CO2 enrichment in a maturing pine forest: are CO2 exchange and water status in the canopy affected?


D. S. Ellsworth Fax: (919) 684–8741, E-mail:


Anet, leaf net CO2 assimilation
ca, CO2 concentration of air surrounding a leaf
ci, leaf intercellular CO2 concentration
Δ, 13C isotope discrimination
δ13C, relative stable carbon isotope content
ɛ, ratio of Anet at ca = 560μmol mol–1 to Anet at ca = 360 μmol mol–1
FACE, free-air CO2 enrichment
gw, stomatal conductance to water vapour
Πi, initial leaf osmotic potential
Rt, relative water content at incipient turgor loss
Ψl, xylem water potential of leaves
Ψm, soil matric potential

Elevated CO2 is expected to reduce forest water use as a result of CO2-induced stomatal closure, which has implications for ecosystem-scale phenomena controlled by water availability. Leaf-level CO2 and H2O exchange responses and plant and soil water relations were examined in a maturing loblolly pine (Pinus taeda L.) stand in a free-air CO2 enrichment (FACE) experiment in North Carolina, USA to test if these parameters were affected by elevated CO2. Current-year foliage in the canopy was continuously exposed to elevated CO2 (ambient CO2+200μmol mol–1) in free-air during needle growth and development for up to 400 d. Photosynthesis in upper canopy foliage was stimulated by 50–60% by elevated CO2 compared with ambient controls. This enhancement was similar in current-year, ambient-grown foliage temporarily measured at elevated CO2 compared with long-term elevated CO2 grown foliage. Significant photosynthetic enhancement by CO2 was maintained over a range of conditions except during peak drought.

There was no evidence of water savings in elevated CO2 plots in FACE compared to ambient plots under drought and non-drought conditions. This was supported by evidence from three independent measures. First, stomatal conductance was not significantly different in elevated CO2 versus ambient trees of P. taeda. Calculations of time-integrated ci/ca ratios from analysis of foliar δ13C showed that these ratios were maintained in foliage under elevated CO2. Second, soil moisture was not significantly different between ambient and elevated CO2 plots during drought. Third, pre-dawn and mid-day leaf water potentials were also unaffected by the seasonal CO2 exposure, as were tissue osmotic potentials and turgor loss points. Together the results strongly support the hypothesis that maturing P. taeda trees have low stomatal responsiveness to elevated CO2. Elevated CO2 effects on water relations in loblolly pine-dominated forest ecosystems may be absent or small apart from those mediated by leaf area. Large photosynthetic enhancements in the upper canopy of P. taeda by elevated CO2 indicate that this maturing forest may have a large carbon sequestration capacity with limiting water supply.


Native plants grown at elevated atmospheric CO2 nearly always exhibit a sustained stimulation in photosynthesis due to effects of CO2 on leaf biochemical processes (Sage 1994; Drake, Gonzàlez-Meler & Long 1997). Elevated CO2 responses have been most often quantified under conditions of optimal water and nutrient availability even though such conditions are uncharacteristic of those of most native plants and may lead to unrealistic projections of CO2 effects on natural vegetation (Amthor 1995; Curtis 1996). In natural environments, factors such as temperature, water and nutrient supply, and competitive interactions among trees interact simultaneously to impose stresses on plants. It is important to evaluate whether environmental limitations will diminish the magnitude of physiological responses to elevated CO2 in forest trees, which usually occur on sites with significant resource limitations to productivity. In order to make inferences about CO2 effects on trees in natural environments, the free-air CO2 enrichment (FACE) approach has been developed to permit experimentation with elevated CO2in situ without impacting the forest microclimate (Hendrey et al. 1998).

In forest ecosystems, water availability is considered to be a major limitation to net primary production world-wide (Woodward 1987; Kozlowski & Pallardy 1996) and according to some global change scenarios drought periods may increase in warm temperate regions (Rind et al. 1990; Gregory, Mitchell & Brady 1997). Enhancements in the ratio of leaf net CO2 assimilation (Anet) to evapotranspiration (E) in enriched CO2, often termed instantaneous water-use efficiency (Eamus 1991), could be advantageous to tree growth and survival when water supply is limiting (Tschaplinski et al. 1995; Knapp et al. 1996). These enhancements are mediated through both a direct stimulation in rates of net photosynthesis by elevated CO2, and through partial stomatal closure. The stimulation response of Anet/E to CO2 enrichment, along with responses such as changes in leaf area, root water access, hydraulic conductivity and canopy-atmosphere coupling, will determine species performance with rising atmospheric CO2 concentration, particularly in water-limited situations. Thus, drought–CO2 interactions may affect tree carbon balance and survival during dry periods in an elevated CO2 atmosphere (Miao, Wayne & Bazzaz 1992). While this would suggest that photosynthetic enhancements by elevated CO2 would be greater under drought than under well-watered conditions as is often observed (Morison 1993), not all species will respond similarly in this respect (Dixon, LeThiec & Garrec 1995; Tschaplinski et al. 1995; Knapp et al. 1996). A greater understanding of how water availability may differentially affect plant responses to elevated CO2 underlies the ability to understand CO2 effects on forest productivity under different climate conditions and along soil moisture gradients in the landscape.

