Effects of long-term elevated [CO2] from natural CO2 springs on Nardus stricta: photosynthesis, biochemistry, growth and phenology

Authors


Andrea C. Cook Lawrence Livermore National Laboratory, Center for Accelerator Mass Spectrometry, PO Box 808, L397 Livermore, CA 94551–9900, USA. Fax: 925 423 7884; e-mail: acook@llnl.gov

Abstract

Plants of Nardus stricta growing near a cold, naturally emitting CO2 spring in Iceland were used to investigate the long-term (> 100 years) effects of elevated [CO2] on photosynthesis, biochemistry, growth and phenology in a northern grassland ecosystem. Comparisons were made between plants growing in an atmosphere naturally enriched with CO2 (≈ 790 μmol mol–1) near the CO2 spring and plants of the same species growing in adjacent areas exposed to ambient CO2 concentrations (≈360μmol mol–1). Nardus stricta growing near the spring exhibited earlier senescence and reductions in photosynthetic capacity (≈25%), Rubisco content (≈26%), Rubisco activity (≈40%), Rubisco activation state (≈23%), chlorophyll content (≈33%) and leaf area index (≈22%) compared with plants growing away from the spring. The potential positive effects of elevated [CO2] on grassland ecosystems in Iceland are likely to be reduced by strong down-regulation in the photosynthetic apparatus of the abundant N. stricta species.

INTRODUCTION

Short-term elevated [CO2] experiments have shown that most plant species initially exhibit higher rates of photosynthesis and enhanced growth with exposure to elevated levels of atmospheric [CO2] (Strain & Cure 1985; Ceulemans & Mousseau 1994). Extending the duration of elevated [CO2] exposure to weeks or years, however, often leads to ‘down-regulation’, that is, a reduction or loss of the initial photosynthetic response (Tissue & Oechel 1987; Sage, Sharkey & Seemann 1989; Jacob, Greitner & Drake 1995); but this does not always occur (Sage 1994). Because the extent of photosynthetic down-regulation has been shown to vary considerably with species, developmental stage, temperature, resource availability and internal factors, such as source–sink balance (Bowes 1991; Long 1991; Stitt 1991), there is great uncertainty associated with making predictions about the long-term effects of elevated [CO2] on plant growth and ecosystem function.

Despite the fact that wild plants contain 99% of the global pool of biomass carbon [only about 1% is stored in agricultural crops and managed pastures; Acock & Allen (1985)], very few elevated [CO2] experiments have focused on wild plants growing in natural ecosystems. To date, only five experiments have exposed natural ecosystems to elevated [CO2] for more than two seasons. These include studies of the Arctic tundra (Grulke et al. 1990; Oechel et al. 1994), alpine grassland (Körner et al. 1996), serpentine and sandstone grasslands (Jackson et al. 1994; Field et al. 1996), saltmarsh (Drake & Leadley 1991; Jacob et al. 1995) and tallgrass prairie (Owensby et al. 1993; 1996). Gaps remain in our understanding of the long-term response of natural systems to rising [CO2], especially at time scales greater than 2 years. Responses at the leaf level through the ecosystem level still need to be addressed.

Long-term experimental manipulations of [CO2] designed to address long-term acclimation responses (e.g. over many decades) are not feasible. An alternative is to use naturally occurring high [CO2] treatments as analogs for long-term elevated CO2 experiments. CO2 springs, where geological CO2 is released at surface vents or mineral springs, are one example of such a natural analog, as they continuously emit CO2 into the atmosphere and expose the surrounding vegetation to elevated [CO2]. Although not controlled manipulations, CO2 springs expose natural ecosystems to elevated [CO2] for extended periods of time and provide information concerning plant responses to elevated [CO2] otherwise unobtainable.

