C3–C4 composition and prior carbon dioxide treatment regulate the response of grassland carbon and water fluxes to carbon dioxide

Authors


†Address to whom correspondence should be addressed. E-mail: wpolley@spa.ars.usda.gov

Summary

  • 1Plants usually respond to carbon dioxide (CO2) enrichment by increasing photosynthesis and reducing transpiration, but these initial responses to CO2 may not be sustained.
  • 2During May, July and October 2000, we measured the effects of temporarily increasing or decreasing CO2 concentration by 150–200 µmol mol−1 on daytime net ecosystem CO2 exchange (NEE) and water flux (evapotranspiration, ET) of C3–C4 grassland in central Texas, USA that had been exposed for three growing seasons to a CO2 gradient from 200 to 560 µmol mol−1. Grassland grown at subambient CO2 (< 365 µmol mol−1) was exposed for 2 days to an elevated CO2 gradient (> 365 µmol mol−1). Grassland grown at elevated CO2 was exposed for 2 days to a subambient gradient. Our objective was to determine whether growth CO2 affected the amount by which grassland NEE and ET responded to CO2 switching (sensitivity to CO2).
  • 3The NEE per unit of leaf area was greater (16–20%) and ET was smaller (9–20%), on average, at the higher CO2 concentration during CO2 switching in May and July. The amount by which NEE increased at the higher CO2 level was smaller at elevated than subambient growth concentrations on both dates, but relationships between NEE response and growth CO2 were weak. Conversely, the effect of temporary CO2 change on ET did not depend on growth CO2.
  • 4The ratio of NEE at high CO2 to NEE at low CO2 during CO2 change in July increased from 1·0 to 1·26 as the contribution of C3 cover to total cover increased from 26% to 96%. Conversely, in May, temporary CO2 enrichment reduced ET more in C4- than C3-dominated grassland.
  • 5For this mesic grassland, sensitivity of NEE and ET to brief change in CO2 depended as much on the C3–C4 composition of vegetation as on physiological adjustments related to prior CO2 exposure.

Introduction

Plants in most ecosystems respond initially to an increase in atmospheric CO2 concentration with faster rates of photosynthesis and growth and a decrease in stomatal conductance (Drake, Gonzàlez-Meler & Long 1997; Long et al. 2004). However, initial responses to CO2 enrichment may not be sustained, complicating the task of predicting CO2 effects on carbon (C) and water exchange. Nutrient limitations and increased carbohydrate levels in plant leaves (Long et al. 2004) are among the factors that can induce downward regulation of canopy-level photosynthesis (Baker et al. 1990; Oechel et al. 1994; Drake et al. 1996) and transpiration at elevated CO2 (Dugas et al. 2001). Conversely, an increase in leaf area or a shift in plant composition to more responsive species may reinforce or sustain stimulatory effects of CO2 enrichment.

Photosynthesis at a given CO2 concentration is often lower for plants grown at elevated than ambient CO2 concentrations (acclimation or downregulation; Baker et al. 1990; Oechel et al. 1994; Drake et al. 1996). When expressed per unit of above-ground biomass, for example, net ecosystem exchange of CO2 (NEE) was reduced by 50% in a calcareous C3 grassland exposed for 2 years to elevated CO2 (Stocker, Leadley & Körner 1997).

The response of stomatal conductance (Polley, Johnson & Mayeux 1997) and of plant transpiration rates (Dugas et al. 2001) to CO2 also may depend on the CO2 concentration at which plants were grown. When measured at a given CO2 concentration, transpiration per unit of leaf area of Acacia farnesiana plants was more than halved by growth at 980 compared with 385 µmol mol−1 (Dugas et al. 2001).

However, CO2 enrichment need not alter plant sensitivity to CO2. CO2 enrichment had no effect on photosynthetic potential of field-grown soybean (Campbell, Allen & Bowes 1990), wheat (Kimball et al. 1995), or rice (Baker et al. 1997). Similarly, CO2 stimulation of NEE persisted for 6 years in a scrub-oak ecosystem (Hymus et al. 2003) and from years 3 to 17 following enrichment in a salt marsh (Rasse, Peresta & Drake 2005).