It is generally expected that elevated CO2 will promote reductions in stomatal conductance that can ameliorate the negative effects of drought in many species through decreased water use and/or enhanced instantaneous and long-term water-use efficiency (Morison 1993). Moreover, reductions in water use as a result of partial stomatal closure could indirectly affect other important ecosystem processes and delay the onset of water stress during drying cycles (Field, Jackson & Mooney 1995; Hungate et al. 1997). However, growing experimental evidence suggests that many forest tree species show small or non-significant changes in stomatal conductance under long-term elevated CO2 (Eamus & Jarvis 1989; Bunce 1992; Curtis 1996; Saxe, Ellsworth & Heath 1998), particularly conifers (Ellsworth et al. 1995; Teskey 1995; Picon, Guehl & Ferhi 1996). If long-term experiments confirm the hypothesis that elevated CO2 has small or non-significant effects on stomatal conductance of coniferous forest trees, then there may be substantial errors in global modelling efforts that assume a larger response exists in conifer-dominated ecosystems and regions (Henderson-Sellers, McGuffie & Gross 1995; Sellers et al. 1996; Pan et al. 1998). On the other hand, significant long-term CO2 responses of conifer stomata have been observed in some studies (Dixon et al. 1995; Tissue, Thomas & Strain 1997). Do conifers show a partial stomatal closure in response to year-long elevated CO2 exposure, and if so, will this response affect tree–water relations? To date virtually no studies have comprehensively examined multiple aspects of stand–water relations in forest plots under elevated CO2 beyond leaf-level gas exchange, and this hypothesis has remained largely untested.

I hypothesized that year-long CO2 enrichment would have minimal effects on conductance to water vapour and water relations in the upper canopy of Pinus taeda L. (loblolly pine), a dominant species in managed and unmanaged forests covering nearly 20 million ha of land in the south-eastern United States. As in many warm temperate regions, potential evapotranspiration in the growing range of P. taeda may exceed precipitation for extended periods. A strong summer drought in central North Carolina USA, site of the first forest FACE experiment, provided an opportunity to test (1) whether forest water relations and physiological function are improved in the canopy under elevated CO2, and (2) whether photosynthetic enhancement by CO2 is larger under water-limited conditions compared with well-watered conditions. The first year of elevated CO2 exposure in this stand provided an opportunity to examine stomatal responses to elevated CO2 prior to the onset of potential treatment differences in leaf area in the 15-year-old trees in the same stand, and hence separate these two potentially important mechanisms of altering water-use in elevated CO2. The magnitude of leaf-level CO2 and H2O exchange responses to elevated CO2 in conifers is important to increase understanding of possible feed-backs between climate and forest ecosystems affecting carbon sequestration at the forest stand scale.


Study site and FACE treatment

The study was conducted in the Blackwood Division of Duke Forest in Orange County, NC, USA (35°58' N, 79°05' W). The site was described in Ellsworth et al. (1995) and consists of a 32-ha parcel of even-aged loblolly pine forest on a clay loam soil. Site elevation is 174 m with a nominal atmospheric pressure of 99·2 kPa for the location, so all molar ratios for CO2 in air (μmol CO2 mol–1) expressed here can be considered approximately equivalent to the partial pressure of CO2 in air (Pa) × 10. The P. taeda overstory trees were approximately 12–13 m tall in the summer of 1997 at age 15 years. Within this stand, six circular plots were established, three replicates representing the CO2 treatment and three representing experimental controls. The three treatment plots each consisted of a 30-m-diameter free-air CO2 enrichment ring that was constructed and began treatment exposure in August 1996 (designated ‘Elevated CO2’ or FACE plots hereafter). In addition to the three CO2 treatment rings, rings were constructed at the same time in the three control plots and outfitted with blowers, vent pipes and towers configured and operated identically to the treatment rings but lacking CO2 injection (‘Ambient’ control plots). Each ring had a central tower and a vertical personnel lift for access to the canopy. The FACE rings employed the design described in Hendrey et al. (1998), but were operated continuously (24 h a day) throughout the exposure period except when ambient temperature was below 5 °C for more than 1 h. Furthermore, instead of a constant set-point as used in the previous forest FACE experiment (Ellsworth et al. 1995), the treatment rings tracked ambient CO2 concentration 24 h a day with a target of ambient + 200 μmol CO2 mol–1.

The average 24 h CO2 concentration at the centre of each FACE ring near the top of the canopy was 575, 574 and 573 μmol mol–1 for the respective treatment rings during May–October 1997. This was the period of the growing season during which the current-year foliage developed and matured. Average daytime CO2 concentration at 11–12 m height in each ring centre was 568, 564 and 567 μmol mol–1 for the three treatment rings, respectively, between sunrise and sunset. Standard deviations of the 1 min exposure CO2 concentration measured continuously over the season were ± 50–54 μmol mol–1, and the FACE system reliability was 99·8% of planned hours when treatment was applied. The corresponding ambient CO2 concentration during the growing season was 380 μmol mol–1 for 24 h, and 368 μmol mol–1 for daytime. Each ring was located at least 90 m from any other ring and contamination of the ambient signal for controlling the FACE system was avoided by using the minimum of four monitored ambient signals in different cardinal directions.

The first full year of FACE operation included a dry summer. Over the main portion of the growing season (May–September), 38 cm of precipitation was recorded at the National Weather Service station 7 km from the study site, 30% lower than the 30 year normal for this period (National Weather Service, US Department of Commerce, unpublished results). Less than 3 cm of precipitation (80% below normal) was recorded over the 6 week period prior to 10 September when a drought-breaking rain occurred. Over the growing season, volumetric soil moisture was calculated on the basis of measurements of the soil dieletric constant in the upper 30 cm of the soil profile using modified time-domain reflectometry techniques (Topp, Davis & Annan 1980) with waveguides (CS615: Campbell Scientific, Ogden UT, USA). The upper 30 cm of soil at the study site comprised more than 90% of the stored soil water accessed by overstory P. taeda trees (Oren et al. 1998). Within each ring the signals were recorded every half hour as an average for four locations within each plot. Continuous seasonal data were only available for two treatment and control plots (n = 2). Soil matric potential at pre-dawn was calculated according to Campbell (1985) using the moisture content data measured at pre-dawn and soil textural characteristics, and agreed well with the in situ soil moisture release curve parameters presented in Oren et al. (1998).

