Species-specific tree growth responses to 9 years of CO₂ enrichment at the alpine treeline

Summary

1. Using experimental atmospheric CO₂ enrichment, we tested for tree growth stimulation at the high-elevation treeline, where there is overwhelming evidence that low temperature inhibits growth despite an adequate carbon supply. We exposed Larix decidua (European larch) and Pinus mugo ssp. uncinata (mountain pine) to 9 years of free-air CO₂ enrichment (FACE) in an in situ experiment at treeline in the Swiss Alps (2180 m a.s.l.).

2. Accounting for pre-treatment vigour of individual trees, tree ring increments throughout the experimental period were larger in Larix growing under elevated CO₂ but not in Pinus. The magnitude of the CO₂ response in Larix ring width varied over time, with a significant stimulation occurring in treatment years 3–7 (marginal in year 6).

3. After 9 years of treatment, leaf canopy cover, stem basal area and total new shoot production were overall greater in Larix trees growing under elevated CO₂, whereas Pinus showed no such cumulative growth response. The Larix ring width response in years 3–7 could have caused the cumulative CO₂ effect on tree size even if no further stimulation occurred, so it remains unclear if responsiveness was sustained over the longer term.

4. Larix ring width was stimulated more by elevated CO₂ in years with relatively high spring temperatures and an early snowmelt date, suggesting that temperatures were less limiting in these years and greater benefit was gained from extra carbon assimilated under elevated CO₂. The magnitude of CO₂ stimulation was also larger after relatively high temperatures and high solar radiation in the preceding growing season, perhaps reflecting gains due to larger carbon reserves.

5. Synthesis. Contrasting above-ground growth responses of two treeline tree species to elevated CO₂ concentrations suggest that Larix will have a competitive advantage over less responsive species, such as co-occurring Pinus, under future CO₂ concentrations. Stimulation of Larix growth might be especially pronounced in a future warmer climate.

Introduction

Atmospheric CO₂ concentration is now higher than it has ever been during the last 25 million years (e.g. Pearson & Palmer 2000), and models using coupled carbon-climate cycle simulations predict CO₂ levels to reach between 730 and 1020 μmol mol⁻¹ by the year 2100 (IPCC 2007). Over the last 25 years, researchers have experimentally manipulated the atmospheric CO₂ concentration to study the effects on plant growth and productivity from the genetic level to the whole ecosystem scale. Tree responses have been studied with particular interest because forests are major terrestrial biomass carbon stores and hence play an important role in the global carbon cycle (Schimel 1995; Jarvis 1998; Huang et al. 2007). Results from studies of older trees in systems with complete plant–soil coupling indicate high interspecific differences in growth responses and overall lower responsiveness in biomass production than initially found in chamber experiments with young trees and otherwise optimal growth conditions (Nowak, Ellsworth & Smith 2004; Norby et al. 2005; Körner 2006). Further, it has become apparent that the availability of resources other than carbon plays a large role in the CO₂ response (Spinnler, Egli & Körner 2002; Finzi et al. 2006; Körner 2006). Although the field is developing rapidly, long-term experimental studies about growth responses of pole stage and mature trees in natural environments are still rare and tend to focus on systems dominated by a single species.

It has become clear from long-term CO₂ enrichment studies that an appropriate experimental duration is essential for understanding the dynamics of plant responses to elevated CO₂, particularly for long-lived plants like trees (Körner 2006). Long-term CO₂ enrichment can have a negative feedback on plant growth if greater nutrient sequestration into organic matter during CO₂-induced growth enhancement is not met with sufficient replenishment of nutrients via mineralization (Luo et al. 2004; Reich, Hungate & Luo 2006; Millard, Sommerkorn & Grelet 2007; Pepper et al. 2007). Early studies of young trees also attributed a decline in the CO₂ response to downward adjustment of photosynthetic capacity under longer-term exposure to elevated CO₂ (e.g. Medlyn et al. 1999); however, studies of mature trees growing in near-natural conditions have not provided much evidence for such an effect (Nowak, Ellsworth & Smith 2004; Zotz, Pepin & Körner 2005). Positive feedbacks are possible over the longer term because elevated CO₂ can enhance plant nutrient use efficiency and increase nutrient acquisition by stimulating mycorrhizal and root growth (Norby et al. 2004; Treseder 2004). Long-term manipulation experiments are particularly important because temporal variation in climate or resources might influence treatment effects, and these patterns are only observable over several years. Changes in the responses to elevated CO₂ might also occur as trees become older because CO₂ responsiveness might be greatest in young trees, when growth is vigorous and nutrient supply is relatively high (Wang 2007), but few CO₂ manipulation experiments have lasted long enough to investigate this question adequately. In natural ecosystems, shifts in the responses of trees might additionally result from CO₂-mediated biotic interactions between trees and understorey species or herbivores (Zvereva & Kozlov 2006).

The current understanding of how trees will respond to increasing atmospheric CO₂ concentrations is based almost entirely on low-elevation forest sites, where direct competition for light, space, water and nutrients potentially regulates the CO₂ response under conditions of a steady-state leaf area index (LAI) (Körner 2006). In contrast, no previous in situ CO₂ enrichment experiments on trees have involved conditions where low temperature is thought to be a major limiting factor for growth. Several dendrochronological studies of high-elevation conifers have led to the argument that, along with climate warming, rising atmospheric CO₂ concentrations might have contributed to increasing tree ring width over the last 150 years (Graybill & Idso 1993; Nicolussi, Bortenschlager & Körner 1995). However, the confounding and possibly interacting effects of these two factors, along with several other biotic and abiotic variables, complicate the interpretation of growth trends from dendrochronological records (Graumlich 1991; Huang et al. 2007). Experimental studies aimed at understanding how trees growing at treeline respond to rising CO₂ concentrations and how that response depends on temperature can provide a valuable complement to dendrochronological studies for predicting future changes in these ecosystems.

