Environmental and stomatal control of photosynthetic enhancement in the canopy of a sweetgum (Liquidambar styraciflua L.) plantation during 3 years of CO2 enrichment


Correspondence: Carla A. Gunderson. Fax: 1-865-576-9939, E-mail: gundersonca@ornl.gov


Light-saturated photosynthetic and stomatal responses to elevated CO2 were measured in upper and mid-canopy foliage of a sweetgum (Liquidambar styraciflua L) plantation exposed to free-air CO2 enrichment (FACE) for 3 years, to characterize environmental interactions with the sustained CO2 effects in an intact deciduous forest stand. Responses were evaluated in relation to one another, and to seasonal patterns and natural environmental stresses, including high  temperatures, vapour pressure deficits (VPD), and drought. Photosynthetic CO2 assimilation (A) averaged 46% higher in the +200 µmol mol−1 CO2 treatment, in mid- and upper canopy foliage. Stomatal conductance (gs) averaged 14% (mid-canopy) and 24% (upper canopy) lower under CO2 enrichment. Variations in the relative responses of A and gs were linked, such that greater relative stimulation of A was observed on dates when relative reductions in gs were slight. Dry soils and high VPD reduced gs and A in both treatments, and tended to diminish treatment differences. The absolute effects of CO2 on A and gs were minimized whenever gs was low (<0·15 mol m−2 s−1), but relative effects, as the ratio of elevated to ambient rates, varied greatly under those conditions. Both stomatal and non-stomatal limitations of A were involved during late season droughts. Leaf temperature had a limited influence on A and gs, and there was no detectable relationship between prevailing temperature and CO2 effects on A or gs. The responsiveness of A and gs to elevated CO2, both absolute and relative, was maintained through time and within the canopy of this forest stand, subject to seasonal constraints and variability associated with limiting air and soil moisture.


Understanding leaf-level responses to CO2 enrichment, and the impact of environmental variation on those responses, remains essential for projecting long-term changes in forest ecosystem function and structure. Considerable uncertainty surrounds such potential secondary and tertiary effects of increasing atmospheric CO2 as altered productivity and water use, feedbacks to the carbon, nitrogen and water cycles, and patterns of forest succession (Eamus & Jarvis 1989; Ceulemans & Mousseau 1994; Field, Jackson & Mooney 1995; Johnson & Ball 1996; Drake, Gonzàlez-Meler & Long 1997; Saxe, Ellsworth & Heath 1998; Norby et al. 1999), yet the question of higher scale consequences is moot without sustained foliar responses.

The CO2 enrichment of seedlings and saplings has repeatedly been shown to increase light-saturated photosynthesis (A), but the degree of stimulation has varied widely, not only among studies (e.g. from none to +180%), but on different dates within each study (reviewed in Ceulemans & Mousseau 1994; Gunderson & Wullschleger 1994; Curtis 1996; Curtis & Wang 1998; Saxe et al. 1998; Medlyn et al. 1999; Norby et al. 1999). Taxonomic differences, duration of the experiments, developmental stage of the trees, temperature differences, and environmental stresses such as drought have all been proposed as significant sources of variability (Gunderson & Wullschleger 1994; Curtis 1996; Saxe et al. 1998). The latter two are most likely to explain within-experiment variability, notwithstanding differences related to leaf age (Rey & Jarvis 1998; Turnbull et al. 1998). It has been hypothesized that the relative effect of CO2 enrichment on A will be greater under water-limited conditions (Idso & Idso 1994), and at higher temperatures (Long 1991; Idso & Idso 1994), yet the evidence for such patterns in mature tree canopies remains quite limited. Indeed, when considering the ecosystem-scale impacts of elevated CO2, the absolute differences in assimilation under stressful conditions may be more meaningful than the relative responses.

Despite the well-known responsiveness of stomatal aperture to [CO2], documenting a sustained stomatal response to CO2 enrichment in woody plants has not been straightforward, and reported impacts range from none (even increases) to reductions of 60% (reviewed in Morison 1985; Gunderson & Wullschleger 1994; Field et al. 1995; Drake et al. 1997; Curtis & Wang 1998; Saxe et al. 1998; Norby et al. 1999; Medlyn et al. 2001). Taxonomic, ontogenetic, and environmental factors have been proposed to account for some of the differences. The hypotheses most pertinent to within-experiment variability involve interactions between CO2 exposure and water limitations (Morison 1993; Beerling et al. 1996). It has been hypothesized that CO2 enrichment reduces relative stomatal sensitivity to drought, vapour pressure deficit (Heath 1998), or to CO2 itself (discussed in Sage 1994; Assmann 1999), but again, absolute differences may be more important for predicting stand-level water balance. Because photosynthesis and stomatal conductance (gs) are tightly coupled, environmental influences on one cannot be considered in isolation from the other. Environmentally induced variation in stomatal sensitivity to CO2 could thus affect photosynthetic responses, in the same way that species differences in stomatal sensitivity have been proposed to explain differences in photosynthetic response to CO2 (Idso 1991).

Because the ultimate goal of CO2 enrichment studies is to be able to predict and model changes in intact forest ecosystems, the focus of research has moved toward multi-year CO2 exposures of large-stature trees, using methods such as free-air CO2 enrichment (FACE) (Hendrey et al. 1999). The challenge of current research is to relate variation in leaf-level CO2 responses to variations in atmospheric and soil conditions, and to integrate this information to the canopy scale, providing a link to forest growth and ecosystem processes.

In this article we report on 3 years of foliar gas exchange in a closed-canopy stand of deciduous trees (Liquidambar styraciflua L., or sweetgum) exposed to CO2 enrichment in a FACE facility (Norby et al. 2001; Wullschleger & Norby 2001). The objectives included quantifying both the sustained responses to elevated CO2 and the environmental sources of variation in photosynthetic and stomatal responses, focusing on temperature, vapour pressure deficit, and soil moisture availability. We also examined the interdependence of stomatal and photosynthetic responses to CO2 enrichment, and compared responses in upper and mid-canopy foliage. The following hypotheses were addressed: (1) that photosynthetic enhancement and stomatal sensitivity would persist, and not decrease through acclimation to CO2 enrichment; (2) that the relative effect of CO2 on photosynthesis would be greater under environmental stress such as low soil moisture, high vapour pressure deficit, and high temperatures; and (3) that moisture stress would limit the impact of CO2-induced stomatal closure.

