Carbon allocation in calcareous grassland under elevated CO2: a combined 13C pulse-labelling/soil physical fractionation study



  • 1 To test whether plant–soil C fluxes in natural grassland increase under elevated atmospheric CO2 concentration, intact calcareous grassland monoliths exposed to ambient or elevated CO2 were pulse-labelled and the dynamics of the 13C label followed throughout the rest of the growing season.
  • 2 The experiment revealed no increased fluxes of C to soils at elevated CO2. The only changes found were relatively small shifts towards increased C allocation to roots by the end of the growing season. This effect was probably because wetter soil under elevated CO2 prolonged the growing period. At elevated CO2, plant C pools increased below ground (+28%) at the end of the season, resulting in slightly increased root : shoot ratios. Plant 13C pools increased significantly below ground. There were no effects of CO2 enrichment on 13C in soil microbes, fine roots or earthworms.
  • 3 Elevated CO2 caused a shift in soil particle size distribution towards smaller aggregate sizes, but had no effect on the total C and 13C content of low- and high-density soil fractions.
  • 4 The absence of effects of CO2 on the labelling of soil microbial biomass, and of C and 13C accumulation in low-density macro-organic fractions, suggest that there is no significant effect of elevated CO2 on root exudation or turnover, agreeing with published labelling studies, but conflicting with CO2-exchange budgets.


The global increase in atmospheric CO2 concentration may increase photosynthesis and primary productivity (Woodward, Thompson & McKee 1991) and change ecosystem energetics and carbon (C) cycling. Increased net primary production under elevated CO2 can increase soil C inputs from plants through enhanced above-ground litter production and rhizodeposition (Darrah 1996; Gorissen 1996; Norby 1994). This will lead to soil C accumulation if these inputs are not offset by concomitant increases in C exports. Soil C accumulation could mitigate the current atmospheric CO2 increase.

This paper focuses on effects of elevated CO2 on C dynamics in natural and seminatural grasslands. The extent to which C sequestration will increase in such systems under elevated CO2 is unknown because (i) few experimental assessments of C balance under elevated CO2 have been made for such systems; (ii) scaling-up C storage observed in elevated CO2 experiments is difficult because biome-scale ecosystem C storage is dominated by disturbance (harvesting, fire, climatic events) rather than by net ecosystem production (IGBP 1998); and (iii) projecting C sequestration data to longer time scales is complicated because the step increase in experimental CO2 concentrations causes transitory imbalances in soil C pools and nutrient cycles that will not equilibrate over experimental time scales (even in multiyear studies).

Nonetheless, long-term studies of atmospheric CO2 effects on C allocation in ecosystems and calculations of the ecosystem C budgets are required to predict potential C sequestration. Estimates of rhizodeposition at elevated CO2– the rate of organic C losses from roots to the soils including root turnover – are crucial because, in many natural ecosystems, rhizodeposition is the major fraction of soil C input (Stanton 1988).

Rhizodeposition per unit root mass or length may increase at elevated CO2. Pregitzer et al. (1995) reported increased fine root production rates and reduced longevity in Populus saplings grown at elevated CO2. Similarly, Fitter et al. (1996) reported increased rates of root birth and death in mixed Juncus/Nardus swards. Concentrations of non-structural carbohydrates in root tissue increase under elevated CO2 (Rogers & Runion 1994; Wong 1990). If this results in increased cytoplasmatic concentrations of soluble C compounds, this could enhance exudation from roots because of the greater cytoplasm–soil solution gradient (Darrah 1996).

Support for increased rhizodeposition per unit root mass at elevated CO2 appears to arise from CO2 exchange/biomass comparisons. In many natural or seminatural grasslands, large ecosystem C gains have been attributed to faster net photosynthesis (Woodward et al. 1991) and small or negligible changes in dark respiration (Drake et al. 1996b; Stocker, Leadley & Körner 1997). Ecosystem C balances often suggest annual C sequestration at elevated CO2 of at least several hundred g C m−2 (Niklaus et al. 2000). CO2-induced stimulation of plant biomass in natural grasslands is commonly much smaller than this (salt marsh, Drake et al. 1996a; calcareous grassland, Leadley et al. 1999; alpine grassland, Körner et al. 1996; annual Mediterranean grassland, Hungate et al. 1997), and therefore grassland plants would have to dissipate excess C via increased root turnover and exudation. Consequently a dramatic shift in plant C allocation must occur if CO2 balances are correct. This shift could greatly affect the biosphere’s C balance. However, such biomass/CO2 exchange comparisons have to be interpreted cautiously, because those at elevated CO2 may be systematic overestimates (see Lund et al. 1999; Niklaus et al. 2000 for detailed analyses).

