Responses of net ecosystem CO2 exchange in managed grassland to long-term CO2 enrichment, N fertilization and plant species


Urs Aeschlimann. Fax: 0041016321153; e-mail:


The effects of elevated pCO2 on net ecosystem CO2 exchange were investigated in managed Lolium perenne (perennial ryegrass) and Trifolium repens (white clover) monocultures that had been exposed continuously to elevated pCO2 (60 Pa) for nine growing seasons using Free Air CO2 Enrichment (FACE) technology. Two levels of nitrogen (N) fertilization were applied. Midday net ecosystem CO2 exchange (mNEE) and night-time ecosystem respiration (NER) were measured in three growing seasons using an open-flow chamber system. The annual net ecosystem carbon (C) input resulting from the net CO2 fluxes was estimated for one growing season. In both monocultures and at both levels of N supply, elevated pCO2 stimulated mNEE by up to 32%, the exact amount depending on intercepted PAR. The response of mNEE to elevated pCO2 was larger than that of harvestable biomass. Elevated pCO2 increased NER by up to 39% in both species at both levels of N supply. NER, which was affected by mNEE of the preceding day, was higher in T. repens than in L. perenne. High N increased NER compared to low N supply. According to treatment, the annual net ecosystem C input ranged between 210 and 631 g C m−2 year−1 and was not significantly affected by the level of pCO2. Low N supply led to a higher net C input than high N supply. We demonstrated that at the ecosystem level, there was a long-term stimulation in the net C assimilation during daytime by elevated pCO2. However, because NER was also stimulated, net ecosystem C input was not significantly increased at elevated pCO2. The annual net ecosystem C input was primarily affected by the amount of N supplied.


Grasslands play an important role in the global carbon cycle because they cover 24% of the terrestrial surface (Sims & Risser 2000) and have a high capacity to sequester carbon (C); they therefore contribute to the residual terrestrial carbon sink (Prentice et al. 2001). Additionally, land-use change from annual crops to grassland may favour carbon sequestration (Watson et al. 2000). In humid temperate regions agricultural grasslands are often sown and persist for several years. These systems are usually dominated by perennial ryegrass (Lolium perenne) and white clover (Trifolium repens), both species which are well-adapted to the typical management with fertilization and frequent defoliation.

Since photosynthesis is directly affected by the partial pressure of atmospheric carbon dioxide (pCO2) (Drake, Gonzalez-Meler & Long 1997), its investigation is crucial for an understanding of plant responses to elevated pCO2. The effects of elevated pCO2 on leaf photosynthesis of grassland species have been investigated in many studies (Stirling et al. 1997; Rogers et al. 1998; Clark, Newton & Barker 1999; Davey et al. 1999; Isopp et al. 2000). Across many FACE (Free Air Carbon dioxide Enrichment) experiments, leaf photosynthesis in C3 plants increased by more than 30% at elevated pCO2 (Long et al. 2004). In L. perenne, leaf photosynthesis has been shown to increase by 40% (Rogers et al. 1998), with this higher rate persisting for several years (Ainsworth et al. 2003). The CO2 response of leaf photosynthesis was affected by nitrogen (N) availability: at low N supply, sink limitation led to an acclimation of photosynthesis, whereas at high N supply there was no acclimation (Fischer et al. 1997; Rogers et al. 1998).

Although N availability has little direct effect on photosynthetic rates, N is often the most limiting resource for plant growth. Consequently, the CO2 response of harvestable plant biomass was smaller than that of leaf photosynthesis and depended on N supply (Schneider et al. 2004). Obviously, under low N conditions a higher proportion of the additional photosynthates assimilated at elevated pCO2 is allocated to roots and stubble (Daepp, Nösberger & Lüscher 2001; Suter et al. 2002; Schneider 2003). Together with higher exudation, this may stimulate the C input to the soil compartment. On the other hand, higher C losses by the ecosystem due to enhanced respiration at elevated pCO2 probably reduce the net soil C input and the CO2 response of biomass. Thus, leaf level measurements alone are not sufficient to predict the CO2 response of an ecosystem.

Besides assimilation, the other important component of ecosystem CO2 exchange is ecosystem respiration, which includes both autotrophic plant respiration and heterotrophic microbial respiration. Since about half of the C assimilated by photosynthesis of terrestrial plants is released as CO2 during subsequent plant respiration (Amthor 1997), the rate of autotrophic respiration has a major effect on the carbon balance of ecosystems. Microbial respiration in soil is also an important carbon source, with about 10% of the atmosphere's CO2 passing through soils each year (Raich & Potter 1995).

There have been several studies of the effects of clipping and grazing (Bremer et al. 1998), ecophysiology (Craine, Wedin & Chapin 1999) and environmental conditions (Mielnick & Dugas 2000) upon soil CO2 efflux in grassland ecosystems. However, much less is known about respiration at the ecosystem level in grasslands (Franzluebbers, Franzluebbers & Jawson 2002), and effects of elevated pCO2 on ecosystem respiration have scarcely been investigated in any terrestrial ecosystems.