Materials and methods

Site description

The experimental site is in a 1·7 ha plantation of sweetgum trees (Liquidambar styraciflua L) established in 1988 from 1 year bare-root seedlings on the Oak Ridge National Environmental Research Park in Roane County, Tennessee (35°54′ N, 84°20′ W). The soil is classified as an Aquic Hapludult with a silty clay loam texture, pH approximately 5·5–6·0, and is moderately well drained (van Miegroet, Norby & Tschaplinski 1994; Norby et al. 2001). Seedling spacing in the 185 m by 70 m section of the plantation used for the FACE study was 2·3 m × 1·2 m. Herbicide application was used during establishment (1989 and 1990), but no fertilizer was applied. The canopy has been closed since at least 1996, and the trees are in a linear growth phase (for additional details, see Norby et al. 2001). When the CO2 exposure began in the spring of 1998, the trees were 12–13 m in height, with live crown beginning at 8–9 m above the ground. Each year the trees have added approximately 1 m in height, and by the next spring have lost lower branches so that the live crown remains approximately the same depth (4–5 m). In the spring of 2000, the trees were approximately 15–16 m tall.

FACE facility and treatments

The FACE facility at Oak Ridge National Laboratory consists of five circular plots, 25 m in diameter, within the sweetgum plantation. In each of two ‘elevated CO2’ plots, the air is enriched with CO2 dispensed from surrounding vent pipes, according to wind direction. The target CO2 concentration within the plot is regulated based on the design, equipment and software of Hendrey et al. (1999). Details of the Oak Ridge FACE facility and operation may be found in Norby et al. (2001). Three ‘ambient CO2’ plots serve as controls for the experiment, two surrounded by the same towers, vent pipes, and blowers as the elevated CO2 plots, but receiving only ambient air, and a third ambient CO2 plot without towers or blowers. No differences in gas exchange (data not shown) or other response variables have been detected between blower and non-blower controls (Norby et al. 2001).

The CO2 treatment began in April 1998, prior to leaf-out, and CO2 regulation was fully operational by 11 May 1998. Fumigation was terminated after leaf senescence each autumn and re-initiated each spring before new leaves emerged. The treatment set-point for 1998 was a constant 560 µmol mol−1 CO2, approximately 200 µmol mol−1 above global average [CO2]. This was modified in 1999 and 2000 to incorporate a natural pattern of diurnal variation, with daytime set-point of 565 µmol mol−1 and a higher night-time set point of 645 µmol mol−1. Actual mid-day CO2 concentrations coincident with gas exchange sampling (0900 to 1600 h) averaged 548, 556 and 555 µmol mol−1 in 1998, 1999, and 2000, respectively, at the centre of the plots. Means include periods when fumigation was suspended for experimental purposes, or because of high winds, equipment failure, or delayed CO2 delivery. Mid-day [CO2] in the control plots averaged 362, 365, and 366 µmol mol−1 in 1998, 1999, and 2000, respectively.

Canopy access and leaf selection

Canopy access was achieved using hydraulic personnel lifts (Model UL48; UpRight, Inc., Selma, CA, USA) positioned at the centre of each plot. The aerial work platforms extended up to 15·5 m, providing access to multiple crown positions. Lifts became operational in July 1998, and gas exchange measurements began at that point, ending in mid-September. Measurements for 1999 and 2000 were conducted between May and October.

Leaves selected for measurements were representative of general canopy conditions, fully expanded, and near the tips of the branches in either the upper or mid-canopy. Foliage from four to six trees in each plot was accessible from the lift, and data from all leaves measured at a given canopy position (four to eight leaves per plot on each date) were used to derive mean values for the plot.

Upper canopy leaves were chosen from branches exposed to mid-day full sunlight, in the top 2–3 m of the canopy, i.e. from the upper 20–30% of canopy foliage. Upper canopy measurements were taken on five to seven dates each season.

Mid-canopy (i.e. mid-height) measurements were taken approximately 3–4 m down into the canopy, on fully expanded leaves near the branch tips. Leaves within this portion of the canopy received 20–50% of full PPFD (estimated each summer after full canopy development, using a 1 m line quantum sensor (model LI-191SA; Li-Cor, Inc., Lincoln, NE, USA). Mid-canopy leaves were measured in July of 1998 and 1999, and four times in 2000.

Gas exchange methods

Gas exchange responses were evaluated at the nominal treatment [CO2], under saturating light conditions. Measurements were conducted between 0900 and 1600 h, on a single date in some cases, or over 2 to 4 d in other cases, when light or CO2 response curves were being conducted in the same sampling period.

Data from 21 July 1998 onward were obtained using the LI-6400 steady state photosynthesis system (Li-Cor, Inc.) using the 6 cm2 cuvette. Measurements were taken at saturating irradiance (1800–2000 µmol m−2 s−1 PAR) provided by red/blue LED light source (model LI 6400–02B; Li-Cor, Inc.).

Cuvette temperatures were set based on mid-afternoon temperatures forecast for each measurement period. Resulting cuvette air temperatures were highly correlated with prevailing air temperatures, and means were ± 1–2 °C of mid-day site air temperatures during the measurements (Fig. 1a; R2 = 0·83 for the correlation). Heat generated by the light source increased leaf temperatures by 1–2·8 °C, but leaf temperatures (Fig. 1a) were still correlated with prevailing air temperatures (R2 = 0·82). Cuvette humidity was not controlled, except as needed to avoid condensation on rare occasions when the relative humidity of the cuvette exceeded 80%. Atmospheric vapour pressure deficits (VPD) at the site varied considerably among measurement dates (Fig. 1b), but also changed markedly within a day, mostly as a function of increasing air temperatures throughout the day. For example, between 0900 and 1600 h, VPD increased from 0·9 to 2·3 kPa on day 208, 1999, and from 1·4 to 2·8 kPa on day 258, 1999. Nevertheless, mean leaf-to-air vapour pressure deficits in the cuvette tracked mid-day atmospheric VPD within 0·7 kPa (Fig. 1b), notwithstanding the counteracting influences of elevated leaf temperatures and higher humidity of the cuvette environment (R2 = 0·52 for the correlation). There were no significant differences in VPD and leaf temperature between treatments.

