Atmospheric CO₂ enrichment of alpine treeline conifers

Summary

• •  
  Experimental CO₂ enrichment of mature Larix decidua and Pinus uncinata trees and their understory vegetation was used to test the carbon limitation hypothesis of treeline formation at the alpine treeline in Switzerland.
• •  
  Forty plots (each 1.1 m²) were established; half of them were exposed to elevated (566 ppm) atmospheric CO₂ using a free air CO₂ enrichment (FACE) system releasing pure CO₂, and the other half were treated as controls at current ambient [CO₂].
• •  
  Reliable and adequate CO₂ control was achieved, with 63% and 90% of 1-min averages having a [CO₂] within ±10% and ±20% of the target value, respectively, which is comparable to previous FACE systems. Both tree species showed higher net photosynthesis, lower stomatal conductance, and increased accumulation of nonstructural carbohydrates in response to CO₂ in the first year of treatment. Quite unexpectedly, shoot length increment increased significantly at elevated CO₂ (up to 23%) compared with controls in both species.
• •  
  The pure CO₂ release technology proved suitable for CO₂ enrichment of native trees on this remote mountain slope. Our results suggest an improved C balance and growth of treeline trees in response to elevated CO₂. However, it is unclear whether this initial growth stimulation will persist in the longer term.

Introduction

The occurrence and causal explanations of the alpine treelines around the world have been discussed intensively for over a century (Brockmann-Jerosch, 1919; Griggs, 1946; Wardle, 1971; Tranquillini, 1979; Grace, 1989; Körner, 1998; Sveinbjörnsson, 2000). Among the many hypotheses related to stress, disturbance, reproduction, nutrient limitation, growth limitation, and carbon limitation, the latter two have provided the most plausible suites of explanations for causes of treeline formation from a global perspective (Körner, 1998).

The carbon limitation hypothesis states that low temperatures and short vegetation periods (in extratropical areas) increasingly prevent a positive carbon balance of the tree life form burdened with a large proportion of nonproductive tissues with increasing altitude or latitude. Climatic control of CO₂ exchange can be seen as the immediate cause of a reduced carbon balance and growth at treeline. However, different studies have illustrated that photosynthetic carbon uptake in treeline trees is relatively insensitive to temperature (Pisek & Winkler, 1958; Slatyer & Ferrar, 1977; Häsler, 1982; Goldstein et al., 1994). Pisek & Winkler (1958) in addition found only very small altitudinal differences in photosynthetic capacity or respiratory carbon losses at the leaf level. In a wider sense, the carbon limitation hypothesis also accounts for tissue damages and losses that can considerably reduce net carbon acquisition in the long term and can significantly impair tree growth in the hazardous treeline ecotone (Sveinbjörnsson, 2000). In support of the carbon limitation hypothesis, higher needle mortality and slower rates of shoot elongation were observed in white spruce trees at treeline compared to trees growing at lower altitudes in Alaska (Sveinbjörnsson, 2000). A recent study across an altitudinal gradient within the treeline ecotone (2080–2230 m asl) of the Alps, showed large variability in total tree biomass (up to 10-fold differences among individuals), but no correlation with altitude in equal-aged trees of three conifer species (Bernoulli & Körner, 1999), rather questioning the hypothesis of a carbon shortage in treeline trees. The alternative growth limitation hypothesis (Körner, 1998), assuming a sink rather than a source limitation of tree growth at treeline, states that tissue formation is temperature limited. If low temperatures restricted carbon investments more than C-acquisition, then an increasing accumulation of nonstructural carbohydrates with increasing altitude might be expected as it was indicated for Pinus cembra in the Central Alps (Hoch et al., 2002).

The continuing rise in atmospheric CO₂ concentration directly improves carbon balance and stimulates growth in many tree species (Ceulemans & Mousseau, 1994; Curtis & Wang, 1998; Norby et al., 1999) irrespective of any possible indirect effects of greenhouse warming on plant performance. It has been argued that trees at high altitudes may be particularly sensitive to CO₂ enrichment, because they live in an atmosphere of lower partial pressure of CO₂ (LaMarche et al., 1984). In line with this reasoning, gas exchange studies of herbaceous plant species (Körner & Diemer, 1987) and global comparisons of stable carbon isotope data (Körner et al., 1991), both gave clear evidence for a greater efficiency of CO₂ uptake at any given CO₂ partial pressure in plants from high altitudes, suggesting potentially greater sensitivity to CO₂ enrichment as compared to plants from lower altitudes. Several dendrochronological studies reported increasing tree ring widths in upper montane conifers over the last 150 yr (La Marche et al., 1984; Graybill & Idso, 1993; Nicolussi et al., 1995; Rolland et al., 1998; Paulsen et al., 2000), and rising CO₂ concentration was discussed as a possible explanation. Nonetheless, it remains difficult to tease apart all the potentially influential and interacting factors, such as changes in forest management, nitrogen deposition and recently increased temperatures (Kienast & Luxmoore, 1988). For example, the data set of La Marche et al. (1984) was found to be confounded with rainfall history (Graumlich, 1991).

The in situ exposure of treeline trees to a higher atmospheric CO₂ concentration may be the most promising approach to test whether or not the growth of these trees is stimulated by elevated [CO₂]. Such an experiment would also provide a direct test of the carbon limitation hypothesis of treeline formation.

Using the technique of pure CO₂ enrichment (Miglietta et al., 2001; Okada et al., 2001) we designed and constructed a free-air CO₂ enrichment (FACE) system for the exposure of trees and their understory vegetation in the alpine treeline ecotone at Stillberg, Davos in Switzerland (2180 m asl). The main hypotheses to be tested in this ongoing project are tree growth at treeline is limited by carbon assimilates, and the CO₂ response differs among species with likely consequences for biodiversity and structure of treeline ecotones. Here we report on the performance of the custom-made CO₂ supply and regulation system and the responses of the studied tree species to elevated CO₂ after the first growing season.