In this paper, we quantify some of the important biochemical, physiological, and morphological responses of a sub-Arctic grass after long-term exposure to elevated [CO2]. Biochemical responses, such as changes in Rubisco content and activity, have been shown to be important in controlling individual plant response to short-term exposure to elevated [CO2] (Stitt 1991; Sage 1994). However, their importance in controlling the long-term (>100 years) response of plants to elevated [CO2] has not been demonstrated. This study has the advantage of evaluating long-term physiological effects in an ecosystem context. Ultimately, our goal is to extrapolate our understanding of leaf level processes to describing the functioning of the entire ecosystem.

MATERIALS AND METHODS

Site description

The research site has been described previously in detail (Cook, Oechel & Sveinbjornsson 1997). A single CO2 spring in Iceland, emitting ≈ 99% CO2 (Arnorsson & Barnes 1983), was used to study the long-term effects of elevated [CO2] on naturally occurring sub-Arctic vegetation. While the precise date of first CO2 emission at the study site has not been determined, it is thought to coincide with the last major volcanic eruption in the area, which occurred some 2500 years ago (Brandsson, personal communication). The CO2 spring is located on the Snæfellsnes Peninsula of western Iceland near the village of Ólafsvík. Measurements focused on Nardus stricta, an abundant sub-Arctic grass, which comprises ≈ 50% of the plant cover (Cook, unpublished). Grasslands dominated by N. stricta are predicted to expand under warmer and wetter climate change scenarios (Yurtsev 1996).

Our approach has been to compare plants growing in the elevated [CO2] environment of the spring with plants of the same species growing in two adjacent ‘control’ sites subjected to ambient [CO2]. The control sites are referred to throughout this paper as ‘upper control’ and ‘lower control’ due to their topographical position relative to the CO2 spring, but controls 1 and 2 or controls A and B would have been equally appropriate names. The two control sites are meant to be viewed simply as replicate examples of the sub-Arctic ecosystem type under investigation. Multiple control sites are helpful because they provide an estimate of ecosystem variation, and when plotted, allow the CO2 spring site to be compared with more than one control site. In the end, the control data were grouped (as the two control sites did not differ statistically in any of the parameters measured therein), such that probability values shown are for the orthogonal contrast where the mean of the CO2 spring site is compared with the means of the two control sites taken together. As there was only one CO2 spring in the ecosystem, it could not be replicated. All sites are located within 50 m of each other and have similar environmental conditions (Cook et al. 1997).

Maps showing the distribution of [CO2] in the vegetation near the CO2 spring on both calm and windy days have been published, along with a complete table showing the variability in both daily mean and instantaneous [CO2] levels (Cook et al. 1997). These measurements were made with an infrared gas analyzer (LiCor, Lincoln, Nebraska, USA) and instantaneous CO2 concentrations ranged from ≈ 2300 to 360 μmol mol–1 while daily averages ranged from 1800 to 360 μmol mol–1, depending on wind speed, wind direction, and distance from the spring. These data indicate that plants growing near the CO2 spring experience high levels of CO2 that fluctuate widely (much more than the global ambient [CO2]). Over time, the plants surrounding CO2 springs have presumably adapted to both average high and fluctuating [CO2]. This may have consequences for optimal physiology, biochemistry and morphology, as CO2 spring plants must be able to operate under both high and low [CO2] conditions.

In this paper, we focus our attention on the average high [CO2] which can be calculated using carbon isotopes (described below) and assume that CO2 spring plants readily integrate over the rapidly fluctuating [CO2]. Because the average lifetime high CO2 concentrations are within the range of those predicted for the middle of the 21st century (Houghton et al. 1996), the results from this research should be relevant to predicting the long-term response of northern ecosystems to elevated [CO2]. It should be noted, however, that selection under fluctuating conditions could differ from selection under a continuously elevated [CO2].