Leaf CO2 assimilation and conductance measurements

At the end of the 1996 growing season, south to west-facing upper crown branches (approximately 12 m height above the ground) in all plots which had been produced during the growing season were selected for measurements. The selected trees were within 10 m of the centre of each plot. In the following growing season, net CO2 assimilation (Anet) and conductance to water vapour (gw) were measured at the growth CO2 concentration (ca) for the first cohort of current-year foliage on these branches. Daily maximum Anet at light-saturation was followed at 2–3 week intervals on clear, sunny days through the growing season beginning when current-year foliage was 80% elongated in early July. Elevated CO2 treatment foliage had developed continuously under elevated CO2 since the time when needle primordia were determined at bud set in the year previous to the reported measurements.

The measurements of Anet were made on single fascicles of foliage at ambient temperature and vapour pressure in the upper canopy (12–13 m height). Mid-day Anet measurements were designed to reflect the daily maximum Anet under ambient conditions, light saturation in natural sunlight (quantum flux density > 1500 μmol m–2 s–1), and natural leaf angles. Data from 10 full diurnal courses of leaf gas exchange were used to ensure that Anet was maximal around noon local time. Mid-day Anet measurements in all rings were completed within an hour of noon, except during peak drought when the measurements were made earlier in the morning to avoid mid-day stomatal closure. Measurements were made with a portable infra-red gas analyzer system for CO2 and water vapour (CIRAS-1; PP-Systems, Hitchin, UK) operated in the open-flow mode with a 5·5-cm-long leaf chamber and an integrated gas CO2 supply system. The chamber was modified with an attached Peltier cooling device to maintain chamber temperature near ambient air temperatures. Gas exchange rates are reported on a unit surface area basis calculated using a geometric approximation of the needle surface.

On a subset of sampling dates, gas exchange of ambient and elevated CO2 trees was measured at a common ca by temporarily changing the CO2 supply delivered to the leaf chamber to produce the ca level appropriate for the opposite treatment, immediately following the measurement of light-saturated Anet at growth ca. Gas exchange measurements in full sunlight were continued once the ca level was stable and the gas analyzers had been internally matched, typically about 3 min after the step change in ca. For measurements carried out in this way the coefficient of variation over time for Anet was generally less than 2%. Chamber temperature was controlled to maintain constant ambient levels during the step change in ca. The Anet measurements at the growth ca for ambient trees will be referred to as A360 and those at the appropriate ca similar to the elevated CO2 treatment will be denoted A560. Only Anet data from the reciprocal ca measurements (A360 and A560) were used in the data analysis since the step change in ca was not long enough to permit possible adjustments in gw under elevated CO2. Photosynthetic enhancement ratio (e) is calculated as the instantaneous ratio of A560/A360 for any given plot (ambient or elevated CO2). The enhancement ratio for FACE Anet at growth ca to ambient Anet at growth ca is denoted ɛ'.

Leaf stable carbon isotopes

At the end of September, a sample of current-year needles was collected from a branch marked for gas exchange measurements from one tree at the centre of each ambient and elevated CO2 ring (n = 3 per treatment). The foliage was collected at mid-day after peak daily light-saturated Anet was measured. Foliage was situated less than 3 m from a continuously monitored port used for measuring CO2 concentration so the source air ca and air δ13C could be calculated (see Appendix). The foliage was stored on ice until oven-dried, then finely ground and homogenized and stored until analysed for stable carbon isotope ratio (δ13C as defined by Farquhar, Ehleringer & Hubick 1989). Stable carbon isotope ratio was expressed as the ratio of 13C/12C relative to the Pee-dee Belemnite (PDB) standard, in per mil units. The δ13C determinations were done on 2 mg subsamples from each tree within a plot and analysed at the Stable Isotope Ratio Facility for Environmental Research, University of Utah. From these data, the ratio ci/ca was calculated as described in the Appendix.

Xylem pressure potential measurements

Pre-dawn and mid-day measurements of xylem water potential were made during the season on single fascicles using the pressure chamber technique (Scholander et al. 1965). Previous year foliage was measured until current-year foliage was fully elongated in August. All foliage sampled was obtained from upper canopy positions at the third to fourth whorl from the top of the tree, where gas exchange measurements were made. Upon collection, fascicles were placed in a plastic bag with moist filter paper and taken to the field laboratory at the site for Ψl measurements. Fascicles were placed in the pressure chamber with the cut end protruding, and the cut end was swabbed with 70% v/v alcohol and dried prior to measurements to prevent resin from obscuring the xylem under pressure (Schulte & Henry 1992). Samples were discarded in cases where the balance pressure could not be determined because of exuded resin.