Figure 1.

Environmental variation over the three growing seasons in (a) air and leaf temperatures during the measurements; (b) vapour pressure deficits at the site (mid-day) and in the cuvette; (c) soil water potential in the two treatments, and total precipitation for each month. Means (from n= 2 or 3 plots) are shown; bars indicate standard error of the mean. Symbols are as indicated on the figure, and in (c), the open square connected by a dotted line indicates SWP measured between plots.

The CO2 concentrations in the cuvette were regulated using a CO2 mixer and injector system (Model LI-6400–01; Li-Cor, Inc.) and cartridges of compressed CO2. Inlet air CO2 concentrations were set to 360–365 µmol mol−1 for measurements in the ambient CO2 plots, and 560–565 µmol mol−1 in the elevated CO2 plots, resulting in mean cuvette CO2 concentrations (Ca) of 346 and 541 µmol mol−1 in the ambient and elevated CO2 treatments, respectively. Some of the gas exchange data were obtained while developing CO2 response curves, using 10 inlet CO2 concentrations between 0 and 1500 µmol mol−1 (Sholtis et al. MS in preparation). The relationships between A and the intercellular CO2 concentration (Ci) in these curves were also used to assess stomatal limitation.

In addition to the LI-6400 data, on 7–9 July 1998, measurements were obtained using the LI-6200 photosynthesis system (Li-Cor, Inc.), using prevailing light, temperatures, humidity, and [CO2], in both upper and mid-canopy positions. Mean cuvette conditions for the upper canopy measurements were 1213 µmol m−2 s−1 PAR, 37 °C, and CO2 concentrations of 336 ± 9 and 550 ± 25 µmol mol−1 in ambient and elevated CO2 treatments, respectively. Mean mid-canopy PAR was 687 ± 210 µmol mol−1, at or near saturating for both A and gs in mid-canopy sweetgum leaves (data not shown).

Data obtained as part of the gas exchange measurements included area-based light-saturated net photosynthetic CO2 assimilation (A), stomatal conductance to water vapour (gs), leaf and air temperature, relative humidity, leaf-to-air vapour pressure deficit (VPD), intercellular CO2 concentration (Ci), and leaf-level photosynthetic water use efficiency, or instantaneous transpiration efficiency (ITE), which was calculated as assimilation/transpiration.

Soil moisture measurements

Soil water content (%, v/v, integrated from 0 to 20 cm soil depth) was measured periodically with a time domain reflectometer (TDR; Soil Moisture Equipment Corp., Santa Barbara, CA, USA) following the procedure of Topp & Davis (1985). Six pairs of stainless steel rods were installed in each plot, to a depth of 20 cm, providing a total of 12 and 18 soil water content observations for the elevated and ambient CO2 treatments, respectively. Volumetric soil water content was converted to soil water potential (SWP) using bulk density determinations and a moisture release curve constructed using thermocouple psychrometry (True Psi; Decagon, Pullman, WA, USA). Soil water potentials from all rod positions in a plot were averaged to produce plot means for each date that measurements were taken. Treatment SWP values (Fig. 1c) were calculated from the plot means (n= 2 elevated CO2 plots and n= 3 ambient CO2 plots). SWP data paired with the September 1998 gas exchange measurements were obtained by treatment-specific interpolation during the extended drought period, as only limited measurements were taken during that period. A trend toward slightly higher SWP in the elevated CO2 plots was apparent during dry periods, particularly in 2000, but differences were not statistically significant.

Statistical analysis

Treatment effects were evaluated using the mean values of gas exchange and atmospheric and soil conditions from each plot for each measurement period. Two-way analysis of variance (CO2 treatment × date) with plot as the experimental unit (n= 2 and n= 3) was used to assess treatment effects on gas exchange.

Environmental influences

Light-saturated upper canopy gas exchange relationships were evaluated with respect to variation in individual environmental variables, using both linear and non-linear models. Least-squares regression techniques were applied to A and gs, and to treatment differences and ratios of rates (elevated/ambient) as a function of VPD, SWP, and temperature. Mid-canopy data were insufficient for analysis in relation to environmental variation.

Differences between significant regression lines were evaluated with an F-test, based on the principle of conditional error, as described by Neter & Wasserman (1974). Where differences were apparent, tests for homogeneity of slopes were also performed (Cody & Smith 1997).

Multiple linear regression analysis using the stepwise selection technique was applied to upper canopy A, gs and ITE (and to the ratios and differences between CO2 treatments for each of these) as a function of VPD, SWP, and temperature (Cody & Smith 1997).

Relationships between photosynthetic and stomatal responses

The relationships between photosynthetic and stomatal responses in each treatment were evaluated by linear and non-linear least-squares regression using three empirical approaches. First, mean A in each treatment group was described as a logarithmic function of mean gs. Second, the relative enhancement of A on each date was described as a linear function of the relative change in gs (Idso 1991). In the third approach, using the steady-state (LI-6400) data, A and gs within each treatment were related using the Ball et al. model (Ball, Woodrow & Berry 1987) which describes gs as a linear function of the index (A × hs/Ca), where hs is the relative humidity (as a fraction) at the leaf surface. This model describes the interdependent covariance of A and gs in response to humidity and [CO2] (Aphalo & Jarvis 1993); it is used to relate A and gs in models of CO2 and water vapour exchange at canopy and ecosystem scales (e.g. Harley & Baldocchi 1995; Dang, Margolis & Collatz 1998). Differences between regressions were evaluated as described above.

Stomatal limitation of A in relation to moisture stress

Stomatal limitation and its influence on photosynthetic response to CO2 were evaluated by comparing the Ci/Ca ratios for each treatment and by estimating relative stomatal limitation of A (RSL) from upper canopy A-Ci curves generated on 10 dates. Because there were no significant treatment differences in model parameters for the A-Ci curves (Sholtis et al. MS in preparation), a composite curve was constructed for each date. Composite curves were constructed using means of the parameters generated by least-squares regressions for individual leaves using the empirical function describing A as a rectangular hyperbolic function of Ci (Olsson & Leverenz 1994; Photosyn Assistant software; Dundee Scientific, Dundee Scotland, UK). RSL was then calculated as (A0 − A)/A0, where A was the mean assimilation rate for the treatment, and A0 the assimilation rate that would occur if resistance to CO2 diffusion were zero, that is, A at Ci equal to the treatment Ca (Farquhar & Sharkey 1982). A paired comparisons t-test (pairing by sample date) was used to evaluate treatment differences.