Materials and Methods

Study site and plant material

The experimental area is located within the long term research site at Stillberg, Davos in the Central Alps maintained by the Swiss Federal Institute for Snow and Avalanche Research (SLF), Davos as part of the Swiss Federal Research Institute WSL, Birmensdorf. The NE-exposed Stillberg research site was established in the late 1950s with the aim to explore afforestation techniques and the interaction with avalanche dynamics in the treeline ecotone. A large experiment was started in 1975 when a total of 92 000 individuals of high elevation provenances of three treeline species (Pinus uncinata Ramond, Pinus cembra L., Larix decidua L.) were planted across an area of 5 ha spanning an altitudinal range of 2080 and 2230 m asl. The design of the plantation allows the recognition of each individual seedling. Microclimate, snow dynamics, vegetation composition, gas exchange, tree growth and survival has been measured extensively over the last three decades (Kuoch & Amiet, 1970; Schönenberger, 1975; Häsler, 1982; Turner et al., 1982; Schönenberger & Frey, 1988; Senn & Schönenberger, 2001). These established trees of the same age (P. uncinata: 29-yr-old, L. decidua: 27-yr-old at the start of our experiment) provide a unique experimental setup for the study of CO₂ effects on treeline trees.

An area of approximately 2500 m² at 2180 m asl, that is, at the upper end of the long-term research site, was selected for our study. This particular area was chosen because it is at or maybe slightly above the actual natural treeline (uppermost native adult tree at 2180 m asl), it is characterized by a rather homogenous microrelief, slope (25–30°), exposure (NE) and understory species composition (see below), and power and CO₂ supply can be made available at a relatively close distance.

The long-term average annual precipitation at the study site is 1050 mm with a mean maximum snow depth of 1.46 m, and the average temperature is −5.8°C in January and 9.4°C in July (Schönenberger & Frey, 1988). The growing season starts approximately on 15 June with bud break of larch and ends 25 September with needle senescence of larch (i.e. c. 110 days). Temperature, wind speed, precipitation and snow depth during the experimental year are shown in Table 1. The soil is classified as a Ranker (U.S. system: Lithic Haplumbrept) with a 10-cm-deep organic top soil underlain by siliceous bedrock (Paragneis, Schönenberger & Frey, 1988).

Table 1.  Temperature, wind speed, precipitation and snow depth during the experimental year 2001 (meteorological station of the Stillberg research site at 2090 m asl, A. Streule)

Climatic parameter	Entire year	April	May	June	July	Aug	Sept	Oct
All values are based on hourly means. Air temperature was measured 2 m above soil surface (4 m during winter), soil temperature at 0.1 m below the ground, and wind speed at 10 m above the ground. Sums of precipitation and monthly means of snow depth are shown (* maximum snow depth as it was measured in 2001).
T air mean (°C)	2.3	−1.8	6.9	6.7	10.5	11.8	3.1	7.4
T air min (°C)	−19.2	−13.0	−2.7	−5.0	0.7	1.9	−4.1	−0.3
T air max (°C)	21.7	14.6	17.2	19.9	21.5	21.7	13.2	18.4
T soil mean (°C)	2.9	0.2	1.8	5.9	9.0	9.6	4.1	3.6
T soil min (°C)	−  0.3	0.2	0.2	1.8	5.8	6.4	2.3	1.0
T soil max (°C)	12.3	0.2	9.7	10.6	11.7	12.3	7.5	6.1
Wind mean (m s^−1)	1.4	1.8	1.8	1.6	1.7	1.5	1.3	1.1
Precipitation (mm)	1520	109	61	279	216	187	158	70
Snow depth (m)	1.69*	1.45	0.51	0.02	0	0	0.07	0

Except for the trees planted in 1975, there are no other trees within the study area. Survival of the planted trees has varied considerably such that within the area, L. decidua became the most abundant tree species. There are fewer P. uncinata individuals and only three individuals of P. cembra. This tree species composition is characteristic for the entire Stillberg research site and is explained by considerably higher mortalities in P. uncinata and especially in P. cembra due to fungal attacks (Senn & Schönenberger, 2001). The trees are not taller than approx. 1.5 m and are widely spaced, forming an open canopy with dense understory vegetation composed of 33 different plant species. The understory is dominated by the dwarf shrubs Vaccinium myrtillus, Vaccinium uliginosum, and Empetrum hermaphroditum. Vaccinium vitis-idaea, Loiseleuria procumbens, and Rhododendron ferrugineum are also present, but are less abundant. Gentiana punctata, Homogyne alpina, and Melampyrum pratense are the most common herbaceous species.

Experimental design

A total of 40 trees, 20 of each of the two species L. decidua and P. uncinata, together with their understory vegetation were selected for the experiment in early June 2001. The trees had to meet the following criteria for selection: intact terminal leading shoot; no signs of serious herbivory and/or diseases; total height between 0.8 and 1.5 m; and no more than one close (> 80 cm in distance) neighboring tree.