Pressure–volume analysis was applied to data from dehydration isotherms for determining tissue water relations parameters (Koide et al. 1989) and test for possible effects of growth in elevated CO2 on these parameters. Individual current-year fascicles of P. taeda were collected from the upper canopy at pre-dawn in mid-October toward the end of the growing season. Foliage was not rehydrated when collected due to concerns over artifacts (Schulte & Henry 1992), but samples were collected following a rainfall event so the initial portion of the pressure–volume curve could be resolved. Fascicles were allowed to transpire outside the pressure chamber and periodical measurements of fascicle mass and Ψl were recorded as dehydration progressed. Pressure–volume curve parameters were derived as described in Koide et al. (1989). Initial tissue osmotic potential (Πi) was predicted from the linear regression and extrapolation of 1/Ψl below turgor loss versus foliage relative-water content with regression fits of r2 = 0·93 or better. Three pressure–volume curves were constructed per treatment with each replicate sample originating from a different treatment ring.

Data analysis

A repeated measures analysis of variance (ANOVA) model was employed to test for CO2 effects on individual leaf gas exchange and water potential parameters across multiple sampling dates. For the repeated measures ANOVA, a probability level of 0·10 was considered significant due to the small number of replicates, whereas for other tests P < 0·05 was considered significant. Individual plots (e.g. rings) were considered as replicates for the purposes of the statistical analyses. Tukey's studentized range test was used to test for differences between treatment means for gas exchange parameters. For gw, differences of at least 23% between means could be detected for the treatment versus control with a plot-to-plot coefficient of variation of 19% and the desired probability level. Differences between means for tissue water relations parameters were tested using Student's t-test. Soil moisture was analysed by fitting an exponential decay function to the time series of the soil drying cycle for each ring and testing for differences in the fitted parameters using Student's t-test. There was a complete seasonal record for two ambient and two elevated CO2 rings. The model fits had r2 = 0·95 or better in all cases. All data were analysed in the SAS statistical package (Release 6·12, SAS Institute, Cary, N.C.) and were normally distributed (Shapiro–Wilk W test).


There was a progressive development of drought during the growing season at the site culminating in volumetric soil moisture values reaching 0·15 m3 m–3 in the upper 30 cm of the soil in early September (Fig. 1). For soils with a high clay content such as that of the Duke Forest site, most soil water at low moisture contents is held at high matric potential at low moisture contents and hence is unavailable for plant use (Fig. 1; Campbell 1985). Pre-dawn Ψl declined throughout the dry period in August in parallel with the declines in soil moisture, and recovered following the rain events in mid-September. There was no significant effect of CO2 treatment on soil moisture (P > 0·1, t-test) or pre-dawn Ψl (P > 0·1, repeated-measures ANOVA) which is often considered a plant-integrated measure of soil moisture in the rooting zone. The CO2 treatment also did not affect mid-day Ψl (P > 0·1). Despite large variations in soil moisture content and pre-dawn Ψl, there was no apparent seasonal variation in mid-day Ψl, and values remained between –2·2 and –2·5 MPa on warm sunny days (Fig. 1).

Figure 1.

. Seasonal course of soil moisture and mean xylem pressure potentia (Ψl, measured at pre-dawn or midday) in the study plots in Duke Forest, North Carolina, USA during the 1997 growing season. Data shown are treatment means for ambient (•) and elevated CO2 (•) plots (‘rings’) in the stand n = 2–3 for Ψl and n = 2 for soil moisture content, and bars indicate ± 1 standard error. Soil matric potential (Ψm) data are shown for both treatments combined. For soil moisture content data, the average standard error bar across all days is shown.

The water potential at turgor loss (Ψt) was not significantly different between FACE and ambient foliage (P > 0·1; Table 1). There was also no treatment difference in the water content at turgor loss (Rt) or tissue osmotic potential at full hydration (Πi). The Ψt-values in Table 1 were similar to mid-day Ψl measured throughout the growing season (Fig. 1).

Table 1.  . Tissue water relations parameters derived from pressure-volume analysis of current-year loblolly pine foliage developed under elevated CO2 in FACE or under ambient CO2. Rt is the relative water content at the turgor loss point, Πi is the initial tissue osmotic potential from the portion of the curve near full saturation (MPa), and Ψt is the water potential at the point of incipient turgor loss (MPa). Data are for n = 3 plots per treatment Thumbnail image of

Seasonal patterns in maximum Anet under ambient conditions and gw at maximum Anet in current-year foliage generally followed the pattern of soil moisture during July to October. The seasonal maximum Anet and gw corresponded to a brief period of increased soil moisture following heavy rain in late July (Figs 1 and 2). With rapidly decreasing soil moisture following this rainfall, a large progressive decrease in maximum Anet and gw at maximum Anet is apparent (Fig. 2). At peak drought in late August–early September, Anet was reduced to less than 60% of the seasonal maximum in both ambient and elevated CO2 foliage, although there was recovery in Anet less than 1 week later with the increase in soil moisture after a drought-breaking rain event (Fig. 2). The decline and recovery of gw during the drought cycle was similar in foliage in ambient versus elevated CO2. The seasonal decrease in gw with drought to 40% of the summer maximum was larger than that for Anet in both ambient and elevated CO2 foliage. The declining gw in July and August indicates that progressive stomatal limitations to Anet developed during the drying cycle. At peak drought the foliage ci/ca ratio (at growth ca) was 0·42 ± 0·08 in elevated CO2 trees and 0·49 ± 0·07 in ambient trees compared with 0·62 ± 0·02 in trees in both treatments during well-watered periods (data not shown). Stomatal closure was also evident when comparing the diurnal course of gas exchange between a sunny day during peak drought and later in the same month under well-watered conditions (Fig. 3). During drought, daily maximum gw was lower than under well-watered conditions and there was also a larger apparent decline in gw through the day. The diurnal decrease in gw during drought restricted the majority of CO2 assimilation to times before solar noon.

