Elevated CO2 increases tree-level intrinsic water use efficiency: insights from carbon and oxygen isotope analyses in tree rings across three forest FACE sites


Author for correspondence:

Giovanna Battipaglia

Tel: +39 0823274647

Email: giovanna.battipaglia@unina2.it


  • Elevated CO2 increases intrinsic water use efficiency (WUEi) of forests, but the magnitude of this effect and its interaction with climate is still poorly understood.
  • We combined tree ring analysis with isotope measurements at three Free Air CO2 Enrichment (FACE, POP-EUROFACE, in Italy; Duke FACE in North Carolina and ORNL in Tennessee, USA) sites, to cover the entire life of the trees. We used δ13C to assess carbon isotope discrimination and changes in water-use efficiency, while direct CO2 effects on stomatal conductance were explored using δ18O as a proxy.
  • Across all the sites, elevated CO2 increased 13C-derived water-use efficiency on average by 73% for Liquidambar styraciflua, 77% for Pinus taeda and 75% for Populus sp., but through different ecophysiological mechanisms.
  • Our findings provide a robust means of predicting water-use efficiency responses from a variety of tree species exposed to variable environmental conditions over time, and species-specific relationships that can help modelling elevated CO2 and climate impacts on forest productivity, carbon and water balances.


Terrestrial plants play a significant role in the global carbon cycle and in the control of the carbon dioxide (CO2) concentration in the atmosphere. The responses of plants to an increasing concentration of atmospheric CO2 will depend on their ability to use water and nutrient resources efficiently under a changing climate. While the potential growth response of trees to increased atmospheric CO2 concentration is relatively well understood, there is still uncertainty regarding the process and environmental constraints that modulate the responses of photosynthesis and stomatal conductance to elevated CO2 (Wullschleger et al., 2002; Herrick et al., 2004; Ainsworth & Long, 2005; Keel et al., 2006), and the tree-level expression of those responses as intrinsic water use efficiency (WUEi: the amount of carbon acquired per unit of water lost) in the long term (Ainsworth & Rogers, 2007; Gagen et al., 2011).

Elevated atmospheric CO2 (ca) is expected to affect plant carbon–water relationships, as a decline in stomatal conductance (gs) is predicted when plants are exposed to elevated CO2 (Field et al., 1995). If a decline in gs occurs in conjunction with an increase in carbon assimilation (owing to CO2 supply), this can change the ca to internal leaf CO2 concentration (ci) gradient, which in turn determines the 13C isotope composition of assimilated carbon. it is known that discrimination against 13C (Δ13C) during photosynthesis in C3 plants is related to the ci /ca ratio, where the ci/ca level is determined by the relationship between photosynthesis (A) and gs (WUEi = A/gs ~ ca-ci). As the stomata tend to close under elevated CO2, the mechanism for saving water often results in an improvement of the WUEi. When the stomatal limitation of photosynthesis is strong, the plants exhibit a low ci and a reduced discrimination, along with improved WUEi (McCarroll & Loader, 2004; Saurer et al., 2004). Thus, discrimination against 13C is a proxy for the integrated response of A and gs, and can be used to infer average plant WUEi over the time when the plant organic matter was formed. In tree rings, carbon-isotope based WUE-estimates reflect a growing season (photosynthesis-weighted) average, that is difficult to assess with instantaneous gas-exchange measurements (Picon et al., 1997).

The relative oxygen isotope ratio (δ18O) in tree rings can provide a different insight into the causes of variation of WUEi. It has been suggested that δ18O in tree rings can be used as a proxy for gs (Farquhar et al., 1993; Barbour et al., 2000). Variations in gs influence the δ18O of plant material through different competitive effects on: (1) the evaporative cooling of the leaf; (2) the diffusive resistance of water vapour; and (3) the plant transpiration and the subsequent gradient in H218O enrichment in leaf water, also known as the Péclet effect (Barbour et al., 2004). When gs decreases under elevated CO2 and photosynthetic CO2 demand is unchanged, ci/ca should decrease and δ18O in leaf water, as well as in the organic matter and tree rings, would increase (Roden & Ehleringer, 2000).

The combined analyses of carbon and oxygen isotopes in tree rings can therefore suggest whether gs or A contribute most to the variation of WUEi in response to elevated CO2 (Scheidegger et al., 2000).

Several studies, using Δ13C in trees growing in natural conditions, have shown wide variability among forests species in their WUEi responses to rising ca (Penuelas et al., 2011 and references therein). In most cases WUEi was improved, but there is a lack of understanding of the extent to which the elevated CO2 increased WUEi of different species and which mechanism (i.e. stomatal closure or A increases) drove the differences (or similarities) between WUEi in the various species and sites (Gagen et al., 2011; Linares & Camarero, 2012). Free-air CO2 enrichment (FACE) field experiments have played a key role in addressing uncertainties in forest responses to elevated ca (reviewed in Ainsworth & Long, 2005), but with contrasting results regarding the principal mechanisms involved in the increase in plant WUEi (Norby & Zak, 2011).