RSL, Ci/Ca ratios, and relative photosynthetic enhancement by CO2 enrichment (as A560/A360) were also evaluated as functions of total moisture stress, using gs as an indicator of combined atmospheric and soil moisture stress.


CO2 effects on photosynthesis

Light-saturated photosynthetic rates (A) in the sweetgum upper canopy were highest in mid-summer and declined late in the growing season. The decline was especially pronounced in 1998 (Fig. 2a) during a prolonged drought, when the July to mid-September rainfall totalled only 124 mm (Fig. 1c). Seasonal patterns were similar in both treatments, however, and A remained higher in elevated CO2 over the 3 year period. Differences between treatments were significant, averaging 5·0 µmol m−2 s−1 higher in elevated CO2 (Table 1). The relative effect of CO2 enrichment on A, calculated as a ratio of elevated/ambient (A560/A360), was remarkably stable during most of the growing season (mean 1·46, or 46% higher, Table 1), but varied late in the 1998 and 1999 growing seasons, when rates in ambient CO2 were low. For example, on 10 September 1998, when gs and A were sharply reduced by drought (A ≤ 3 µmol m−2 s−1 in either treatment, Fig. 2a), the treatment differences were necessarily small (e.g. A560 − A360 = −1·1 µmol m−2 s−1), and A560/A360 dropped to 0·65. Five days later, the rates remained low and differences small, but positive (+1·8 µmol m−2 s−1). The result was a rebound in the relative enhancement to 1·33. This suggests that the apparent loss of enhancement on 10 September was not an indication of acclimation to the CO2 treatment, but the result of variability obscuring small treatment differences. In September and October of 1999, by contrast, A560 was more than double A360 (Fig. 2a). As in 1998, this was associated with low rainfall and a fairly small denominator (A360 < 8 µmol m−2 s−1), which tends to magnify relative differences.

Figure 2.

Light-saturated gas exchange of upper canopy sweetgum foliage, measured at the nominal treatment [CO2], during 3 years of CO2 enrichment: (a) net assimilation; (b) stomatal conductance to water vapour; and (c) instantaneous transpiration efficiency. Means (from n= 2 or 3 plots); bars indicate standard error of the mean.

Table 1.  Light-saturated gas exchange, ratios, and differences in upper and mid-canopy foliage, measured at the treatment [CO2]. Values are 3 year means and standard errors (in parentheses) for each treatment (cfFig. 2)
  1. Significance probabilities are given for anova associated with CO2 treatment and time, using plot as the unit (n= 2 elevated CO2 plots, n= 3 ambient plots). Variables are photosynthesis (A,µmol m−2 s−1), stomatal conductance (gs, mol m−2 s−1), instantaneous transpiration efficiency (ITE = A/transpiration, µmol CO2 mmol−1 H2O), and ratios of intercellular to ambient [CO2] (Ci/Ca).

Upper canopy (18 dates)
 Elevated15·9 (1·2) 0·17 (0·02)5·85 (0·37)  0·64 (0·01)
 Ambient10·9 (0·8) 0·23 (0·03)3·44 (0·18)  0·67 (0·01)
 Elevated/ambient1·46 (0·31) 0·76 (0·03)1·69 (0·06)  0·94 (0·01)
 Elevated–ambient 5·0 (0·5)−0·06 (0·01)2·41 (0·23)−0·04 (0·01)
 Significance probabilities
 CO2 treatment<0·01<0·01<0·01  0·03
 Treatment × dateNSNS<0·01NS
Mid-canopy (six dates)
 Elevated13·1 (2·0)  0·16 (0·02)6·33 (0·58)  0·65 (0·03)
 Ambient 9·0 (1·6)  0·19 (0·03)3·56 (0·33)  0·69 (0·03)
 Elevated/ambient1·47 (0·10)  0·86 (0·09)1·78 (0·32)  0·94 (0·02)
 Elevated–ambient 4·1 (0·6)−0·04 (0·02)2·76 (0·31)−0·04 (0·02)
 Significance probabilities
 CO2 treatment<0·01  0·06<0·01  0·04
 Date  0·01<0·01<0·01<0·01
 Treatment × dateNSNSNSNS

Mean light-saturated A in mid-canopy foliage (Fig. 3a) was 15–20% lower than A in upper canopy foliage, but the effect of CO2 enrichment was almost identical, and A560/A360 averaged 47% higher in elevated CO2 (Table 1). There was no decline in enhancement over the 3 year period, nor within the 2000 growing season.

Figure 3.

Light-saturated gas exchange of mid-canopy sweetgum foliage, measured at the nominal treatment [CO2], during 3 years of CO2 enrichment: (a) net assimilation; (b) stomatal conductance to water vapour; and (c) instantaneous transpiration efficiency. Bars indicate standard error of the mean.

CO2 effects on stomatal conductance

Upper canopy gs (Fig. 2b) followed the same seasonal patterns as did A, with highest gs in July, and reductions in late summer. Rates were almost always lower in elevated CO2 foliage than in ambient CO2 (mean difference −0·06 mol m−2 s−1, Table 1), and overall, treatment differences were highly significant (Table 1). The relative reductions in gs averaged 24% over the three growing seasons (Table 1). As was true for A, the ratio (gs-560/gs-360) was most variable late in 1998 and 1999 when gs was reduced by drought and differences were small, varying from 0·48 (September 1998) to 1·06 (October 1999).

Mid-canopy gs (Fig. 3b) tended to be lower than upper-canopy gs (Fig. 2b) for comparable time periods, and the reduction with elevated CO2 averaged only 14% (P = 0·06, Table 1). As in the upper canopy, differences were minimal on the dates when gs was low, i.e. near 0·1 mol m−2 s−1.