Figure 2.

. Seasonal patterns in light-saturated daily maximum net CO2 assimilation (Anet) and stomatal conductance at maximum Anet (gw) in current-year foliage in the canopy of P. taeda under ambient blower control plots and elevated CO2 plots. Anet was measured under ambient conditions and growth ca on sunny days. Foliage was 95% fully elongated by mid-July. Data shown are treatment means for trees in ambient control (•) and elevated CO2 treatment (•) plots with ± 1 standard error

Figure 3.

. Diurnal course of leaf-level net CO2 assimilation (Anet) and stomatal conductance (gw) in current-year foliage of P. taeda at 12 m height in the canopy under ambient (•) and elevated (•) CO2 treatments. Environmental parameters (Tleaf, leaf temperature; PAR, quantum flux density in the photosynthetically active radiation wavebands) are shown in the upper two panels and physiological measures are shown in the bottom panels. Data are for clear, sunny days at peak drought (left panels; DOY 249) and under well-watered conditions (right panels; DOY 273). For pooled ambient and elevated CO2 trees Ψl was –1·4 ± 0·1 MPa at pre-dawn on DOY 249, – 2·58 ± 0·1 MPa at mid-day on DOY 249, and –0·8 ± 0·1 MPa on DOY 273. Data are for foliage from n = 3–5 trees per treatment (± SE) at each time point within representative plots of each treatment.

Effects of CO2 treatment on leaf gas exchange

Elevated CO2 treatment had large apparent effects on Anet with instantaneous ɛ at mid-day varying between 1·45 and 1·65 across the period shown (Fig. 2). However, in Fig. 2 there was not complete replication of experimental units (plots) across the entire seasonal time series, and hence a more limited data-set with full replication was used for statistical tests. For the smaller data-set representing sampling dates with treatment replication (Fig. 4), the daily maximum Anet at growth ca was significantly higher (P < 0·05; studentized range test) in FACE treatment plots compared with the ambient control plots for four out of five sampling dates. The sampling date where this contrast was not significant was toward the peak of the drought in late August. The overall difference between Anet at growth ca in ambient versus elevated CO2 plots was marginally significant (P < 0·08; repeated-measures ANOVA) across all dates and significant at P < 0·05 across all dates eliminating the data collected around peak drought in late August. When Anet at growth ca was significantly different between treatments, indicated by * in Fig. 4, the ratio of A560 of elevated CO2 foliage to A360 in ambient foliage (ɛ') varied between 1·5 and 1·8.

Figure 4.

. Treatment effects on gas exchange parameters [Anet at growth ca and light saturation, upper panel; leaf conductance (gw), middle panel; instantaneous photosynthesis to transpiration (Anet/E) ratio, bottom panel] of upper canopy pine foliage over five sampling dates during the growing season. Data shown are treatment means ± SE for ambient control (dark bars) and elevated CO2 (FACE) treatment plots (hatched bars) in the stand. Treatment comparisons within a sampling date were not significant (NS) or significant at the P = 0·05 level (*; Tukey range test) for each date as indicated.

Integrated over the day, cumulative photosynthetic performance was enhanced by 84% by elevated CO2 at peak drought and by 69% on a well-watered day (Fig. 3). The daily sum of photosynthetic enhancement was somewhat larger than the instantaneous ɛ at mid-day on these days. Using the integrated daily Anet values for drought versus well-watered conditions, an estimate of the potential loss in photosynthesis due to drought can be calculated and compared between ambient and elevated CO2 conditions. The two days in Fig. 3 were comparable in maximum air temperature and vapour pressure deficit, although duration of sunny conditions was approximately 45 min less in late September due to decreasing day-length which would reduce daily Anet by less than 5%. Considering this, the estimated reduction in integrated daily Anet attributable to drought was 45% for upper canopy foliage under ambient CO2 and 40% for the elevated CO2 treatment.

In contrast to the large effect of elevated CO2 on Anet, there was no evidence of differences in gw in elevated CO2 compared with ambient CO2 either seasonally (Fig. 2) or with time of day under conditions of different water availability (Fig. 3). There was no significant difference in gw in FACE treatment versus ambient controls across all sampling dates (P > 0·10; repeated-measures ANOVA), or on any individual sampling date (Fig. 4). Thus enhancement in Anet alone was responsible for highly significant differences in photosynthesis to transpiration ratio (Anet/E) across the season (P < 0·003; repeated-measures ANOVA). Leaf temperatures and leaf–air vapour pressure difference for ambient and elevated CO2 measurements were not significantly different (P > 0·10), so no treatment bias in Anet/E attributable to measurement conditions other than ambient ca existed.

Differences in the relationship between Anet and gw between the ambient and elevated CO2 treatments (Fig. 5) are consistent with enhanced Anet/E under elevated CO2. The slopes of the lines shown are significantly different (P < 0·05). While the regression relationship may imply that ɛ will vary with the magnitude of gw, ɛ calculated in this way is biased by the non-zero intercepts of the relationships. The ratio of the slope for the elevated CO2 regression line to that of the ambient was 1·5, suggesting an overall seasonal photosynthetic enhancement of this magnitude for a similar range in gw between ambient and elevated CO2 foliage. Despite this enhancement, the long-term ratio of ci/ca for the ambient foliage was apparently maintained in the elevated CO2 treatment since neither the ci/ca ratio nor the 13C isotope discrimination (Δ) was different between treatments (Table 2). The short-term stimulation of current-year foliage developed under ambient CO2 and measured at ca = 560 μmol mol–1 (e.g. ɛ for ambient foliage) was 1·57 when averaged over the season (Fig. 6a). There was no significant difference in the slope of this relationship between foliage from ambient and FACE plots (P > 0·1), so the photosynthetic enhancement ɛ of foliage that developed under elevated CO2 in FACE was similar to that of ambient foliage (Fig. 6a). The relationship between A560 and A360 was also similar when values from paired Elevated CO2 and ambient plots were used rather than instantaneous ɛ values, although the variance was larger (data not shown). The enhancement ratio for single leaves is approximately constant over the seasonal range in Anet values from July to September with a limited range in temperature (Fig. 6a).