High-elevation treelines follow a global isotherm of 6.7 ± 0.8 °C (mean ± SE growing season soil temperature; Körner & Paulsen 2004), and the explanation for this existential limit has been debated extensively by the scientific community (e.g. Tranquillini 1979; Körner 2003; Smith et al. 2009). The most plausible explanations for treeline formation from a global perspective are the sink and source limitation hypotheses (Körner 1998). The former states that low temperature restricts the rate at which carbon can be used for structural growth more than it limits the rate of net photosynthesis. The latter proposes that low temperatures and frequent damage and disturbance (at high latitudes) cause a shortage of photoassimilates and a negative carbon balance over the long term. Photosynthetic rates in treeline trees are relatively insensitive to temperature, which casts doubt on assimilation limitation at the leaf level (Pisek & Winkler 1958; Häsler 1982). Further, concentrations of non-structural carbon reserves in trees have consistently been found to increase with increasing elevation at locations across the globe, suggesting that restricted carbon investment, rather than acquisition, limits tree growth (e.g. Hoch & Körner 2003; Shi, Körner & Hoch 2008). The sink limitation hypothesis has also been supported by evidence that wood formation was only active when the minimum daily temperature was above 2–4 °C in Larix decidua, Picea abies and Pinus cembra growing at treeline in the Italian Alps (Rossi et al. 2007). Likewise, root growth in seedlings of the three conifer species Picea abies, Pinus cembra and Pinus sylvestris, given optimal levels of other resources, ceased at temperatures below 4–5 °C (Alvarez-Uria & Körner 2007). For L. decidua saplings growing at our own treeline research site in the Swiss Alps, the rates of root and shoot elongation were exponentially related to temperature, with a distinct reduction in above- and below-ground growth below 5–7 °C (Häsler, Streule & Turner 1999).

Experimental manipulation of atmospheric CO₂ concentration at the high-elevation treeline provides the unique opportunity to test directly whether CO₂ enrichment stimulates tree growth in an environment where there is overwhelming evidence that low temperature inhibits growth despite an adequate carbon supply. We exposed L. decidua (European larch) and Pinus mugo ssp. uncinata (mountain pine) growing at the alpine treeline to elevated CO₂ concentrations and studied tree growth over 9 years. During the first 4 years of CO₂ enrichment, Pinus showed low responsiveness to the enhanced carbon supply whereas Larix showed sustained above-ground growth stimulation under elevated CO₂, contrary to predictions based on the sink limitation hypothesis (Hättenschwiler et al. 2002; Handa, Körner & Hättenschwiler 2005, 2006). In this study, we present new growth data from 2005 to 2009, the final 5 years of the long-term experiment, and re-analyse results from earlier years of the study in the context of the full experimental period. We aimed to understand (i) whether trees show sustained growth stimulation under elevated CO₂ over several years; (ii) if co-occurring Larix and Pinus respond differently to long-term CO₂ enrichment; and (iii) how interannual variability in climate conditions influences the growth response to CO₂ enrichment.

Materials and methods

Study site and experimental setup

The study site is located at Stillberg, Davos in the Central Alps, Switzerland (9°52′ E, 46°46′ N). The free-air CO₂ enrichment (FACE) experiment covers an area of 2500 m² and is situated at or slightly above the natural climatic treeline (2180 m a.s.l.) on a NE-exposed 25 30° slope (Hättenschwiler et al. 2002; Handa, Körner & Hättenschwiler 2006). The site is located within a 5-ha long-term afforestation research area where seedlings of three treeline species, Larix decidua L., Pinus cembra L. and Pinus mugo ssp. uncinata Ramond, of high-elevation provenances were planted into the intact dwarf shrub community in 1975 by the Swiss Federal Institute for Forest, Snow and Landscape Research (WSL). Long-term annual precipitation at this location is 1050 mm and average temperature is −5.8 °C in January and 9.4 °C in July (Schönenberger & Frey 1988). Soil types are sandy Ranker and Podzols (Lithic Cryumbrepts and Typic Cryorthods), derived from siliceous Paragneis parent material and dominated by an organic Humimor layer of 5–20 cm (Schönenberger & Frey 1988; Bednorz et al. 2000).

For the FACE experiment, 40 hexagonal 1.1-m² plots were established in early June 2001, 20 with a Pinus uncinata individual in the centre and 20 with a L. decidua individual in the centre. Trees selected for the experiment were separated by a distance of at least 2 m and fit the following additional criteria: one dominant stem, no serious signs of disease or herbivory, not more than one neighbouring tree within 80 cm, and total tree height of 0.8–1.5 m (Hättenschwiler et al. 2002). These trees are now 38 (Pinus) and 36 (Larix) years old, with average heights of 1.4 m (Pinus) and 2.3 m (Larix) and a stem basal diameter of 5–10 cm. The trees are sparsely distributed and do not form a closed canopy, with a dense cover of understorey plants in each experimental plot surrounding the tree base. Each plot therefore consists of a single tree and the typical understorey vegetation of dwarf shrubs (deciduous Vaccinium myrtillus and V. uliginosum, evergreen V. vitis-idaea and Empetrum nigrum ssp. hermaphroditum) and herbaceous species (e.g. Gentiana punctata, Homogyne alpina, Leontodon helveticus, Melampyrum pratense). The 40 plots were assigned to 10 groups of four neighbouring plots (two Larix and two Pinus 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 (target concentration 550 μmol mol⁻¹) whereas the remaining groups served as controls and received no additional CO₂ (c. 380 μmol mol⁻¹). The resulting experimental setup for multi-year analysis was a split–split-plot design with (i) 40 ‘split–split-plots’ (individual tree, unit upon which repeated measures were made) nested within (ii) 20 ‘split-plots’ (containing either two Pinus or two Larix individuals) nested within (iii) 10 ‘whole-plots’ with either ambient or elevated CO₂ concentration (four trees each).