Effects of elevated pCO2 on plant respiration are mainly indirect, resulting from changes in biomass accumulation and protein content (Amthor 1997). Specific leaf respiration tends to be reduced at elevated pCO2 (Schapendonk et al. 1997) whereas at the whole plant level or on a ground area basis, respiration often increases due to higher biomass (Casella & Soussana 1997). Little is known about the interacting effects of pCO2, N and species on the ecosystem respiration.

The few studies which have investigated CO2 exchange in grassland on fertile soil have either been conducted under controlled conditions at the whole-plant level (Ryle, Powell & Tewson 1992) or under semi-field conditions at the ecosystem level (Casella & Soussana 1997; Schapendonk et al. 1997). Estimates of CO2 effects on the C balance of grasslands are mainly based on model studies (Schimel et al. 2000; Watson et al. 2000; Leifeld, Bassin & Fuhrer 2003). There have been no long-term field investigations to determine whether increases in net ecosystem CO2 exchange at elevated pCO2 persist for several years. As it had been exposed to elevated pCO2 (60 Pa) for 9 years, the Swiss FACE experiment offered us the first opportunity to investigate such long-term responses in a grassland ecosystem on fertile soil. The aim of the work described here was to quantify the effects of pCO2 and N on net ecosystem CO2 exchange after long-term CO2 enrichment. This information is important in order to determine whether grasslands will act as an additional C sink under conditions of rising pCO2.

We investigated midday net ecosystem CO2 exchange (mNEE) and night-time ecosystem respiration (NER) of a managed grassland ecosystem. The interactions between pCO2 and two levels of N supply were studied in monocultures of Lolium perenne and Trifolium repens. Based on the CO2 exchange measurements an annual C balance was estimated.


Experimental site

The experimental site is located at Eschikon (8°41′ E, 47°27′ N) near Zurich at an altitude of 550 m above sea level. Free Air Carbon dioxide Enrichment (FACE) technology (Hendrey 1992) was used to investigate the long-term effects of elevated pCO2 on a fertile grassland ecosystem in the field. Prior to the FACE experiment, the site was grown for several years with annual crops. The experiment was arranged in three blocks, each consisting of two circular areas (18-m-diameter), one CO2-enriched (60 Pa pCO2) and the other an ambient control (36 Pa pCO2). The CO2 enrichment began in May 1993 and lasted each year for the whole growing season (from March to November) during the daylight hours. Because, plant growth is slow and CO2 effects are weak at low temperatures (Long 1994), CO2 enrichment was stopped when the temperature fell below 5 °C and was recommenced when the temperature rose above 6 °C. A more detailed account of the experimental set-up is given by Hebeisen et al. (1997b).

Mean temperatures during the growing seasons 2000, 2001 and 2002 were 14.0, 13.7 and 13.5 °C, respectively. Total rainfall during the growing season was 772 mm in 2000, 953 mm in 2001 and 882 mm in 2002. Average PAR was 31.0 mol m−2 d−1 in 2000, 35.5 mol m−2 d−1 in 2001 and 34.9 mol m−2 d−1 in 2002.

Experimental treatments

Monocultures of Lolium perenne cv. Bastion and Trifolium repens cv. Milkanova, two important species of managed grassland in temperate and humid climates, were sown in 5.3 m2 plots in 1992. Since 1993, the swards were cut five times each year at a height of 5 cm and the harvested material removed. All plots were fertilized each year with 5.5 g P m−2 and 24.1 g K m−2 to ensure that a shortage of these nutrients did not limit plant growth (Daepp et al. 2000). To examine the effects of N availability on the CO2 response of the ecosystem, two levels of nitrogen (N) fertilization (14 and 56 g N m−2 year−1) were applied. The N fertilizer was applied as liquid NH4NO3 at the beginning of each regrowth period. The amount applied was divided between the five successive regrowth periods in the proportions 30, 20, 20, 15 and 15%, these percentages corresponding to the expected yields at the end of each period.

Measurements and data collection

Net ecosystem CO2 exchange was measured during the growing seasons of 2000, 2001 and 2002 using an open-flow chamber system. For our purposes, eddy-flux measurements would not have been an alternative, because of the proximity of the different treatments in the FACE array. Our open-flow system consisted of two identical but independent units, each including two chambers, so that simultaneous measurements could be made of both N treatments and both pCO2 levels. The chambers remained installed on the same experimental treatment for 1 to 5 d, and were then moved between vegetation types and blocks. The two units were regularly exchanged between controls and fumigated areas, in order to prevent possible biases in the data.