Figure 5.

. Relationship between leaf conductance (gw) and daily maximum Anet at growth ca for upper canopy foliage in ambient (•) and elevated (•) CO2 plots. The lines shown are Y = 1·94 + 0·15 ×X + 5·8 × 10–4×X2 (r2 = 0·84) for elevated CO2 foliage and Y = 0·12 + 0·13 ×X + 5·5 × 10–4×X2 (r2 = 0·86) for ambient foliage.

Table 2.  . Carbon isotope discrimination and ci/ca ratio of current-year, upper-canopy loblolly pine (Pinus taeda) foliage developed under ambient and elevated CO2. Time-integrated ci/ca was calculated from foliar and air stable carbon isotope data (Appendix). Data are for n = 3 plots per treatment Thumbnail image of
Figure 6.

. (a) Relationship between light-saturated A560 (leaf net CO2 assimilation rate at ca = 560 μmol mol–1) and light-saturated A360 (CO2 assimilation rate at current ambient ca) for current-year canopy foliage in ambient(•) and elevated (•) CO2 plots during the growing season. Data shown are means ±SE for each treatment on a given sampling date. The line shown is Y = 1·57 ×X. (b) Composite relationship between light-saturated leaf net CO2 assimilation rate (Anet) and calculated intercellular CO2 concentration (ci) for current-year canopy foliage of P. taeda in ambient (closed symbols) and FACE plots (open symbols) under drought and well-watered conditions. Measurements were made at ca = 360 μmol mol–1 (triangles) and ca = 560 μmol mol–1 (circles). Data are daily maximum Anet from July to September, and leaf temperature varied between 29 and 32 °C for the measurements. The mean CO2 supply function corresponding to foliage from each treatment for well-watered conditions (13 September) is indicated by the parallel dashed lines showing the CO2 gradient from ca to ci.

To evaluate the possibility of an intrinsic physiological change under elevated CO2 that would reduce carboxylation efficiency or photosynthetic enhancement (Sage 1994), I compared Anet measurements in ambient and FACE plots made at a common ca. For the sampling dates shown in Fig. 4, there were no significant differences between treatments for Anet measured at a common ca (data not shown). The composite Anetci curve across sampling dates also indicates that all the data follow the same intrinsic relationship over the range of ci values (Fig. 6b). For the CO2 supply function lines shown in Fig. 6(b) illustrating the same gw among treatments, the ci/ca ratios were also the same for ambient and elevated CO2 foliage. The ci/ca ratio depicted (0·63) was similar to values from the δ13C data in Table 2.


Elevated CO2 exposure in free-air had substantial impacts on canopy leaf carbon assimilation in the upper canopy but little apparent effect on water relations or water supply over a season with a strong drought event. Theoretical predictions of enhanced water supply under elevated CO2 are supported by observations in crop ecosystems (Pinter et al. 1996) and in grasslands (Jackson et al. 1994; Knapp et al. 1996; Owensby et al. 1996) but have rarely been tested in forest ecosystems. A general prediction supported by these studies is that partial stomatal closure elicited by increased intercellular CO2 under elevated CO2 (Morison 1987; Mott 1988) improves plantwater relations, which may affect ecosystem function during water-limited conditions and soil drying cycles (Morison 1993; Field et al. 1995). However, for the dominant forest species P. taeda examined in this study, a lack of significant differences in gw between ambient and elevated CO2-grown foliage at daily (Fig. 3) and seasonal time scales (Figs 2, 4) after more than 12 months of CO2 exposure was probably responsible for the similarity in soil moisture and leaf–water relations parameters between CO2 treatments (Fig. 1, Table 1). Thus a major primary mechanism for elevated CO2 effects on ecosystem water exchange with the atmosphere via stomatal appears to be relatively unimportant in the short term or on the seasonal time scale for this conifer forest. If the responses of conifer forest in the field differ significantly from those established in field-based grassland studies, then such differences must be considered in models projecting future forest–climate CO2 and H2O feed-back and exchange processes (Sellers et al. 1996). There is a paucity of field data from elevated CO2-exposed forest stands for making a general comparison with grasslands. However, many current modelling efforts consider relative stomatal responses to elevated CO2 to be similar among diverse herbaceous and tree species (Henderson-Sellers et al. 1995; Haxeltine, Prentice & Cresswell 1996; Pan et al. 1998) and may misrepresent the magnitude of CO2 responses in some ecosystems with a wide geographic range.