Previous studies (Table 1) have reported reduced gs in response to elevated CO2, but the magnitude of the response varied widely across sites and species (Warren et al., 2011a,b) and appeared to be influenced by other environmental and growth-related factors (Leakey et al., 2009; Norby & Zak, 2011). Those studies estimated the responses of stomatal conductance to rising CO2 from short-term, leaf ‘exposure’ analysis, which cannot take account of acclimatization that may alter stomatal responses over longer time-scales (Gagen et al., 2011). Analyses of tree-rings are ideal proxies for assessing physiological and growth-changes over time and under different environmental conditions and allow assessment of pretreatment behaviour.

Table 1. Summary results of the effects of elevated CO2 for different sites regarding net primary productivity (NPP), stomatal conductance (gs) and photosynthetic activity (A)
SiteSpeciesNPP g s A
  1. Responses are indicated as increases (+), decreases (−) no change (nc) or no reported information (nr). L. s, Liquidambar styraciflua; P. t, Pinus taeda; P. a, Populus alba; P. n, Populus nigra; P. e, Populus  x euramericana.

  2. a

    Norby et al. (2010).

  3. b

    Gunderson et al. (2002).

  4. c

    Sholtis et al., (2004).

  5. d

    Herrick et al. (2004).

  6. e

    Springer & Thomas (2007).

  7. f

    McCarthy et al. (2010).

  8. g

    Ellsworth et al. (2012).

  9. h

    Gielen et al., (2005).

  10. i

    Tricker et al. (2005).

  11. j

    Bernacchi et al., (2003) .

ORNLL. s+24% (2001–2003) + 9 (2008)a−14–44% (1999–2000)b+44% (1999–2000)c nc (2008–2010)a
DukeL. snr−28% (1998–2001)d+41–52%e
DukeP. t+22–30% (1997–2004)fncg+35–65%g
POP-EUROFACEP. a; P. n; P. e+21–26%h−19 to 24%i+38%j

This study examines the temporal relationships between variations in tree ring δ13C-derived WUEi and δ18O with ca, climate and tree growth, across five species in three FACE sites in Italy and the USA. They represent a wide range of climate, biological conditions, stand development history and dominant tree species, with a broadleaved species, Liquidambar styraciflua, growing in two FACE sites (ORNL and Duke), a coniferous species, Pinus taeda, growing in a mixed fumigated forest (Duke) and three fast-growing poplar species (Populus sp.), planted in a Mediterranean environment. Hence, they offered us the possibility to test the following hypotheses: (1) the mechanism driving the expected improvement in WUEi, under elevated CO2 is more strongly related to species than to site conditions; (2) CO2 fertilization will interact with climate controls on WUEi, influencing tree responses to temperature and precipitation; (3) the combined tree-ring width, carbon and oxygen isotope analysis enables us to understand why increased WUEi sometimes does not lead to increased growth. To test those hypotheses, we applied a dendro-ecological approach at three FACE sites to investigate whether any observed patterns extend across species, and we analysed Liquidambar styraciflua, present in two different sites, to assess whether CO2 responses are site dependent. Further, we analysed tree rings of individual trees before and after the beginning of FACE experiments to test for legacy effects (i.e. translation of trees before treatment condition and life history) into responses to elevated CO2.

The simultaneous measurements of isotopes and growth allow us to estimate species-specific changes in WUEi and can be useful to predict the influence of future climate change on the productivity and water consumption of major trees used in forest plantations.

Materials and Methods

Site descriptions

Three FACE experiments in fast-growing temperate-zone tree plantations were sampled for our study. In particular, two experiments were in established monoculture plantations: a deciduous sweetgum (L. styraciflua L.) stand at ORNL (Oak Ridge National Laboratory) FACE, Tennessee (USA) and an evergreen loblolly pine (P. taeda L.) stand at Duke FACE, North Carolina (USA). The third site was a poplar plantation (Populus alba L., Populus nigra L., Populus x euramericana Dode Guiner), established on a former agricultural field, in Tuscania (Italy).

All three experimental sites used FACE technology and full detailed descriptions are provided for Duke FACE by Hendrey et al. (1999), for ORNL FACE by Norby et al. (2001) and for the POP-EUROFACE by Miglietta et al. (2001).

In the ORNL site the L. styraciflua plantation was established in 1988 and the CO2 fumigation started in 1998. Our data refer to two fumigated and two control plots and cover the period 1988–2004.