Determining the mean [CO2]

Carbon isotope measurements were used to determine lifetime mean [CO2] exposure levels for the N. stricta plants used in this study. The isotope method was appropriate because the CO2 emitted from the spring near Ólafsvík is of geological origin and has a very distinctive isotopic signature (0‰14C) that can be clearly detected in the plants that assimilate CO2 from the spring. While the isotopic signature of a plant depends upon the rate of photosynthesis and carbon turnover times, it is indicative of how much of the CO2 fixed is from the CO2 spring and yields an integrated measure of the mean [CO2] to which a plant has been exposed over its lifetime. In simplest form, the [CO2] to which a plant has been exposed can be calculated using Cs = Ca*Ma/Mp where Cs = mean [CO2] to which the plant has been exposed, Ca = mean ambient atmospheric [CO2], Ma = 14C content of N. stricta tissue growing away from the spring in an area known to be of ambient [CO2]) and Mp = 14C content of N. stricta tissue growing in the elevated [CO2] environment of the spring [equation from Geyh & Scheider (1990)]. The calculation of Ma and Mp accounts for discrimination occurring during photosynthesis and assumes that 14C is discriminated against exactly twice as strongly as 13C [demonstrated by Craig (1954)]. Carbon isotope measurements (13C, 14C) were made in the Center for Accelerator Mass Spectrometry (CAMS) at Lawrence Livermore National Laboratory (LLNL).

The carbon isotope measurements indicated that, as a group, plants growing in the elevated [CO2] of the spring and selected for photosynthetic ACi curves in 1993 were exposed to an average [CO2] of 792 (± 79 μmol mol–1, with a range of 534–1194 μmol mol–1, n = 9), while those plants measured in 1994 were exposed to an average [CO2] of 788 (± 82 μmol mol–1, with a range of 519–1179 μmol mol–1, n = 8). Isotopic measurements made on plants from the two control sites confirmed that they had been exposed to only ambient levels of CO2 (358 ± 0·9 μmol mol–1, with a range of 356–361 μmol mol–1, n = 5) and were not fixing CO2 from the spring.

Gas exchange

To investigate the photosynthetic response of N. stricta, leaf gas exchange measurements were conducted using an open-path photosynthesis system, the PACsys 9900 (Data Design Group, San Diego, CA, USA). The relationship between assimilation (A) and internal [CO2] (Ci) was measured in the field under steady-state conditions over a range of CO2 concentrations (100, 200, 300, 400, 500, 600, 800, 1000 and 1200μmol mol–1 in 1993; 50, 100, 200, 300, 400, 500, 800, 1000 and 1200μmol mol–1 in 1994). Leaves were allowed to adjust to each new [CO2] for a minimum of 15 min before a measurement was taken. During all measurements, the photosynthesis system maintained the cuvette at humidity and leaf temperature conditions typical for summer days in Iceland, 10 °C with the vapour pressure gradient from leaf to air held between 0·4 and 0·5 kPa. All measurements were made under saturating photosynthetic photon flux density (PPFD) 1500μmol m–2 s–1 using supplementary lighting. Photosynthesis was expressed on a projected leaf area basis.

ACi curves were measured during two successive summers. In 1993, measurements were made from mid-July until late August whenever weather permitted. By rotating between the sites randomly, 20 plants were measured including nine CO2 spring plants, six upper control plants, and five lower control plants. In 1994, the photosynthetic measurements were more evenly distributed throughout the entire growing season with measurements being made during the months of June, July and August. Three ACi curves were made per site per month (mid-month, weather permitting), enabling within-season as well as year to year comparisons to be made. The layout of the CO2 spring and control plots is given in detail in Cook et al. (1997).

Photosynthesis model

A biochemical model of photosynthesis was used to interpret the ACi curves. This model (Lewis et al. 1994), which is based on previous models (Farquhar, von Caemmerer & Berry 1980; Sharkey 1985; Harley & Sharkey 1991; Harley et al. 1992), was used to estimate three parameters of the ACi curve: (1) Vcmax, the maximum rate of CO2 fixation by ribulose-1,5-bisphosphate carboxylase (Rubisco); (2) Jmax, the maximum capacity for ribulose bisphosphate (RuBP) regeneration from the Calvin cycle which is mediated by the electron transport capacity of the thylakoid reactions; and (3) PiRC, the phosphate regeneration capacity from starch and sucrose synthesis [also known as triose phosphate utilization (TPU) in Harley et al. (1992)]. The model was parameterized and run holding Rd (rate of non-photorespiratory CO2 efflux occurring in the light) constant at the mean for all treatments (12·5 ± 2·9 μmol m–2 s–1 in 1993, and 8·2 ± 2·4 μmol m–2 s–1 in 1994) during the iteration of Vcmax, Jmax, and PiRC (Lewis et al. 1994). Glycerate re-entry into the chloroplast during photorespiration was assumed to be 75% (Harley & Sharkey 1991).