We measured the effects of temporarily altering the CO2 concentration by 150–200 µmol mol−1 on the daytime net ecosystem CO2 exchange (NEE) and evapotranspiration (ET) of C3–C4 grassland exposed for more than three growing seasons to a continuous gradient in CO2 from subambient to elevated concentrations (200–560 µmol mol−1). CO2 enrichment substantially increased biomass production and net C uptake of this grassland (Mielnick et al. 2001; Polley, Johnson & Derner 2003), but also altered the sensitivity of leaf C and water fluxes to CO2. Leaf photosynthesis of the dominant C4 grass and C3 forb (Anderson et al. 2001) and stomatal conductance of the dominant forb displayed up-regulation in plants grown at subambient compared with ambient CO2 (Maherali et al. 2002). These changes should diminish the sensitivity of NEE and ET in CO2-enriched grassland to short-term variation in CO2. However, biomass of C3 species increased at the expense of C4 grass biomass, irrespective of CO2 treatment (Mielnick et al. 2001; Polley et al. 2003). This shift towards physiologically more-responsive C3 than C4 plants (Anderson et al. 2001; Maherali et al. 2002) would be expected to sustain grassland response to CO2 enrichment. Our objective in temporarily altering CO2 was to determine whether a multiyear CO2 treatment affected the amount by which grassland NEE and ET responded to CO2 change (sensitivity to CO2). Consistent with previous results, we predicted that sensitivity of NEE and ET to CO2 would depend more on the relative cover of C3 plants than on the CO2 concentration at which the grassland had been maintained for 3 years. We know of no previous study in which responses of NEE and ET to short-term change in CO2 concentrations have been examined in an intact ecosystem exposed for several years to both subambient and elevated CO2 concentrations.

Materials and methods

experimental facility

We used elongated field chambers to expose a C3–C4 grassland in central Texas, USA (31°05′ N, 97°20′ W) to a continuous gradient in CO2 from 200 to 560 µmol mol−1 (Johnson, Polley & Whitis 2000). The CO2 gradient was maintained during growing seasons (March–November) of 1997–2000 on grassland dominated by the C4 perennial grass Bothriochloa ischaemum (L.) Keng and C3 perennial forbs such as Solanum dimidiatum Raf. Annual precipitation at the site averages 879 mm (89 years mean).

The CO2 facility consisted of two transparent, tunnel-shaped chambers, each with 10 consecutive compartments 1 m wide × 1 m tall × 5 m long (Johnson et al. 2000). During daylight, pure CO2 was injected into air introduced into the south end of one chamber (elevated chamber; compartments 1–10) to initiate a 560–350 µmol mol−1 gradient in CO2. Ambient air was introduced into the south end of the second chamber (subambient chamber; compartments 11–20) to initiate a 365–200 µmol mol−1 gradient in CO2. Night-time CO2 concentrations were regulated at about 150 µmol mol−1 above daytime values along each chamber. Ambient air was introduced into the north end of the subambient chamber at night. Pure CO2 was injected into air blown from the north of the superambient chamber at night to increase the initial CO2 concentration to 500 µmol mol−1. Desired CO2 concentration gradients were maintained by automatically varying the rate of air flow through chambers in response to changes in photosynthetic (daylight) or respiration rates (night). Air temperature and vapour pressure deficit were regulated near ambient values by cooling and dehumidifying air at 5-m intervals along chambers. Irrigation equivalent to rainfall was applied to the chambered grassland through a metered surface irrigation system.

flux measurements

Daytime totals of NEE and ET were calculated for each 5-m compartment for three periods during the 2000 growing season using the CO2 or water vapour gradient measured each 20 min in each compartment with an infra-red gas analyser (Model 6262, Li-Cor, Inc., Lincoln, NE, USA) and the volumetric rate of air flow (Mielnick et al. 2001). The rate of air flow through each compartment was a linear function of the number of revolutions of the fan that moved air through chambers (r2 > 0·99). Net fluxes of CO2 and ET calculated every 20 min were summed for each daylight period. Fluxes for compartments located at the north end of the chambers (air exit; compartments 10 and 20) were highly variable because the air exit was poorly buffered from the influence of winds. Data from these compartments were excluded.