Each chamber, covering a square area of 0.49 m2 and 0.6 m high, consisted of an aluminium framework covered with Teflon (PTFE) film, this material is highly transparent to PAR and infra-red, so that the light quality in the chambers was not substantially altered and heating was of minor importance. One side, where the gas inlet and outlet were placed, was made of Plexiglas. To achieve a complete seal of the chamber at ground level, it was fixed to a steel frame that was inserted into the soil. The air that was to be passed over the vegetation was sampled with a vertical tube 7 m above the ground level in order to minimize rapid fluctuations in pCO2. Fans then blew the air through flexible PVC tubes into the chambers. The air-flow was measured with a mass-flow meter (Accu-Flo 600; Sierra Instruments, Monterey, CA, USA) placed in the tubes. A small fan was placed in each chamber in order to ensure a thorough mixing of the air. The air within the chambers was exchanged up to four times per minute during the day and about once per minute at night. The outlet of the chamber was sufficiently large, so that the positive pressure in the chambers was less than 1 mm of water-column. Nevertheless, this may have reduced soil respiration to a small extent, but since this potential error was similar in all treatments, the conclusions were not affected. In the fumigated areas, the sampled air was enriched with CO2 during the day time, so that pCO2 in the chambers reached a level similar to the FACE conditions of about 60 Pa. At night, as in the FACE experiment, pCO2 within all chambers was ambient, but because of net respiration, ambient pCO2 at night was higher than at day time.

CO2 exchange within the chambers was measured by sampling air at both the inlet and the outlet of the chambers. This air was pumped through flexible polyethylene tubes to an infra-red gas analyser (Binos 100 4P; Fisher-Rosemount, Hasselroth, Germany) where the difference in the CO2 concentration was measured. In order to determine the air humidity, the air samples were pumped through heated PTFE tubes to a dewpoint sensor (MTR 2.0; IL Metronic, Ilmenau-Unterpörlitz, Germany).

Incident PAR was measured continuously by a light sensor (BF2; Delta-T devices, Cambridge, UK) in one FACE area. Relative light transmission through the canopies was measured at approximately weekly intervals during the growing season with a Sunfleck Ceptometer (SF-40; Decagon, Pullman, WA USA). The multiplication of these two measured parameters allowed us to calculate the intercepted PAR which was used as surrogate measure of leaf area during regrowth.

For biomass determination at harvest, the cut plant material was oven-dried at 65 °C for 48 h prior to weighing. A subsample was ground into powder and analysed for total C content.

Calculations and statistical analysis

For statistical analysis, the mNEE data for the growing seasons 2000 and 2001 were combined. The mNEE rates, averaged for 10-min intervals, were integrated for the 4-h period from 1100 to 1500 h. Averaged for the growing season this period accounted for about 50% of total net CO2 uptake during the photoperiod. Data for intercepted PAR were loge-transformed before statistical analysis in order to obtain linear dependencies with mNEE. NER data were loge-transformed in order to obtain a normal distribution.

NER rates were integrated for the 7-h period between 2200 and 0500 h. During this interval, respiration in full darkness was relatively constant so that mean values could be used to investigate trends between treatments and across the growing season. NER was measured during six periods (spring 2001 and 2002, summer 2000 and 2001, autumn 2000 and 2001).

The statistical analyses were carried out using the procedure Mixed of SAS 8.02 (SAS Institute Inc., Cary, NC, USA). The model was a split-plot with pCO2 as the main plot factor, thus block and block × pCO2 were tested as random effects. Since block × pCO2 has only two degrees of freedom, the split-plot model requires a high F-value for the main plot factor pCO2 to be significant. Denominator degrees of freedom were adjusted according to the method of Kenward–Rogers (Littell et al. 1996). For the analysis of the mNEE data, the variable ‘midday intercepted PAR’ was used as a covariate. The effect of mNEE on NER was tested in the subset of NER data for which mNEE on the preceding day was available, using mNEE as a covariate.

The net ecosystem C input, estimated for 2001, is the result of the daytime net C uptake minus the C respired during night and the C removed with the harvested material. Daytime net C uptake was modelled for each day using the fits of mNEE to intercepted PAR (Fig. 1) and summed over the growing season. The amount of respired C during night results from mean values (Table 1) multiplied by the length of the night period and the number of days. The harvested C corresponds to the annual sum of C in plant material removed.

Figure 1.

Midday net ecosystem CO2 exchange (g C m−2 h−1) with respect to cumulated midday intercepted PAR (mol m−2). (a) Lolium perenne at low N supply (14 g N m−2 year−1); (b) L. perenne at high N supply (56 g N m−2 year−1); (c) Trifolium repens at low N supply (14 g N m−2 year−1); (d) T. repens at high N supply (56 g N m−2 year−1).

Table 1.  Means of night-time ecosystem respiration (2200 to 0500 h) in six periods of measurement, total means, standard errors (SE) and number of measurements (n)
(g m−2 year−1)
pCO2 (Pa)Respiration (g C m−2 h−1) n
  1. Significance probabilities of statistical analysis for the total means. The ‘spring’ period covers the first two regrowth periods (April to mid June), ‘Summer’ covers the third and forth regrowth period (mid June to end of August) and ‘Autumn’ covers the fifth regrowth period (until end of October).