Leaf properties and biochemistry

Leaves (≈ 2·2 cm2) were removed from each plant following photosynthesis measurements, immediately frozen in liquid nitrogen and stored until analysed. Harvests were made midday under ambient light, temperature and [CO2]. Rubisco activity was measured at 25 °C. The initial and total (fully activated) activity of Rubisco were analysed spectrophotometrically by measuring the rate of disappearance of the reduced form of the coenzyme nicotinamide adenine dinucleotide (NADH), as described by Tissue, Thomas & Strain (1993). The activation state of Rubisco was calculated as the ratio of initial activity to total activity. Rubisco content was determined by allowing 14C-carboxyarabinitol bisphosphate (14C-CABP) to bind to the catalytic sites of Rubisco, precipitating the 14C-CABP-bound Rubisco with Rubisco-specific antibodies, and quantified using liquid scintillation counting (Sharkey, Seeman & Berry. 1986).

Chlorophyll content was determined by grinding tissue in liquid nitrogen, double extracting with 80% (v/v) acetone, centrifuging for 1 min, and measuring absorbence of the supernatant at 646·6 and 663·6 nm (Porra, Thompson & Kriedemann 1989). Nitrogen content was determined on leaves dried at 60 °C, ground in liquid nitrogen, digested using a microKjeldahl technique, and measured with an Technicon Autoanalyser II (Tarrytown, New York, USA) (Lowther 1980). For sugars, starch, and total non-structural carbohydrate (TNC), plant material was dried at 60 °C, ground into a fine powder using a Wiley mill, then extracted three times using a methanol:chloroform:water solution (12:5:3 v/v) to separate the soluble sugars from the pellet fraction containing starch (Tissue & Wright 1995). The pellet was digested with perchloric acid (35% v/v) for 1 h to hydrolyze the starch. Soluble sugar and starch concentrations were determined colorimetrically using the phenol–sulphuric acid method of Dubois et al. (1956). TNC was calculated as the sum of soluble sugar and starch. Specific leaf area was calculated on dry tissue as the ratio of leaf area to leaf mass.

Growth

Growth was measured non-destructively by regularly monitoring the above-ground leaf area index (LAI = total projected leaf area/ground area) within 4 cm diameter rings placed in the centre of random N. stricta tussocks. The LAI for the tussocks was calculated by measuring the length of each leaf within a ring, summing the total length of all leaves within each ring, and multiplying this total leaf length by the average leaf width. The rings were left in place throughout the entire growing season. The N. stricta plants used in the growth analyses were interspersed with those used to measure photosynthesis and leaf biochemistry. Above-ground growth was measured three times during the 1993 growing season on 10 plants from each of two sites; the CO2 spring site and the upper control site. Because of time constraints, the lower control site could not be measured in 1993. In 1994, growth measurements were made on all three sites and repeated five times throughout the growing season on 10 plants from each site. The sampling dates are presented by Julian day in the 1993 and 1994 figures for LAI.

Phenology

Throughout the growing season, the phenological development of N. stricta plants was monitored and the degree of leaf senescence scored with the following scale: 0 ≥ 50% dead, 1 = 41–50% dead, 2 = 31–40% dead, 3 = 21–30% dead, 4 = 11–20% dead, 5 = 6–10% dead and 6 ≤ 5% dead. The phenological measurements were made on the same plants that were used to determine growth trends.

Statistical analyses

Because of variations in the pattern of sampling, the statistical analyses used in 1993 differed slightly from those used in 1994. In 1993, a one-way analysis of variance (ANOVA) was used to test the effect of site on the photosynthetic model parameters (Vcmax, Jmax, and PiRC) and the biochemical properties of the leaves. Because only one control site was measured for LAI and phenology, simple t-tests were used to test these parameters.

