Spatial and temporal scaling of intercellular CO2 concentration in a temperate rain forest dominated by Dacrydium cupressinum in New Zealand


David Tissue. Fax: +1-806-742-2963; e-mail:


Seven methods, including measurements of photosynthesis (A) and stomatal conductance (gs), carbon isotope discrimination, ecosystem CO2 and water vapour exchange using eddy covariance and the use of a multilayer canopy model and ecosystem Keeling plots, were employed to derive estimates of intercellular CO2 concentration (Ci) across a range of spatial and temporal scales in a low productivity rain forest ecosystem dominated by the conifer Dacrydium cupressinum Lamb. in New Zealand. Estimates of shoot and canopy Ci across temporal scales ranging from minutes to years were remarkably similar (range of 274–294 µmol mol−1). The gradual increase in shoot Ci with depth in the canopy was more likely attributable to decreases in A resulting from lower irradiance (Q) than to increases in gs due to changes in air saturation deficit (D). The lack of marked vertical gradients in A and gs at saturating Q through the canopy and the low seasonal variability in environmental conditions contributed to the efficacy of scaling Ci. However, the canopy Ci estimate calculated from the carbon isotope composition of respired ecosystem CO2 (δ13CR; 236 µmol mol−1) was much lower than other estimates of canopy Ci. Partitioning δ 13CR into four components (soil, roots, litter and foliage) indicated root respiration as the dominant (> 50%) contributor to δ 13CR. Variable time lags and differences in isotopic composition during photosynthesis and respiration make the direct estimation of canopy Ci from δ 13CR problematic.


fractionation of 13CO2 due to diffusion (4.4‰)


rate of net photosynthesis


effective fractionation of 13CO2 due to carboxylation (taking into account the mesophyll conductance

27‰); Ca

ambient CO2 concentration


intercellular CO2 concentration


CO2 concentration at the sites of carboxylation


CO2 concentration at a surface


rate of net CO2 exchange


ambient air saturation deficit


air saturation deficit at a surface


rate of transpiration


mesophyll conductance


stomatal conductance to water vapour transfer


stomatal conductance to CO2 transfer


residual stomatal conductance at light compensation point


a parameter relating Ci to stomatal conductance






CO2 concentration at compensation in the presence of dark respiration

Γ *

CO2 concentration at compensation in the absence of dark respiration


ratio of diffusivities of water vapour and CO2 in air


plant discrimination against 13C , relative to source air


volumetric root-zone water content


carbon isotope ratio of source air

δ 13CR

ecosystem-respired CO2

δ 13Cs

carbon isotope ratio of shoots


height above ground. The subscripts s, z and c refer to shoot, layer number in the canopy and canopy scales, respectively.


Ecosystem CO2 exchange, and its response to changing climate variables, is strongly regulated by physiological processes that occur at the shoot and canopy scale. To use measurements of A and gs on individual shoots to estimate net CO2 uptake for canopies and ecosystems, we need to scale the processes regulating CO2 exchange both spatially and temporally (Field & Ehleringer 1993). There has been considerable effort to test appropriate scaling approaches for CO2 exchange (Lloyd et al. 1995; Lavigne, Ryan & Anderson 1997; Law, Ryan & Anthoni 1999) using different methods (Ruimy et al. 1995; Flanagan, Brooks & Ehleringer 1997; Katul, Ellsworth & Lai 2000; Dore et al. 2003; Dawson, Ward & Ehleringer 2004). One of the major challenges is to use data from short-term observations at different spatial and temporal scales with models to estimate long-term effects (Weltzin et al. 2003). When scaling spatially, care is needed to deal appropriately with non-linearity in parameter values (Raupach 1995) and the increase in uncertainty with increasing number of parameters (Wang et al. 2001). The use of multilayer models is appropriate when scaling processes of CO2 exchange from shoots to canopies (de Pury & Farquhar 1997; Baldocchi & Meyers 1998) and this approach has been shown to be successful when tested against direct measurements of CO2 exchange using eddy covariance measurements (Dang, Margolis & Collatz 1998; Wilson, Baldocchi & Hansen 2001).