L. perenne14360.1670.1690.1080.2280.0620.1170.1360.009 99
600.2210.2060.1420.3550.0880.1520.1890.014 83
600.2330.1920.2420.3890.1790.2110.2430.012 81
T. repens14360.2600.1960.0930.3260.1040.1580.1760.011 88
600.3700.2770.1700.5160.1300.2630.2450.019 82
56360.2700.2190.2170.3470.1280.1790.2120.011 96
600.2750.1980.2400.4590.1550.2840.2450.014 85
CO2        P < 0.01 
N        P < 0.0001 
species        P < 0.0001 
period        P < 0.0001 
CO2 × species        NS 
CO2 × N        P < 0.05 
CO2 × period        NS 


Midday net ecosystem CO2 exchange

Midday net ecosystem CO2 exchange (mNEE) includes plant photosynthesis as well as plant and heterotrophic respiration. mNEE was strongly correlated with the sum of midday intercepted PAR (P < 0.0001), and mNEE data were therefore plotted against this variable (Fig. 1). Intercepted PAR is dependent upon leaf area and incident PAR, and thus differences between the fitted curves were not caused by differences in these variables but by treatment effects.

mNEE was significantly higher in L. perenne and T. repens monocultures which had been exposed for nine growing seasons to continuous CO2 enrichment than in controls (P < 0.0001; Fig. 1). The absence of CO2 × N and CO2 × species interactions shows that this increase occurred irrespective of N supply and species.

Intercepted PAR influenced the relative effect of pCO2 on mNEE. This can easily be demonstrated by comparing the fitted curves at 5 and 15 mol m−2 of intercepted PAR (Fig. 1). When intercepted PAR was low (5 mol m−2), elevated pCO2 increased mNEE of L. perenne by 20% at low N and by 24% at high N supply. However, with increasing interception of PAR the relative differences between the CO2 treatments decreased, especially at low N supply. Thus, when intercepted PAR was 15 mol m−2, elevated pCO2 increased mNEE by 12% at low N and by 23% at high N supply.

For T. repens at low N supply, elevated pCO2 increased mNEE by 30% at 5 mol m−2 of intercepted PAR. This stimulation decreased to 24% at 15 mol m−2 of intercepted PAR. At high N supply, intercepted PAR of 5 and 15 mol m−2 resulted in a stimulatory effect of pCO2 on mNEE of 32 and 10%, respectively.

The N fertilization affected the light-response curves of mNEE (P < 0.05). The species × N interaction (P < 0.01) shows that the response of mNEE to N fertilization differed between L. perenne and T. repens. In L. perenne, high N compared to low N supply decreased the curve of mNEE by up to 10% when the interception of PAR was low. In contrast, when the interception of PAR was high, high N increased the curve of mNEE up to 17% compared to low N supply. In T. repens, the N effect on the curve of mNEE was negative, ranging from −10 to −20%.

Night-time ecosystem respiration

Mean night-time ecosystem respiration (NER) ranged between 0.062 and 0.459 g C m−2 h−1 (Table 1). The period of measurement had a significant effect on NER (P < 0.0001), with highest values of NER being recorded in summer 2001 and lowest in autumn 2000 (in seven out of eight treatments). NER in T. repens was consistently higher than in L. perenne (Table 1, P < 0.001), the difference, with a mean value of 18%, ranged from 1 to 30% according to treatment. The difference in NER between T. repens and L. perenne was influenced by the period of measurement (P < 0.01), and varied between 7% in summer 2000 and 45% in spring 2001.

N fertilization strongly affected NER (Table 1, P < 0.0001). In L. perenne, high N supply stimulated NER by 36% at ambient and by 29% at elevated pCO2. In T. repens, high N supply only increased NER at ambient pCO2 (by 20%). These results show that NER of the legume and the grass ecosystems differed considerably in their responses to N supply (N × species: P < 0.01). The period of measurement also affected the N response of NER (N × period: P < 0.0001), indicating that it was affected by environmental conditions and plant growth. The N response of NER in L. perenne was small or absent (−7 to 12%) in the spring periods, and largest (> 100%) during autumn 2000.

Elevated pCO2 markedly stimulated NER (Table 1, P < 0.01), the response being greater at low than at high N supply (CO2 × N: P < 0.05). In L. perenne, elevated pCO2 led to a mean stimulation of NER of 39% at low N and of 31% at high N supply. In T. repens, elevated pCO2 increased NER by an average of 39% at low N and by 16% at high N supply. These results demonstrate that NER of both species responded comparably to elevated pCO2 (CO2 × species: not significant). Remarkably, the CO2 response of NER did not vary significantly between the periods of measurement (CO2 × period: not significant).

The NER of both species was positively correlated with mNEE of the preceding day (P < 0.0001, Fig. 2), the effect being stronger for T. repens than for L. perenne (P < 0.0001).