In 1994, data were collected throughout the growing season so a two-way ANOVA was used to test the effect of site and season on the photosynthetic model parameters and the biochemical properties of the leaves. Because the interaction terms were not significant at the 0·05 probability level, the seasonal data were pooled (each measurement was independent and taken on a different plant) and tested using a one-way ANOVA (d.f. = 2, 23). Seasonal trends are noted when less than the 0·05 probability level. Because two control sites were measured for LAI in 1994, this parameter was also tested using a one-way ANOVA.

As the measured variables for the two control sites did not differ statistically, the probability values shown are for the orthogonal contrast where the mean of the CO2 spring site is compared with the mean of the two control sites taken together. Essentially, this means that the control data were grouped before statistical comparisons were made and probability values calculated. To facilitate year to year comparisons, data for 1993 and 1994 were presented side by side in matching figures. Whenever available, data from the two control sites were plotted separately, aiding visual comparisons. All analyses were conducted using the computer package SuperANOVA (Abacus Concepts, Berkeley, CA, USA) and type III sums of squares.

RESULTS

The ACi curves demonstrate that over the range of Ci, N. stricta plants growing in the elevated [CO2] of the spring have lower rates of photosynthesis than plants growing in ambient [CO2]. Representative curves are shown (Fig. 1) from individual N. stricta plants growing in each site. While these curves were selected because they approximate 1993 and 1994 site means, they provide a visual display of the ACi data only and make no statistical statement. The photosynthesis model was used to test whether the curves are statistically different.

Figure 1.

. Representative photosynthetic response (ACi) curves for Nardus stricta growing near and away from a natural CO2 spring in Iceland. Measurements were taken during (a) August 1993 and (b) July 1994. Statistical comparisons are available in Table 1.

The photosynthesis model shows that the ACi curves are statistically different in at least two of the three estimated parameters. Nardus stricta plants growing at the spring consistently showed reductions in Vcmax (– 25·1% in 1993, – 34·8% in 1994) and Jmax (– 21·5%, – 28·1%, respectively), but differences in PiRC were only significant in 1994 (– 36·6%) (Table 1). There was a significant downward shift in the rate of photosynthesis for CO2 spring plants at all CO2 concentrations.

Table 1.  . Least-squares estimates of model parameter values based on analyses of ACi curves from Nardus stricta. Values are means and standard errors (parentheses). The probability values (P) provided are for the orthogonal contrast where the mean of the CO2 spring is compared with that of the two control means taken together. The upper and lower control sites were not statistically different from each other at the P = 0·05 level Thumbnail image of

In terms of leaf properties and biochemistry, substantial reductions in Rubisco content (–19·3%, –32·3%), Rubisco activity (–35·9%, –44·5%), Rubisco activation state (–29·0%, –16·4%), and chlorophyll content (–32·8%, –32·4%) were found in 1993 and 1994, respectively, when N. stricta plants growing in the elevated [CO2] of the spring were compared with those growing in the two control sites (Fig. 2). Pooled data are shown. Season was a significant component in the two-way ANOVA for Rubisco content, Rubisco activity and chlorophyll content, but the interaction terms were not significant at the 0·05 probability level.

Figure 2.

. Biochemical properties of Nardus stricta leaves collected near and away from a natural CO2 spring in Iceland. Values are means ± SEs, sample sizes vary (1993: n = 3–4; 1994: n = 7–8). P values are for the orthogonal contrast where the mean of the CO2 spring is compared with the means of the control sites taken together.

Leaf nitrogen concentration was reduced at the CO2 spring relative to the control sites in 1993 (– 8%) and significantly in 1994 (– 16%). Starch, sugar and TNC concentrations were not different between the sites, except for one instance in 1993 when starch content was higher (+ 8·8%) near the CO2 spring. Specific leaf area was lower at the CO2 spring site in 1994 (– 8·6%) despite no change in leaf starch content, suggesting thicker leaves (Table 2).