Forty hexagonal plots covering a surface area of 1.1 m² each and with one tree in the center were established, beginning with snowmelt on 12 June 2001. The comparatively small, single-tree plot size was chosen because trees are typically widely spaced with limited interactions among individuals in the treeline ecotone. Furthermore, statistical and technical considerations both clearly favored single-tree plots. Due to the much lower abundance of pine trees than larch trees, single-tree plots enabled us to include sufficient and equal numbers of individuals of both species which would not have been possible with larger and consequently fewer plots. Given the steep and uneven terrain of the study site, it would have become also increasingly difficult to set up technical constructions and maintain a homogenous CO₂ concentration with larger plot sizes. A hexagonal stainless steel frame, held in a horizontal position by three wooden posts, was used to mount a ring of 24 vertical plastic tubes (15 cm apart from each other), individually cut according to plot-specific variations in microrelief, surrounding each plot (Fig. 1). The commercially available plastic tubes, designed originally for conventional drip irrigation systems (Drip Store Inc., Escondido, CA, USA), have an inner diameter of 4.3 mm and laser-drilled holes of uniform diameter (0.5 mm, one every 15 cm on one side). Through these holes, oriented to the center of the plot, CO₂ jets were injected. To keep the tubes straight and rigid, a stainless steel rod of 3 mm in diameter was inserted into each tube and the sealed end of the tube was inserted 2–4 cm into the soil. These relatively light constructions around each plot would make it easy to adjust plot height and width in accordance with tree growth, if needed during the course of the experiment.

Figure 1. Experimental plots at the research site Stillberg, Davos, Switzerland, 2180 m asl. The construction for support of CO₂ emitting tubes surrounding a tree in the center of each plot and its understory vegetation is shown in the two pictures on top. Tubes for CO₂ emission extend from the top (slightly above the tree top) all the way down to the soil surface to ensure complete CO₂ exposure of all the vegetation within the plot (control plots look the same). A schematic view from the top of an experimental plot is shown below. The hexagonal metal frame on top of the plot holding all tubing, arrangement of CO₂ supply and CO₂ emitting tubes along each cardinal direction, and the position of CO₂ sampling lines within the plot are shown.

Half of all plots were randomly assigned to an elevated [CO₂] atmosphere (growing season average of 566 ppm, that is, roughly twice the preindustrial concentration) and the other half served as control plots at a current ambient [CO₂] (c. 367 ppm), resulting in a replication of 10 plots per CO₂ treatment and per tree species. At the altitude of our research site, the treatment CO₂ concentrations correspond to a partial pressure of CO₂ of 45 Pa (elevated) and 29 Pa (ambient). Initial tests of CO₂ enrichment started on 16 June 2001 and CO₂ exposure of all 20 plots began on 26 June, just before budbreak of the trees, and continued until 20 September 2001, when larch needles started to turn yellow.

Carbon dioxide consumption during the first month of the experiment was considerably higher than initially calculated due primarily to strong winds at the experimental area. Although this did not markedly affect the performance of the CO₂ regulation system and the maintenance of the target [CO₂], budgetary considerations forced us to construct wind shields facing the main wind direction (N) in order to reduce costs. The 40 wind shields (treatment and control plots) were made of transparent 0.75 mm thick polyethylene film (Melinex 400, Hifi Industrial Film Ltd, Al Hoorn, NL, USA) that is neutral to light transmission in the visible spectrum as well as to UV. These screens were attached directly to the horizontal steel frame and the vertical CO₂ emitting tubes and covered one third of the total plot circumference from the top to c. 0.2 m above the soil surface. Thus, the understory vegetation was not screened and free access for pollinators and ground-living insects were guaranteed from all sides of the plots.

Pure CO₂ enrichment system

The relatively new technology of free-air CO₂ enrichment (FACE) using pure CO₂ injection was chosen as the technically and logistically most suitable method for CO₂ enrichment in the uneven terrain and remote location of this alpine treeline ecotone. The most apparent difference of the ‘pure CO₂ release’ technology compared to the traditional FACE technology is the omission of blowers and the provision of premixed CO₂-enriched air of a particular set point [CO₂]. Free-air CO₂ enrichment with pure CO₂ injection is successfully being used in at least three ongoing long-term and large-scale experiments with rice (Okada et al., 2001), a poplar plantation (Miglietta et al., 2001), and a mature deciduous forest (Pepin & Körner, 2002). These studies reported highly reliable CO₂ control and a similar performance of the pure CO₂ system compared to the traditional FACE system, with the additional advantage of comparatively low construction and maintenance costs. An effective mixing of CO₂ with the bulk air is achieved because CO₂ is released under high pressure (5 bar) producing a rapidly distributing CO₂ jet (Miglietta et al., 2001) and because of the large number of small laser-drilled holes along the plot edges (in our case an average of 215 CO₂ emitting holes per 1.4 m³ volume of air).

Six ‘batteries’ of 12 single bottles of totally 480 kg liquid CO₂ (i.e. 2880 kg CO₂ in sum, with a total freight weight of 9 tons) were stored on a helicopter platform adjacent to the research station at 2230 m asl. Because a hiking trail is the only access to the research site, the ‘CO₂-batteries’ had to be brought in by helicopter (Rotex AG, Liechtenstein) from the nearest road at 1600 m asl biweekly. Carbon dioxide is delivered to an electric heat-exchanger, vaporized and supplied to an array of two-way normally closed solenoid valves (EVT317, SMC Pneumatik, Engelsbach, Germany). These solenoid valves are part of the custom-made CO₂ control and measurement system in the center of the experimental area about 100 m in distance from the helicopter platform where the CO₂ batteries are stored (Fig. 2). Four CO₂ supply lines were used per plot allowing independent CO₂ injection rates for each cardinal direction. The injection valves were actuated separately by the control program run on an industrial PC placed in the field (Fig. 2) via 24-volt DC solenoids. A pulse-width modulation routine adjusted the duration of the pulse (max. 2 s) used to drive the injection solenoids as a function of measured [CO₂] in the tree canopy in each cardinal direction (CO₂ sampling approximately 0.25 m from the center of the plot and 50 cm above the ground, Fig. 1). The target [CO₂] was set at 550 ppm.