The FACE system released pure CO₂ through laser-punched drip irrigation tubes during daytime hours only (10–14 h per day). From 2001 to 2006, the tubes hung vertically from a hexagonal frame surrounding each plot. From 2007 to 2009, tubes were woven into the tree crown in a manner similar to the web-FACE approach (Pepin & Körner 2002) in addition to a ring of vertical tubes from ground level to 50-cm height. This modified system provided more efficient CO₂ delivery to the entire canopy as trees continued to grow both taller and wider and helped avoid structural problems with larger frames in the steep, rocky terrain. Detailed descriptions of the original and revised setup and of the temporal and spatial variability of the CO₂ enrichment facility are given in Hättenschwiler et al. (2002) and in Hagedorn et al. (2010). Over the 9 years of the experiment, the seasonal mean CO₂ concentrations in enriched plots during the CO₂ dispensing period ± 1 SD, ranged from 545 ± 45 to 595 ± 62 μmol mol⁻¹ (based on daily averages). Accounting for interruptions in CO₂ delivery because of adverse weather conditions and technical failures, trees received CO₂ enrichment for 73–87% of the potential treatment period each season (daytime only). The carbon isotope composition of the supplied pure CO₂ gas came from fossil fuel sources depleted in ¹³C compared to CO₂ in ambient air (−29.9‰ vs. −10.1‰). Therefore, the isotopic ratio of needle tissue provided evidence of how much supplementary CO₂ trees received. The 9-year mean isotopic ratio in needles of trees exposed to elevated CO₂ was shifted by −5.71‰ (averaged across the two species), indicating a mean canopy CO₂ concentration of 535 μmol mol⁻¹. The adapted CO₂ supply system used in 2006–2009 did not change the needle isotopic ratio compared with the earlier years, demonstrating that enrichment was effective even as the trees grew larger. Further, isotope values from the ambient CO₂ plots corresponded to the natural ¹³C abundance, verifying that the enrichment system did not alter the CO₂ concentration in these control plots (von Felten et al. 2007). CO₂ concentration and δ¹³C values for individual years are presented in Table S1 in Supporting Information.

Climate

A long-term meteorological station maintained by the Swiss Federal Institute for Forest, Snow and Landscape Research is located at 2090 m a.s.l., c. 100 m below the FACE site, and provided climate information for the Stillberg research site both before and during the experimental period. Daily (24 h) mean soil (10-cm depth) and air (2-m height) temperature and solar radiation (310–2800 nm) were used to calculate means for each individual month of the year (temperature only) and over the entire growing season (c. late May to early October). The beginning and end of the growing season were defined as the first date when daily mean soil temperature rose above 3.2 °C in spring and fell below 3.2 °C for more than two contiguous days in autumn, respectively (Körner & Paulsen 2004). Cumulative precipitation was calculated over the entire growing season. Maximum snow depth was determined for each winter, and snowmelt date was defined as the date in spring when the ground surface temperature rose sharply from values near zero during periods with snow cover.

Tree ring width

We collected microcores of each tree stem between 21 October and 21 November 2009. Trees were sampled at the base of the crown (approximate height 50 cm in Larix and 30 cm in Pinus), avoiding the (often) curved section at the base of the tree. A 2-mm-diameter increment puncher with a cutting length of 35 mm (TREPHOR, Università degli Studi di Padova, S. Vito di Cadore, Italy; Rossi, Anfodillo & Menardi 2006) was used to extract two microcores, one from each side of the tree facing perpendicular to the slope. With this minimally invasive tool, we were able to collect at least one microcore from each tree that extended back to 1997, thus including 4 years before CO₂ enrichment started. The width of each tree ring was measured using an MS5 stereomicroscope at ×40 magnification (Leica Microsystems GmbH, Wetzlar, Germany). Individual rings showing reaction wood (round, highly lignified cells throughout the ring) were excluded from the analysis because ring width tends to be inflated in these rings (1.2% of all measured rings). Three pines died during the study after becoming infected with the fungus Gremmeniella abietina in the early years of the experiment, reducing the experimental replication for this species to n = 8 for ambient and n = 9 for elevated CO₂ treatment.

Annual lateral shoot length increment

The terminal shoot on five mid-canopy lateral branches was measured after the first growing season of CO₂ enrichment (2001), along with the length of the terminal shoot produced in the preceding pre-treatment year (2000). In each of the four subsequent years (2002–2005), five lateral branches were again randomly selected and the terminal shoot was measured at the end of the growing season (Handa, Körner & Hättenschwiler 2005, 2006). The length of the terminal shoot and all new lateral shoots was measured on 10–15 tagged mid-canopy lateral branches in 2006–2009 to obtain a more representative mean length measurement, as trees had grown substantially during the experiment. The three pines that died during the study were removed from all years of the shoot length analysis because they were already visibly unhealthy in the early years of the experiment.

Tree size

We used multiple measurements from the final 2 years of the experiment to estimate the cumulative effects of several years of CO₂ enrichment on overall tree size and shoot production. (i) Total tree height was measured in autumn 2009, following the method used for the regular census of trees at Stillberg since they were planted in 1975 (Schönenberger & Frey 1988). (ii) Percentage leaf canopy cover was estimated using hemispherical photographs taken in August 2008 during the seasonal peak of leaf area (Nikon Coolpix camera with a fisheye lens attachment; Memphis, TN, USA). One photo was taken in each plot, with the camera positioned c. 20 cm downslope of the tree trunk and 20 cm above the ground. The tree canopy excluding the main trunk was isolated in the image, and the percentage of a standard image size covered by canopy was calculated using image processing software (ImageJ version 1.43k; Rasband 1997). (iii) The number and length of all new shoots on each tree was measured at the end of the season in 2008 and 2009. The resulting sum of all new shoots (averaged over the 2 years) represented gross annual shoot production per tree. (iv) Larix stem diameter was measured in autumn 2009 (mean of 40 and 80 cm above-ground) and used to calculate basal area at the end of the experiment. Many of the Pinus individuals branched close to the ground and did not have one main stem; consequently, stem basal area was not considered representative of tree size in this species, leaving canopy cover and shoot production as the best available proxies for cumulative CO₂ effects on above-ground growth.