For shoots, the Ci reflects both A and gs, and thus the effects of physiological and environmental variables on CO2 exchange (Farquhar, Ehleringer & Hubick 1989; Le Roux et al. 2001; Dawson et al. 2002; Niinemets, Sonninen & Tobias 2004). At the canopy scale, Ci reflects more complex processes, including the effects of canopy architecture and the contributions from different layers of foliage within the canopy. For example, shoot Ci is generally lower in the upper canopy relative to the lower canopy due to higher A at higher Q in the upper canopy, which subsequently reduces shoot Ci (Le Roux et al. 2001). These complexities are recognized in multilayer models that incorporate coupling between A and gs, recognize the separate contributions of sun and shaded foliage and include changes in energy balance with depth in canopies (Collatz et al. 1991; Leuning et al. 1995). Thus, values of Ci obtained using different methods can be used to test the suitability of spatial and temporal scaling approaches (Lloyd & Farquhar 1994) and the success of models (Katul et al. 2000). This also allows comparison of methods employing measurements of A and gs with the use of stable carbon isotopes (Farquhar et al. 1989; Dawson et al. 2002).

Unlike gas exchange measurements, which provide short-term estimates of A, and hence Ci, under current environmental conditions, measurements of carbon isotope discrimination provide an integrated measure of the relativity between CO2 supply (i.e. boundary layer, gs and gm) and demand (i.e. rate of photosynthesis) throughout the lifetime of the plant material. The temporal integration period of carbon isotope discrimination depends on the carbon fraction being analysed and on whether or not recently assimilated carbon (i.e. soluble carbohydrates) exchanges with structural carbon (i.e. lignin and cellulose) that was laid down when the foliage expanded. If carbon exchange is minimal, and structural carbon comprises the largest fraction of foliage carbon, then carbon isotope discrimination will reflect processes that occurred during foliage expansion (i.e. structural carbon) and the last few weeks prior to sampling (i.e. soluble carbon). Interpretation of the carbon isotope ratio of shoots (δ13Cs) gives an A-weighted integration of the balance between CO2 supply and demand, reflecting integrated Ci (with consideration of gm; Farquhar et al. 1989). To compare these estimates with those derived from measurements of gas exchange, it is important that integration of the latter incorporates weighting both spatially and temporally in relation to the contribution to A at the time of the measurement. δ 13Cs often varies considerably within a forest canopy, and while most of this variation is related to variation in Ci (Brooks et al. 1997; Buchmann, Kao & Ehleringer 1997), some variation is also a result of variation in δ 13C of the source air (δ 13Ca). In studies where δ 13Ca is measured, δ 13Cs may be expressed as Δ so that variation may be interpreted with respect to Ci alone (again, allowing for appropriate consideration of gm). Variation in Δ can therefore be used as an integrative record of shoot Ci over the time period when the carbon was fixed.

Recent work suggests that the carbon isotope composition of δ 13CR, estimated using Keeling plots (Keeling 1958), may reflect canopy Ci to some extent (Bowling et al. 2002; McDowell et al. 2004). Following work demonstrating rapid links between canopy photosynthesis and soil-respired CO2 (Ekblad & Högberg 2001; Högberg et al. 2001), Bowling et al. (2002) showed a strong positive relationship between δ 13CR and D measured 5–10 d before CO2 sampling occurred. An increase in D was suggested to have resulted in lower gs, which reduced Ci, so carbon fixed during this time was less depleted (δ 13Cs became less negative and Δ decreased). The less-depleted carbon was then respired as CO2 by the ecosystem 5–10 d later (Bowling et al. 2002). However, many components other than shoot respiration contribute to whole ecosystem respiration, including litter, soil and root processes (Dawson et al. 2002), thereby complicating the calculation of canopy Ci from δ 13CR. A unique partitioning solution from a linear isotope mixing model is not possible with more than two contributing sources for a single isotope ratio. However, Phillips & Gregg (2003) have published a (IsoSource) model that calculates all possible combinations of partitioning to achieve isotope mass balance, so that interpretation of δ 13CR in terms of canopy Ci may be possible.

In this study, our objective was to compare the different approaches for scaling Ci from shoots to canopy in a mixed conifer-broadleaved rain forest dominated by Dacrydium cupressinum Lamb. (rimu) (Podocarpaceae). This was an ideal system to test the validity of temporal and spatial scaling because of the dominance of one species and the temperate coastal climate with only small seasonal variations in temperature and water availability. Seven methods were used to estimate Ci across a range of spatial and temporal scales. Shoot Ci was estimated from instantaneous measurements of A and gs using gas exchange (minute) and measurements of carbon stable isotope discrimination (year). Upper canopy Ci was estimated from independent gas exchange measurements at different heights in the upper canopy, analysed by weighting A spatially (day) or by applying Leuning (1995) coupled photosynthesis-stomatal conductance model (day) that incorporates the response of gs to D. Canopy scale estimates of Ci were generated using gas exchange data at different heights throughout the canopy in a multilayer canopy model with both spatial and temporal weighting for A (year; Whitehead et al. 2004). Eddy covariance measurements for estimating net ecosystem CO2 exchange were adjusted using independent estimates of evaporation and respiration from the forest floor (month) and used to calculate canopy Ci. Shoot and canopy Ci were then compared with canopy Ci estimated from measurements of δ 13CR (decade). The proportional contribution of shoot, litter, soil and root respiration to ecosystem respiration was considered, as well as the physiological basis for observed canopy gradients in shoot Ci.


