significant CO2 × tree species × year interaction for V. myrtillus; Tables 1, S5). Leaf N concentration of both species varied among the four years and tended to be lower in plots with pine than in those with larch (Table 1).

Soil inorganic N pool size increased strongly with soil warming (F1,35 = 9.5, < 0.01; Fig. 4), with a more pronounced response in 2009 (140% greater in warmed plots than in unwarmed plots, standardized to 2004 values and pooled across tree species identities and CO2 concentrations) than in 2007 (+52%). Warming also had a positive influence on leaf N concentration of both Vaccinium species (Fig. 4), although this effect declined over the 3 yr for V. myrtillus (marginally significant warming × year interaction; F1,35 = 2.8, = 0.07). Increased shoot increment length of V. myrtillus was weakly associated with increased leaf N concentration (< 0.05 in 2007 and 2009) and with increased soil inorganic N in 2009 (= 0.03) but not 2007. By contrast, there was no indication of enhanced shoot growth associated with increased N availability in V. gaultherioides or in E. hermaphroditum (see Tables S6, S7 for relevant statistical results).

Figure 4.

Effects of soil warming on nitrogen (N) availability. (a) Mean inorganic N pool size in the soil and (b, c) leaf N concentration of (b) Vaccinium myrtillus and (c) V. gaultherioides± 1 SE for each warming treatment group, pooled across plot tree species identities and CO2 treatment groups (= 20). Leaves and soil were each sampled once per vegetation period. Pre-warming values are shown in the shaded region. Open squares, unwarmed; closed squares, warmed.

Understorey vegetation composition

After accounting for 2005 differences, the total number of vascular and nonvascular (moss and lichen) species per plot in 2009 was influenced by both experimental treatments, with a marginally significant CO2 × tree species interactive effect (F1,31 = 3.8, = 0.06) and negative soil warming effect (F1,31 = 3.3, = 0.08). Analysis of vascular species alone indicated that changes in richness over time were influenced by elevated CO2, again interacting with tree species identity (F1,31 = 4.6, = 0.04), but not by warming. Both total and vascular species richnesses were lower in plots with taller V. myrtillus ramets (< 0.07) but showed no influence of canopy shading, average understorey vegetation height or soil moisture (Table S4). Separate analysis of plots with each tree species indicated a marginally significant negative CO2 effect in plots with larch (F1,15 = 3.9, = 0.07), with a mean loss (± 1 SE) of 1.3 ± 0.6 vascular species over the 4 yr in elevated CO2 plots but no change in ambient CO2 plots (Fig. 5). Plots with pine showed an opposite but nonsignificant pattern (F1,15 = 1.1, = 0.30), with somewhat more pronounced vascular species losses in plots with ambient CO2 concentrations (Fig. 5). Nonvascular species richness declined overall from 2005 to 2009, with more species lost from warmed plots (= 0.02), especially those also exposed to elevated CO2, than from unwarmed plots (Fig. 6; see Table S8 for statistical results). Plots with lower soil moisture were associated with more nonvascular species losses (= 0.07; Table S4). Individual species lost most frequently from the plots are presented in Table S9.

Figure 5.

Changes in understorey vascular plant species richness from 2005 to 2009. The mean number of species per plot is shown for each CO2 treatment and tree species combination, ± 1 SE (pooled across temperature treatments; = 10). Open circles, ambient CO2; closed circles, elevated CO2.

Figure 6.

Changes in understorey nonvascular (moss and lichen) plant species richness from 2005 to 2009. The mean number of species per plot is shown for each CO2 and temperature treatment combination, ± 1 SE (pooled across tree species identities; = 10). Open squares, unwarmed; closed squares, warmed.


Species-specific growth responses to simulated atmospheric and climate change

In this study of three dominant ericaceous dwarf shrub species growing at the treeline in the Swiss Alps, V. myrtillus growth was stimulated by both CO2 enrichment and soil warming and generally showed a stronger response than V. gaultherioides or E. hermaphroditum.

The CO2 effect on each dwarf shrub species was relatively consistent over several years of enrichment, providing a clear indication that these species differ in their responsiveness, irrespective of annual variations in climatic conditions. In contrast to our first hypothesis, the shoot growth response of V. myrtillus to elevated CO2 was sustained over the full experimental period. This result suggests that any increase in competition for nutrients or light that occurred over time with enhanced plant growth and microbial activity did not constrain the longer term above-ground growth response of this species. CO2 enrichment did not lead to a significant decline in the soil inorganic N pool or in leaf N concentration, providing evidence that the treatment did not cause N to become limiting.