Statistical analysis

Ring width and lateral shoot length were tested with Type I analysis of covariance, using repeated-measures linear mixed-effects models to incorporate data from all treatment years. We first fitted a full model: CO₂ treatment (ambient, elevated), tree species (Larix, Pinus), and their interaction were between-subject fixed factors; treatment year (categorical variable, 1–9) and all two- and three-way interactions with year were within-subject fixed factors. Mean shoot length measured in the pre-treatment year 2000 was included as a covariable in the shoot length model and the mean ring width of the four pre-treatment years 1997–2000 was used as a covariable in the ring width model, thus accounting for any differences between treatment groups that existed prior to the experiment in both species. All two-way interactions between the covariable and the three main effects were also tested.

Based on sub-models of each individual treatment year, we determined that the random effects associated with split-plot- and whole-plot-specific intercepts could be omitted for both ring width and shoot length models (West, Welch & Galecki 2007; Zuur et al. 2009). Consequently, we used repeated-measures models that only included random effects for each individual tree. Full shoot length and ring width models both indicated heterogeneity of residual variance across the 9 years, which we addressed by applying a heterogeneous residual variance structure. We accounted for violation of independence of residuals from different treatment years by applying a residual auto-correlation structure (auto-regressive model of order 1 (corAR1); Pinheiro et al. 2008). We then applied backward selection using maximum likelihood (ML) to remove any interactions between fixed factors that did not contribute significantly to the model fit. We refit the reduced final models using the restricted maximum likelihood (REML) estimation method (Zuur et al. 2009). The full models showed a strong tree species effect on both ring width and shoot length, and we completed repeated-measures tests for Larix and Pinus separately to compare the CO₂ effect in the two species. A strong year effect and marginally significant CO₂ × year interaction in Larix ring width prompted us to test the CO₂ effect on that species in individual treatment years.

The effect of elevated CO₂ on total tree height, leaf canopy cover, total shoot production and basal area (Larix only), each measured during the final 2 years of the experiment, was tested separately for the two tree species using a statistical approach parallel to that used for ring width and shoot length. Tree height in 1995 was used as a covariable for testing the effect of elevated CO₂ on final tree height. Stem diameter and leaf canopy cover were not measured before the experiment started, but total tree height, ring width and lateral shoot length all indicated that, on average, the vigour of Larix trees was very similar in the two treatment groups prior to CO₂ enrichment. We could thus confidently test CO₂ effects on Larix final basal area and leaf canopy cover without relating these measurements to pre-treatment values. Pinus mean tree height was also similar in the two CO₂ treatment groups prior to the experiment, but pre-treatment ring width and shoot length were somewhat greater in the elevated CO₂ group. The test of the CO₂ effect on the leaf canopy cover in Pinus was therefore interpreted with caution, as any differences between CO₂ treatment groups might have been at least partially due to pre-treatment differences in vigour.

A one-time defoliation event applied in June 2002 influenced the CO₂ effect to some extent in both species during 2002 and 2003 (Handa, Körner & Hättenschwiler 2005, 2006), and we therefore included only undefoliated trees for tests of the CO₂ effect on shoot length and ring width for these 2 years (n = 5 for Larix and 4 for Pinus). Defoliation treatment had no influence on the CO₂ effect in subsequent years, and we therefore included all trees. Further, a soil warming treatment was applied to half of the plots during the snow-free period in 2007–2009, in a crossed manner with both the CO₂ treatment and the previous defoliation treatment. Heating cables arranged on the ground surface of the plots increased the growing season mean soil temperature by 4°C (5-cm depth) and near-ground air temperature by 1°C (20 cm above-ground), but the treatment had no effect on temperatures in the tree canopy (Hagedorn et al. 2010). Important distinctions between the warming treatment and naturally warm growing seasons were that air temperature in the tree canopy remained at control levels, the heated soil volume (area 1.1 m², depth < 20 cm) was smaller than tree rooting zones found previously at the Stillberg site (Bernoulli & Körner 1999), and snowmelt date was not altered. Analysis of variance tests for parameters measured in the last 3 years of the experiment revealed no significant interaction between CO₂ and soil warming treatments for either species. We therefore pooled warmed and unwarmed trees for the statistical analysis of the long-term CO₂ effects presented in this paper in order to maintain the replication used in 2004–2006 (n = 10 for Larix, n = 8 for ambient CO₂Pinus and 9 for elevated CO₂Pinus). Soil warming had a slight positive effect on Pinus (but no effect on Larix) growth in 2008 and 2009, which influenced the mean ring width and shoot length values in those years but did not alter the CO₂ effect.

We used ordinary least squares regression to determine if interannual differences observed in the magnitude of the CO₂ effect on Larix ring width (mean of all elevated CO₂ trees standardized to pre-treatment ring width divided by the mean of all ambient CO₂ trees standardized to pre-treatment ring width) could be explained by climate conditions in the current or preceding year. As Pinus ring width showed no significant response to elevated CO₂ in the multi-year analysis or in any individual year, an investigation of how climate conditions influenced the CO₂ effect was not relevant for this species.

Assumptions of normality and homoscedasticity of the residuals in all final models were verified visually using diagnostic plots and, when necessary, response variables were log-transformed to improve homoscedasticity. For all statistical tests, effects were considered significant at P < 0.05. Due to relatively low replication and therefore statistical power, we also designated P-values ≥ 0.05 but < 0.10 as marginally significant. All analyses were performed using R version 2.8.1, and mixed models were run using the NLME package (Pinheiro et al. 2008; R Development Core Team 2008).