We altered the CO2 concentration of air input to chambers for 2 days during early season (May) and late-season in 2000 (October) and for 4 days during mid-season (July) to determine the grassland response to a short-term change in CO2. Grassland maintained (grown) for 3 years at subambient CO2 (< 365 µmol mol−1) was exposed for 2 days during July (days 206–207) and October 2000 (days 277–278) to an elevated CO2 gradient (> 365 µmol mol−1; Fig. 1a). Plants in each compartment of the subambient chamber thus were briefly exposed to a CO2 level that was 150–200 µmol mol−1 greater than the ‘growth’ concentration (Fig. 1b). Grassland grown at elevated CO2 was exposed for 2 days during May (days 137–138) and July 2000 (days 208–209) to a subambient gradient. Grassland in each compartment of the elevated chamber thus was briefly exposed to a CO2 level that was 150–180 µmol mol−1 lower than the growth concentration. The CO2 was altered (switched) when clear skies were predicted for several days.

Figure 1.

(a) Depiction of the application of CO2 switching treatments. Grassland grown at subambient CO2 concentration [CO2] was exposed for 2 days in October and July 2000 to an elevated CO2 gradient. Grassland grown at elevated CO2 was exposed for 2 days in May and July 2000 to a subambient gradient. (b) Average daytime CO2 concentration on days 206 and 207 (solid bars) and 208 and 209 (open bars) in consecutive 5-m long compartments of CO2 chambers during CO2 switching in July 2000. Grassland had been exposed to a gradient in elevated CO2 (compartments 1–9) and in subambient CO2 (compartments 11–19) for three seasons (indicated by bars with no arrows) when CO2 concentration was increased in compartments 11–19 on days 206 and 207 (indicated by upward arrow) and reduced in compartments 1–9 on days 208 and 209 (indicated by downward arrow). CO2 also was increased in compartments 11–19 for 2 days in October 2000 and reduced in compartments 1–9 for 2 days in May 2000.

The influence of prior CO2 exposure on physiological sensitivity to CO2 usually is assessed at a common CO2 concentration. This approach was not feasible in elongated chambers. Instead, we standardized the absolute difference between the lowest and highest CO2 levels to which grassland in each chamber compartment was exposed during each of the three periods of CO2 switching. In the absence of acclimation, the absolute change in NEE or ET for a given differential in CO2 should be similar irrespective of the absolute flux rate provided fluxes respond approximately linearly to CO2.

Daily totals of photosynthetic photon flux density (PPFD) during CO2 switching varied from 45 to 54 mol m−2 day−1 on days 135–138, 42–49 mol m−2 day−1 on days 206–209, and 38–42 mol m−2 day−1 on days 275–278. Diurnal variation in PPFD likely affected the amount by which CO2 switching changed NEE and ET. Variation in PPFD should not have compromised our ability to detect effects of prior CO2 exposure on fluxes, however, because on a given day, the grassland in treatment and control chambers was exposed to the same light environment and CO2 gradient (Fig. 1a).

plant and environmental measurements and analyses

Rates of NEE also vary as a function of within-day change in PPFD. In order to determine whether the response of NEE to CO2 switching depended on within-day changes in PPFD, we calculated the effect of short-term CO2 treatment on parameters derived from regressions fit to relationships between NEE and PPFD. For each day and for each 5-m compartment, we fit a hyperbolic regression to the NEE-PPFD relationship (Ruimy et al. 1995),

image

where α is the initial slope of the light–response curve, NEEmax is maximum NEE, and R is respiration rate (NEE at PPFD = 0). For each hyperbolic equation, we calculated NEE at 800 and 1600 µmol (quanta) m−2 s−1 (NEE800, NEE1600). The absolute differences in parameter values between high and low CO2 levels then were calculated for each period of CO2 switching.