Figure 2.

Linear regression between midday net ecosystem CO2 exchange (mNEE) and loge-transformed night-time ecosystem respiration (NER).

Harvestable biomass

The three-year average (2000–02) of annual harvestable biomass of L. perenne was considerably increased at high N compared to low N supply (P < 0.01) at both pCO2 levels. At low N, harvestable biomass of L. perenne did not respond to CO2 enrichment, being 356 ± 29 g m−2 year−1 (mean ± standard error) at ambient and 348 ± 33 g m−2 year−1 at elevated pCO2. At high N supply, elevated pCO2 increased the harvestable biomass of L. perenne by 28% (544 ± 50 g m−2 year−1 at ambient pCO2 and 698 ± 58 g m−2 year−1 at elevated pCO2), but this difference was not statistically significant.

Elevated pCO2 increased the 3-year average of annual harvestable biomass of T. repens by 11–14%, irrespective of N supply but this difference was also not statistically significant. The N supply had no effect on the harvestable biomass of T. repens. Harvestable biomass at ambient pCO2 was 533 ± 63 g m−2 year−1 at low N and 523 ± 22 g m−2 year−1 at high N supply. At elevated pCO2, the annual harvestable biomass was 606 ± 101 g m−2 year−1 at low N and 578 ± 50 g m−2 year−1 at high N supply.

Net ecosystem C input

The net ecosystem C input in 2001, calculated from the data for CO2 flux and harvested plant material, was strongly positive (210–631 g C m−2 year−1) in all treatments (Table 2). This means that at the end of the growing season, more C was present in the ecosystem than in spring when growth started.

Table 2.  Estimated carbon balance for the 2001 growing season, mean values ± standard errors (n = 3) and significance probabilities of statistical analysis
(g m−2 year−1)
Daytime net C-uptake
(g C m−2 year−1)
Night-time respiration
(g C m−2 year−1)
(g C m−2 year−1)
Annual net C-uptake
(g C m−2 year−1)
  1. Interactions not listed showed no significance.

Lolium perenne14361061 ± 29344 ± 25155 ± 2562 ± 41
601316 ± 46523 ± 72162 ± 15631 ± 95
 5636 924 ± 109435 ± 41277 ± 26212 ± 60
601215 ± 9552 ± 105309 ± 49354 ± 159
Trifolium repens14361072 ± 89489 ± 45220 ± 54363 ± 6
601335 ± 40795 ± 101173 ± 36367 ± 88
 5636 969 ± 8533 ± 35226 ± 7210 ± 49
601175 ± 53633 ± 67228 ± 35314 ± 111
CO2  P < 0.0001P < 0.01NSNS
N  P < 0.01NSP < 0.01P < 0.01
Species  NSP < 0.01NSP = 0.05
N × Species  NSNSP < 0.05NS

In L. perenne and T. repens, the net ecosystem C input at low N was markedly increased compared to high N supply (P < 0.01). Elevated pCO2 tended to increase net ecosystem C input but this difference was not statistically significant. The net ecosystem C input was slightly higher in L. perenne than in T. repens (P = 0.05), especially at low N supply.


For the first time, the effects of elevated pCO2 on the net CO2 exchange of a managed grassland on fertile soil have been investigated at the level of the ecosystem in a field-grown sward fumigated with CO2 for nine growing seasons. Because of the long duration of this FACE experiment, the results may be indicative of long-term effects of elevated pCO2 and N supply on ecosystem processes.

Midday net ecosystem CO2 exchange increased at elevated pCO2

Elevated pCO2 increased mNEE by up to 32% in L. perenne and T. repens ecosystems after long-term CO2 enrichment (Fig. 1a–d). Thus, net assimilation at elevated pCO2 in the field increased not only in the short term but remained considerably stimulated for 9 years, showing that even over a long period photosynthetic acclimation to elevated pCO2 may be of minor importance. In accordance, Ainsworth et al. (2003) found at the same experimental site no evidence for a decrease in the stimulation of photosynthesis in L. perenne leaves across 10 years of CO2 enrichment.

N limitation has been reported to promote photosynthetic acclimation (Drake et al. 1997). However, in our experiment, elevated pCO2 stimulated mNEE also when the N supply was low. This is most probably because the frequent biomass removal by cutting restored the strength of the carbon sink in the plants (Rogers et al. 1998), and eliminated photosynthetic acclimation that may have occurred at the end of a regrowth period (Fischer et al. 1997; Isopp et al. 2000). This finding demonstrates the important influence of management on the photosynthetic CO2 response.

The effect of elevated pCO2 on mNEE in our study was lower than that on leaf photosynthesis (Ainsworth et al. 2003). This reduction of the CO2 effect was attributed to increased root biomass (Soussana, Casella & Loiseau 1996; Daepp et al. 2001), with increased root respiration as well as higher ecosystem respiration at elevated pCO2 (Table 1). Additionally, elevated pCO2 most probably enhanced photosynthesis at the top of the canopy more than in older, shaded foliage lower down. As leaf photosynthesis is usually measured on the uppermost leaves, extrapolation from such measurements results in an overestimate of the CO2 effect on net assimilation at the canopy level.