CO2 effects on water use efficiency

Leaf-level photosynthetic water use efficiency, or instantaneous transpiration efficiency, ITE (A/transpiration) was always higher in elevated CO2, in both canopy positions (Figs 2c & 3c). ITE declined in both treatments late in the season, and in association with high VPD (when transpiration was higher). The CO2-induced difference in ITE averaged 2·41 and 2·76 µmol CO2/mmol H2O in upper and mid-canopy foliage, respectively (Table 1) with differences smaller during the drought of September 1998 (Fig. 2c). The mean relative increases were 68% in the upper canopy, and 78% in mid-canopy (Table 1), larger than relative effects on either A or transpiration alone. The ratio of A to gs can also be used to assess the contrasting impacts of elevated CO2 on carbon gain and water loss. This index, which removes the influence of VPD on transpiration, still varied by a factor of two over time (data not shown). As with ITE, A/gs was always higher in elevated CO2, by a mean of 92% (SE, 6) in the upper canopy, and 75% (SE, 9) in mid-canopy foliage.

Effects of environmental variation on gas exchange responses

The range of natural environmental variation observed in this experiment (Fig. 1) is potentially sufficient to alter both A and gs through a combination of biochemical and stomatal mechanisms. In order to quantify the impacts, we first evaluated the individual relationships between upper canopy gas exchange and VPD, SWP, and leaf temperature, and then used multiple linear regression to assess the relative impacts of each variable.

Photosynthesis and gs both decreased at high VPD (Fig. 4), which occurred at various times during the experiment (Fig. 1b). Of three equations commonly used to describe the response of gs to VPD, the best fit was achieved using an exponential function (Fig. 4a). The curves were significantly different in an F-test (P = 0·04), although the differences were less pronounced at high VPD. Fits of linear and logarithmic regressions were also acceptable over this VPD range (R2 = 0·46–0·64).

Figure 4.

Upper canopy gas exchange as a function of leaf-to-air vapour pressure deficit: (a) stomatal conductance; (b) photosynthetic CO2 assimilation. Curves (dashed lines for ambient CO2; solid lines for elevated CO2) represent the least-squares regressions using (a) exponential and (b) linear models.

The best fit for A was as a linear function of VPD (Fig. 4b). The lines were significantly different (P < 0·01), but the slopes were not, indicating that the increase in A with elevated CO2 was similar across all levels of VPD. Similar decreases in ITE were observed with increasing VPD (data not shown), and variation in VPD explained 62 and 51% of the variation in ITE in ambient and elevated CO2, respectively.

Reduced gas exchange was also associated with drought and drying soils (Fig. 5), which occurred late in the growing season, particularly in 1998 and 1999, when July, August and September rainfall was unusually low (Fig. 1c). The variation in SWP in the upper 20 cm of soil explained 41% (ambient) and 34% (elevated) of the total variation in gs using linear regression (Fig. 5a). Across the range of soil moisture contents, gs was lower in elevated CO2, and the regressions for the two treatments were significantly different (P = 0·01), although the slopes were not. Variation in SWP explained more of the total variation in A (Fig. 5b), approximately 60%. The two regressions were significantly different (P < 0·01), as were the slopes (P < 0·01), so that the lines converged at low soil water potentials. The relationships between ITE and SWP in the treatments (not shown) were similar to those observed between A and SWP (R2 = 0·41 and 0·36 in ambient and elevated CO2, respectively).

Figure 5.

Upper canopy gas-exchange as a function of soil water potential from 0 to 20 cm: (a) stomatal conductance; (b) photosynthetic CO2 assimilation. Lines (dashed lines for ambient CO2; solid lines for elevated CO2) represent the least-squares linear regressions.

Variation in leaf temperature alone explained very little of the total variation in A or gs with either a linear or parabolic fit (R2 = 0·01–0·12, data not shown). There was a slightly stronger relationship between ITE and temperature (R2 = 0·12 and 0·17), because higher temperatures tend to increase transpiration, thus decreasing ITE.

Multiple linear regression was also used to assess the interacting effects of environmental variation on steady-state A, gs, and ITE, and on CO2 effects, as treatment differences and ratios. Variables were selected using the stepwise method (Cody & Smith 1997) to account for correlation among variables. For example, high VPD tended to occur more often during drought, or at high temperatures. In most cases, after accounting for variation in the most significant individual environmental factor, adding the others did not result in an improvement in the explained variance (Table 2). Variation in VPD was the best single regressor for many of the responses, and higher VPD tended to decrease both rates and treatment differences. After accounting for the effects of VPD on gs and A360, leaf temperature explained a small but significant additional portion of the variation (Table 2), and indicated that A and gs tended to increase with temperature. Reductions in A560, and diminished treatment differences in A, were best explained by decreasing SWP. Leaf temperature explained only 20–30% of the variability in CO2 treatment effects on ITE, and no other factors were significant. None of the environmental factors explained a significant portion of the variation in relative CO2 effects on A or gs, i.e. the variation in A560/A360, or gs-560/gs-360.

Table 2.  Coefficients, intercepts, and significance probabilities in the stepwise multiple linear regressions of steady-state light- saturated upper canopy gas exchange (and CO2 effects on gas exchange) as functions of vapour pressure deficit (VPD), soil water potential (SWP) and leaf temperature (T-leaf)
  1. NS indicates that the remaining variables did not explain a significant portion of the residual variability (i.e. P > 0·15) after accounting for the most significant factors (those listed). None of the factors explained a significant portion of the variability in the relative CO2 effects on A or gs (as elevated/ambient).

 Ambient CO2−7·84<0·01NS0·3430·0715·4  0·01
 Elevated CO2NS19·2<0·01NS27·0<0·01
 Elevated–ambientNS3·20·06NS 7·2<0·01
Stomatal conductance
 Ambient CO2−0·192<0·01NS0·0100·05 0·2770·06
 Elevated CO2−0·133<0·01NS0·0130·03 0·0310·83
 Ambient − elevated−0·0450·02NSNS 0·137<0·01
Instantaneous transpiration efficiency
 Ambient CO2−1·40<0·01NSNS 6·2<0·01
 Elevated CO2−2·45<0·01NSNS10·8<0·01
 Elevated − ambientNSNS−0·2100·03 9·0<0·01
 Elevated/ambientNSNS−0·0420·09 3·0<0·01

Relationships between stomatal and photosynthetic effects

Rates of A and gs were tightly coupled, and A at elevated CO2 was always higher than at ambient CO2 for an equivalent gs (Fig. 6a). Within each treatment, A could be characterized as a logarithmic function of gs (Fig. 6a). These curvilinear relationships are consistent with restriction of Ci by partial stomatal closure in response to the environment, rather than with a stomatal response to variation in photosynthetic capacity (Wong, Cowan & Farquhar 1979). Logarithmic relationships similar to those in Fig. 6a were also observed among A and gs of individual leaves within a sampling date (data not shown), indicating a robust association between the two variables and suggesting that variability in gs is responsible for much of the within-treatment variability in A.