Leaf area index (LAI) in each 5-m compartment was measured on days 145 (early season), 199 (mid-season), and 285 (late-season) using a SunScan canopy analysis system (Delta-T Devices Ltd, Burwell, Cambridge, UK), and averaged 2·5, 3·4 and 3·2 at elevated CO2 and 2·4, 3·1 and 2·6 at subambient CO2 during early, mid and late season, respectively. Percentage ground cover by plant species was estimated visually from photographs of the centre 4 m of each 5-m long compartment each day on which LAI was measured. Cover of C3 grasses and litter plus standing dead material was near zero in all compartments (results not shown), so we used the ratio of C3 cover (mostly forbs or broadleaf herbaceous plants) to total [C3 plus C4 (grasses)] cover to characterize vegetation differences among compartments. Absolute responses of NEE and ET (each expressed per unit of leaf area) and of parameters derived from fitted light–response curves to CO2 change were analysed as a function of the growth CO2 concentration and C3/total cover using regression (Sigma Plot 2000, SPSS Inc., Chicago, IL, USA). Relationships between fluxes measured prior to each period of CO2 switching and growth CO2 were also subject to regression analysis.

Results

nee and et measured at growth co2

Daytime NEE per unit of leaf area increased sharply at higher growth CO2 concentrations during the 6 days prior to CO2 switching in May (DOY 129–134; Fig. 2). As estimated from regression analysis, increasing growth CO2 from 235 to 550 µmol mol−1 increased early season NEE by an average of 145% (from 12·3 to 30·2 g CO2 m−2 leaf area day−1). As the growing season progressed and the contribution of C3 species to plant cover decreased from an average of 79% in May to 69% in July and 58% in late September, NEE per unit of leaf area declined and linear relationships between daytime NEE and growth CO2 weakened (Fig. 2). By contrast, there was no relationship between growth CO2 and daytime ET measured prior to either the early season (May), mid-season (July) or late-season (late September) period of CO2 switching (not shown; F < 0·34, P > 0·56, n = 6; mean ± SE = 921 ± 70 g H2O m−2 leaf area day−1 in May and 1783 ± 127 g H2O m−2 leaf area day−1 in July).

Figure 2.

Relationships between means of daytime net ecosystem CO2 exchange (NEE) over 6 days in May (DOY 129–134), July (DOY 200–205), and late September 2000 (DOY 270–274) prior to CO2 switching and the CO2 concentration to which grassland had been exposed for three growing seasons (growth CO2). Data from May were fit with a logarithmic function (NEE = −102·3 + 21·0 × ln(growth CO2), r2 = 0·44, P = 0·002). Data from July and September were fit with linear functions (NEE = 5·883 + 0·016 × growth CO2, r2 = 0·11, P = 0·09 in July; NEE = 3·788 + 0·009 × growth CO2, r2 = 0·11, P = 0·10 in September).

response of nee and et to co2 switching: relationship to growth co2

Temporarily increasing CO2 usually increased NEE and reduced ET. The ratio of daytime NEE at the higher compared to lower CO2 concentration during CO2 switching averaged 1·20 in May across plots grown at elevated CO2 (n = 9), 1·29 in October across plots grown at subambient CO2 (n = 9), and 1·16 in July across plots grown at subambient and elevated concentrations (n = 18). The ratio of ET at high CO2 to ET at low CO2 during switching was ≤ 1·0 at all growth concentrations in May and July (data not shown) and averaged 0·80 (May; n = 9) and 0·91 (July; n = 18) across long-term treatments. The ratio of ET at high compared with low CO2 during switching averaged 0·97 in October (n = 9). During early to mid-season therefore temporarily increasing CO2 by 150–200 µmol mol−1 increased NEE per unit of leaf area by 16–20% and reduced ET per unit of leaf area by 9–20%. Yet, the absence of relationships between ET and growth CO2 and the weakening of NEE-growth CO2 relationships during mid- and late-season (Fig. 2) imply that the sensitivity of fluxes to CO2 differed among growth concentrations during at least part of the growing season.