Study site and species

Measurements were made in an extensive, mixed conifer-broadleaved rain forest at Okarito Forest, South Westland, New Zealand (latitude 43.2 S, longitude 170.3 E). This lowland (50 m elevation) terrace forest was dominated by the canopy-emergent conifer D. cupressinum, but other tree species were present (James & Norton 2002). Mean tree age was ≈ 200 years and D. cupressinum comprised 73% of the basal area with a mean canopy height of 20 m. The half-surface area index at the site was 4.8 m2 m−2, consisting of 3.7 m2 m−2 for foliage and 1.1 m2 m−2 for branches (Walcroft et al. 2005). Foliage of D. cupressinum consisted of small, keeled, imbricate scales, 0.5–1 mm wide, surrounding the pendulant branches.

Annual rainfall is high (≈ 3400 mm) at the site and evenly distributed throughout the year. Precipitation primarily occurs as intense rainfall events, rather than as continuous rainfall with constant cloud cover, such that there are many clear days during the year. Based on long-term measurements made in 2001 and 2002, the mean diffuse fraction of global radiation was 0.54, although most days (70%) were either clear (diffuse fraction < 0.4; 38% of days) or overcast (diffuse fraction > 0.9; 31% of days). Mean annual air temperature is 11.3 °C, with frosts occurring rarely. The soil taxonomy is described as Entisols that have evolved to Inceptisols or Spodosols (Soil Survey Staff 1994). The soils have a high organic matter content (≈ 30%), low permeability and porosity and are frequently waterlogged. Soil nitrogen concentration at an adjacent site was 633 µmol g−1 and soil phosphorus was 12 µmol g−1 (Richardson et al. 2004), indicating very low soil phosphorus availability.

Measurements of air temperature, D and Q were made above the forest canopy and recorded at half-hourly intervals throughout the period of field data collection. The vertical profile of Q through the canopy was determined from the mean of eight measurements of photosynthetically active radiation (PAR; 400–700 nm), made on a horizontal plane at 12 heights from the canopy access tower using a quantum sensor (Model LI-190, Li-Cor Inc., Lincoln, NE, USA), on two clear days during summer (January). Values were normalized by dividing by Q, which was measured above the canopy during each measurement period, and mean Q was calculated for each height.

Theory for calculating Ci

For a shoot, Ci can be described in relation to the ambient CO2 concentration at the foliage surface (C′), the rate of photosynthesis (A) and stomatal conductance to CO2 transfer (gsc) based on the diffusion of CO2 through stomata as


Similarly, gsc can be described by


where gs is the stomatal conductance to water vapour diffusion, γ is the ratio of the molecular diffusivities of water vapour and CO2 in air (= 1.6), D′ is the air saturation deficit at the foliage surface, and E is the rate of transpiration.

The same equations can be applied to canopies if sunlit and shaded leaves are treated separately (de Pury & Farquhar 1997). Canopy fluxes and conductance are then defined as the product of mean values of leaf-level fluxes and conductance and leaf area index of sunlit and shaded fractions. In well-mixed conditions, with small leaves, it can be assumed that boundary layer and eddy diffusive conductances are sufficiently large such that C′ ≈ Ca and D′ ≈ D, where Ca and D are the CO2 concentration and the air saturation deficit in the ambient air surrounding the shoot or above the canopy, respectively (Jarvis & McNaughton 1986). Calculations from measurements of stomatal conductance and estimates of transpiration at shoot and tree scales in this forest (Barbour & Whitehead 2003) showed that values for the decoupling coefficient, Ω, were consistently less than 0.2, confirming that these assumptions are valid (Jarvis & McNaughton 1986).

Seven different methods were used to calculate average Ci at scales from shoots to the whole canopy, as outlined below. Where possible, standard errors (SE) and 95% confidence intervals were also calculated. Differences in estimated Ci using the seven methods were tested for significance using a Student's t-test.