Figure 6.

(a) Upper canopy assimilation as a function of stomatal conductance to water vapour in ambient and elevated CO2 foliage, measured on 18 dates over three growing seasons. Curves are logarithmic regressions of photosynthesis on conductance. (b) Relative effect of elevated CO2 on photosynthesis (as the ratio of elevated rate/ambient rate), plotted as a function of relative stomatal effect, shown with the linear regression combining upper canopy (solid triangles) and mid-canopy (open triangles) data.

The relative effect of [CO2] on gs on any given date was also tightly coupled with the relative enhancement of A on that date. Photosynthetic enhancement by elevated CO2 (A560/A360) was greater on dates when high CO2 caused the least relative reduction in gs (Fig. 6b), and this was true in both upper and mid-canopy foliage.

The relationships between A and gs were also compared using the Ball et al. (1987) model which incorporates stomatal sensitivity to humidity and Ca. In this analysis, mid-canopy relationships were not significantly different from those in the upper canopy, and data from both positions were used in the regressions (Fig. 7). Regressions for the two CO2 treatments were significantly different (P < 0·01) because gs was lower in elevated CO2 for an equivalent index A × hs/Ca. The slopes were not different, suggesting that stomata were not significantly more or less sensitive to variation in Ca or hs. This was confirmed by individual regressions of gs on hs, A, and A/Ca. The pairs of regressions (0·55 < R2 < 0·85) remained significantly different overall for the two treatments (P < 0·05), and the slopes (apparent sensitivity of gs to each factor) similar.

Figure 7.

Stomatal conductance as a function of the Ball et al. (1987) index. Lines represent the best-fit linear regressions for each CO2 treatment, combining upper and mid-canopy data (dashed lines for ambient CO2; solid lines for elevated CO2).

Stomatal limitation of photosynthesis

As CO2-mediated decreases in gs appeared to proportionally reduce photosynthetic enhancement (Fig. 6b), we examined the CO2 effects on stomatal control of photosynthesis by two methods. First, we evaluated the effects of reduced gs on the Ci/Ca ratio (Sage 1994). Although elevated CO2 reduced gs by approximately 24%, the effect on the Ci/Ca ratio was small but significant (Table 1). The Ci/Ca ratio at elevated CO2 averaged 94% of that in ambient CO2 leaves. The effect was fairly consistent, and almost identical in the upper canopy and mid-canopy foliage. A CO2-mediated reduction in the Ci/Ca ratio is described by Sage (1994) as a conservative pattern of stomatal acclimation, beneficial if water limits growth more than does carbon supply. If the lower Ci/Ca ratio in these leaves is the result of conservative stomatal acclimation, however, the degree of adjustment did not respond to variation in the water limitations that reduced gs. That is, although the Ci/Ca ratio tended to decrease in response to integrated moisture stress, as reflected by decreasing gs (Fig. 8a), this was true in both treatments, and the relative reduction in Ci/Ca by elevated CO2 did not change. During the most severe drought, in September 1998, direct limitations to A appeared to outweigh stomatal constraints such that the Ci/Ca ratio increased in both treatments (points inside the dotted ellipse in Fig. 8a), and remained lower in elevated CO2.

Figure 8.

Gas exchange responses expressed using stomatal conductance to water vapour to integrate the effects of interacting environmental stresses. Stomatal control of A as indicated by (a) the ratio of intercellular to ambient [CO2] and (b) relative stomatal limitation of photosynthesis (see text). (c) Relative photosynthetic enhancement (elevated/ambient) in upper and mid-canopy foliage as a function of gs in the ambient treatment on that date. The solid line is the predicted enhancement, calculated from ratios of the logarithmic relationships of Fig. 6a, assuming that gs is 24% lower in elevated CO2. Dotted ellipses enclose data from September 1998, during the most severe drought, when non-stomatal factors predominated.

The relative stomatal limitation to photosynthesis (RSL) was also assessed by the method of Farquhar & Sharkey (1982), using data from A–Ci curves. The mean RSL in ambient CO2 was 26·9% (SD, 5); mean RSL was actually lower in elevated CO2, by 3–7%, with a mean of 22·8% (SD, 6). The differences, although small, were significant in a paired comparisons t-test (P < 0·01, pairing by sample date). RSL in both treatments tended to increase when gs was low (Fig. 8b), again with the exception of the September 1998 data (inside ellipse), when both gs and A were strongly limited by drought, and mesophyll limitations were relatively greater than stomatal limitations. The treatment difference in RSL on that date was maintained, however.

As the environmental factors controlling gas exchange interact in a complex fashion, patterns in the responses of Ci/Ca (Fig. 8a), RSL (Fig. 8b), and A (Fig. 6) were clearer when expressed relative to the concurrent gs than when expressed as a function of the underlying environmental variation (cf, for A, Figs 4b & 5b versus Fig. 6). Accordingly, A560/A360 was evaluated using gs-360 to integrate atmospheric and soil moisture stress of the site (Fig. 8c). Overlaid on the data is the theoretical prediction of A560/A360 generated using ratios of the logarithmic regressions from Fig. 6a, using the assumption that gs-560 is 24% lower than gs-360. The responses of upper and mid-canopy foliage were similar, and fell close to the predicted curve, except at low gs, when CO2-induced stomatal closure was most variable (Fig. 2b). The point substantially below the prediction line, at gs = 0·05 mol m−2 s−1 and A560/A360 = 0·65, is one of two points from the September 1998 drought, which, as noted previously, induced additional non-stomatal limitations of A.