The effect of a temporary change in CO2 on fluxes did not depend consistently on the growth concentration, however. The amount by which NEE was greater at the higher CO2 concentration during switching was smaller at elevated than subambient growth concentrations in both May and July, but relationships between NEE response and growth CO2 were weak (Fig. 3). The relationship between NEE response to switching and growth CO2 was significant at the P = 0·07 level in May (F = 4·59, n = 9) and declined significantly in July only following exclusion of an outlying data point at the highest growth concentration (F = 5·27, P = 0·04, n = 17). Even after excluding this outlier, the effect of growth CO2 on change in NEE depended on a single data point at the lowest growth concentration without which the regression of NEE response on growth CO2 was not significant (F = 1·77, P = 0·20, n = 16). There was no relationship between change in NEE during late-season switching (October) and growth CO2 (F = 3·7, P = 0·10, n = 9), nor did the decline in ET at the higher CO2 level during switching depend on growth CO2 on any sampling date (F = 0·18, 0·06, 0·16; P = 0·69, 0·80, 0·70; n = 9, 18, 9 in May, July and October, respectively).

Figure 3.

The difference in the daytime total of net ecosystem exchange (NEE) between days at elevated CO2 concentrations and days at subambient concentrations (E-S) during CO2 switching in May and July 2000 as a function of the growth CO2 concentration. Data from May were fit with a linear function [NEE(E-S) = 13·070 – 0·018 × growth CO2, r2 = 0·31, P = 0·07]. Data from July were fit with an exponential function following deletion of the outlying point at the highest growth concentration [NEE(E-S) = 5·8762 × inline image r2 = 0·21, P = 0·04].

During CO2 switching in both May and July, NEE-PPFD relationships were strongly curvilinear whether measured at high (average r2 of hyperbolic functions = 0·88) or at low CO2 concentrations (average r2 = 0·87; Fig. 4). The trend in July for NEE to increase more following a short-term increase in CO2 concentration at subambient than elevated growth concentrations (Fig. 3) resulted largely from a similar trend in the response of NEE1600 to CO2 change (Table 1). Growth CO2 concentration did not affect the response of other parameters from light–response curves (NEE800, R, α) in July to CO2 change (F < 0·09, P > 0·77, n = 18) or the response of any parameter from curves in May to CO2 change (F < 0·54, P > 0·48, n = 9).

Figure 4.

Net rate of ecosystem CO2 exchange (NEE) as a function of photosynthetic photon flux density (PPFD) for grassland grown at the current CO2 concentration (360 µmol mol−1) and exposed for 2 days in July 2000 to an elevated CO2 concentration (556 µmol mol−1). Illustrated are data from single days at ambient and elevated CO2 that were fit with hyperbolic functions, where NEE = [(α × b × PPFD)/(b + α × PPFD)] − R and α = 0·0158 and 0·0262, b = 16·2161 and 17·5054, R = 1·0861 and 2·5058, and r2 = 0·95 and 0·85 at ambient and elevated CO2, respectively (P < 0·0001).

Table 1.  Significant regressions between the change in gas exchange from elevated to subambient CO2 concentrations (E-S) during CO2 switching and either the growth CO2 concentration (CO2) or the ratio of C3 plant cover to total cover (C3 cover). NEE800, NEE1600 = Rates of daytime net ecosystem exchange (µmol CO2 m−2 leaf area s−1) at 800 and 1600 µmol (quanta) m−2 s−1 as calculated from light–response curves in July (n = 18). ET = daytime rate of evapotranspiration in May (g H2O m−2 leaf area day−1; n = 9). Linear (y = ax + b) or exponential (y = a × e(–bx)) functions were fit to relationships between dependent and independent variables
Dependent variableMonthIndependent variableModel typeSlope or a-valueIntercept or b-valuer2P-value
NEE1600 (E-S)JulyCO2Exponential4·595  0·0030·200·04
NEE800 (E-S)JulyC3 coverLinear−0·691  2·1080·350·006
ET (E-S)MayC3 coverLinear−751·15427·630·390·04