The relatively strong response of V. myrtillus to CO2 enrichment compared with the other two dwarf shrub species might be related to different leaf traits. For example, higher SLA in V. myrtillus compared with V. gaultherioides (20–30% in all years measured) and E. hermaphroditum (> 50%; Zumbrunn, 2004) means that there is a larger amount of leaf area displayed per unit mass invested (Poorter et al., 2009) in V. myrtillus than in the other two species. Efficient light capture and high photosynthetic capacity associated with higher SLA could have led to larger assimilation gains under elevated CO2 (Roumet & Roy, 1996) in V. myrtillus compared with the other two species. For V. gaultherioides, the CO2 effect on annual shoot growth and total ramet height depended on which tree species grew in the experimental plot. This response could not be explained by differences in light conditions (canopy shading), snowmelt date, plot topography, soil moisture, leaf N concentration (slightly higher in plots with larch than in those with pine) or soil inorganic N pool size (Tables S1, S4). Reasons for the shoot growth response of V. gaultherioides remain unclear, but differences in litter production and quality or in below-ground competition, for example the availability of nutrients other than N, might have played a role. In general, lower V. gaultherioides ramet height in plots with larch than in those with pine could indicate less favourable growth conditions for this species in the understorey of larch trees.

Consistent with our findings, species-specific responses were observed in a 3-yr mesocosm CO2 enrichment study at the Abisko research station in northern Sweden that included the same three dwarf shrub species as in our experiment (Gwynn-Jones et al., 1997). Similar to our results, V. myrtillus was the only species in this subarctic experiment to show a positive growth response to CO2 enrichment. However, whereas we observed a tree species-specific CO2 response in V. gaultherioides (group V. uliginosum agg.) and no effect in E. hermaphroditum, at the Abisko site V. uliginosum showed no CO2 response and there was a significant negative effect for E. hermaphroditum in one treatment year (Gwynn-Jones et al., 1997). Enhanced growth was also observed for V. myrtillus exposed to elevated CO2 for a single season in a glasshouse study in low-elevation heathlands of the Netherlands (Arp et al., 1998). The consistent growth responses to CO2 enrichment observed for V. myrtillus across multiple studies under various growth conditions indicates an inherent CO2 responsiveness of this species and suggests that V. myrtillus growth and abundance might increase in a future CO2-enriched atmosphere.

Deciduous V. myrtillus was also the only species to show a significant positive shoot growth response to the warming treatment, with an average stimulation over twice the size of the mean CO2 effect. Vaccinium gaultherioides, which is also deciduous, showed no response to warming, suggesting that factors other than leaf type had a greater influence on the responses of individual species. In the Alps, V. myrtillus has a lower elevational distribution compared with V. gaultherioides and E. hermaphroditum, both of which extend to > 3000 m asl (Landolt et al., 2010). Vaccinium myrtillus might therefore be better adapted and more responsive to warmer temperatures. As in our study, V. myrtillus had a more pronounced shoot growth response to soil warming than E. hermaphroditum (stimulation only with additional air warming) or V. uliginosum (no growth response) in a 5-yr study in Abisko, Sweden (Hartley et al., 1999). Positive shoot growth responses were similarly observed for dwarf shrubs after the second and third seasons of warming by OTCs in the Swedish subarctic heath (Parsons et al., 1994), although there all three species that also occur in our experiment responded positively. Finally, warming by OTCs at a temperate alpine site in northern Japan had no effect on vegetative growth of V. uliginosum (Kudo & Suzuki, 2003), whereas Empetrum nigrum var. japonicum shoot elongation was strongly stimulated.

Different responses to warming for the same species might be attributable to genetic differences between regions, especially for E. nigrum and V. uliginosum, which are highly heterogeneous species complexes (Bell & Tallis, 1973; Jacquemart, 1996). Lower atmospheric pressure (and therefore lower CO2 partial pressure), contrasting day–night solar radiation and temperatures during summer, and generally higher precipitation in temperate alpine environments compared with arctic regions might also have contributed to the different findings (Körner, 2003). Finally, different heating techniques might have played a role in the divergent results: passive warming by OTCs generally results in smaller increases in air and soil temperature than warming by heating cables on the ground surface (Rustad et al., 2001). Further, warming by OTCs is confounded to some extent with shelter effects, that is, increased humidity and reduced wind speed and night-time radiative cooling, whereas the heating cables in our study slightly reduced air humidity near the ground surface and did not affect the latter two parameters (Hagedorn et al., 2010).

Soil warming led to a strong increase in soil mineral N content that was still evident after three growing seasons. This result supports our prediction that warming would at least initially accelerate N cycling and lead to an enhanced N supply (Melillo et al., 2002). In a study in open birch (Betula pubescens ssp. tortuosa) forest in northern Sweden including the same dwarf shrub species as our study, N mineralization rates were doubled compared with controls in plots with 5°C soil warming in the second year of treatment, although no effect was observed in the fifth year (Hartley et al., 1999). Corresponding to the larger inorganic N pool in the soil, the leaf N concentration of both Vaccinium species showed a short-term increase in warmed plots in our study. Warming effects on V. myrtillus leaf N concentration were only apparent in the first year of treatment when the growth response was smallest. It is possible that greater stimulation of shoot growth in 2008 and 2009 diluted the soil warming-induced increase in total N uptake, yielding an overall larger N pool in leaf biomass but no effect on concentrations (Weih & Karlsson, 2001). Hartley et al. (1999) observed no effect of soil warming on foliar N concentrations in V. myrtillus or V. uliginosum in any year of their study, despite increased N mineralization rates. By contrast, 8 yr of warming by tents and then OTCs in another experiment in northern Sweden had a large positive effect on leaf and shoot N concentration of V. myrtillus (73%) but a negative effect on that of V. uliginosum (19%) (Richardson et al., 2002). We observed weak but positive correlations between V. myrtillus shoot increment length and soil/leaf N, suggesting that increased N availability might have been partially responsible for stimulation of shoot growth but that other mechanisms (i.e. direct effects of temperature on photosynthesis) were probably also important.