The Duke FACE site was established in a mixed forest containing an overstory of P. taeda planted in 1983 and several deciduous understory trees, including L. styraciflua, Acer rubrum, Ulmus alata, Cercis canadensis and Cornus florida. The fumigation started in 1996 and our data covers the period 1983–2004.

In the POP-EUROFACE the poplar plantation (Palba, P. nigra and P. x euramericana) and the FACE facility were established in 1999, coppiced in 2001 and the FACE experiment ended in 2004. For this study we analysed the tree rings from stems grown during the second coppiced rotation (2002–2004), including the growing season after the end of the fumigation (2005).

Full description of the sites and of the atmospheric CO2 enrichment supply is provided in the Supporting information, Notes S1.

Meteorological data

We used the climate data reported by Riggs et al. (2003) for the ORNL FACE (Fig. S1), which measured at the site since April 1997 (http://face.env.duke.edu/database/login.cfm) for the Duke FACE (Fig. S2), and that recorded by the site meteorological station for the POP-EUROFACE (Fig. S3). Meteorological data were also obtained from the Southeastern Regional Climate Center of the Department of Natural Resources (http://www.sercc.com/sco) and from the Climate Explorer (http://climexp.knmi.nl) to calculate monthly anomalies of precipitation and temperature for the duration of the experiment relative to long-term mean values (1900–2009) (see Notes S2, Figs S1–S3).

Tree core sampling

Eight trees of each species per each experimental plot were sampled during a field campaign in the USA and Italy in spring–summer 2005. At the ORNL FACE and at the POP-EUROFACE overstory trees were cored. At the POP-EUROFACE site, characterized by a multistem system during the second rotation cycle, dominant shoots on each stump were sampled. For DUKE FACE, we sampled the overstory P. taeda as well as understory L. styraciflua. Two cores were collected with a 5 mm diameter borer (Suunto, Finland), near the base of each tree, following standard methods (Schweingruber, 1988).

Tree ring width

Ring-width measurements were made with a resolution of 0.01 mm on each of the cores, using LINTAB measurement equipment (Frank Rinn, Heidelberg, Germany) fitted with a Leica MS5 stereoscope and analysed with TSAP software package. All cores were cross-dated according to standard procedure (Fritts, 1976; Schweingruber, 1996). We derived the series of yearly basal area increments (BAI) from raw ring width assuming concentrically distributed tree-rings; BAI were used instead of ring-width directly, because BAI is less dependent on age and thus avoids the need for detrending (Biondi, 1999), which could also remove low frequency variability.

Isotope analysis

Tree rings from all cores were split into earlywood and latewood, and only latewood was used in this study (Lipp et al., 1991; Hill et al., 1995). The latewood of the two cores per tree was pooled by year, and the samples were ground with a centrifugal mill (ZM 1000, Retsch, Germany) using a mesh size of 0.5 mm to ensure homogeneity. Cellulose was extracted with a double-step digestion (Boettger et al., 2007; Battipaglia et al., 2008). The carbon and oxygen stable isotope compositions were measured at the CIRCE laboratory (Center for Isotopic Research on the Cultural and Environmental heritage, Caserta, Italy) and at the PSI isotope-lab (Villigen, Switzerland), respectively, by continuous-flow isotope ratio mass spectrometry (Finnigan Mat, Delta S, Bremen, Germany) using 1 mg of dry matter for 13C measurements and 1.5 mg for 18O determinations.

We report isotope values in the delta notation for carbon and oxygen, where δ13C or δ18O = (Rsample/Rstandard – 1) (‰), relative to the international standard, which is VPDB (Vienna Pee Dee Belemnite) for carbon and VSMOW (Vienna Standard Mean Ocean Water) for oxygen. Rsample and Rstandard are the molar fractions of 13C/12C and 18O/16O for the sample and the standard, respectively. The standard deviation for the repeated analysis of an internal standard (commercial cellulose) was better than 0.1‰ for carbon and 0.2‰ for oxygen. The calibration vs VPDB was done by measurement of International Atomic Energy Agency (IAEA) USGS-24 (graphite) and IAEA-CH7 (polyethylene) and vs VSMOW by measurement of IAEA-CH3 (cellulose) and IAEA-CH6 (sucrose).

Data analysis

As the series of consecutive observations are collected from the same tree (annual rings from an increment core), data autocorrelation may arise (Monserud & Marshall, 2001). Thus, to determine and remove long-term trends without altering or modifying the existing data, the response ratio (RR, the ratio of means for a measured variable between the elevated CO2 treatment group and the control group) was used as an index of the estimated magnitude of the elevated CO2 effect. The RR is commonly used as a measure of experimental effects or manipulations (Hedges et al., 1999) and has often been used in ecology to test the response of biomass or nitrogen (N) concentrations to increased CO2 levels (Gunderson & Wullschleger, 1994; Cotrufo et al., 1998).