Estimates of shoot Ci using gas exchange

Measurements of A and gs were conducted on shoots distributed at ≈ 2 m intervals through the depths of the canopies on three D. cupressinum trees. Access to the shoots was available from a 25-m-tall canopy access tower at the site. Gas exchange measurements were taken on fully expanded mature foliage maintained at saturating Q (1000 µmol m−2 s−1), ambient atmospheric CO2 concentration (Ca, 370 µmol mol−1) and constant temperature (20 °C) during the summer (January 2000) using four portable photosynthesis systems (Model 6400, Li-Cor Inc.) equipped with CO2 control modules and a red-blue light source (Model 6400-02B). Air saturation deficit in the cuvettes was generally low and close to ambient conditions, between 7 and 13 mmol mol−1, during the measurements. Two to three shoots from at least two separate branches were measured at each canopy height for each tree. The shoots were placed in the cuvette and measurements were recorded after A and gs had equilibrated, typically after 1 min. Values of shoot Ci were calculated using Eqns 1 and 2(Table 1). The saturating Q used for these measurements is typical for sunlit shoots (either shoots at the top of the canopy or shoots in sunflecks lower in the canopy) on clear days (diffuse fraction was < 0.4), which represent 40% of days in the year.

Table 1.  Summary of the calculations needed to estimate Ci at different spatial and temporal scales Thumbnail image of

Estimates of projected foliage area were determined from digital photographs using image analysis software (Scion Image Beta 3b, Win, Scion Corporation, Frederick, MD, USA). Projected surface area was converted to half-surface area using an allometric multiplier of 1.66 determined by Carswell et al. (2005).

Estimates of shoot Ci using carbon isotope discrimination

Three or four samples of current year, fully expanded, foliage were collected throughout the canopy at heights similar to those used for measurements of A during the growing season (December to March) in 2001 and 2002. Samples were dried at 70 °C to constant weight and ground to a fine powder using a ball mill, then analysed on a Dumas elemental analyser (Europa Scientific ANCA-SL, Europa Scientific Ltd, Crewe, UK) interfaced to a stable isotope mass spectrometer (Tracermass, Europa Scientific Ltd). Isotope ratios are reported relative to the PDB standard (i.e. fossil belemnite from the Pee Dee formation in South Carolina), such that


where Rs and Rst are the 13C/12C ratios of the sample and the standard, respectively. Foliar carbon isotope ratios are presented as the degree of discrimination relative to ambient air (Δ), following Farquhar & Richards (1984) as


where δ 13Ca is the carbon isotope ratio at ambient CO2 concentration and δ 13Cl is the carbon isotope composition of leaf material. Values for Ci were calculated assuming that the fractionation of 13CO2 due to diffusion (a) and carboxylation (b) were 4.4‰ and 27‰, respectively (Farquhar et al. 1989) and Ca is the ambient CO2 concentration, such that


Measurements of the changes in δ 13Ca and Ca with canopy height were made to allow the calculation of Δ and thus Ci through the canopy (Table 1). We did not measure gm or include it in the calculation of Ci. However, published values of gm (Warren et al. 2003), have shown that it does not appreciably affect Ci at values of A and gs typical for D. cupressinum, if allowance for gm is made with an appropriately reduced value of ribulose 1.5-bisphosphate carboxylase/oxygenase fractionation, b (see Appendix). Air was sampled at six canopy heights (0.2, 1, 4.5, 9, 18 and 23 m above ground level) and at approximately 10 m above the canopy (36 m above ground level) in January and March 2002. Using an automated sampling system (Xu, Matista & Hsiao 1999), air was drawn for 2 min (there was a 9 s delay in recording a change in CO2 concentration) using a single pump from inlets attached to the canopy access tower and dried by passing it through a magnesium perchlorate trap before sampling into pre-evacuated flasks. Gas was then analysed for CO2 concentration using a gas chromatograph (Model HP 5890 Series II, Hewlett-Packard, Avondale, PA, USA) and δ 13C using a stable isotope mass spectrometer (Model Finnigan, MAT, Bremen, Germany) as described by Ferretti et al. (2000). Measurements of D at the six canopy heights were conducted at the same time by drawing air through an infrared gas analyser (Model 6262, Li-Cor Inc.) prior to passing the air through the water vapour trap.