Extra soil C fluxes at elevated CO2 are notoriously difficult to detect due to the huge and variable natural background of soil C. Isotope labelling and soil physical fractionation techniques can overcome this problem. Isotope labelling allows recently assimilated C to be tracked. Soil fractionation separates fresh plant-derived macro-organic fractions originating from root turnover from a large background of organic C. While these techniques have been applied to CO2-enriched agricultural systems (Van Kessel et al. 2000), few such studies have been conducted in natural or seminatural grassland under elevated CO2.Ross et al. (1996) labelled intact grassland turf with 14CO2 under ambient and elevated CO2; and Hungate et al. (1996) tracked incorporation of the CO2-C used for atmospheric enrichment into soils (but had no tracer at ambient CO2). The present study aimed to provide this critical data, and is the first to combine isotope [13C] labelling with soil physical fractionation in natural grassland. It aimed specifically to semiquantitatively separate rhizodeposition into (i) labile plant-derived C such as exudates and (ii) more recalcitrant, particulate, plant-derived C (such as the recalcitrant, lignified fraction of root litter).

Materials and methods

Grassland monoliths and experimental design

The grassland monoliths used in this study originated from a species-rich calcareous grassland in north-western Switzerland (47°33′ N 7°34′ E, 520 m a.s.l.). On the same site, a 5·5-year long-term in situ CO2 enrichment study was conducted from 1994 to 1999 (Leadley et al. 1999). The plant community on this pasture is dominated by Bromus erectus Huds., and has been described in detail by Huovinen-Hufschmid & Körner (1998). The growing period starts in mid-March and ends in late October. Mean annual precipitation is ≈800 mm, and mean daily air temperatures are ≈1 °C in January and 20 °C in June. The 15–20 cm topsoil has a pH ≈7–8, underlain by calcareous debris. In the top 10 cm, the horizon where most of the fine roots occur, organic C and N contents are ≈3·9% and 0·33%, respectively (Niklaus 1998).

In March 1997, 38 monoliths (25 × 25 cm, × 15 cm depth) were excavated and transferred to containers. A synthetic mat overlain by marl at the bottom of the container ensured adequate drainage. To minimize edge effects, green nylon mesh was fixed at the borders of the container and its height adjusted with canopy development. Ambient (356 µl l−1) and elevated (600 µl l−1) CO2 treatments were randomly assigned to the monoliths, which then were grown in two climate-controlled daylight greenhouses per CO2 treatment. The monoliths were randomized weekly within chambers. Each month the CO2 treatments were switched between chambers. Water supply was 8 mm week−1 from April to mid-June 1997, and 10–16 mm week−1 until October.

13c pulse labelling

On July 30 and 31 1997, 32 monoliths were 13C pulse-labelled at their growing CO2 concentration for two photoperiods. Labelling was done in two polyester film-covered chambers (Melinex 401, 0·075 mm, ICI Chemicals, Opfikon Glattbrugg, Switzerland) holding 16 monoliths each. 13CO2 (>99%13C, Cambridge Isotope Laboratories, Andover, USA) was applied with syringes connected to a three-way valve between a non-pressurized 13CO2 reservoir and the chambers. The total CO2 concentration (12CO2 + 13CO2) and 13CO2/12CO2 were measured continuously using an infrared gas analyser (IRGA; Li-Cor 6252, Li-Cor Inc., Lincoln, NE, USA) and a quadrupole mass spectrometer (QMG 421, Balzers AG, Balzers, Liechtenstein). The differential sensitivity of the IRGA for 12CO2 and 13CO2 was corrected using the 13C/12C data. The leak rate of the chambers was measured with synthetic N2O (molecular weight equal to CO2) before 13C labelling, and was negligible (≈10−6 g CO2 s−1). Photosynthetic photon flux density inside the chambers was usually 800–1200 µmol m−2 s−1. A large ventilated heat exchanger cooled the air inside the chambers so that the temperature never exceeded 26 °C.