The relative CO2 effect on mNEE in L. perenne was highest when the amount of intercepted PAR was low, especially at low N supply. This may be due to the fact that when the sward height was very low, shortly after cutting, the measured intercepted PAR could not fully account for differences between treatments in the leaf area remaining on plants after defoliation (residual leaf area). The mass of pseudostems and leaves below cutting height (Daepp et al. 2001) and residual leaf area (F. Stadelmann, written communication) were increased at elevated pCO2. Furthermore, in the first weeks after cutting the number of tillers was higher at elevated pCO2 (Suter, Nösberger & Lüscher 2001), leading to a higher number of small leaves. Consequently, shortly after the harvest, there was a higher residual leaf area at elevated pCO2, which was not reflected in intercepted PAR, but probably enabled a higher mNEE.

The CO2 response of mNEE is larger than that of harvestable biomass

The remarkable increase in mNEE at elevated pCO2 was associated with a relatively small increase in the harvestable biomass of L. perenne at high N and no increase at low N supply (Schneider et al. 2004). This is because, especially when N limitation was severe, the additional assimilated carbon was not transformed into harvestable biomass. Although N availability has little direct effect on photosynthetic rates, N is often the limiting resource for plant growth, especially in species with a high growth rate (e.g. L. perenne). Elevated pCO2 strengthens the N limitation, which causes an accumulation of non-structural carbohydrates, because not all the photosynthates can be transformed into structural biomass (Stitt & Schulze 1994). The non-structural compounds are exported to the base of pseudo-stems near the soil surface and roots. These reserves are not necessarily used later on for additional production of harvestable biomass but they may be respired or lost by senescence. Such changes in the source–sink relations at elevated pCO2 have been shown to lead to considerable increases in root biomass (Soussana et al. 1996; Daepp et al. 2001), root-to-shoot ratio (Suter et al. 2002) and the mass of pseudostems below cutting height (Schneider 2003). The result is a higher CO2 effect on total plant biomass than on harvestable biomass above cutting height, which helps to explain the difference between the CO2 responses of mNEE and the harvestable biomass.

As the large increase in mNEE did not lead to an increase in harvestable biomass under N poor conditions, our data indicate that elevated CO2 changed the allocation of photosynthates and C fluxes in grasses especially at low N supply. This is in accordance with Schneider (2003) who found in the same FACE experiment that elevated pCO2 stimulated the mass of pseudostems below cutting height more when N supply was low.

Intercepted PAR and species affected the effects of N supply on mNEE

In L. perenne the effect of N fertilization on mNEE depended on the amount of intercepted PAR. Low intercepted PAR reflects most often a low leaf area index (LAI) and under these conditions below-ground respiration is relatively more important. Therefore, decreased below-ground (Van Ginkel, Gorissen & Van Veen 1997) and total respiration at low N supply (Table 1) contributed to higher mNEE under N-poor conditions. Furthermore, greater residual leaf area at low N supply than at high N supply (Schneider 2003), which was not represented in measured intercepted PAR, may also have contributed to a higher mNEE shortly after the cut, when measured interception of PAR was low.

When incident PAR and LAI were high (i.e. high amounts of PAR were intercepted), high N supply had a positive effect on the mNEE of L. perenne. Under these conditions, above-ground biomass, which was higher at high than at low N supply, was the determining factor, with residual leaf area and below-ground respiration being relatively unimportant. Additionally, with an increasing LAI towards the end of a regrowth period, sink limitation and a related decrease in photosynthesis may have become important at low N supply (Fischer et al. 1997; Isopp et al. 2000).

The legume T. repens responded differently to N fertilization from L. perenne because higher symbiotic N2 fixation at low N supply compensated for the smaller amount of fertilizer N (Zanetti et al. 1996). As a consequence, the production of harvestable biomass of T. repens did not depend on the level of N fertilization. A negative photosynthetic response to high N fertilization in legumes, as observed in our study, is consistent with the results obtained by Lee et al. (2001). A higher C sink in the poorly fertilized and highly nodulated plants may have stimulated photosynthesis (Schulze, Adgo & Merbach 1999).

Elevated pCO2 strongly increased night-time ecosystem respiration

Elevated pCO2 increased NER by up to 39% in both species and at both low and high N supply (Table 1), although the effect was more pronounced at low N supply.

The observed large increase of NER at elevated pCO2 suggests, in accordance with Amthor (1997), that indirect effects of pCO2 on respiration are more important than direct effects which would reduce plant respiration.

Our results are in line with Casella et al. (1997), who showed that at the whole-plant level or on a ground-area basis, total respiration was increased at elevated pCO2. Higher plant biomass at elevated pCO2 was most probably the key factor leading to the increased NER.