For ci/ca, WUEi and δ18O, the statistical analyses were conducted on the natural logarithm of the two parameters:

display math(Eqn 1)

where, for the response to elevated atmospheric CO2, X1 corresponds to values of ring width and δ18O in FACE plots, and X0 to the respective value in controls plots. Log-transformed data were used in order to linearize the metrics and to have a normal distribution even with a small number of samples (Hedges et al., 1999).

For 13C we could not directly use the δ13C of control and FACE wood, because the CO2 used at DUKE, ORNL FACE and POP_EUROFACE was depleted in 13C (Notes S1), while the δ13C ratio of the background atmospheric CO2 is c. −8.0‰. To account for these differences in the source isotope signal, 13C isotope discrimination (Δ13C) was calculated, according to Farquhar et al. (1982):

display math(Eqn 2)

where δm = δ13C of the tree ring and δa = δ13C of the atmospheric CO2. The RR was calculated from those values.

In FACE plots, for the years after the start of fumigation, eqn (2) becomes:

display math

where δx = δ13C of atmospheric CO2 in FACE plots and the following mass balance was applied to calculate δx:

display math

where [CO2]x is the total concentration of CO2 in each fumigated plot, δa and [CO2]a is, respectively, the δ13C and the concentration of atmospheric CO2 and δF, and [CO2]F is the 13C value and concentration of CO2 used for fumigation. [CO2]F was the growing season mean of the continuous series of CO2 concentration available online for each plot at each site (http://public.ornl.gov/face/ORNL/ornl_data_co2weather.shtml for ORNL and http://face.env.duke.edu/database/for Duke). We used the equations above to calculate discrimination at all sites and plots. Plant Δ13 C values are then used to estimate WUEi using the following two equations (Farquhar et al., 1989):

display math

where a is the fractionation resulting from diffusion (4.4‰), and b is the fractionation associated with carboxylation by Rubisco (c. 27‰). Note that Δ13 C should be directly related to the CO2 concentration in the chloroplast (cc) rather than ci. As a result, using ci may create complications if mesophyll conductance to CO2 is not constant (Seibt et al., 2008). WUE is estimated from ci and ca as follows:

display math

where 1.6 is the ratio of diffusivities of water and CO2 in air. A/gs values estimated here are strongly dependent on the parameter assumptions of the model and that mesophyll conductance does not limit A.

BAI, Δ13C and δ18O raw data are reported in Figs S4, S5.

For statistical analysis, BAI, Δ13C and δ18O measurements were normalized to account for pretreatment differences by subtracting each tree's pretreatment mean value from the values for each year in the tree's time-series. Data across years, period of fumigation and treatments were analysed by a nested analysis of variance. Species and CO2 treatments were tested at a significance level of < 0.05.

Partial correlations between RR_WUEi vs δ18O and climate variables were calculated to identify the climate parameters that best explained variation in tree growth and physiology. Partial correlation is a procedure that allows us to better understanding the relationship between variables (Kleinbaum et al., 2008).

Pearson and correlation coefficients were calculated using precipitation, temperature and soil moisture data from each site. Linear regressions were performed in order to explore the relationship between WUEi vs the other parameters.

Correlation, regression and time-series analyses were carried out using the SPSS 16.0 statistical package (SPSS, Chicago, IL, USA).


FACE effects and temporal dynamics of ci/ca and WUEi

Measurement of carbon isotopes indicated a significant decrease in ci/ca in all the fumigated plots compared with controls for all the tree species analysed (Figs 1, 2). The declining trend in ci/ca at ORNL and DUKE started with the start of the FACE treatment and reached minimum values for L. styraciflua growing at Duke during 1996–2002 (Fig. 1b). Post-fumigation measurements were taken at the POP-EUROFACE in 2005, and they showed that there were no longer significant differences in ci/ca between post FACE-treatment and control plots (Fig. 2a–c).

Figure 1.

Natural logarithm of response ratio (RR) of ci/ca (closed bars), intrinsic water use efficiency (WUE)i (tinted bars), δ18O (open bars) related to Liquidambar styraciflua at ORNL (a), L. styraciflua (b) and Pinus taeda (c) at Duke. The vertical lines point to the beginning of fumigation for ORNL (1998) and Duke FACE (1996). Bars are standard deviation (n = 8). Statistical significance is indicated as: *,  0.05; **,  0.01; ***,  0.001. RR_L. s_ORNL, response ratio of Liquidambar styraciflua at ORNL; RR_L.s_DUKE, response ratio of Liquidambar styraciflua at DUKE FACE; RR_P. t_Duke, response ratio of P. taeda at DUKE FACE.

Figure 2.