Results

Tree ring width and elevated CO₂

Ring width in both tree species and both CO₂ treatment groups increased substantially over the 13 years measured in the stem microcores sampled in 2009 (Fig. 1a,b), apparently an age-dependent pattern because no clear trends were visible in the climate data (Fig. 1c,d). Repeated-measures analysis including both species and all years of the experiment showed an overall significant CO₂ effect on tree ring width (F_1,33 = 7.7, P = 0.009; Fig. 1). Rings were wider in Larix than in Pinus (F_1,33 = 10.0, P = 0.003), and ring width varied across the 9 years of the experiment (F_8,244 = 15.5, P < 0.0001). The pre-treatment covariable did not have a significant effect on ring width during the experiment (F_1,33 = 1.8, P = 0.193), and none of the two- or three-way interactions between the main effects or between the main effects and the covariable contributed significantly to the model fit. Separate repeated-measures analyses of the two tree species revealed that CO₂ treatment (F_1,17 = 4.4, P = 0.052) and the pre-treatment covariable (F_1,17 = 4.3, P = 0.054) had a marginally significant effect on Larix ring width (35% median stimulation under elevated CO₂, standardized to 4 years of pre-treatment growth; Fig. 1a). In contrast, elevated CO₂ did not have a significant effect on Pinus ring width (F_1,14 = 1.3, P = 0.268) and pre-treatment ring width was not significant as a covariable (F_1,14 = 1.5, P = 0.242; Fig. 1b). The interaction between the covariable and CO₂ treatment was not significant in models of either species, indicating that the strength of the relationship between ring width and pre-treatment growth did not change under elevated CO₂. Ring width varied significantly across treatment years in both Larix (F_8,120 = 21.3, P < 0.0001) and Pinus (F_8,108 = 5.6, P < 0.0001). The CO₂ × year interaction was marginally significant in Larix only (F_8,120 = 1.9, P = 0.068), and analysis of individual years revealed that the CO₂ treatment effect in Larix was significant in years 2003, 2004, 2005 and 2007, marginally significant in 2006, and not significant in the first two or last two years of the experiment (see Table S2 and S3 for ancova results).

Figure 1.  Mean ring width in Larix (a) and Pinus (b), ± 1 SE (2002 and 2003: n = 5 for Larix, n = 4 for Pinus; all other years: n = 10 for Larix, n = 8 for ambient CO₂Pinus and n = 9 for elevated CO₂Pinus). Years 1997–2000 show pre-treatment differences in vigour, and the mean ring width of these years was used as a covariable in statistical tests. For Larix, * indicates a significant CO₂ effect (P < 0.05) and (*) indicates a marginally significant effect (0.05 ≤ P < 0.10). The bottom two panels show mean air and soil temperature (c) and cumulative precipitation (d) for each growing season.

Lateral shoot length and elevated CO₂

Mean annual lateral shoot length in Larix showed a temporal pattern similar to ring width, with an overall increase over the experimental period and reduced growth in the final treatment year (Fig. 2a). Pinus lateral shoot length did not follow this pattern and instead decreased somewhat over the years (Fig. 2b). Lateral shoot length did not show a significant CO₂ effect when fit after the pre-treatment growth covariable in a Type I analysis of covariance model including both species (F_1,33 = 0.1, P = 0.813; Fig. 2). Pre-treatment shoot length influenced growth in individual trees throughout the experiment in both species (F_1,33 = 221.9, P < 0.0001) and showed a significant interaction with treatment year (F_8,234 = 7.4, P < 0.0001). Treatment year and the treatment year × tree species interaction were both highly significant (P < 0.0001), whereas none of the other two- or three-way interactions contributed significantly to the model fit. There was a strong tree species effect (F_1,33 = 278.5, P < 0.0001), with longer lateral shoots in Larix than in Pinus, but separate tests of the two species confirmed that CO₂ did not significantly affect lateral shoot length in either species (Larix: F_1,17 = 1.8, P = 0.194; Pinus: F_1,14 = 1.0, P = 0.343). Although not significant, there was a small but consistent trend of longer lateral shoots under elevated CO₂ relative to pre-treatment shoot length in Larix only (median stimulation of 11% over the 9 years; Fig. 2a). The interaction between CO₂ treatment and treatment year did not contribute significantly to the model fit for either species (see Table S2 for ancova results).

Figure 2.  Mean annual growth of shoots on randomly selected mid-canopy lateral branches in Larix (a) and Pinus (b), ± 1 SE (2002 and 2003: n = 5 for Larix, n = 4 for Pinus; all other years: n = 10 for Larix, n = 8 for ambient CO₂Pinus and n = 9 for elevated CO₂Pinus). Year 2000 shows pre-treatment differences in vigour and was included as a covariable in statistical tests.

Cumulative CO₂ effect on tree size after nine years of enrichment

Nine years of exposure to elevated CO₂ led to a cumulative effect of increased tree size and shoot production in Larix but no such response in Pinus (Fig. 3). In the final 2 years of the experiment, Larix trees growing under elevated CO₂ were not significantly different in height compared with trees growing under ambient conditions (2009; F_1,17 = 1.7, P = 0.216; Fig. 3) but had a marginally significantly greater leaf canopy cover (27 ± 13% difference in 2008; F_1,18 = 3.4, P = 0.082; Fig. 3). The lack of response in tree height probably reflects damage to leader shoots from snow and animals that occurs frequently at the treeline and the resulting nonlinear canopy structure of many trees. There was no CO₂ effect on either tree height or leaf canopy cover in Pinus (P > 0.80; Fig. 3). In 2008 and 2009, Larix trees growing under elevated CO₂ produced on average 42 ± 18% more new shoots than trees experiencing ambient conditions (F_1,18 = 5.5, P = 0.031), leading to a stimulation in the total length of all new shoots produced per tree of 62 ± 25% (F_1,18 = 5.1, P = 0.037; Fig. 3). During these last 2 years of the experiment, Pinus did not produce a significantly different number of shoots (F_1,15 = 0.9, P = 0.361) or total shoot length per tree (F_1,15 = 1.4, P = 0.257) when grown under elevated CO_2. The basal area of Larix, calculated from 2009 diameter measurements, indicated a marginally significant stimulation of 49 ± 19% (F_1,18 = 4.2, P = 0.056; Fig. 3), largely a result of the combined stimulation of radial stem growth in years 3–7 of the 9-treatment years (see ring width results above). Results of these statistical tests are summarized in Table S4.