In L. perenne, the response pattern of NER to elevated pCO2 and N supply followed rather that observed for root and stubble biomass below cutting height in the same FACE experiment (Schneider 2003) than that of harvestable or total plant biomass. This indicates that, as in a tallgrass prairie where a large fraction of ecosystem respiration was found to originate from roots and soil (Ham et al. 1995), NER in L. perenne was dominated by processes associated with roots and stubble.

Microbial biomass, which increased in our Swiss FACE experiment at elevated pCO2 (Sowerby et al. 2000) may also have contributed to the higher respiration.

In starch-storing species the export of carbohydrates can account for a large fraction of nocturnal respiration (Bouma et al. 1995). It is likely therefore that in T. repens enhanced night-time translocation of degraded starch, as found in the Swiss FACE experiment by Fischer (1998), also contributed to the higher NER at elevated pCO2

We found no effect of the period of measurement on the CO2 response of NER (Table 1), most probably because our site did not experience extreme conditions, which have been shown to influence the effect of elevated pCO2 on respiration (Luo et al. 1996).

NER was markedly affected by N supply

We observed a species-specific response of NER to N supply (Table 1), which was probably due to the symbiotic N2 fixation of T. repens, whose growth and hence growth-related respiration are largely independent of N fertilization. In contrast, high N supply increased total biomass and N content of L. perenne (Daepp et al. 2001), which most probably promoted respiration.

Leaf turnover is faster at high N because growth is scarcely N limited (Craine & Reich 2001) and it seems reasonable to assume that root turnover is also faster at high N supply. Faster turnover of plant biomass would favour increased rates of shoot and root respiration when N supply is high, as shown by Van der Werf et al. (1993) in several grass species. Furthermore, in our experiment high N supply may have increased microbial activity and biomass shortly after application of fertilizer, leading to temporarily higher heterotrophic respiration.

The effect of N on NER was strikingly small in spring. Possibly, a high decomposition of roots during reproductive growth (Troughton 1957) led to a larger contribution of below-ground respiration during this period, which could have reduced the effect of N on NER, because root biomass and its senescence are little affected by N supply.

NER was species-specific

Mean NER across all treatments was 2–56 mg C m−2 h−1 higher in T. repens than in L. perenne (Table 1), demonstrating that plant species composition can affect ecosystem level processes in a grassland. Various factors may contribute to this difference.

First, tissue N and protein concentrations are higher in T. repens, which may increase specific maintenance respiration (Wullschleger, Norby & Gunderson 1992). Second, symbiotic N2 fixation in T. repens is more energy demanding than nitrate reduction in grasses and consequently is associated with higher specific root respiration than in L. perenne. Furthermore, in T. repens degradation of starch and subsequent translocation is a highly energy demanding, nocturnal process (Fischer 1998), whereas fructan translocation in grasses requires less energy and is not strongly associated with the night period.

MNEE was affecting NER

The positive association of NER with mNEE of the preceding day (Fig. 2) suggests that respiration is linked to the availability of recently assimilated carbon. This finding is in line with other studies; Whitehead et al. (2004) found a strong relationship between total night-time respiration and total photosynthesis during the previous day in Quercus rubur leaves. In spring wheat, total below-ground CO2 efflux was closely coupled with photosynthesis (Kuzyakov & Cheng 2001) and in a Minnesota grassland the soil CO2 flux was largely controlled by the availability of photosynthates to the roots (Craine et al. 1999). However, the relatively weak correlation between mNEE and NER in our results shows, in accordance with McCutchan & Monson (2001), that the respiration of perennial herbaceous species is not exclusively controlled by the availability of carbohydrates.

The management of the ecosystem is an additional factor that affects respiration (Bremer et al. 1998). Throughout the growing season, the cutting regime in a managed grassland causes strong dynamics in growth rates and standing biomass, and these in turn determine growth and maintenance respiration, explaining part of the variability in NER.

The results discussed so far, show that the nutrient status, the species composition, and the management of an ecosystem are important factors affecting respiratory C losses. In the context of land-use change, this conclusion has important implications for C losses from terrestrial ecosystems.

Elevated pCO2 did not significantly increase net ecosystem C input

Elevated pCO2 had only a marginal effect on the net ecosystem C input, because the increased net CO2 uptake during daytime at elevated pCO2 was mostly compensated for by a higher night-time respiration.

If the tendency towards increased net C input at elevated pCO2 in 2001 (69 and 142 g C m−2 year−1 in L. perenne at low and high N supply, respectively) were representative for an average year, the soil C content would have been significantly enhanced after several years of CO2 enrichment when compared with ambient pCO2. However, no significant changes in soil organic C were found after our FACE experiment had been running for 6 and 8 years (Van Kessel et al. 2000; Van Groenigen et al. 2002). This suggests that there may be a tendency to higher net C input at elevated pCO2 in some years, such as 2001, but this is most probably not a consistent trend. However, CO2 exchange measurements are very sensitive and may detect small differences in the C balance that are not observed when alternative methods were used. Fluxes of gases other than CO2, which were not measured here, could also be affected by pCO2, e.g. CH4 (Ineson, Coward & Hartwig 1998), although the amounts involved are too small to make a significant contribution to the ecosystem C balance.