Estimates of upper canopy Ci using gas exchange

Measurements of A and gs were conducted at heights of 18 m (mid-canopy) and 22 m (upper canopy) above the ground on 10 shoots spread across two mature D. cupressinum trees. Measurements were taken under a wide range of ambient conditions in summer (January) using two portable photosynthesis systems (Model LI-6400, Li-Cor Inc.) equipped with cuvettes with clear chamber tops. Air temperature, D and Q were allowed to vary with ambient conditions and measurements were made at approximately two-hourly intervals on the same shoots for three clear days, during which the diffuse fraction was between 0.2 and 0.3 and total daily irradiance was between 30 and 34 MJ m−2. There were no significant differences in gs between the two trees, therefore the data were combined.

Two approaches were used to analyse these data. Firstly, mean Ci for the 10 shoots was calculated for each measurement throughout the day, and then these values were weighted by the mean proportion of total A for the day at the time of measurement. The mean of these values provided estimates of daily mean Ci for shoots in the two canopy positions (Table 1). The second approach used a coupled model to describe the relationships between gsc (= gs), A, Ca and D (Leuning 1995) as


where, gsc0 is the residual stomatal conductance at light compensation point, Do is a parameter describing the sensitivity of gsc to D when D > Dmin where Dmin is the minimum value of D below which gsc is constant at its maximum value (Table 1). Γ is the CO2 concentration at compensation in the presence of dark respiration, and α is related to the intercellular CO2 concentration such that 1/α = (1 − Ci/Ca). The value for α can be obtained from the slope of a plot of gsc as the ordinate against the abscissa comprising the right-hand term in Eqn 6 excluding α.

Estimate of whole canopy Ci using a model

A one-dimensional, multilayer model was used to estimate the distribution of Ci throughout the D. cupressinum canopy at the site. The model calculated the direct and diffuse components of radiative transfer, energy balance, transpiration and photosynthesis for sunlit and shaded foliage in 20 layers through the canopy (Leuning et al. 1995) and combined this with water balance (Whitehead, Leathwick & Walcroft 2001) to provide daily estimates of canopy photosynthesis (Table 1). The model required values for parameters describing A (Farquhar, von Caemmerer & Berry 1980; Farquhar & Wong 1984), the coupling of A with gs and D as given by Eqn 6 (Leuning 1995), respiration and their changes with depth in the canopy. A full description of the model and values for the parameters for D. cupressinum at the field site, based on data reported by Tissue et al. (2005) and in Fig. 1 of this paper, was given in Whitehead et al. (2004). Each day, hourly estimates of Ci for sun and shade foliage in the 20 layers in the canopy were weighted by A then meaned to give daily values for a year. These daily values were then further weighted by daily A and meaned to give an annual estimate of Ci for the canopy (Table 1).

Figure 1.

Relationships between (a) stomatal conductance to water vapour transfer (gs) and air saturation deficit (D) and (b) stomatal conductance to CO2 transfer (gsc) and the right-hand term in Eqn 6 excluding α over several days for shoots distributed in the upper canopy of two Dacrydium cupressinum trees. The value for α is obtained from the slope of the line. Measurements in (a) were conducted on three days in summer (January) under different ambient environmental conditions. The same measurements were used to generate the relationship shown in (b), but data were confined to conditions when irradiance (Q, 400–700 nm) was greater than 500 µmol m−2 s−1. The relationship in (a) is given by Eqn 6 and values for the parameters, gsmax = 182 mmol m−2 s−1 and D0 = 10.02 mmol mol−1 with Dmin set at 7 mmol mol−1, r 2 = 0.33, P < 0.001. The slope (α) and the intercept (gsc0) of the linear relationship in (b) are 5.08 and 16.19 mmol m−2 s−2, respectively, with r 2 = 0.67 and P < 0.001. gsmax, maximum stomatal conductance; Dmin, minimum value of air saturation deficit below which stomatal conductance is maximum; D0, sensitivity of decreasing stomatal conductance to increasing air saturation deficit.

Estimate of canopy Ci using eddy covariance and soil surface exchange

The exchange of CO2 and water vapour between the atmosphere and the ecosystem was measured continuously during late summer (January and February 2002) at the site. An eddy covariance system (Hunt et al. 2002) was mounted 36 m above the ground and air was pumped through a closed-path infrared gas analyser (Model LI-6262, Li-Cor Inc.). The mean covariance between fluctuations in the vertical wind speed and the scalar of CO2 or water vapour was determined and used to calculate the flux density following the corrections on the effects of storage below the height of the sonic anemometer and for spectral losses (Table 1). This procedure was described fully by Hunt et al. (2002).