Figure 3.  Summary of measurements of tree size at the end of the experiment. Mean values for ambient and elevated CO₂ treatment groups ± 1 SE (n = 10 in Larix, n = 8 in ambient CO₂Pinus and n = 9 in elevated CO₂Pinus). Basal area was calculated for Larix only, based on diameter measurements at c. 50 cm above ground in autumn 2009; percent canopy cover was measured using hemispherical photographs during 2008; total tree height was measured along the main stem in autumn 2009; and shoot growth is the sum length of all new shoots at the end of the season, averaged over 2008 and 2009. Asterisk indicates a significant CO₂ effect (P < 0.05) and (*) indicates a marginally significant effect (0.05 ≤ P < 0.10). Effects on Larix basal area were largely due to stimulation during years 3–7 of the 9-year treatment.

Effects of climate on the magnitude of the CO₂ response

Growing season climate conditions varied during the 4-year pre-treatment period and the 9 years of experimental CO₂ enrichment (Fig. 1c,d). Over these 13 years, the seasonal average of daily mean soil (10-cm depth) and air temperature ranged from 7.1 to 8.7 and from 8.3 to 10.6 °C, respectively (Fig. 1c). The widespread European heat wave in 2003 (e.g. Rebetez et al. 2006) resulted in the highest temperatures during the experimental period. The 2006 growing season was also one of the warmest, with particularly high temperatures during July (13.6 °C mean daily air temperature). Cumulative precipitation during the snow-free season showed substantial interannual variation, with a range of c. 300 mm, largely because of the exceptionally wet season in 2001 (Fig. 1d). Growing season solar radiation showed a strong negative correlation with precipitation, and only regressions relating ring width and precipitation are presented below (correlations between climate variables are presented in Table S5). Maximum snow depth varied from 107 to 197 cm during the experimental period, and snowmelt date ranged from 7 May to 6 June.

The magnitude of the CO₂ effect on Larix ring width in each year was influenced by some of the measured climate variables. The mean CO₂ effect was significantly greater in years with less accumulated snow in winter (r² = 0.58, P = 0.018). Similarly, the CO₂ effect was larger in years with an early snowmelt date (i.e. high June soil temperatures), but only when 2004 was excluded (unusually late snowmelt date despite a low to moderate snow pack; r² = 0.70, P = 0.010; Fig. 4a). The CO₂ effect was also larger when the preceding growing season was characterized by relatively high soil temperatures (r² = 0.66, P = 0.008; Fig. 4b), with a particularly strong relationship when the unusually high mean soil temperature in 2003 was excluded (r² = 0.83, P = 0.002; Fig. 4b). The effect of air temperature during the same period was also statistically significant but weaker, possibly because of the slightly greater interannual fluctuation of this parameter (Table S5). When daily mean temperatures averaged over individual months were tested, soil and air temperature in July of the preceding year (soil: r² = 0.71, P = 0.004) and air temperature in November (r² = 0.47, P = 0.042) of the preceding year both had a positive influence on the CO₂ effect size. Cumulative precipitation during the growing season did not (negatively) co-vary significantly with temperature during the same period, and yet the magnitude of the Larix CO₂ effect was also positively influenced by low precipitation in the preceding year, but only when the exceptionally wet year 2001 was excluded (r² = 0.74, P = 0.006; see Table S5 for all regression results).

Figure 4.  Relationship between the CO₂ effect on Larix ring width in each year and (a) snowmelt date in the current year and (b) daily mean soil temperature (10-cm depth) averaged over the preceding growing season. The CO₂ effect was calculated as the ratio of the standardized means of all elevated CO₂ trees (n = 10) to the standardized means of all ambient CO₂ trees (n = 10). For each of the nine treatment years (represented by a single point), ring width in an individual tree was standardized to the mean 1997–2000 pre-treatment ring width. The linear regression equation, coefficient of determination (r²), and error probability of the regression slope (P value) are shown for each relationship. Snowmelt date in 2004 was exceptionally late despite a low-to-moderate accumulation of snow, and this year was removed from the linear regression (circled point in a). Regressions are shown both including (dotted line) and excluding (solid line) soil temperature during the exceptionally warm year 2003 (circled point in b).

Discussion

Species-specific responses during nine years of CO₂ enrichment

Nine years of FACE at the treeline in the Swiss Alps led to an overall stimulation of above-ground growth in Larix but no significant growth response in Pinus. Tree ring width measurements confirmed the species-specific responsiveness to elevated CO₂ observed in an independent set of tree cores during the first four treatment years (Handa, Körner & Hättenschwiler 2006) and indicated that this pattern continued in the following 5 years. Our longer-term results for mean lateral shoot length were also consistent with earlier results (Handa, Körner & Hättenschwiler 2005, 2006), with a consistent but non-significant trend of longer shoots in Larix exposed to elevated CO₂ but not in Pinus. At the whole-tree level, species-specific responses were apparent in leaf canopy cover, total new shoot growth and basal area, indicating a cumulative effect of increased tree size in Larix after 9 years of exposure to elevated CO₂ but no effect in Pinus.