response of nee and et to co2 switching: relationship to c3 cover

Flux responses to CO2 switching did not depend consistently on growth CO2, implying that factors other than or in addition to physiological acclimation explained the limited sensitivity of NEE (Fig. 2) and ET to growth CO2, especially during mid- and late-season. In this grassland, CO2 response depended partly on the photosynthetic pathway of dominant plants. The amount by which NEE was stimulated at the higher CO2 level during switching in July increased linearly as the contribution of C3 species to plant cover rose from 26% to 96% (Fig. 5). The greater response of NEE to CO2 enrichment in C3- than C4-dominated plots resulted largely from a similar trend in the response of NEE800 to CO2 change (Table 1). By contrast, there was no relationship between C3/total cover and the change in other parameters from light–response curves (NEE1600, R, α) following a CO2 change in July (F < 2·25, P > 0·14, n = 18). Nor, did C3/total cover affect the response of any parameter from light–response curves in May to CO2 change (F < 0·31, P > 0·60, n = 9). Surprisingly, higher CO2 during switching reduced ET more in C4- than in C3-dominated grassland in May. The amount by which ET declined at higher CO2 in May decreased linearly from 580 g H2O m−2 leaf area day−1 when C3 plants comprised 40% of plant cover to 330 g H2O m−2 leaf area day−1 when C3 cover approached 100% (Table 1). Effects of CO2 change on ET did not depend on C3 cover in either July (F = 0·06, P = 0·80, n = 18) or October (F = 2·88, P = 0·13, n = 9).

Figure 5.

The difference in the daytime total of net ecosystem CO2 exchange (NEE) between days at elevated CO2 concentrations and days at subambient concentrations (E-S) during CO2 switching in July 2000, vs. the ratio of C3 plant cover to total cover. Data were fit with a linear function [NEE(E-S) = −0·743 + 3·822 × (C3/total cover), r2 = 0·24, P = 0·02, n = 18].

Daytime NEE responded more to CO2 change in C3- than C4-dominated grassland during mid-season (Fig. 5), implying that compartment-to-compartment differences in C3 cover were at least partly responsible for weak relationships between grassland NEE and long-term CO2 treatment in July and September (Fig. 2). To lessen the influence of variation in C3 cover, we averaged fluxes measured at growth CO2 levels across either elevated or subambient concentrations (average CO2 difference = 195 µmol mol−1). C3 plants contributed a similar fraction to total cover in grassland grown at elevated and subambient CO2 (mean of 64% and 73% of total cover, respectively, in July and 55% and 63% of total cover, respectively, in September). Over the 6 days prior to CO2 switching, the ratio of NEE at elevated growth CO2 to NEE at subambient growth CO2 averaged 1·40 in July and 1·34 in September Excluding two extreme values of NEE in July (> 20 g CO2 m−2 leaf area day−1; Fig. 2) reduced the ratio of NEE at elevated to subambient CO2 to 1·18, a value similar to the average ratio of 1·16 derived by dividing the NEE measured in each compartment at high CO2 by the NEE measured at low CO2 during CO2 switching. When CO2 was switched in early October, the ratio of NEE at high CO2 to NEE at low CO2 averaged 1·29. This value is similar to the average of 1·34 derived by dividing NEE measured at elevated growth concentrations by NEE measured at subambient growth concentrations prior to switching (Fig. 2). The sensitivity of NEE to CO2 thus apparently depended more on C3–C4 composition of vegetation than on physiological acclimation during mid- and late-season.