Evidence from species with high photosynthetic rates such as crops, ruderal herbs and grasses suggests that the guard cells of stomata generally respond to intercellular CO2 concentration (Mott 1988). The lack of such a response in other C3 plants may be considered exceptional, although until recently stomatal responses of native woody species have been largely overlooked (Mansfield, Hetherington & Atkinson 1990). In literature comparisons, Field et al. (1995) found that the stomatal responses of trees were generally smaller than those documented for crops, although Curtis (1996) concluded that there were small but significant CO2 effects on stomatal conductance based on a meta-analysis of 16 studies on unstressed trees. Even so, there are exceptions where a lack of a stomatal response to CO2 has been noted (Eamus & Jarvis 1989; Ellsworth et al. 1995; Heath & Kerstiens 1997). Robinson (1994) and Saxe et al. (1998) speculated that some plant groups might be less responsive in this regard than those with high rates of metabolism, especially woody species and coniferous trees. The hypothesis that coniferous trees show small or non-significant stomatal responses to elevated CO2 is supported by results from field studies on different species of Pinaceae studied in chambers by Beerling & Woodward (1996), Dixon et al. (1995), Hogan et al. (1996) and Picon et al. (1996) (see also references in Saxe et al. 1998). Studies on P. taeda itself have yielded inconsistent results in this respect for young plants (cf. Tolley & Strain 1985; Tissue et al. 1997), although mature individuals have consistently not shown significantly elevated CO2 effects on gw with long-term CO2 exposures of isolated branches (Teskey 1995; Murthy, Zarnoch & Dougherty 1997) or entire trees in FACE (Ellsworth et al. 1995; this study). Experiments employing containerized or root-restricted young trees with limited rooting space may artificially induce large stomatal responses to elevated CO2 in seedlings as a result of size-dependent water stress as was shown for P. taeda seedlings (Will & Teskey 1997). Thus apart from a mechanistic interpretation of CO2-induced partial stomatal closure, growth conditions and plant size can also have a strong indirect effect on the magnitude of stomatal responses to elevated CO2 (Talbott, Srivastava & Zeiger 1996; Will & Teskey 1997). Results from studies that utilize artificial plant growth conditions such as root restrictions or chamber artifacts need to be interpreted carefully against the context of actual field conditions.

In addition to effects of elevated CO2 on plant–water relations that are mediated by stomata, other response mechanisms can be considered. For instance, indirect effects of elevated CO2 on plant–water relations may occur via CO2-induced differences in plant leaf area or leaf area to water-absorbing surface (Eamus 1991; Heath & Kerstiens 1997). In mature trees exposed to elevated CO2 as in this study, such indirect effects are negligible in the first year of CO2 exposure compared with the size-induced differences between ambient and elevated CO2 trees that are common in seedling studies (Saxe et al. 1998). Elevated CO2 may also directly alter leaf dehydration tolerance in cases where CO2-induced excess carbohydrates serve as osmotica (Morse et al. 1993; Tschaplinski, Norby & Wullschleger 1993). There was no evidence of CO2 effects on tissue water relations parameters in P. taeda in this study (Table 1) nor in chamber-grown seedlings (Tschaplinski et al. 1993) which again supports the contention that water relations and dehydration tolerance in this conifer species were largely unaffected by CO2 exposure. Results from P. taeda have important implications for understanding forest CO2 responses under water-limited conditions, which are predicted to intensify in its growing region under climate radiative-forcing scenarios (Rind et al. 1990) and with increasing urban water demands in the region.

A tendency for similarity in ci/ca ratios between ambient and elevated CO2 foliage has been noted by Morison (1987), Sage (1994) and Drake et al. (1997). This may suggest that stomatal responses to elevated CO2 are largely not independent of responses in Anet (Sage 1994). Well-watered P. taeda maintained ci/ca ratios in elevated CO2 although at peak drought the lower ci/ca ratios in elevated CO2 versus ambient trees suggests that there are greater stomatal limitations during severe stress in elevated CO2. However, the seasonally integrated ci/ca ratios indicated by 13C isotope discrimination remained identical in ambient and elevated CO2 (Table 2). As long as ci is maintained within the nearly linear region of the initial portion of the Anetci relationship (Fig. 6b), a decrease in gw at elevated CO2 is not necessary to maintain a constant ci/ca with increasing ca and adequate water availability. The constancy of ci/ca ratios forms the basis for empirical models for predicting stomatal conductance (Jarvis & Davies 1998) but does not necessarily presuppose partial stomatal closure with long-term exposure to elevated CO2 in species where gw is intrinsically low.

CO2 assimilation in canopy foliage under elevated CO2 and drought

In spite of the lack of observable CO2 effects on water balance, photosynthetic CO2 assimilation was consistently enhanced in elevated CO2-grown foliage over the course of a growing season with drought (Figs 2, 3 & 4). The photosynthetic enhancement ratio ɛ in upper canopy foliage that developed under continuous exposure to ambient + 200 μmol mol–1 CO2 in FACE was large (approximately 1·5 to 1·8 on different days, mean of 1·57; Figs 4, 6a). This was similar to the ɛ of 1·65 reported in a single-season FACE exposure in Ellsworth et al. (1995) in the same forest stand, using a similar exposure CO2 concentration in FACE with daytime-only fumigation. There is considerable evidence for a lack in photosynthetic adjustments to elevated CO2 in previous experiments lasting one to several growing seasons (Ellsworth et al. 1995; Teskey 1995, Tissue et al. 1997). This is supported by the maintenance of similar A560/A360 ratios in FACE and ambient foliage (Fig. 6a) and by the similarity in the Anetci curve, composited across sampling dates (Fig. 6b). Myers, Thomas & DeLucia (1999) presented evidence that there were no biochemical adjustments in photosynthetic capacity during summer in ambient-grown foliage of the previous year's cohort in this stand when exposed to free-air CO2 enrichment. Single-season or single-year CO2 exposures are unlikely to be sufficiently long to induce physiological adjustments in the photosynthetic apparatus in trees (Ceulemans et al. 1997; Drake et al. 1997), particularly those in a mature stand that developed under current ambient CO2 (Körner 1995).