Figure 2. Experimental setup of the supply, control, monitoring and field enrichment of CO₂. The control and monitoring system in the field contained the infrared gas analyzer (IRGA) for CO₂ measurement, the industrial control PC and the solenoid valves for CO₂ flow control (shown in detail on the right). The field-based control computer was connected to a second PC within the Stillberg research facility (Hüttiboden) via fibre optic cable connection. Four enriched plots are supplied with CO₂ according to actual CO₂ concentrations measured within a reference plot (a total of five reference plots for 20 enriched plots). Single bold letters indicate cardinal directions measured for and supplied with CO₂ individually.

Control signals and data logging were implemented using a custom control program run on the industrial PC in the field. This CO₂ control system in the field was located in the center of the research area and contained the CO₂ release and monitoring systems, an infrared gas analyzer (IRGA, LI-800, Li-Cor Inc. Lincoln, NE, USA), two vacuum pumps, and digital boards driving the solenoid valves (Fig. 2). The field-based system was connected to a second computer located in the research station about 160 m from the research area by a fibre optic serial cable (Fig. 2). This second PC was used for data storage, viewing and processing, for modifying the control program and for communication with the Botanical Institute in Basel via modem.

Temporal and spatial variability of concentrations within experimental plots was determined using a customized 24-port sequential sampler in connection to the IRGA. Twenty sampling lines were used to monitor [CO₂] in each cardinal direction of five reference plots and to drive the injection solenoids. The reference plots were carefully selected to represent another three CO₂-enriched plots that were supplied with CO₂ by the same 1 cm I.D. polyurethane tubing (connected by manifolds to one common solenoid valve per cardinal direction, Fig. 2). The remaining four sampling lines were regularly moved within and among plots during the first three weeks of the experiment in order to fine-tune the system performance. Afterwards they were installed in the center of four plots to continuously measure [CO₂].

Air from CO₂-enriched plots was continuously drawn by a vacuum pump (YP-70VC, ASF Thomas, Wuppertal, Germany) through all the sampling lines (6 mm I.D. polyurethane tubing). Sampled air was then pumped sequentially from each port through a manifold of 24 three-way solenoid valves at a flow rate of 1 l min⁻¹ (YP-40VC, ASF Thomas, Wuppertal, Germany) and routed through the gas analyzer. Each channel was monitored during 15 s, allowing sufficient time to purge the measurement system. After each measurement cycle an additional measurement of a calibration gas (391 ppm CO₂) was taken to verify the stability of the IRGA. A given sampling line was scanned approximately once every 6 min. Carbon dioxide readings from the IRGA were monitored at 1-s intervals, and only the last reading was recorded.

Plant sampling and data analysis

Length of the current-year leading shoot and of five mid-canopy lateral shoots was measured monthly in all CO₂-enriched and control trees of both species beginning in July and ending in late September 2001. Total length attained by the same shoots at the end of the previous year (2000) was additionally measured and treated as a covariable in the statistical tests of treatment effects on shoot length increment. Numbers of current-year needles per unit shoot length, projected area and needle dry mass (oven dried at 80°C) were determined in one fully mature, lateral shoot of each individual tree harvested on 24 August 2001 between 18:00 and 20:00 h local time.

The same needle material from the harvest described above was ground and used for chemical analyses. Nitrogen and carbon concentrations (% of dry mass) were determined with a CHN-analyzer (Model 900, LECO Instruments, St. Joseph, Michigan, USA). Nonstructural carbohydrates (NSC = starch, sucrose, glucose and fructose) were analyzed using an enzymatic starch digestion and a spectrophotometric glucose test after invertase and isomerase addition (Körner & Miglietta, 1994).

Gas exchange of intact current-year shoots (the same shoots that were harvested thereafter, see above) of all trees was measured during three consecutive days in late August (22nd to 24th) 2001 between 9.30 and 15.30 h. Gas exchange was measured at treatment CO₂ concentrations and saturating light levels (natural full sunlight > 1200 µmol photons m⁻² s⁻¹) using the Li-Cor-6400 photosynthesis system (Li-Cor Inc., Lincoln, Nebraska, USA) and the conifer chamber (LI-6400–05). Total shoot length enclosed in the gas exchange chamber ranged between 35 and 60 mm in larch and between 20 and 40 mm in pine, with a total of 111–383 individual needles with a projected leaf area of 14.3–45.7 cm² in larch and 53–136 needles with a projected leaf area of 30.0–54.6 cm² in pine. For all measurements the internal Li-Cor-6400 CO₂ control was used to achieve treatment CO₂ concentrations. Conifer cuvette temperature (21.1 ± 0.4°C) and air humidity (77.3 ± 6.3% RH) were kept constant. Leaf temperature during measurements ranged between 20°C and 24°C.

The overall model for data analyses was a 2 × 2 model I analysis of variance to test for differences between species and CO₂ treatments with a replication of n= 10 plots. Multivariate repeated measures analysis of covariance was used to test for the effects of species, CO₂, and shoot length attained during the last year's growing season (covariable) on shoot length increment over time (three dates in 2001). To meet the requirement of normal distribution, percentage data (e.g. N concentration) were transformed with arcsine [square root (y)] before analyses.