After labelling with a total of 19·7 g 13C (sum of ambient and elevated CO2 monoliths), we added 12CO2 to the chambers for several hours to ensure a maximum uptake of the applied 13CO2. 13CO2/12CO2 during labelling was ≈0·6 due to recycling of plant- and microbially respired 13C in this closed system (Fig. 1).

Figure 1.

13CO2 concentration of the atmosphere in ambient and elevated CO2 treatments during labelling.

The six unlabelled monoliths served to determine the 13C contents of the analysed C pools in the absence of labelling.

Plant sampling

Plants were clipped at 5 cm above ground on June 17 and at ground level on October 27. On August 25 and September 9, root biomass was estimated in soil cores (2·3 cm diameter, 10 cm depth). At the final harvest in October, root biomass was determined in ≈500 g fresh soil. Fine root productivity was assessed in ingrowth cores incubated from August 4 to September 1; from August 25 to September 22; and from September 22 to October 27. Ingrowth cores were installed by taking soil cores (2 cm diameter, 10 cm depth) and filling the holes with sieved, root-free soil. After the ≈4 week incubation period, 1·5-cm-diameter cores were sampled from the same position and the newly formed fine roots washed free of soil.

Soil sampling

Soils were sampled on August 4, 8, 15 and 29, and October 2, corresponding to 4, 8, 15, 29, 63 and 88 days after labelling. Except for the final harvest, when we took ≈800 g soil per monolith, one soil core (2·3 cm diameter, 10 cm depth) per monolith was used. The soil cores taken throughout the experiment and a subsample from the final harvest were kept frozen until analysis for microbial C and 13C by chloroform fumigation–extraction. The rest of the soil from the final harvest was used fresh for microbial biomass determination by substrate-induced respiration and for physical fractionation.

Microbial biomass c and 13c

Microbial C and 13C were estimated by chloroform fumigation–extraction (Vance, Brookes & Jenkinson 1987). We combined soil samples from two monoliths each, reducing n from 32 to 16. Microbial 13C was analysed in freeze-dried extracts. Substrate-induced respiration measurements (n = 32) were done on fresh sieved soil adjusted to 50% water-holding capacity and preincubated for 3 days at 22 °C. After amendment with a glucose/talcum mixture, CO2 release was measured for 10 h and microbial biomass C was calculated (Anderson & Domsch 1978).

Soil physical fractionation

Fresh soil samples corresponding to 150 g dry matter were suspended in 170 ml 0·5% sodium hexametaphosphate and shaken gently for 2 h. Hexametaphosphate improves dispersion and binds Ca2+ which otherwise would precipitate with the density separation agent (polytungstate). The suspension was washed through a stack of sieves (2000, 1000, 500, 250 and 125 µm) with 3 l of water. The fraction <125 µm was centrifuged for 10 min at 2000 g and the supernatant discarded. Small stones in the particle size fraction >2000 µm were removed by hand. This fraction did not contain any high-density organomineral aggregates, and was therefore not further separated by density.

The 125–250, 250–500 and 500–1000 µm fractions were stirred into 100 ml aqueous sodium polytungstate (SOMETU, Falkenried, Germany) solution with a density of 1·6 g cm−3. The suspension was allowed to settle for 10 min and the low-density macro-organic supernatant removed, vacuum-filtered (GF9 glass fibre filter, Schleicher und Schüll, Dassel, Germany), washed with 0·1 m hydrochloric acid to remove carbonates, washed with distilled water and dried at 80 °C for 48 h. The organomineral fraction with a density >1·6 g cm−3 was treated in the same way, except that it was removed from the filter before drying.


To detect possible transfer of plant-derived C into soil macrofauna, a subsample of earthworms was taken in October from each monolith, shock-frozen in liquid N2, freeze-dried and analysed for C and 13C.