Photosynthetic enhancement of upper canopy foliage was largely maintained despite drought conditions during a portion of the growth period (Figs 2, 4). A lack of significant CO2 effects on Anet at growth ca toward peak drought was observed, suggesting greater sensitivity of Anet to drought in elevated CO2-grown trees (Fig. 4). However, both ambient and elevated CO2 foliage both showed similar and large instantaneous enhancements in photosynthesis, even when daily maximum Anet was low due to drought (Figs 3, 6a). Thus there is not sufficient evidence to conclude that photosynthetic enhancement by elevated CO2 is intrinsically smaller under drought than under well-watered conditions in mature P. taeda. The effects of CO2 treatment on the magnitude of water-stress induced repression of Anet at a given ca were not large, since the instantaneous photosynthetic enhancement was nearly constant across the range of Anet values over well-watered and drought conditions (Fig. 6a). The constant ɛ for a given temperature can be expected if intrinsic leaf biochemical processes in photosynthesis are relatively unaffected during drought (Fig. 6b), and stomatal conductance largely limits photosynthesis with decreasing soil moisture (Fig. 2). Given the lack of strong curvature in the Anetci curve relationship below calculated ci values of 400 μmol mol–1 (Fig. 6b), near-constant ɛ values are expected for the CO2 supply function corresponding to a typical gw for pine. At summertime temperatures, light-saturated Anet in P. taeda operates in the responsive initial portion of the Anetci curve, which explains why photosynthetic enhancement by elevated CO2 is large.

Much of the seasonal and diurnal decrease in Anet with drought was apparently controlled by stomata (Figs 2, 3) as would be expected in drought sensitive species like some Pinus species (Green & Mitchell 1992; Picon et al. 1996). The maintenance of consistent mid-day Ψl over the season with declining pre-dawn Ψl in both treatments during the drought period (Figs 1, 2) indicates that as drought progressed, the opportunity for photosynthesis over the course of a day would decrease in proportion to the available drop in Ψ to a threshold value corresponding to the turgor loss point (Table 1). Due to a lack of partial stomatal closure with elevated CO2, the relative reductions in Anet at growth ca by water-stress induced stomatal closure in ambient and FACE trees are similar. The overall results from this FACE experiment support earlier evidence that maturing P. taeda canopies have a large capacity for CO2 responses in photosynthesis even under drought, but low stomatal responsiveness to elevated CO2. This observation points to the need for reconsideration of elevated CO2 effects on water relations in conifer-dominated forest ecosystems in warm temperate regions, and highlights an important difference from warm temperate grassland ecosystems. While evidence from other coniferous forest ecosystems and longer-term experiments is needed, the results suggest that in some forests, feed-backs to the climate system in an elevated CO2 atmosphere may not occur as has been predicted by the current generation of coupled biosphere–atmosphere models (Sellers et al. 1996). Understanding the mechanisms of short-term and long-term regulation of stomatal conductance, stomatal density and leaf area per unit stem basal area in woody plant species in elevated CO2 underlie our ability to predict forest function and its feed-backs with climate and rising atmospheric CO2.


This research is part of the Forest–Atmosphere Carbon Transfer and Storage (FACTS-1) project at Duke Forest. The FACTS-1 project is supported by the US Department of Energy, Office of Health and Environmental Research, under DOE contract numbers DE-AC02–98CH10886 at Brookhaven National Laboratory and DE-FG05–95ER62083 at Duke University. Partial support from the US Department of Energy through the Inter-agency Terrestrial Ecology Initiative is also gratefully acknowledged. R. Oren and K. Schaefer generously provided the soil moisture data for Fig. 1. I thank Professor S. Long and E. Naumburg, D. Myers and two anonymous reviewers for critical reviews of an earlier draft of the manuscript. I am grateful to G. R. Hendrey for conception and support of the FACE system concept in the USA and to K. Lewin, J. Nagy and A. Palmiotti for dedication in constructing and operating the FACE system.


Calculation of ci/ca from leaf δ13C

Stable C isotope discrimination (Δ) can be calculated from leaf δ13C values (in ‰) as


where δa is the δ13C of the source air and δp is the δ13C of the plant tissue (Farquhar et al. 1989). The source air (δa) was assumed to be approximately –8·5‰ for the daytime CO2 concentration of 368 μmol mol–1 for ambient rings. In FACE rings, a two-ended mixing ratio model was used to calculate the isotope ratio of the source air in elevated CO2 (δe) using the 1 min mean CO2 concentration of the sampling port near the foliage and the equation


where camb is the air CO2 concentration outside the FACE plot (equivalent to ca in ambient plots) and δc is the stable isotope signature of the pure CO2 added. The calculation was performed using daytime CO2 concentration data only from the point of leaf emergence in April to the end of September when foliage was collected. All CO2 used for enrichment in the FACE rings was from a constant fossil-fuel source with a δc of –43·7 ± 0·6‰ over the study period (J. A. Andrews, unpublished results). From Δ, the ratio of ci to ca can be calculated by solving the equation


where a is the coefficient for diffusion through the stomatal pore (4·4‰) and b is the fixation of gaseous CO2 with respect to ci/ca (27‰) as described in Farquhar et al. (1989).

Rearranging (3) and (1) to solve for ci/ca yields


which can be solved for ci/ca of ambient and elevated CO2 foliage. Mean δp was 27·4 and 39·3‰ for foliage from ambient and elevated CO2 plots, respectively.