Discussion

Plant and ecosystem responses to short-term changes in CO2 concentration may depend on the CO2 concentration at which plants have been grown (Stocker et al. 1997; Dugas et al. 2001). We varied CO2 by 150–200 µmol mol−1 on C3–C4 grassland that had been exposed for three seasons to a 200–550 µmol mol−1 CO2 gradient. The amount by which NEE increased at high compared with low CO2 during switching declined at higher growth concentrations on two dates (May, July), but relationships between NEE response and growth CO2 were weak. For only one date (July) could we identify an explanation from light–response curves for reduced sensitivity of NEE to CO2 at higher growth concentrations. Similarly, we found no relationship between change in ET following CO2 switching and growth CO2. If physiological acclimation occurred in this grassland, it was difficult to detect at the canopy level amid other factors regulating CO2 response. During the mid- and latter-part of the growing season, the C3–C4 composition of vegetation proved a more important regulator of NEE and ET sensitivity to CO2 than growth CO2.

It often has been reported that canopy-level gas exchange is more sensitive to CO2 enrichment in C3 than C4 plants (Drake & Leadley 1991). Elevated CO2 had little effect on NEE of C4-dominated tallgrass prairie until late in the growing season when CO2 enrichment stimulated C uptake by delaying senescence (Ham et al. 1995), for example, but CO2 enrichment increased the seasonal total of NEE per unit of above-ground biomass of a C3 Scirpus olneyi (Grey) stand by 48% during the second year of treatment (Drake & Leadley 1991). At least initially following CO2 treatment, biomass production of the dominant C4 grass (Bothriochloa ischaemum) in our study was highly responsive to growth CO2 (Polley et al. 2003). Leaf net photosynthesis of B. ischaemum remained a linear function of growth CO2 throughout the experiment (Anderson et al. 2001), but biomass of this short-statured grass decreased sharply during the final 2 years of CO2 exposure as taller grasses and forbs proliferated (Polley et al. 2003). As a fraction of C4 grass production, biomass of B. ischaemum also declined from 94% in 1997 to 50% in 2000 across CO2 treatments. Our finding that daytime NEE responded little to short-term variation in CO2 in C4-dominated grassland implies B. ischaemum was replaced by less-responsive C4 grasses.

Neither the ratio C3 to total (C3 + C4) plant cover nor the CO2 treatment at which grassland had been grown consistently affected the CO2 response of ET, implying that short-term CO2 enrichment reduced ET in both C3 and C4 species. Indeed at the May CO2 switching, the higher CO2 concentration reduced ET more in C4- than C3-dominanted grassland. CO2 enrichment reduced stomatal conductance of both C3 and C4 species in this grassland, but the absolute decline in conductance from subambient to elevated concentrations was greater by more than a factor of 2 for the C3 grass Bromus japonicus and by more than an order of magnitude for the C3 forb Solanum dimidiatum than for the C4 grass Bothriochloa (Anderson et al. 2001; Maherali et al. 2002). ET rates have been measured in relatively few CO2 studies on intact ecosystems, but CO2 enrichment has been shown to reduce rates of soil water depletion, and by inference to reduce ET, in both C3- and C4-dominated ecosystems (Fredeen et al. 1997; Owensby et al. 1997; Nelson et al. 2004).

In this mesic grassland, carbon and water fluxes remained responsive to short-term change in CO2 after three growing seasons at different CO2 levels. NEE was greater (16–20%) and ET was smaller (9–20%), on average, at high than low CO2 during CO2 switching in May and July. However, the amount by which NEE and ET responded to CO2 change depended on both growth CO2 and the C3–C4 composition of vegetation. Thus, our results imply that sensitivity of this grassland to CO2 depends both on management and other influences on C3–C4 dynamics (e.g. C4 dominance may be increased by spring fires and reduced by grazing ungulates) and on prior CO2 exposure.

Acknowledgements

Ron Whitis operated CO2 chambers. Anne Gibson and Catherine Miller participated in data analysis. Laurie Anderson, Jeff Baker and Phil Fay provided helpful reviews.

Ancillary

Advertisement