Analyses of c and 13c contents

Carbon contents of plants, earthworms and soil samples were determined by dry combustion (CHN-900 or CHN-932, LECO Instruments, St Joseph, MI, USA). The 13C/12C of samples was measured with an isotope-ratio mass spectrometer (Finnigan MAT delta S IRMS coupled to a Finnigan MAT CHN, Finnigan MAT, Bremen, Germany). 13C enrichment (percentage labelled C) was calculated as:


where RPDB = 0·011237, 13Csample is the 13C content of the sample (‰) relative to the PDB standard, and percentage 13Cunlabelled reference is the percentage of 13C of total C in unlabelled monoliths.

Statistical analysis

Data were analysed by anova. The only factor was the CO2 treatment. Time series were analysed using the manova repeated measures procedure of systat (Wilkinson, Hill & Vang 1992). Data were log-transformed to meet the assumption of normal distribution. Error estimates given in the text and error bars in figures are standard errors of the means. In figures, *, P ≤ 0·1; **, P ≤ 0·05; ***, P ≤ 0·01.


Plant biomass

Above-ground plant biomass (>5 cm) in June was 26% greater at elevated CO2 than in ambient monoliths (F1,30 = 8·80, P < 0·01; ambient CO2: 80 ± 5 g m−2; elevated CO2: 100 ± 5 g m−2). This increase originated from greater graminoid biomass (+29%, F1,30 = 7·13, P = 0·02), while legumes and non-legume forbs did not respond to CO2 enrichment. In October, total above-ground biomass was 10% greater at elevated CO2 (F1,30 = 6·08, P = 0·02). The layer above 5 cm responded more strongly (+15%, F1,30 = 5·53, P = 0·03) than the stubble layer (+7%, n.s.). Root biomass did not differ between CO2 treatments with the exception of the bulk harvest in October (+33%, F1,30 = 4·06, P = 0·06). Fine root production as assessed by ingrowth cores did not respond to elevated CO2 in any of the three incubation periods (60 ± 2 and 0·029 ± 0·003 g 13C m−2, averages across the three incubation periods and both CO2 treatments).

Plant tissue C concentrations did not differ between CO2 treatments, and consequently plant C stocks closely followed the biomass responses (Fig. 2). About two-thirds of plant C was below ground, but 13C enrichment per unit plant C was much lower below than above ground, resulting in more than two-thirds of the plant label being above ground. Root 13C was marginally significantly increased at elevated CO2 (+30%, F1,30 = 3·75, P = 0·07).

Figure 2.

(a) Carbon per unit land area; (b) 13C enrichment in percentage of plant C; (c) 13C enrichment per unit land area in plants at the final harvest (white bars, ambient CO2; black bars, elevated CO2). Note that root C was not increased by elevated CO2 until the end of the season for which data are shown here.

Soil biota

Microbial C did not differ between CO2 treatments, regardless of whether they were measured by chloroform fumigation–extraction (Fig. 3a) or by substrate-induced respiration (final harvest: 1378 ± 42 and 1375 ± 30 µg C g−1 soil at ambient and elevated CO2, respectively). Microbial 13C labelling declined from the first sampling date (Fig. 3b), suggesting that maximum labelling occurred before the first measurement. Microbial 13C never differed between CO2 treatments.

Figure 3.

Kinetics of (a) microbial biomass C (Cmic) and (b) 13C (δ13Cmic; bars start at δ13Cmic of unlabelled monoliths). Because no significant differences were found on the other dates, the samples taken on October 2 were not analysed for reasons of cost. n = 16 for analysis of Cmic and 13Cmic because each analysed sample consisted of the combined soils from two monoliths (white bars, ambient CO2; black bars, elevated CO2).

13C enrichment of earthworms did not depend on CO2 treatment (0·053 ± 0·008 and 0·061 ± 0·007% at ambient and elevated CO2, respectively). In the field experiment from which the monoliths were excavated, earthworm biomass did not depend on CO2 enrichment (118 g m−2; Zaller & Arnone 1998). Assuming equal earthworm biomass in the monoliths, 13C enrichment in earthworms approximates to 0·024 and 0·027 g 13C m−2 at ambient and elevated CO2, with no significant difference between CO2 treatments.