Natural logarithm of response ratio (RR) of ci/ca (closed bars), intrinsic water use efficiency (WUE)i (tinted bars), δ18O (open bars) related to Populus alba (a), Populus nigra (b), Populus  x euramericana (c) at POP-EUROFACE. The vertical lines point to the end of fumigation for POP-EUROFACE (2005). Bars are standard deviation (n = 8). Statistical significance is indicated as: *,  0.05; **,  0.01; ***,  0.001. RR_P. a, response ratio of Populus alba; RR_P. n, response ratio of Populus nigra; RR_P. e, response ratio of P. x euramericana.

A novel component of our work is the analysis of physiological characteristics of trees at Duke and ORNL FACE and control plots before the beginning of the FACE study. Before fumigation started, ci/ca was not statistically different in control and experimental plots, ranging, on average, between 0.65 and 0.78 for all plots.

Based on our estimates of ci/ca, we calculated ci at ambient and elevated atmospheric CO2, which increased under elevated CO2 in all tree species (Table 2).

Table 2. Intercellular partial pressure of CO2 (ci) and atmospheric CO2 (ca) for elevated (ci/ca_E) and control (ci/ca_C) plots with the corresponding per cent increase during the fumigation period analysed
  1. Data are for 1998–2004 for ORNL; 1996–2004 for Duke and 2002–2004 for POPEUROFACE. L. s, Liquidambar styraciflua; P. t, Pinus taeda; P. a, Populus alba; P. n, Populus nigra; P. e, Populus  x euramericana.

ORNLL. s0.570.63+ 25+ 39
DukeL. s0.630.70+ 35+ 50
DukeP. t0.700.75+ 41+ 50
POPEUROFACEP. a0.630.65+ 53+ 57
POPEUROFACEP. n0.700.72+ 53+ 57
POPEUROFACEP. e0.700.75+ 45+ 57

Similarly, in all the experimental years and at each site, we observed a pronounced and significant increase in WUEi under elevated CO2, compared with control. In the POP-EUROFACE site, the increase in WUEi was reduced but still present in 2005, a year after the end of fumigation (Fig. 2). The linear regression (Fig. 3a) between WUEi in control plots (WUE_C) and WUEi in fumigated plots (WUE_E) during the CO2 exposure was largely consistent across the broad range of species and sites, with an average WUEi increase of 75 ± 13% under elevated CO2. The linear regression is above the 1 : 1 line (< 0.001) indicating a significant stimulation effect of CO2 on WUEi. The coefficient of determination (r2) is 0.55, < 0.01, the slope is significantly < 1 and the intercept is significantly different from 0 and positive, showing that, overall, as WUEi increases the CO2 stimulation declines.

Figure 3.

Regressions of intrinsic water use efficiency (WUEi)_E (a) and δ18O_E (b) as a function of WUEi_C and δ18O_C, respectively. The dotted line represents the 1 : 1 relationship. Each data point represents the mean value under elevated CO2 and control sites for each species investigated during the fumigation period: 1998–2004 for ORNL, 1996–2004 for Duke and 2002–2004 for POP-EUROFACE. L. s_Duke, Liquidambar styraciflua at DUKE; L. s_ORNL, Liquidambar styraciflua at ORNL; P. t, Pinus taeda; P. a, Populus alba; P. n, Populus nigra; P. e, Populus x euramericana.

FACE effects and temporal dynamics of δ18O

For L. stryraciflua growing at ORNL, in the 9 yr before the FACE experiments, the δ18O of tree rings in the future FACE plots were significantly lower than that of trees in control plots, in seven out of the nine years (Fig. 1a). However, when comparing δ18O values recorded in tree ring of FACE plots before (1988–1997) and during (1998–2004) fumigation, we observed a significant difference (< 0.001), with higher δ18O during CO2 enrichment.

At the Duke FACE, δ18O varied between 26.1‰ and 30.7‰ across the entire time-period (Fig. S4b,c). The fumigation led to a slight increase in δ18O of P. taeda during 2000–2004, while this increase was insignificant in the other years under CO2 fumigation (Fig. 1c). In L. styraciflua the increase in δ18O was significant (< 0.05) during 1999–2003 (Fig. 1b). Moreover, when we compared the δ18O values of the years before and after fumigation in FACE plots, we found small but statistically significant differences (< 0.05) for both species (Fig. 1b,c).

At the POP-EUROFACE, elevated CO2 increased the δ18O of tree rings in all three genotypes and years of fumigation compared with control (Fig. 2a–c), but the difference was highly significant only in 2003 and 2004. In 2005, after the end of the fumigation, the tree ring δ18O in FACE plots decreased, and no significant differences between FACE and control trees were recorded.

The relationship between δ18O at elevated (δ18O_E) and at ambient CO218O_C) was significant across all the sites and species (r2 = 0.78, < 0.01; Fig. 3b). The linear regression is not significantly different from the 1:1 line, showing only a slight increase of δ18O as a result of the elevated CO2 in some species and some experimental years. The positive intercept is trigged mainly by POP-EUROFACE results and implies that δ18O increases as the CO2 response declines.