Results

CO₂ control in the treeline FACE system

The FACE system using pure CO₂ injection was operational during daytime from 26 June to 20 September 2001, that is, 95% of the growing season. Records of CO₂ concentrations within high [CO₂] exposed plots throughout the growing season demonstrated that the system could adequately maintain CO₂ concentrations close to the target concentration of 550 ppm. We measured an average daytime CO₂ concentration of 566 ppm (partial pressure of 45 Pa) in the elevated CO₂ plots over the entire growing season compared to an ambient [CO₂] of 367 ppm (partial pressure of 29 Pa). Representative diurnal curves of instantaneously measured [CO₂] within the high [CO₂] exposed area of the five reference plots showed comparatively small fluctuations around the target concentration (Fig. 3). The diurnal averages of [CO₂] ranged from 550 to 575 ppm with relatively few individual measurements below 450 ppm or over 800 ppm (Fig. 3). The frequency distribution of [CO₂] during that same period from 22 to 28 July 2001 showed more than 70% of all readings within the range of 450 and 650 ppm and a median [CO₂] of 548 ppm (Fig. 4). Carbon dioxide concentrations lower than 450 ppm were recorded in 10% of all 1-s IRGA readings, and less than 4.5% of all 1-s IRGA readings were above 800 ppm (Fig. 4). Short-term excursions of concentrations exceeding 1000 ppm were rarely measured (< 0.1%). Over the entire growing season, 11.1% of all readings were below 450 ppm, and 5.5% were higher than 800 ppm (mean across the five reference plots). Sixty-nine percent of these readings were within the range of 450 and 650 ppm. Based on these measurements, the CO₂ regulation system is estimated to control CO₂ levels within ±10% of the target concentration for 42% of the exposure time, and within ±20% for 74% of the total exposure time. This estimate is based on instantaneous 1-s readings of [CO₂] taken every 15 s (grab samples). From four such 1-s readings consecutively measured at each cardinal direction within a given plot, we calculated ‘one-minute averages’. These 1-min averages of [CO₂] were for c. 63% of the total exposure time within ±10%, and for c. 90% of the total exposure time within ±20% of the target concentration.

Figure 3. Diurnal course of CO₂ concentrations (enriched for 12 h from 7 am to 7 pm) measured within five CO₂-enriched plots over a representative mid summer week in July 2001 (measurements are recorded every six minutes). The mean of five reference plots (black hairline) ± the standard deviation (grey area) are shown. Bold numbers at the bottom of the graph indicate mean daytime CO₂ concentrations (± SD).

Figure 4. Frequency distribution of daytime CO₂ concentrations (1 s readings) measured within the five reference plots during a representative week in July (22.–28.7. 2001). Bars represent mean values ± SE of 5 plots.

Carbon dioxide sampling lines in the center of the plots served to further explore within and among plot variation in [CO₂]. The characterization of the variation among plots was particularly important in assessing the suitability of CO₂ regulation for four plots based on CO₂ measurements at the four peripheral positions within just one of these plots (reference plot). Representative diurnal curves of [CO₂] indicated that similar CO₂ atmospheres could be maintained among plots (Fig. 5). Although the time course of [CO₂] differed somewhat among plots, mean [CO₂] varied little and daytime minimum and maximum values were comparable (Fig. 5). Not surprisingly, there was some gradient of [CO₂] from the periphery towards the center of the plots. Mean CO₂ concentrations in the center were roughly 50 ppm lower than those measured at the periphery used for CO₂ control and regulation. Increasing wind speed resulted in somewhat higher spatial and temporal variation in [CO₂] within plots, but had no detectable influence on the daytime mean [CO₂]. Maintaining the target CO₂ value during very windy conditions, however, markedly increased the overall CO₂ consumption which was the reason for the construction of transparent wind shields facing the main wind direction (N). The relative insensitivity of the CO₂ regulation system to changes in wind speed regardless whether the small sized wind shields were present or not, may be explained by the low stature of the experimental trees, and more importantly, by the comparatively small plot size.

Figure 5. Diurnal course of CO₂ concentrations measured in the center of three CO₂-enriched plots over two days in July 2001. The solid black line shows the CO₂ concentrations measured in the center of the reference plot used to control and regulate the CO₂ concentrations for the other plots. The grey and dashed lines represent CO₂ concentrations for two additional plots regulated according to measurements made in the reference plot. Numbers within the figure indicate the mean, maximum and minimum CO₂ concentrations for all three plots.

To maintain [CO₂] close to our target of 550 ppm within a total of 20 CO₂ enriched plots of a total volume of 20 × 1.1 m² × 1.3 m (average height of CO₂ emitting tubes) which is 28. 6 m³, we used an average of about 20 kg CO₂ per hour. This corresponds to a total CO₂ consumption of 18.5 tons per growing season (c. 105 days).

Responses of treeline conifers to elevated [CO₂]

Light-saturated net photosynthesis (A_max) of fully mature current-year shoots developed under treatment conditions was significantly higher in larch than in pine trees (Tables 2 and 3). Both species showed a significantly increased photosynthetic CO₂ uptake of current-year needles when grown in a CO₂ enriched atmosphere (Tables 2 and 3). In addition, we observed a highly significant species × CO₂ interaction on A_max expressed per unit of needle dry mass (Table 3). This interaction term was significant because larch had a much higher absolute CO₂ response (+0.033 µmol g⁻¹ s⁻¹) compared to pine (+0.010 µmol g⁻¹ s⁻¹), even though the two species showed a similar relative CO₂ stimulation of A_{max (mass)}. Similar to A_max, stomatal conductance (g_s) was also considerably higher in larch than in pine (Tables 2 and 3). Elevated [CO₂] reduced g_s in both species, the difference between CO₂ treatments being more pronounced in larch than in pine (Table 2).