Table 2.  . Properties of Nardus stricta leaves collected near and away from a CO2 spring in Iceland. Values are means and standard errors (parentheses). The probability values (P) provided are for the orthogonal contrast where the mean of the CO2 spring is compared with that of the two control means taken together. N values are given for 1993 and 1994, respectively Thumbnail image of

At the spring, above-ground growth was not enhanced. Nardus stricta plants growing near the CO2 spring produced less above-ground leaf area per unit ground area (LAI) than ambient grown plants (Fig. 3). This trend was evident throughout the entire 1993 (– 25%) and 1994 (– 20%) growing seasons.

Figure 3.

. Leaf area index for Nardus stricta growing near and away from a natural CO2 spring in Iceland. Values are means ± SEs, n = 10. P values ≤ 0·05 are marked *.

Nardus stricta growing at the spring experienced a shift in phenology and became dormant earlier than plants growing away from the spring in an atmosphere of ambient [CO2]. A t-test showed a significant difference in the degree of leaf senescence occurring by 1 September 1993 (P < 0·01; n = 10).

DISCUSSION

Arctic plants have been shown to have similar rates of photosynthesis under ambient and elevated [CO2] (Oberbauer et al. 1986; Tissue & Oechel 1987). Tissue & Oechel (1987) suggested that the homeostatic adjustment observed in the Alaskan Arctic species, Eriophorum vaginatum, after 3 weeks of exposure to elevated [CO2] might be due to either: (1) starch accumulation in the chloroplasts; or (2) decreased Rubisco content and activity. While these mechanisms of down-regulation have been reported in plants from other ecosystems after short-term exposure to elevated CO2 (DeLucia, Sasek & Strain 1985; Sage et al. 1989; Sage 1994), they have not been investigated for northern species exposed to long-term (> 100 year) elevated [CO2].

While carbohydrate concentrations were high for N. stricta plants (16–18%) in this study, they were consistently high in all sites, and the magnitude of the accumulation did not increase in plants grown in elevated [CO2]. Reductions in Rubisco content, Rubisco activity and chlorophyll content were, however, evident. It appears that N. stricta is adjusting to elevated [CO2] by decreasing its investment in the biochemical components of the photosynthetic apparatus, and thereby limiting the production of excess carbohydrates. This mechanism of down-regulation may aid in maintaining whole plant source–sink balance. Tissue & Oechel (1987) also found consistently high carbohydrate concentrations in E. vaginatum (14–20%), and no [CO2] effect. Körner et al. (1996) noted that Arctic and alpine plants typically store large amounts of non-structural carbohydrates in their leaves (15–30% of leaf dry mass), even under ambient [CO2] conditions.

The biochemical analyses agreed well with the model parameters estimated from the photosynthetic ACi measurements and provided direct evidence for the specific mechanisms responsible for the observed down-regulation in photosynthetic capacity. In N. stricta, the photosynthetic down-regulation occurs through reductions in at least Rubisco (associated with model parameter Vcmax) and chlorophyll (associated with model parameter Jmax). Inconsistencies between the results in 1993 and 1994 prevent firm conclusions from being drawn about the importance of the inability to regenerate phosphate and starch accumulation (associated with model parameter PiRC). A significant reduction in PiRC was observed in 1994, but not in 1993, while starch accumulation was observed in 1993, but not in 1994. Certainly, reductions in Rubisco and chlorophyll content contribute to the down-regulation of photosynthesis.

Since Rubisco constitutes 30–50% of leaf protein in C3 plants and is the single largest sink for nitrogen in the photosynthetic apparatus (Seemann et al. 1987), reductions in Rubisco allow for reallocation of nitrogen to other processes. For N. stricta, acclimation to high [CO2] involves the allocation of nitrogen away from Rubisco and the light harvesting pigment chlorophyll to unknown, perhaps below-ground, processes. While we cannot say if below-ground growth is responsive to elevated [CO2] in this Icelandic grassland, others have suggested that below-ground growth will increase under conditions of elevated [CO2] (e.g. Rogers, Runion & Krupa 1994).