Overall responses of A and gs

Relative photosynthetic enhancement in the elevated CO2 treatment was remarkably stable throughout the first 3 years of this study, varying primarily under conditions of moisture stress, either high VPD or low SWP, which caused partial stomatal closure and, during the driest period, additional biochemical limitations to A. The mean value of A560/A360 was 1·48 in the summer of 2000, higher than either the 1998 mean, or the 3 year mean, indicating that the CO2 effect on A was not lost over time. The 1·46 mean value of A560/A360 was comparable with the results from CO2 enrichment studies in other woody species; for example, the mean responses ranging from approximately 1·45 in early pot studies (Gunderson & Wullschleger 1994; Curtis 1996), to 1·66 and 1·51 in seedlings and saplings growing directly in the ground (Norby et al. 1999; Medlyn et al. 1999). The slightly lower enhancement in the current study is consistent with the smaller CO2 enrichment (+200 versus +300 to +350 µmol mol−1 CO2 in many of the earlier studies).

The mean relative enhancement of A in this study thus supports trends observed in younger plants, but is of note because it has been sustained for 3 years in a stand of large trees that has attained a constant leaf area index, and no longer has the additional sink capacity provided by exponential growth (Norby et al. 2001). Moreover, the relative enhancement of light-saturated A was equivalent in mid-canopy and upper canopy foliage, regardless of differences in leaf chemistry and anatomy associated with the development of shade leaves (Gunderson et al. 1999). This is in contrast to the findings of Herrick & Thomas (1999), who reported comparable enhancements in shade leaves, but a surprising 92–166% enhancement in sun leaves.

The effect of elevated CO2 on gs was a sustained reduction, indicating that stomatal sensitivity to CO2 was not lost over time. The 24% mean relative decrease in gs was smaller than the increase in A, and comparable to mean declines of 11–30% reported across multiple woody species, albeit with appreciable variation within and among studies (Field et al. 1995; Curtis 1996; Drake et al. 1997; Curtis & Wang 1998; Medlyn et al. 2001). Treatment differences in this study, if evaluated for individual dates, would not always have been statistically significant, which is a frequent observation in CO2 experiments with trees (Gunderson, Norby & Wullschleger 1993; Beerling et al. 1996; Heath 1998; Rey & Jarvis 1998; Saxe et al. 1998; Ellsworth 1999; Norby et al. 1999). Differences were always significant when evaluated over time, in the context of the relationships between gs and each environmental variable, or between gs and combinations of variables, e.g. the Ball et al. index.

Interactions with environmental stress

In this experiment, as in most field studies, some degree of correlation was observed among environmental variables. Nevertheless, over three seasons, patterns appeared with respect to both VPD and SWP, and multiple regression confirmed the relationships. Overall, VPD remained the most significant environmental influence, explaining more of the variation than either temperature or soil water potential. On the other hand, SWP had a strong effect on A, especially A560, and decreased treatment differences during severe drought, because of limitations to photosynthetic function in addition to the stomatal effects of drought. Leaf temperature alone (within the range of 26–36 °C observed here), explained very little of the within-treatment variation in either A or gs, (cf. Fig. 5, Gurderson, Norby & Wullschleger, 2000) although it was responsible for some of the residual variation after accounting for VPD.

Absolute reductions in gs with elevated CO2 were smaller in magnitude when VPD was high, and in general, whenever gs was low, although the relative decreases, as gs-560/gs-360, were highly variable at low gs. Small stomatal effects of CO2 have previously been noted in connection with inherently low gs, for example, during the dry season, when VPD was high and gs was low (Goodfellow, Eamus & Duff 1997), on warm sunny days with high VPD (Beerling et al. 1996; Heath 1998), and in species having intrinsically lower gs (Morison 1985; Saxe et al. 1998). These observations are all consistent with the Curtis (1996) meta-analysis, which indicated that absolute reductions in gs tended to be less significant in stressed plants. Similar trends with respect to VPD and SWP were apparent in a few of the long-term European tree studies recently reviewed (Medlyn et al. 2001), but a meta-analysis of those data actually detected a greater relative effect size (gs-elevated/gs-ambient) in the studies (n = 5) involving water-stressed trees.

Results from large trees in CO2 enrichment studies have been mixed with respect to photosynthetic enhancement during naturally occurring dry periods involving high VPD or low SWP. Several reported greater relative enhancement during drought (Dixon, Le Thiec & Garrec 1995, in Picea abies only; Scarascia-Mugnozza et al. 1996; Kellomäki & Wang 1996, in the elevated temperature treatment; Goodfellow et al. 1997; Heath 1998, in Fagus sylvatica; Myers, Thomas & DeLucia 1999, in the July data), but others reported reduced enhancement during drought (Dixon et al. 1995, in Quercus rubra; Ellsworth 1999, during peak drought). In the present study, A declined in response to late season drought in both 1998 and 1999, as a result of both stomatal and non-stomatal limitations, and the absolute difference between CO2 treatments was lower under dry conditions. Calculating relative effects, however, can magnify any noise in the signal whenever the denominator is small, as during drought. In many of the studies cited, rates were quite low and treatment differences small during dry periods, and it is likely that much of the variability in drought–CO2 interactions results from the use of relative response ratios.