Table 2.  Density, mass, gas exchange and quality of fully developed current-year needles in the two studied tree species grown at either ambient (Amb.) or elevated (Elev.) atmospheric CO₂ concentrations

	Larix decidua			Pinus uncinata
	Amb. CO2	Elev. CO2	Δ percentage	Amb. CO2	Elev. CO2	Δ percentage
Means ± SE of n= 10 trees at the end of August 2001 are shown along with the relative difference between ambient and elevated CO2.
Needle density(# cm^−1 shoot)	54± 4	50 ± 6	−7%	36± 3	35 ± 2	−3%
Needle mass (100 g of needles)	0.123 ± 0.009	0.133 ± 0.006	+8%	1.086 ± 0.091	1.235 ± 0.068	+14%
SLA (cm^−2 g^−1)	92.2 ± 2.9	89.3 ± 2.7	−3%	37.7 ± 1.7	33.9 ± 1.0	−10%
Amax (area)	10.1 ± 0.3	14.7 ± 0.6	+46%	7.1 ± 0.2	11.1 ± 0.4	+56%
Amax (mass)(µmol g^−1 s^−1)	0.094 ± 0.005	0.127± 0.006	+35%	0.027 ± 0.001	0.037± 0.001	+37%
gs (mmol m^−2 s^−1)	215 ± 15	169 ± 14	−21%	139± 6	117 ± 9	−16%
Nitrogen(% of dry mass)	2.30 ± 0.07	2.11 ± 0.09	−8%	1.21 ± 0.03	1.16 ± 0.02	−4%
NSC(% of dry mass)	18.3 ± 0.5	21.4 ± 1.0	+17%	10.4 ± 0.8	14.3 ± 0.7	+38%
Sugar(% of dry mass)	12.1 ± 0.4	13.1 ± 0.4	+8%	5.1 ± 0.3	5.7 ± 0.1	+12%
Starch(% of dry mass)	6.2 ± 0.6	8.3 ± 0.9	+34%	5.3 ± 0.6	8.6 ± 0.6	+62%

Table 3.  Analysis of variance of needle parameters to test for differences between species, CO₂ concentrations and their interactions

Source of variance	Needle parameter	F	P	Needle parameter	F	P
Species	Needle	16.87	0.0002	gs	31.80	< 0.0001
CO2	density	0.425	0.519		8.696	0.0056
Sp × CO2		0.102	0.752		1.189	0.283
Species	Needle	335.5	< 0.0001	Nitrogen	348.0	< 0.0001
CO2	mass	1.970	0.170		4.356	0.044
Sp × CO2		1.499	0.229		1.002	0.324
Species	SLA	615.8	< 0.0001	NSC	92.56	< 0.0001
CO2		2.797	0.104		21.00	< 0.0001
Sp × CO2		0.137	0.714		1.123	0.296
Species	Amax	68.31	< 0.0001	Sugar	605.3	< 0.0001
CO2	(area)	111.3	< 0.0001		8.287	0.0067
Sp × CO2		0.670	0.419		0.023	0.881
Species	Amax	385.3	< 0.0001	Starch	0.255	0.617
CO2	(mass)	30.31	< 0.0001		15.13	0.0004
Sp × CO2		8.664	0.0059		1.135	0.294

Nitrogen concentrations determined in the same needles used for gas exchange measurements were almost twice as high in larch compared to pine (Table 2). Elevated [CO₂] had a significantly negative effect on N concentration (Table 3), but the effect was comparatively small. Concentrations of nonstructural carbohydrates (NSC) were 50% to 75% higher in needles of larch compared to pine (Table 2). Higher NSC in larch was exclusively due to higher sugar concentrations whereas concentrations of starch were essentially the same in the two species. Concentrations of NSC were significantly higher in needles grown in a CO₂ enriched atmosphere compared to controls, and this effect resulted mainly from starch accumulation with sugar concentration less affected. NSC in branches (pine only, including wood and bark) grown in elevated [CO₂] contained 10.1% NSC of their total dry mass compared to 8.8% in controls (F_1,18 = 3.228, P = 0.093). Roughly two thirds of the overall NSC in branches were sugars that did not respond to elevated [CO₂]. By contrast, starch concentrations increased significantly from 2.9% at ambient [CO₂] to 3.7% at elevated [CO₂] (F_1,18 = 5.630, P < 0.05).

Specific leaf area (SLA) and needle dry mass differed greatly between species (Table 2). The overall CO₂ effect on SLA and needle dry mass was not significant (Table 3). However, testing the CO₂ effect within species with separate one-way ANOVAs, pine showed a lower SLA (F_1,18 = 12.711, P < 0.01) at elevated [CO₂] than at ambient [CO₂], but no difference in needle mass (F_1,18 = 1.560, P= 0.230). Needle density per unit of shoot length was higher in larch than pine, but was not affected by CO₂ in either species (Table 2).

Current-year shoots were longer in larch than in pine (Fig. 6, Table 4) irrespective of CO₂ treatment. Length increment in leading and lateral shoots increased significantly in response to CO₂ enrichment in both species (Fig. 6, Table 4). At the end of the growing season 2001, current-year leading shoots were on average 25 mm longer (+23%) in larch and 11 mm longer in pine (+22%) when grown at elevated [CO₂]. The current-year increment of lateral shoots was on average 13 mm larger (+18%) in larch and 5 mm larger (+13%) in pine in a CO₂ enriched atmosphere compared to ambient [CO₂].

Figure 6. Length increment of leading shoots (top) and lateral shoots (bottom) of Larix decidua (triangles) and Pinus uncinata (circles) grown at either current ambient CO₂ concentrations (open symbols) or at elevated CO₂ concentrations (closed symbols) during the first growing season of treatment conditions. Data points are mean ± SE of 10 measured trees (one leading shoot and five lateral shoots per tree were measured). Shaded areas highlight the differences between CO₂ treatments within each of the two species.