Despite the apparently efficient biochemical readjustments, above-ground LAI was not enhanced in N. stricta growing in conditions of elevated [CO2]. Growth, like photosynthesis, is known to be limited by the availability of resources, especially nutrients. Resource limitations have been shown to greatly reduce or eliminate the positive response of photosynthesis to elevated [CO2] and when nutrients are scarce, biomass increases are reported to be negligible in grassland ecosystems (Williams, Garbutt & Bazzaz 1988; Grulke et al. 1990; Oechel et al. 1994; Field et al. 1996; Körner et al. 1996). While our data support the theory that plants grown under conditions of elevated [CO2] have accelerated development (earlier senescence), this does not equate to greater above-ground leaf area. Others have reported similar shifts in phenology without an increase in canopy leaf area (Mousseau & Enoch 1989; Norby & O’Neill 1991; Norby et al. 1992).

Changes in phenology and biochemistry may be causing litter quality near the Icelandic CO2 spring to decrease, making nutrients generally less available. Green leaves collected from the high [CO2] environment contained less nitrogen per unit dry weight than green leaves collected from ambient [CO2] areas. If this increases the C/N ratio of litter produced near the CO2 spring, the rate of decomposition could be reduced and the amount of nutrient immobilization increased (Moorhead & Reynolds 1993). A change in the rate of nutrient cycling could in turn affect plant physiology and growth. While not quantified in this study, nutrient availability and nutrient mass may be important indicators of how elevated [CO2] affects whole ecosystem function. It appears that northern ecosystems are more limited by factors such as nutrient availability, genetics, light, low temperature, and a short growing season than by [CO2] (e.g. Billings et al. 1984; Shaver, Chapin & Gartner 1986; Grulke et al. 1990).

All long-term growth stimulations observed in natural ecosystems appear to be very modest in comparison with the large growth responses generally obtained in short-term agriculture experiments (e.g. Rogers & Cure 1982). It has not yet been demonstrated that exposure to long-term elevated [CO2] from CO2 springs produces large differences in biomass [temperate aquatic ecosystem (Koch 1993), Mediterranean forests and grasslands (Miglietta & Raschi 1993; Miglietta et al. 1993; Körner & Miglietta 1994)]. In unmanaged ecosystems, the potential positive effects of elevated [CO2] are generally offset by acclimation processes that adjust the rate of CO2 fixation to the availability of other resources. The genetic bases for these adjustments are likely to be important but still need to be investigated.

The consistency between our long-term data and that from short-term elevated [CO2] experiments in the Arctic is striking. Tissue & Oechel (1987) showed that homeostatic down-regulation of photosynthesis occurred after 3 weeks of exposure to elevated [CO2] in E. vaginatum, while in N. stricta we found the reduction to be maintained for the life of the plant. Compensatory mechanisms acting over longer time scales (i.e. changes in nutrient cycling, sink strength, phenology) did not override shorter term responses found in experiments with E. vaginatum (Tissue & Oechel 1987). While not all Arctic or alpine species show such a high degree of photosynthetic compensation after exposure to elevated [CO2] (Körner et al. 1996; Oechel et al. 1997), the dominant E. vaginatum and N. stricta species do, suggesting that the potential benefits from increasing levels of atmospheric [CO2] in northern ecosystems will be at least partially offset by photosynthetic down-regulation.

Acknowledgements

We gratefully acknowledge funding by the United States Department of Energy grant DE-FG03-86ER60479, a doctoral enhancement award from the National Science Foundation INT-9224941, NSF ARCSS LAII flux study OPP-9318527, the Joint Doctoral Program in Ecology at San Diego State University and the University of California at Davis, Sigma Xi, and an award from the San Diego Chapter of Achievement Rewards for College Scientists, Inc. Helpful comments on the manuscript were provided by George Vourlitis, Jim Richards and Stuart Hurlbert.

Ancillary