In contrast to our third hypothesis, we found no evidence of a positive CO2 × warming interactive effect on dwarf shrub growth, suggesting that responses to CO2 enrichment were not constrained by low N availability or low temperature. This result is in contrast to the positive CO2 × warming interactive effect on NPP observed for tussock tundra vegetation in Alaska (Oechel et al., 1994). Although no other combined warming and CO2 enrichment studies have been conducted at high elevation, an alpine grassland community showed no effect of CO2 enrichment on biomass production even when combined with 40 kg ha−1 a−1 of NPK fertilization, indicating that nutrient limitation was not the reason for no biomass response to elevated CO2 (Körner et al., 1997). Similarly, a FACE × N addition experiment on glacier forefield vegetation showed strong stimulation by fertilization over 3 yr but no positive CO2 effect or interaction between treatments (N. Inauen, pers. comm.).

Independent growth responses of V. myrtillus to elevated CO2 and soil warming suggest that this species will have a competitive advantage over the co-occurring dwarf shrub species in the future. However, at our research site both treatments increased the sensitivity of V. myrtillus to damage from early growing season freezing events, whereas neither V. gaultherioides nor E. hermaphroditum was affected (Martin et al., 2010). Further, V. myrtillus was found to be more susceptible than the other two species to negative effects of early snow ablation (Wipf et al., 2009). Therefore, stochastic climate events that consistently impact V. myrtillus more severely than the other dwarf shrub species could counteract increases in its dominance.

Decline in species richness with elevated CO2 concentrations and soil warming

The experimental treatments led to changes in vegetation composition at the plot scale during the final 4 yr of the 9-yr study, with a decline in the number of vascular and nonvascular species in the plots. The observed trend of greater species loss in plots with taller V. myrtillus ramets, although not in those with greater canopy shading, suggests that increased shading within the understorey canopy and/or increased below-ground competition played a role in the decline. The opposite pattern in plots with pine was surprising because the height of V. myrtillus was also enhanced in elevated CO2 plots shared with this tree species. Three years of experimental warming had no detectable effect on vascular plant composition, consistent with results from a 5-yr soil warming treatment in Abisko, Sweden (Hartley et al., 1999). By contrast, 9 yr of warming in glasshouses near Toolik Lake, Alaska led to a decline in species richness (Chapin et al., 1995). Given that the response of V. myrtillus shoot growth to warming was more pronounced than that to CO2 enrichment, sustained shoot growth enhancement of this abundant species could lead to changes in vegetation composition and species richness over the longer term. The warming treatment in our study did result in a loss of moss and lichen species. Increased vascular plant productivity with warming has been associated with reduced abundance of mosses and lichens in several (sub)arctic studies (meta-analysis by Walker et al., 2006). This relationship has previously been attributed to increased shading by vascular plants (Chapin et al., 1995; Cornelissen et al., 2001), although the observed correlation with soil moisture in our study suggests that drying associated with the warming treatment was at least partially responsible for the loss of these species. Overall, negative effects of CO2 enrichment and soil warming on species richness indicate that the ongoing environmental change could lead to less diverse plant communities at the studied alpine treeline.


We are indebted to many colleagues at the SLF, WSL, Paul-Scherrer Institute and University of Basel for their assistance with field and laboratory measurements and technical support. We are especially grateful to E. Amstutz, L. Egli, G. Grun and A. Studer for helping to ensure successful operation of the FACE system and to C. Körner for his feedback throughout this work. Major funding sources included: the Swiss National Science Foundation from 2001 to 2005 (grant 31-061428.00 to S.H.) and from 2007 to 2009 (grant 315200-116861 to C.R.); the Velux Foundation from 2007 to 2009 (grant 371 to F.H.); and an ‘ANR-biodiversité’ grant (Qdiv led by Paul Leadley) to S.H. from 2006 to 2008. Additional financial support was provided by the CCES-ETH-Project ‘MOUNTLAND’, Swiss State Secretariat for Education and Research (COST Action 639, project C07.0032), WSL, University of Basel Botanical Institute, Swiss Federal Office for the Environment and ‘Fonds québecois de recherche sur la nature et les technologies’ (scholarship to I.T.H.).