WUEi interaction with physiological and climatic parameters

The changes in 13C-derived WUEi (RR_WUEi) observed for the various species in the different FACE sites are well related to δ18O (RR_δ18O) during the fumigation period, suggesting that stomatal conductance played a significant role for all the species analysed, except for P. taeda growing at Duke FACE, where no significant relation was found between RR_WUEi and RR_δ18O (Table 3).

Table 3. Linear regression between intrinsic water use efficiency (WUEi) vs δ18O and basal area increments (BAI)
r 2 b r 2 b
  1. r2, slope coefficient (b) and the relative significant P (number of stars) from the F-test are reported. *,  0.05; **,  0.01; ***,  0.001. L. s, Liquidambar styraciflua; P. t, Pinus taeda; P. a, Populus alba; P. n, Populus nigra; P. e, Populus  x euramericana.

WUEiORNLL. s0.65*0.590.0030.50
WUEiDukeL. s0.68*0.480.100.26
WUEiDukeP. t0.120.080.30*0.80
WUEiPOPEUROFACEP. a0.45*0.530.38*0.55
WUEiPOPEUROFACEP. n0.48*0.350.74**0.27
WUEiPOPEUROFACEP. e0.38*0.510.91***0.70

Comparing the elevated CO2-induced changes in WUEi (RR_WUEi) with the correspondent change in BAI values (RR_BAI) for the different species during the fumigation period, we noted a remarkable difference among species. A strong relationship was observed for the three poplar genotypes, with the maximum value for P. x euramericana (r2 = 0.91, < 0.001), a moderate but still significant relationship for the P. taeda and no significant relationship for L. styraciflua, for which the increase in WUEi appeared not to translate into enhanced tree growth (Table 3).

The relative effect of climate vs elevated CO2 on the tree WUEi was evaluated by partial correlations. No significant differences were found between the pre-fumigation and the fumigation periods for temperature and precipitation (Notes S2). For L. styraciflua, the elevated CO2 stimulation of WUEi was enhanced by high temperature and low precipitation, in the summer months, at both the ORNL and Duke FACEs (Table 4). Those results are confirmed by the high WUEi found for this species in warm and dry years such as 2002, 2003 and 2004 (Fig. 1a,b). In contrast, for P. taeda the elevated CO2 stimulation of WUEi was mainly related to soil moisture, and for the three poplar genotypes it was significantly correlated with summer temperature (Table 4).

Table 4. Partial correlation between RR_WUEi and climatic factors
Controlling factorSiteSpecies r 2 Months
  1. Only the months of the year when correlations were significant are reported. The sign (+) indicates a positive correlation between parameters while the sign (−) indicates a negative correlation. RR_WUEi, variation in WUEi. L. s, Liquidambar styraciflua; P. t, Pinus taeda; P. a, Populus alba; P. n, Populus nigra; P. e, Populus x euramericana. *,  0.05; **,  0.01; ***,  0.001.

PrecipitationORNLL. s0.64* (−)June–July
TemperatureORNLL. s0.53* (+)June–July
PrecipitationDukeL. s0.68* (−)July–August
TemperatureDukeL. s0.75** (+)July–August
Soil moistureDukeP. t0.86** (−)June–July–August
TemperaturePOPEUROFACEP. a0.70** (+)May–June–July–August
TemperaturePOPEUROFACEP. n0.74**(+)May–June–July–August
TemperaturePOPEUROFACEP. e0.69* (+)May–June–July–August


Stable isotope analysis was successfully used to confirm our first hypothesis and to clarify the mechanisms responsible for the species-specific ecophysiological responses in terms of trees growth or stomata regulation. Overall, we quantified significantly higher WUEi (inferred from Δ13C) in FACE plots than in controls in all species, with a reduction of stomatal conductance in L. styraciflua (both at ORNL and Duke FACE), and in the Poplar sp. at POP-EUROFACE. The effect of elevated CO2 was more variable for P. taeda at Duke FACE. For all species, seasonal climate variability, mainly temperature and soil moisture, affected tree 13C-derived WUEi.