Table 4.  Multivariate repeated measures analysis of covariance to test for effects of species and CO₂ on length growth in leading and lateral shoots during the 2001 growing season

Source of variance	Leading shoots		Lateral shoots
	F	P	F	P
Shoot length of the previous year was included in the model as a covariable.
Shoot length in 2000	40.30	< 0.0001	48.71	< 0.0001
Species (Sp)	41.26	< 0.0001	49.02	< 0.0001
CO2	5.799	0.022	4.649	0.038
Sp × CO2	2.645	0.113	2.611	0.115
Day (d)	0.028	0.973	2.047	0.145
d × Sp	7.006	0.003	6.785	0.003
d × CO2	0.379	0.687	0.404	0.671
d × Sp × CO2	0.033	0.968	0.167	0.847

Discussion

Performance of the treeline FACE

The design and setup of an experimental system to expose native plants in the alpine treeline ecotone to an elevated atmospheric CO₂ concentration was challenging because of the difficult access (no roads creating CO₂ transport problems), and the steep and uneven terrain. These constraints limited the construction of large sized field installations and made the use of a traditional free-air CO₂ enrichment (FACE) system (Hendrey et al., 1993, 1999) impractical. FACE rings typically have a diameter of > 20 m and use around 2 tons CO₂ per ring and per day. While considerably reducing CO₂ consumption, the use of tall open top chambers (OTCs) would have posed other problems, technical difficulties for field installation, large changes in microclimate to the frequent wind occurrence, and high radiation being the most obvious ones. Hence, we customized the pure CO₂-release technology (Miglietta et al., 2001; Okada et al., 2001; S. Roberts, pers. comm.) for the peculiar situation on a mountain slope. The CO₂ control system described here was able to maintain [CO₂] within ±20% of the target value for 90% of the exposure time (1-min means), similar to the > 91% and 90% in the pure CO₂ FACE systems described, respectively, by Miglietta et al. (2001) and Okada et al. (2001), and to the 92% in the ‘traditionally designed’ Duke forest FACE (Hendrey et al., 1999). One minute average [CO₂] within ±10% of the target value were recorded for 63% of the exposure time in this study which is similar to the 60% reported for the rice FACE (Okada et al., 2001), but somewhat lower than the 69% for the Duke forest FACE (Hendrey et al., 1999) and the 75% for the POPFACE (Miglietta et al., 2001). It should be emphasized that our 1-min averages actually derived from four 1-s measurements every 15 s, and therefore rather underestimate the proportion of exposure time within a certain limit of target compared to the estimates of other FACE systems based on averages of continuous readings.

Spatial distribution of [CO₂] was rather homogenous within the exposed volume of air. Continuous measurements at the periphery and in the center of the CO₂ enriched plots at mid-height of the tree canopies showed rarely differences exceeding c. 60 ppm CO₂. Extensive [CO₂] measurements during the initial phase of system setup and in a prototype system constructed in Basel before the actual experiment started, indicated decreasing [CO₂] from the bottom to the top of the plot (data not shown). This [CO₂] gradient was commonly less than 100 ppm with [CO₂]c. 50 ppm higher than the target at the bottom and c. 50 ppm lower than the target at the top of the plot, resulting in more than 80% of the total exposed volume being within ±10% of the target. Spatial variability of [CO₂] in our system was lower than in other pure CO₂ fumigation systems (Miglietta et al., 2001; Okada et al., 2001; Pepin & Körner, 2002). This is explained by the much smaller plot size chosen here and by the use of a vertical array of small diameter laser-drilled CO₂ emitting holes of a high density (an average of 215 holes per 1.4 m³ volume of air). The four independently acting CO₂ control points within such a small volume of air, enabled us to achieve a highly fine-tuned CO₂ control and regulation. The spatially close arrangement of [CO₂] control points used in the CO₂ regulation algorithm was the reason why measurements of wind speed and direction were not needed for a reliable feedback algorithm, which may be the most evident difference to other FACE systems that need controlling for [CO₂] over an area of up to 650 times the size of our plots.

Short-term variability in [CO₂] and excursions of very high [CO₂] (> 1000 ppm) are likely to occur more frequently in any FACE using pure CO₂ injection than in FACE operated with premixed air using blowers (Pinter et al., 2000). Such fluctuations in [CO₂] can affect physiological processes in plants (Cardon et al., 1995). However, we recorded [CO₂] higher than 1000 ppm less than 0.1% of the time and [CO₂] deviating more than ±20% of the target value lasted seldomly longer than a few seconds. It takes at least one minute to induce changes in stomatal conductance in most plants and even longer in trees (Ellsworth et al., 1995; Hendrey et al., 1997; Saxe et al., 1998). Therefore, it is unlikely that short-term variations in [CO₂] observed here had any significant effect on plant responses to elevated [CO₂].

Using pure CO₂ emitted through fine tubing with very small diameter laser-drilled holes under high pressure provided a useful alternative type of FACE system for the CO₂ enrichment of vegetation in the treeline ecotone. The fine tubing, originally designed for irrigation purposes and first used for atmospheric CO₂ enrichment by Steven Roberts from San Diego State University (pers. comm.) is flexible in its application and might be the only alternative for in situ CO₂ enrichment in some ‘difficult’ environments. It has even been modified for use in 35 m tall old-growth forest canopies (web-FACE, Pepin & Körner, 2002), until recently not believed to be possibly studied in CO₂ research.