Our FACE study was unique in that it included two co-occurring tree species in a well-replicated experimental design, thus allowing us to conclude that the different responsiveness of Larix and Pinus to elevated CO₂ was related to species identity rather than site conditions. Most previous CO₂ enrichment experiments have included only one tree species, making it difficult to distinguish between species and site effects. A FACE experiment at the Swiss Canopy Crane experimental site included several deciduous species and similarly observed species-specific responses, but the ability to compare growth responses among species was limited by the low (or no) replication of individual species (Asshoff, Zotz & Körner 2006). In our study, different responsiveness for the two species could not be explained by the photosynthetic response, as both Larix and Pinus experienced strong photosynthetic stimulation in the first 3-treatment years (Handa et al. 2005) that persisted in the final year of CO₂ enrichment (K. Streit, personal communication). Further, Larix and Pinus growing at our study site had a similar relative growth rate and Pinus actually had greater total biomass than Larix 3 years before the CO₂ experiment began (Bernoulli & Körner 1999). Therefore, the contrasting CO₂ responses in Larix and Pinus observed here cannot be explained by the previously proposed argument that faster-growing species are more responsive to elevated CO₂ (Tangley 2001; Poorter & Navas 2003). Various differences between Larix and Pinus that might have contributed to their distinct CO₂ responses have been described in detail previously (Handa, Körner & Hättenschwiler 2005, 2006). Briefly, responsiveness in Larix but not Pinus might be explained by: deciduous leaves and the associated higher rate of assimilation return per unit carbon investment in Larix than in evergreen Pinus; and potentially larger sink capacity in Larix, because of production of long shoots with indeterminate growth, compared to Pinus which displays only determinate growth.

Previous research on the CO₂ responsiveness of non-seedling trees in cold environments has largely been conducted on a single species using growth chambers, and results from these experiments have varied widely. Long-term stimulation of radial stem growth was observed in a 6-year closed-chamber study of young Pinus sylvestris (c. 14-years old at the start of the experiment) growing in nitrogen-poor sandy soil in Finland (Kilpelainen et al. 2005). In contrast, a model ecosystem study simulating climate conditions at higher montane elevations typically characterized by conifer tree species dominance revealed no stimulation of stem growth in Picea abies saplings over 3 years, regardless of nitrogen availability (Hättenschwiler, Schweingruber & Körner 1996). In a broader comparison, the observed growth stimulation of Larix is consistent with results for young, closed-canopy stands of Pinus taeda growing in a lowland temperate location (Duke Forest FACE), where basal area increment was enhanced during 8 years of CO₂ enrichment (Moore et al. 2006). This growth response contributed to a sustained higher rate of standing biomass accumulation, with soil N availability largely explaining spatial variability in the magnitude of the stimulation in productivity (McCarthy et al. 2010). Woody biomass was also enhanced in a high density coppice plantation including clones of three Populus species (POPFACE; Liberloo et al. 2006) and in a young Populus tremuloides plantation initiated with seedlings (Aspen FACE; Kubiske et al. 2006), both experiments with a high availability of non-carbon resources. In contrast to the above studies, initial enhancement of above-ground stem growth declined substantially after the first year of exposure to elevated CO₂ in a Liquidambar styraciflua forest, and extra carbon was instead allocated to leaf and fine root production (Oak Ridge FACE; Norby et al. 2002, 2004). Transient increases in stem growth were similarly observed in only one out of five deciduous species growing under elevated CO₂ in a mature temperate forest, and in that case, no long-term growth stimulation was observed above or below ground (Asshoff, Zotz & Körner 2006; Bader, Hiltbrunner & Körner 2009).

Non-carbon resource availability and the long-term CO₂ response

It is difficult to compare the positive CO₂ effect on Larix growth with results from other studies because the treeline environment, and our study site in particular, does not easily fit into either an expanding or steady-state category of growth conditions (Körner 2006). The late successional, essentially undisturbed natural dwarf shrub heath provides steady-state consumption and recycling of resources below ground and an overall steady-state LAI of the understorey and tree canopy as a whole. On the other hand, trees are isolated and crown expansion is largely unconstrained with respect to light or space (Handa, Körner & Hättenschwiler 2005). Height and ring width measurements of Larix trees in our experiment, and more broadly throughout the same elevation zone of the Stillberg plantation, indicated an increasing growth rate in years leading up to and during the experiment (Fig. 1a; Rammig et al. 2005; Amos 2007). Similar age-related patterns of increasing radial growth were observed at treeline in 20- to 30-year-old L. decidua and Picea abies in the Austrian Alps (Li, Yang & Kräuchi 2003) and for the first 50–70 years of growth in Pinus cembra in the Swiss Alps (Esper et al. 2008). Due to the expanding tree canopy, the cumulative CO₂ effect observed in Larix tree size after 9 years of treatment could have resulted from stimulation during a few years that was carried over from compound interest or from a continuous positive CO₂ effect on annual growth (Körner 2006). However, the cumulative CO₂ effect on above-ground growth (50% increase in stem basal area after 9 years) was small compared with the exponentially increasing effect on biomass (50 to >250% difference after only 3–4 years) observed previously for seedlings in unrestrictive growing conditions (Körner 2006). The substantially older age of trees at the start of our experiment undoubtedly played a role in the smaller magnitude of the long-term cumulative effect of CO₂ enrichment, but below-ground competition and low growth temperatures might have imposed additional constraints on the response.