A clear net ecosystem C input in all treatments

The C balance for the growing season 2001 indicates that the annual net C input in our Swiss FACE experiment can be considerable (Table 2). Consistent with our finding, the soil C content in the Swiss FACE in 2000 (Van Groenigen et al. 2002) was higher than in 1998 (Van Kessel et al. 2000). In temperate humid regions, a ‘mature’ natural ecosystem usually has an equilibrated C balance, whereas a change in land use may induce enhanced C sequestration and consequently mitigate the increase of atmospheric pCO2 (Lal 2003). Modelling studies of the effects of rising pCO2 on C sequestration in the USA suggest that effects of land-use change may be larger than effects of pCO2 (Schimel et al. 2000). After a change in land use from arable crops to grassland, such as our Swiss FACE site has experienced, the soil C content can increase for a period of 50–70 years (Sauerbeck 2001); thus the net C input observed in our study is likely to reflect the land-use change in 1993 at the start of the FACE experiment. Since in our system land-use change was relatively recent, the soil was most probably not yet saturated with C. This could explain why the estimated net C input exceeded by far predictions made by Watson et al. (2000) and Leifeld et al. (2003) of an annual C sequestration after conversion of arable land to grassland of 50–100 g C m−2. However, we may have overestimated the net C input to a small extent, because an underestimation of respiration due to the small positive pressure in the chambers cannot be excluded. More importantly, the net C input calculated on a growing season basis may overestimate the annual net C input because respiration during winter is neglected. In a calcareous grassland, respiration rates during winter were about 150 g C m−2 (Volk & Niklaus 2002).

High N supply decreased net C input

Net ecosystem C input was considerably higher at low N than at high N supply. This negative effect of high N supply on the net C input is consistent with the trend observed in 2000 (albeit not significant) towards more total soil C at low N supply (Van Groenigen et al. 2002). At ambient pCO2, soil C content was 300 g C m−2 greater at low N than at high N supply and at elevated pCO2 this difference was even 700 g C m−2. These data as well as the C balance presented here both indicate that, at least in the last years of the FACE experiment, more C was sequestered at low N supply.

In L. perenne, where the effect of N supply on the C balance was most pronounced, the increase of net C input at low N supply was a result of higher daytime net CO2 uptake in association with less harvested biomass and a trend to decreased night-time respiration. The fact that annual daytime net CO2 uptake was higher at low N than high N supply was because throughout the growing season mean of midday interception of PAR was 6.27 mol m−2 and under these conditions mNEE was higher at low N supply (Figs 1a & b). The observed N effect on the C balance may as well be related to pCO2-related, long-term changes in the N cycling in the soil (Richter 2003; Schneider et al. 2004). In our FACE experiment, isotopic techniques using 15N detected that the mobilization of N from soil organic matter increased at high N over the 10 years of experimentation. Thus, if N and C mobilization are coupled, long-term high N supply may have reduced net CO2 input.

Above-ground necromass (Blum, Hendrey & Nösberger 1997) and the proportion of plant biomass below cutting height (Daepp et al. 2001) were higher at low N than at high N supply, leading to smaller C exports by harvest at low N supply (Table 2). In addition, carbohydrate concentration in pseudostems was higher at low N than at high N supply (Fischer et al. 1997). It is clear therefore that decomposition of the plant biomass below cutting height constitutes a C flux to the soil and a potential for C sequestration which is especially great at low N supply.

However, these results do not imply that natural, unfertilized vegetations always represent a greater C sink than N-rich agricultural grasslands. On one hand, a severe N limitation in natural grassland may inhibit any CO2 response of assimilation (Stitt & Krapp 1999) and consequently of the C balance. On the other hand, excessive N supply seems to increase C exports disproportionately, by promoting both shoot growth and respiration, resulting in a reduced net ecosystem C input compared with moderate N supply.

Net C input differed between species

The net ecosystem C input was slightly higher in L. perenne than in T. repens, particularly at low N supply (Table 2). This may be due to the higher root biomass of L. perenne, which builds the potential for a C flux to the soil. Additionally, the NER of L. perenne was lower than in T. repens, causing smaller C losses from the ecosystem. As a consequence of this species-specific C balance, CO2 and N-related changes in the species composition (Hebeisen, Lüscher & Nösberger 1997a; Lüscher, Hendrey & Nösberger 1998; Navas et al. 1999) can be expected to affect the net ecosystem C input.

In conclusion, our results convincingly demonstrate that elevated pCO2 increased the C fluxes (CO2 uptake and release) in a managed grassland ecosystem but had little effect on the ecosystem C balance, which was primarily affected by the amount of nitrogen supplied.