display math

Previous estimates of ci/ca showed no significant response to elevated CO2 (Ellsworth, 1999; Ainsworth & Long, 2005), leading to the conclusion that, although gs is reduced by elevated CO2, this by itself does not limit carbon uptake. Our results showed a slight but significant decrease in ci/ca under FACE condition for all species and sites, with the smallest decrease recorded for P. alba and P. nigra and the largest in L. styraciflua_DUKE (Table 2). Previous ci/ca values were often derived from sporadic gas exchange or isotopic measurements on single leaves and during a single growing season (Jackson et al., 1994; Ellsworth, 1999), while our results integrated the physiological responses to elevated CO2 of whole trees and over the entire fumigation period. Saurer et al. (2004) proposed three scenarios that mainly differ in the degree by which the increase in ci follows the increase in ca, as in the case of FACE sites: (1) ci is kept constant thus ci/ca decreases and WUEi can increase strongly; (2) ci increases proportionally to ca and the ratio ci/ca is kept constant, accordingly WUEi still increases; (3) ci increases at the same rate as ca and the ratio ci/ca increases while WUEi does not change. Our study appears to correspond to the first scenario. We observed that a relatively small decrease in ci/ca (Table 2) translated to a significant increase in WUEi (Figs 1, 2) for all species. In particular, we observed an increase of 56% at ORNL, of 77% and 90% at Duke for P. taeda and L. styraciflua, respectively, and of 64%, 72% and 89% at POP-EUROFACE for P. alba, P. nigra and P. x euramericana, respectively.

Elevated CO2 effects on L. styraciflua

A significant relationship was found for L. styraciflua grown at ORNL and at Duke between the FACE-induced variation in WUEi (RR_WUEi) and that in δ18O (RR_δ18O), but there was a lack of correlation between RR_WUEi and the FACE-induced change in BAI (RR_BAI). A and gs are strongly coupled and adjustments in both influence WUEi while δ18O is thought not to be strongly influenced by A. Thus, our results suggest that the increase in WUEi of this species reduced tree water losses rather than increasing assimilation and productivity. Our results confirm previous studies (Gunderson et al., 2002; Schafer et al., 2002; Herrick et al., 2004), where a mean relative decrease in gs of 24% and 28% was reported for L. styraciflua at ORNL and Duke, respectively, throughout the study periods. Broadleaved species generally have stomata that are more responsive to elevated CO2 than coniferous species (Pataki et al., 1998; Ellsworth, 1999; Maier et al., 2002) and L. styraciflua showed stomatal closure, regardless of soil moisture, and limited hydraulic modification under elevated CO2 (Domec et al., 2010). Thus, for this species, the high enrichment in δ18O under FACE is likely caused by increased stomatal closure (Pataki et al., 2003).

The high correlations between the variation of WUEi and summer temperature and precipitation (Table 4) indicate the consistent influence of climatic constraints on inter-annual variation in WUEi. Air temperature is directly linked with water vapour pressure and warming may further reduce gs under elevated CO2 by affecting both assimilation and respiration. The further reduction of gs could reduce foliar evapotraspiration, which may result in higher leaf temperature and, thereby, greater respiratory carbon loss (Korner, 2006). Previous studies conducted at the FACE sites suggested that the magnitude of the reduction in gs for L. styraciflua growing under elevated CO2 varied considerably with environmental factors (Warren et al., 2001; Herrick et al., 2004; Warren et al., 2011a,b) and was most significant during the dry years, accompanied by a consistent 10–16% reduction in transpiration (Warren et al., 2011a). The lack of relationship between RR_WUEi and RR_BAI (Table 3) showed that the increase in WUEi recorded for this species did not necessarily translate into increased plant growth. These results are consistent and supported by the observations that aboveground growth of L. styraciflua at ORNL was significantly greater under elevated CO2 during the first year of treatment, diminished in the subsequent years (although enhancement of net primary productivity was sustained) and became nonstatistically different from the control in the latter years (2006–2008) (Norby et al., 2010). At Duke, hardwood trees stand level growth increased in FACE compared with control plots (McCarthy et al., 2010). However, c. 90% of growth under both ambient and elevated CO2 was contributed by pines, while L. styraciflua's BAI did not significantly differed between treatments (H.R. McCarthy, unpublished data), similar to our results (r2 between ours and McCarthy's set of BAI measurements was 0.82, < 0.001). Moreover, the leaf area index (LAI) for L. styraciflua was relatively constant during the entire fumigation period (Schafer et al., 2002; Norby et al., 2003). When CO2 enrichment does not increase foliar production and, as a consequence LAI, then a reduction in gs can translate to a reduction of transpiration leading to stand level water savings (Warren et al., 2011a), which influences the increase of δ18O.

Elevated CO2 effects on P. taeda

For P. taeda our δ18O results and the lack of relationship between RR_WUEi and RR_δ18O confirmed the limited effects of elevated CO2 on gs for this species (Schafer et al., 2002). Indeed, we observed a slight but significant increase in δ18O in FACE plots, starting 5 yr after the onset of fumigation. This excluded that trees were accessing a different water source than control trees (Barbour, 2007), at least for the first 5 yr. Further, no significant differences were found (Notes S2) between climate data recorded at the site before and after fumigation; thus the increase in δ18O can be assumed to be directly related to the CO2 enrichment.