Soil aggregates

Low-density macro-organic material comprised only a minor fraction of soils (0·28% of soil mass, 2% of soil C; Fig. 4a,c). This material contained more C relative to dry mass (Fig. 4b) and more C relative to N (Fig. 4e) than the denser organomineral aggregates. Elevated CO2 did not change the amount of low-density material in any particle size class. High-density organomineral aggregates were smaller on average at elevated CO2 (Fig. 4a). CO2 had significant effects on mass and C in all fractions (F1,30 > 9·62, P < 0·004; Fig. 4a). Effects on mass and C vanished when the organomineral aggregates of all size classes were combined.

Figure 4.

Low-density macro-organic matter (ρ < 1·6 g cm−3) and high-density organomineral soil aggregates (ρ > 1·6 g cm−3) of differing size classes at the final harvest. (a) mass; (b) C per unit dry mass; (c) C per unit land area; (d) 13C per unit land area; (e) C : N ratios (white bars, ambient CO2; black bars, elevated CO2).

The 13C enrichment per unit C and land area did not change under elevated CO2 (Fig. 4d). 13C enrichment of the combined high-density aggregates was also unaffected. 13C in the fraction <125 µm was increased at elevated CO2 (F1,30 = 12·94, P = 0·001; Fig. 4d), but again this effect originated from the shift in particle size distribution.

Partitioning of total c and 13c

At the end of the experiment, C stocks in plants (above- plus below-ground) were increased at elevated CO2 (+20%, F1,30 = 4·55, P = 0·04; Fig. 5a). Carbon in microbes, earthworms, low-density macro-organic matter and high-density aggregates was not altered by elevated CO2. Due to the relatively large size and variance of the high-density organomineral soil C fraction, the increase in plant C at elevated CO2 did not result in significantly increased total ecosystem C (2961 ± 60 and 3131 ± 75 g C m−2 at ambient and elevated CO2, respectively).

Figure 5.

Partitioning of (a) total C; (b) labelled C among plants, soil microbes, earthworms (‘Lumbr.’), and soil fractions at the end of the experiment. Earthworm C data are based on biomass estimates made in the parallel field study (Leadley et al. 1999) and C and 13C measurements made in earthworms sampled from monoliths (white bars, ambient CO2; black bars, elevated CO2).

Until the final harvest, ≈78% of the initially applied 13C label had disappeared from the system, irrespective of CO2 treatment. Of the remaining 22%, ≈80% was recovered in plants, 7% in soil organisms, 13% in organomineral aggregates and <1% in the macro-organic soil fraction (Fig. 5b).


Total ecosystem uptake of labelled C was slightly stimulated by elevated CO2, but within-system C distribution was not substantially altered. The sizes of the soil pools fed by rhizodeposits did not change significantly (microbes, low-density macro-organic fractions), nor did the kinetics of the label in these fractions differ between CO2 treatments. We found only a slight shift towards more 13C allocation to roots at elevated CO2 in October, which is compatible with the small but significant increase in root : shoot ratio. This end-of-season CO2 effect on root : shoot ratio was probably because the growing period was slightly extended by greater soil moisture availability at elevated CO2 (Niklaus & Körner 1998; see also Jackson et al. 1994; Leadley et al. 1999). Persistently increased soil moisture at elevated CO2 alters drying–rewetting cycles (Niklaus & Körner 1998), possibly the cause of the reduction in organomineral aggregate size found in the monoliths. This alteration of soil structure could affect water and nutrient fluxes (Young & Ritz 2000). Increased soil moisture and reduced aggregate size increase the water-filled pore fraction and decrease O2 diffusivity. The resulting hypoxia may stimulate the efflux of radiatively active trace gases (e.g. N2O, CH4; Arnone & Bohlen 1998). Rillig et al. (1999) reported increased aggregate sizes in two Mediterranean grassland systems exposed to elevated CO2 and in a greenhouse shrub model ecosystem. The authors speculated that this shift was driven by increased mycorrhizal secretion of glomalin, which in turn promoted soil aggregation at elevated CO2. Our study demonstrates that, depending on the conditions, the opposite effect can also occur (less aggregation at elevated CO2).