Stomatal sensitivity

It has been suggested that CO2 enrichment reduces stomatal ‘sensitivity’ to VPD and soil water potential, potentially reducing drought tolerance and survival under stress (Heath 1998), but it is important to evaluate stomatal sensitivity, for example, to VPD (or the resulting water vapour flux; Mott & Parkhurst 1991), in an appropriate context. In a survey of species differences, Oren et al. (1999) evaluated the slope and both intercepts of the line describing gs as a function of the natural logarithm of VPD. The authors defined sensitivity as the slope of that line, that is, change in gs with change in ln VPD. They also described responses in terms of the y-intercept, which is the maximum gs under favourable conditions, that is, when VPD = 1·0 kPa (ln VPD= 0). Lastly, they evaluated the x-intercept, the level of VPD at which the extrapolated gs equals zero. By their definition of sensitivity, the high CO2 sweetgum in the present study could be considered marginally less sensitive to VPD (P = 0·1 for the difference in slopes). However, because gs under favourable conditions was already lower in elevated CO2 (i.e. the y-intercept would be lower), the decrease from a lower maximum to zero (or to a common minimum gs) would necessarily be smaller in magnitude. In fact, across species, Oren et al. (1999) noted a strong correlation between high maximum gs and high sensitivity to VPD (steep slope). Although change in gs with increase in VPD certainly defines stomatal sensitivity in one sense, the inability of partially closed stomata to close in parallel with those in ambient foliage, as observed here and in other species (Goodfellow et al. 1997; Heath 1998) does not indicate a susceptibility to low humidity stress. Parallel slopes, in fact, would require full stomatal closure at a lower VPD than the controls. Using this reasoning, Medlyn et al. (2001) chose to evaluate the effects of CO2 treatment on the x-intercept, or the VPD at which gs would reach zero. Using either a linear fit (as in Medlyn et al. 2001) or a logarithmic fit (Oren et al. 1999) to our data, the extrapolated gs versus VPD lines crossed and thus predicted that a slightly (but not significantly) higher VPD would be required to completely close stomata at elevated CO2. Because gs never reached zero in this study and treatment differences were not distinguishable at high VPD, we conclude, as did Medlyn et al. (2001), that there was not a significant treatment difference in the VPD required to close stomata, and that CO2 enrichment did not result in greater susceptibility to water loss at high VPD. The conclusion that stomatal sensitivity was not altered by elevated CO2 is also consistent with an unchanged slope in the Ball et al. (1987) model (Medlyn et al. 2001)

Stomatal and photosynthetic interactions

In this experiment, environmental factors that limited stomatal conductance, e.g. low rainfall and high VPD, tended to reduce the differences in gs between the two CO2 treatments (gs-360 − gs-560), and consequently also the differences in A. Under stressful conditions, however, gas exchange rates were low in both treatments (as a result of stomatal and non-stomatal limitations in the case of A). Under these conditions, the relative effects of CO2, as the ratio of elevated/ambient, on any given date, could be highly variable, yet those relative effects on A and gs were strongly coupled. Idso (1991) described a general inverse relationship between the relative CO2-induced photosynthetic enhancement and the relative reduction in gs. That relationship was proposed to characterize species differences in CO2 responses, but, because of the tight coupling between A and gs, the relationship is equally valid to describe temporal variation in CO2 responses within a species. Whenever the relative stomatal reduction in this study was small, relative photosynthetic enhancement by CO2 was larger. The inter-relationships between A and gs in the two treatments predict a greater relative enhancement of A by CO2 at very low gs (at or below 0·15 mol m−2 s−1), but again, when gas exchange was low, relative effects on both A and gs were most variable, partly because of potential sampling error (low signal-to-noise ratio), and partly because photosynthesis at extremely low SWP is subject to additional non-stomatal limitations, as evidenced by the altered relationships between Ci/Ca and RSL at the depth of the drought in 1998.


A mean 46% photosynthetic enhancement was sustained in this experiment, without attenuation, through three seasons of CO2 enrichment, as was a 24% reduction in gs. Thus our results support Hypothesis 1, that photosynthetic and stomatal sensitivity to CO2 would persist, subject to transitory impacts of environmental variation. Hypothesis 2, that CO2 would have a greater relative impact on photosynthesis in plants exposed to stress, is theoretically supported, at least for stresses that reduce gs below a threshold, based on the tight logarithmic relationships between A and gs. In practice, small absolute treatment differences, proportionally higher variability, and additional biochemical limitations may reduce the importance of relative enhancement under dry conditions. High temperatures likewise should increase photosynthetic stimulation by CO2, but the stronger influences of VPD and SWP appeared to mute any effects of seasonal temperature variation in sweetgum. As for hypothesis 3, relative stomatal closure did proportionately limit relative photosynthetic enhancement, and that portion of the hypothesis was supported. However, it was the CO2-associated difference in gs that decreased under moisture stress (as when stomata were essentially closed in both treatments), at which time the relative effects were highly variable.

The foliar gas exchange responses of this closed-canopy stand of established trees were generally consistent with trends observed in earlier chamber studies with young seedlings, but contradicted some assumptions regarding environmental interactions. These findings are an important step toward the ultimate use of leaf-level responses to predict forest growth and function in a changing environment. They indicate that CO2 responsiveness is not limited to early seedling growth nor to upper canopy foliage, and they may help explain previous conflicting observations about altered responsiveness late in the growing season, or during stress conditions. Relative CO2 responses have proved valuable in summarizing the long-term responses of a species and in making comparisons across species where long-term records are available. Absolute responses, however, may be more pertinent than relative responses when assessing environmental interactions with CO2, and for scaling up to predict stand-level carbon gain or water use. Only by observing the long-term record of CO2 responses under a sufficient range of environmental conditions, as in this study, is it possible to separate and identify the sources of variability in the CO2 responses of A and gs.


We greatly appreciate the technical assistance provided by the following summer research interns, whose participation was essential to the project: Sarah Schweitzer, Carol Benner, Zeke Duke, Don Yee, Rich Sobiecki, Carrie Pendley, Sally Hileman, and Emily Greer, for gas exchange assistance, and Karen Voiles, Mark Scannell, Phillip Allen, and Kevin Smith for assistance with soil moisture measurements. Student participation was facilitated through the Oak Ridge Institute for Science Education (ORISE), with partial funding provided by the US Department of Energy, Office of Science, through the Energy Research Undergraduate Laboratory Fellowships and the Institute for Biotechnology, Environmental Science, and Computing for Community Colleges. We also thank Jeff Riggs, Roy Freeman and Danny Sluss, who were essential to maintaining the instrumentation for this project and for the entire FACE site, and Don Todd, for help with soil moisture measurements and for keeping the lifts and other site facilities operational. Thanks also to Jerry Tuskan, and FACE colleagues Gerry O’Neill, Tim Tschaplinski, Nelson Edwards, and Tony King for valuable discussions and comments on the manuscript.

This research was supported by the Global Change Research Program and the Program for Ecosystem Research, Office of Biological and Environmental Research, US Department of Energy, and was performed at Oak Ridge National Laboratory (ORNL). ORNL is managed by UT-Battelle, LLC, for the US Department of Energy under contract DE-AC05-00OR22725. This work contributes to the Global Change and Terrestrial Ecosystems Core Project of the International Geosphere-Biosphere Programme.

Received 1 June 2001;received in revised form 16 October 2001;accepted for publication 16 October 2001