The reduced stimulation of Larix ring width in the last 2 years of the experiment raises the question of whether CO₂-induced growth stimulation in Larix disturbed the current balance between growth and soil nutrient supply, thereby leading to a constrained long-term growth response because of nutrient limitation. Increased allocation to roots and higher fine root production are frequently observed responses to elevated CO₂ (Norby et al. 2004; Tingey, Johnson & Phillips 2005) and are commonly interpreted as a sign of increased nutrient limitation, nitrogen in particular (Luo et al. 2004). However, trees in our study did not show enhanced fine root production in response to 4 years of exposure to elevated CO₂ (Handa, Hagedorn & Hättenschwiler 2008), and bulk root samples from the end of the study also showed no indication of greater biomass allocation to roots (F. Hagedorn, unpublished data). Likewise, needles of trees exposed to elevated CO₂ had a somewhat enhanced C/N ratio in treatment years one to three (significant only in 2001; Hättenschwiler et al. 2002; Handa, Körner & Hättenschwiler 2005; Asshoff & Hättenschwiler 2006), but this response did not affect the C/N ratio of the organic layer or cause a decrease in extractable N in the soil in any year of the study (F. Hagedorn, unpublished data). Collectively, these observations suggest that the CO₂-induced growth stimulation in Larix was not large enough to cause a progressive reduction in soil N availability under elevated CO₂.

Climate influence on the CO₂ effect in Larix

The observed correlation between the general climate conditions at our site and the CO₂ effect size in Larix ring width provides a possible explanation for the interannual variability in the CO₂ effect. Tree ring increments showed a stronger CO₂ response in years with a relatively small maximum snow depth and high early season temperatures, which, in combination, led to an early snowmelt date. Larix radial stem growth starts in early to mid-June at our site (M. Dawes, unpublished dendrometer data), confirming previous observations of L. decidua at treeline in the Alps (Rossi et al. 2007; Moser et al. 2010). Therefore, it is possible that the extent to which trees were able to use extra carbon assimilated under elevated CO₂ for early season growth depended on whether temperatures were high enough for those growth processes to be functional (sink limitation hypothesis). Dampening of CO₂-induced growth stimulation under lower temperatures was similarly observed over the short term in seedlings of several boreal species (Tjoelker, Oleksyn & Reich 1998). Temperatures in July, at peak growth activity, and averaged over the entire current growing season did not strongly influence the CO₂ effect, rather suggesting that low temperatures during the early growing season limited the radial stem growth response to elevated CO₂. The importance of early growing season temperatures for radial stem growth of L. decidua has been documented previously in a longer-term analysis at Stillberg (1975–2005; Amos 2007) and in dendrochronological studies conducted elsewhere in the Alps (Carrer & Urbinati 2004; Oberhuber et al. 2008). Substantially reduced snow volume and earlier spring snowmelt are predicted to occur with climate warming for elevations up to 3000 m in the Swiss Alps, irrespective of winter precipitation conditions (Beniston et al. 2003), and our results suggest that this change in climate conditions could amplify the CO₂-induced growth stimulation of Larix. However, earlier snowmelt can expose trees to stochastic early season freezing events, and increased susceptibility to freezing damage observed for Larix growing under elevated CO₂ might offset any potential growth advantage over the longer term (Martin et al. 2010).

The CO₂ effect observed for Larix ring width was also greater in years following a growing season with high temperatures, high solar radiation and low cumulative precipitation. This growth response might have been due to increased photosynthetic CO₂ assimilation with warmer temperatures or to greater CO₂-induced stimulation of photosynthesis at higher temperatures, as theoretically predicted from the Farquhar model (Long 1991). However, it is well-documented that rates of photosynthesis for trees growing at treeline are not lower than rates for the same species located at warmer lower-elevation sites, suggesting that the photosynthetic rate in high-elevation species is relatively insensitive to increases in temperature (Grace, Berninger & Nagy 2002; Wieser et al. 2010). An alternative explanation is that the larger number of days of favourable climate conditions for carbon uptake simply permitted a greater build-up of carbon reserves under elevated CO₂ that were available for growth in the next year, without a temperature-induced increase in the rate of photosynthesis. Irrespective of the mechanism, these conditions might be more frequent in the future, especially as climate warming in Switzerland has been over twice as large as the mean for the Northern hemisphere and summer temperatures have shown a particularly strong increase (Rebetez & Reinhard 2008). Our results therefore suggest that enhanced above-ground growth of Larix with rising CO₂ concentrations will be more pronounced in the future with ongoing climate change.

Conclusions

In this unique 9-year treeline FACE experiment where trees grew at their lower temperature limit, we observed no growth response to elevated CO₂ in Pinus uncinata. This result supports the hypothesis that carbon supply does not limit above-ground growth of this species at the alpine treeline. In contrast, CO₂-induced growth stimulation in L. decidua was sustained through at least the seventh treatment year, indicating that this species benefited from extra carbon despite low growth temperatures. The CO₂ effect on ring width was enhanced under climate conditions in the current and preceding year known from the literature to positively influence radial stem growth in general. The different responsiveness of these two co-occurring species suggests that under future CO₂ concentrations, especially in combination with warmer and sunnier growth conditions, Larix will have a competitive advantage over less responsive species such as Pinus. Consequently, shifts in abundance of these treeline species might occur over the long term, with implications for the associated dwarf shrub heath and the structure of the treeline ecotone.

Acknowledgements

We are indebted to many colleagues at the SLF, WSL, Paul-Scherrer Institute and University of Basel for their assistance with field and lab measurements and technical support. We are especially grateful to E. Amstutz, L. Egli, G. Grun, A. Studer and S. Wipf for helping to ensure successful operation of the FACE system. M. Panayotov, F. Krumm and the WSL Dendrochronology Group contributed valuable advice and field assistance with the microcores. P. Schleppi provided instruction and equipment for the hemispherical photography. Major funding sources included: the Swiss National Science Foundation from 2001 to 2005 (grant 31-061428.00 to Stephan Hättenschwiler) and from 2007 to 2009 (grant 315200-116861 to Christian Rixen); the Velux foundation from 2007 to 2009 (grant 371 to Frank Hagedorn); and an ‘ANR-biodiversité’ grant (Qdiv led by Paul Leadley) to Stephan Hättenschwiler from 2006 to 2008. Additional financial support for this long-term study 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 (BAFU) and ‘Fonds québecois de recherche sur la nature et les technologies’ (FQRNT scholarship to Ira Tanya Handa).