Our study suggests that net plant–soil C fluxes do not increase at elevated CO2. One could argue that a CO2 effect on microbial 13C which occurred before the first assessment, 4 days after labelling, might have been missed. However, this is unlikely because (i) we would expect this to lead to an increased 13C incorporation into microbes still detectable at day 4; (ii) available data suggest no change in microbial C turnover rate (Figure 3); and (iii) continuously increased microbial C turnover would probably increase microbial biomass or activity, but we did not detect any such change during the parallel 5·5-year field study (Niklaus 1998; Niklaus et al. 2000; Stocker et al. 1997; P. Niklaus, unpublished results). Particulate soil C input also does not appear to have increased. In the field study, increased root longevity and resulting reductions in root turnover at elevated CO2 (Arnone et al. 2000) also argue against increased C fluxes to soils by this pathway.

Tracer-based estimates of C diverted to rhizodeposits depend on the season at which the label is applied and the interval between application and recovery of the label (e.g. Swinnen, van Veen & Merckx 1994). Labelling after clipping caused greater-than-average above-ground C allocation, as is evident from the higher root : shoot ratio for C compared to 13C (Fig. 2). Between-organ partitioning of 13C after pulse-labelling depends critically on the season. However, unless CO2 treatment-dependent shifts in phenology occur, comparison of the labelling of specific organs at ambient with that at elevated CO2 should not be impaired, and thus our conclusions remain valid. We applied the 13C label 6 weeks after clipping, because the difference in plant C uptake between CO2 treatments is largest at that time (Niklaus et al. 2000; Stocker et al. 1997), so maximizing the probability of detecting CO2 effects on soil C fluxes. The result is therefore all the more striking. A 3 month labelling-to-harvest time might be too short for substantial transfer of labelled root structural C to soil. However, even if complete eventual transfer of the plant label to the soil is assumed, C inputs to soil would still not be increased per unit plant mass, contrary to expectations (see Introduction).

In accordance with the present experiment, all published studies (e.g. Cotrufo & Gorisson 1997; Gorissen 1996; Ross et al. 1996; Verburg, Gorissen & Arp 1998) suggest that rhizodeposition per unit root mass does not increase at elevated CO2. This accordance across experiments on a variety of species and under widely different conditions implies that our conclusion is generally applicable. The inconsistencies between our 13C tracer data and CO2-exchange budgets, and those of other workers, are not easily resolved. The only explanation compatible with both data sets would be slower decomposition of native, old soil organic matter, but it is not evident how this might be caused if plant–soil C fluxes remain largely unaltered. The most likely explanation is a methodological one: the systematic overestimation of long-term CO2 budgets at elevated CO2 (Niklaus et al. 2000).

In conclusion, C tracer data for woody and herbaceous ecosystems do not indicate increases in soil C inputs beyond the stimulation of plant mass at elevated CO2. Although vegetation biomass in natural grassland often responds less to elevated CO2 than under more fertile conditions, CO2-exchange measurements have nevertheless suggested greatly increased soil C fluxes in natural grassland as well. Our 13C-labelling/soil physical fractionation study indicates that this does not happen, despite strong stimulation of photosynthesis. Consequently, extra soil C inputs at elevated CO2 may be much smaller in these ecosystems than previously anticipated from CO2 budgets. Our study also reveals secondary effects of elevated CO2 on soil structure which may alter nutrient and water fluxes and enhance the release of radiatively active trace gases. The importance of these changes for plant growth and ecosystem processes, however, remains unknown.


We acknowledge two anonymous referees for helpful comments on earlier versions of this manuscript. We also thank Silvio Leonardi who provided the quadrupole mass spectrometer used to control the CO2 C-isotope composition during labelling. This work was funded by the Swiss National Science Foundation (project 5001-035214) and contributes to the Global Change and Terrestrial Ecosystems (GCTE) core research programme.

Received 17 April 2000; revised 14 July 2000; accepted 24 July 2000