Growth of alpine treeline conifers in a CO₂ enriched atmosphere

The physiological responses to elevated [CO₂] observed at the needle and branch level in the two tree species studied, are among the most consistent plant responses to atmospheric CO₂ enrichment (Ceulemans & Mousseau, 1994; Poorter et al., 1997; Norby et al., 1999; Körner, 2000). The mean enhancement in photosynthesis of 51% (on a needle area basis) in response to elevated [CO₂] measured in needles developed after the initiation of the CO₂ treatment, compares well with the mean stimulation of 66% calculated from a number of studies with trees growing in the field (Norby et al., 1999), and even better if only conifers are considered (53%, Norby et al., 1999). The CO₂ responses in stomatal conductance (g_s) reported in the literature are not consistent and range from no differences to comparatively large reductions (Curtis & Wang, 1998; Norby et al., 1999). A frequently confirmed pattern, however, is that conifers show less and often not significant reductions in g_s than do deciduous tree species (Saxe et al., 1998). In comparison to recent studies with different conifer species showing no or only moderate (up to −14%) CO₂ effects on g_s (Dixon et al., 1995; Tissue et al., 1997; Wang & Kellomäki, 1997; Ellsworth, 1999), we found slightly greater and significant responses to elevated [CO₂] in the present study. However, the CO₂ effect on g_s can change somewhat over the course of the season (Egli et al., 1998) which was not assessed here. Moreover, larch as the more responsive of the two studied species functionally compares better with broadleaf deciduous species than with conifers.

The data compilation by Norby et al. (1999) showed an average decrease in leaf nitrogen concentration of 11% in conifers that is a little more than was found here. Lower leaf [N] in larch and pine trees observed here, was exclusively due to a dilution effect of higher nonstructural carbohydrate (NSC) concentrations under elevated CO₂. If expressed on either a leaf area basis or on NSC-free needle dry mass the CO₂ effect on needle [N] disappears (data not shown). These data suggest that CO₂ is unlikely to have had any effect on N allocation during that first year of exposure.

Both tree species showed immediate growth responses upon CO₂ enrichment in the first growing season. While such fast responses to increasing CO₂ are commonly found when starting with small seedlings, particularly when they are not limited by other resources (Ceulemans & Mousseau, 1994; Norby et al., 1999), it was a rather unexpected result in our study with comparatively old trees, growing in the densely vegetated treeline ecotone. Trees competing for resources other than CO₂ and rooting in soils of naturally low fertility showed no or only moderate CO₂ induced growth stimulation in several previous studies (Norby et al., 1992; Hättenschwiler & Körner, 1998; Spinnler et al., 2002). Moreover, current-year shoot elongation in trees with a determinate shoot growth pattern, such as pine and larch, is believed to be largely determined by previous year's carbon balance and bud formation. For that reason, CO₂ effects on shoot growth – if any – were expected to occur in the second year of growth under treatment conditions at the earliest. The up to 23% increase in length growth of current-year shoots exposed to elevated CO₂, thus, is considered a particularly strong response to CO₂ enrichment in the trees studied here.

Higher rates of photosynthetic carbon assimilation, increased accumulation of nonstructural carbohydrates in leaves and branches, and increased shoot growth, all are strong evidence for a significantly improved carbon balance in larch and pine trees growing in a CO₂-enriched atmosphere at treeline. With regard to our first hypothesis (tree growth at treeline is carbon limited), these results suggest that tree growth at the upper alpine treeline might indeed be limited by carbon availability and that rising atmospheric [CO₂] can stimulate tree growth in the treeline ecotone. However, this is a preliminary conclusion based on first-year data and needs verifying in the coming years with an extended assessment of growth including measurements of stem diameter increment and root growth. A long-term perspective is generally important in field experiments due to the variability of climatic factors affecting the CO₂ response (Körner, 2000), and particularly in studies involving long-lived plants such as trees, because diminishing CO₂ responses over time are likely (Körner, 1995; Loehle, 1995; Hättenschwiler et al., 1997; Oren et al., 2001). The first-year data reported here form an important baseline for a multiyear evaluation of the responses of these trees to atmospheric CO₂ enrichment, as necessary for the testing of hypotheses about causes of treeline formation and possible consequences for treeline dynamics in response to global change.

There is little evidence in support of our second hypothesis (tree species differ in their response to CO₂) so far, because larch and pine trees were similarly affected by elevated CO₂. Nonetheless, the significantly greater CO₂ effect on photosynthesis per unit of needle dry mass in larch compared to pine may be indicative of future stronger responses of this deciduous conifer in the long run. Needless to say that this remains speculative because CO₂ effects on rates of photosynthesis and growth rarely correlate (Curtis et al., 1996), and sustained CO₂ stimulation of leaf-level photosynthesis at times may have no detectable influence on growth in the same plants (Hättenschwiler & Körner, 1996; Egli et al., 1998). On the other hand, minute not yet identified differences in the CO₂ responses between species might gain in importance over time, particularly if biotic interactions such as those with understory species, herbivores or mycorrhizae come into play.

Acknowledgements

We thank Steven Roberts for discussions and ideas for setting up the pure CO₂ fumigation system, Steeve Pepin for help in constructing a prototype, Franz Leuenberger and his team for their logistical support, Lukas Zimmermann for software development, Thomas Zumbrunn for the characterization of the understory vegetation and the photographs in Figs 1 and 2, Olivier Bignucolo for chemical analyses, Alain Studer for technical support, and Arnold Streule for providing meteorological data. This research was funded through the Swiss National Science Foundation grant 31–061428.00 to S.H. and a doctoral scholarship (Fonds québecois de la recherche sur la nature et les technologies) to I.T.H.