Possibly, 18O discrimination was affected by structural changes to the tree hydraulic pathway, reported by Domec et al. (2009) for the fumigated trees. Structural modifications could reduce the diffusion of 18O-enriched water from the sites of evaporation to the mesophyll (Pèclet effect) limiting the δ18O dilution effect usually observed at the site of transpiration (Barbour et al., 2000). Hence, in P. taeda the influence of transpiration on WUEi was limited. An important factor that could account for the observed increase in WUEi under elevated CO2 is the increase in soil water availability attributed to increased litter accumulation (Schafer et al., 2002) resulting from increased LAI (McCarthy et al., 2007). Indeed, increases in LAI can result in a decrease in 13C discrimination and a consistent enhancement in tree WUEi (Buchmann et al., 1997). Although changes in LAI are not explicitly included in our calculation, the statistically significant relation between RR_WUEi and RR_BAI (Table 2) indicated that the FACE-induced WUEi enhancement was driven by an enhancement in assimilation (Ellsworth et al., 2012). Furthermore, the positive correlation between soil moisture and RR_WUEi (Table 4) during the fumigation period, indicates that local microclimatic conditions have driven the observed increase in WUEi in P. taeda trees growing under elevated CO2 rather than stomatal closure.

Elevated CO2 effects on Poplar sp

At the POP-EUROFACE, the elevated CO2-induced increase in WUEi for the three species was significantly related to the increase in BAI during the whole study period. Elevated CO2 also increased LAI; the stimulation persisted after canopy closure and was followed by downregulation of photosynthesis in some cases (Calfapietra et al., 2005; Liberloo et al., 2005, 2006). The sustained increase in A and assimilation (Bernacchi et al., 2003) is possibly associated with a reduction in gs, as suggested by the observed changes in δ18O. A reduction in gs of 19–24% across 6 yr was described by Tricker et al. (2005) who suggested a progression in stomatal sensitivity to the ca increase, as also reported by Calfapietra et al. (2005) for P. x euramericana. When canopies were still open, gs decreased, likely because of reduced stomatal density and later, when canopies gradually closed, stomatal closure trigged the response under elevated CO2 leading to an improvement of leaf level WUEi. We also found that the highest δ18O values were in hot and dry years, such as 2003 and 2004, indicating that the response of gs is also strongly dependent on climate (Kubiske et al., 2006).

For all the poplar genotypes, the differences in ci/ca, as well as in gs between the fumigated and control plots were reduced after the fumigation stopped, that is, in 2005. We can suggest a double mechanism involved in those changes. First, the FACE experiment might not have run long enough to reveal negative feedback of CO2 and nutrient limitation on A, and what we observed was basically the initial stimulatory effect (Tognetti et al., 2000; Korner, 2006), which does not allow us to draw conclusions about the long term trend. Second, elevated CO2 could influence the water saving of the plants, without inducing long-term changes in anatomical traits. Yazaki et al. (2005) and Watanabe et al. (2010) analysed various conifer and broadleaved species, showing that cell and lumen diameters, vessel area and other wood structures changed differently in different species under elevated CO2 and those alterations did not permanently affect the plants.


A FACE induced increase of WUEi was observed in all the five tree species investigated and was related to changes in ca and climate. We demonstrated that the ca rise triggered species-specific physiological responses, as well as a change in the climate sensitivity of trees. In particular, in L. styraciflua, warmer conditions seemed to be coupled with a reduction in transpiration, leading to higher δ13C values and therefore increasing WUEi but without a parallel stimulation of tree growth. A reduction of gs was also observed in the fast-growing Populus sp., accompanied by positive tree growth responses, and was partially limited by high temperature during 2003 and 2004. Finally the rise of WUEi in P. taeda was mainly related to soil moisture increases under elevated CO2 and opens new questions about the ability of this isohydric species (with tight stomatal control) to withstand the expected reduction in soil water in combination with an increase in drought.

The simultaneous measurements of tree-ring cellulose isotopic variability and growth uniquely allowed us to estimate changes in tree physiology and productivity in response to elevated CO2. The effects of elevated CO2 on forests' hydrological balance will be highly dependent on forest species composition, and WUEi should be incorporated into conceptual frameworks for assessing the species responses to climate change.


We thank David Frank, Rolf T. Siegwolf and Ram Oren for useful comments and encouragement for this work. We gratefully acknowledge Keith Lewin for sharing with us his experience on the performance of the DUKE FACE facility. We are grateful for logistic and technical support offered by Duke, ORNL and POP-EUROFACE. We thank the editor and the three anonymous referees for their valuable comments. The Duke and ORNL FACE experiments were supported by the United States Department of Energy, Office of Science, Biological and Environmental Research Program. Funding for the sampling campaign was provided by IBIMET CNR.