For the purposes in this paper, half-hourly values of ecosystem flux densities were screened to include data when conditions were sufficiently well mixed (friction velocity > 0.25 m s−1), the canopy was dry and Q was high (> 500 µmol m−2 s−1). These conditions provided the best opportunity to estimate canopy conductance to water vapour without the confounding effect of evaporation from wet surfaces within the canopy (Barbour et al. 2005b). There were 11 complete days when the selection criteria were met. Net canopy CO2 and water vapour exchange were calculated for the tree canopy by subtracting half-hourly estimates of evaporation and adding estimates of respiration from the soil surface (Table 1) using relationships described in DeLucia et al. (2003a). Half-hourly estimates of canopy Ci were estimated using Eqns 1 and 2 and these values were weighted by A and meaned to give daily Ci values. These daily values were further weighted by daily totals of A and meaned to give an estimate of canopy Ci during the measurement period.

Estimate of canopy Ci using carbon isotope ratio of respired CO2

When canopy air is sampled at night, an integrated carbon isotope ratio of CO2 respired by all ecosystem components may be determined from the intercept of a regression of δ 13Ca against the inverse of Ca (Keeling 1958). This technique has been used widely and it produces estimates of ecosystem-respired δ 13CR that compare well with estimates of δ 13C for stored carbon in the ecosystem (Buchmann et al. 1998). Measurements of δ 13C in 14 air samples with a range in Ca were made on each of three nights in summer 2002 and the intercept of the Keeling plot was calculated using geometric mean regression analysis (Pataki et al. 2003). The uncertainty of the regression was reported as the standard error of the model regression intercept. Canopy Ci was calculated from δ 13CR using Eqns 3, 4 and 5 (Table 1).

Recently fixed carbon, with a current canopy Ci is just one of a number of carbon sources contributing to total ecosystem respiration. To allow interpretation of δ 13CR, samples of CO2 respired by four important components were taken for δ 13C analysis. Five replicates each for soil and roots from the upper 200 mm of the soil profile, litter and sunlit canopy foliage were collected and placed in sealed plastic bladders. Between 5 and 10 g (fresh mass) of each component were used for each incubation. Roots were gently separated from the soil, then washed and blotted dry, so some possible wound response was unavoidable. Bladders were tested for leaks and none were found. Air was removed from the bladders, and replaced with CO2-free air then samples were incubated in the dark until the CO2 concentration inside the bladder was greater than 370 µmol mol−1. This required between 25 and 60 min, depending on the type and quantity of material incubated. Air within the bladder was forced through a magnesium perchlorate trap prior to sampling into a pre-evacuated flask and analysed for δ 13C, as described above for Keeling plot samples. By incubating samples in CO2-free air, natural gradients in CO2 concentration and δ 13C composition were removed, resulting in diffusional fractionation (up to 4.4‰) as described by Cerling et al. (1991) and McDowell et al. (2004). The δ 13C of CO2 respired by each component was estimated by subtracting 4.4‰ from measured isotope compositions and were used to partition δ 13CR. The relative contribution of each component to total ecosystem respiration was estimated using IsoSource software (Phillips & Gregg 2003), which calculates all possible combinations of source contributions consistent with isotopic mass balance.


Vertical profiles of Q and D through the canopy

Measurements of Q decreased uniformly with depth in the canopy with less than 10% of the incident Q reaching the ground (Fig. 2a). D within the canopy remained close to the value measured above the canopy down to a height of 9 m above the ground, then decreased with height nearer to the ground surface (Fig. 2b).

Figure 2.

Vertical profiles of (a) mean (± SE) irradiance (Q, 400–700 nm) as a proportion of incident Q and (b) mean (± standard error (SE)) air saturation deficit (D) through the canopy.

Estimates of Ci at the shoot scale

Estimates of Ci at the shoot scale increased with decreasing height in the canopy, based on instantaneous gas exchange measurements of gs and A at ambient Ca, saturating Q and constant temperature of 20 °C. Values of Ci ranged from 288 ± 16 (mean ± SE) µmol mol−1 at the top (22 m above ground; Fig. 3d and Table 2) and 318 ± 11 µmol mol−1 at the bottom of the canopy (10 m above ground; Fig. 3d), consistent with decreasing A with depth in the canopy.

Figure 3.

Vertical profiles of (a) mean δ 13CO2 of daytime canopy air (b) mean CO2 concentration of canopy air (Ca), (c) foliage carbon isotope discrimination (Δ13C) and (d) estimated shoot intercellular CO2 concentration (Ci) based on: (1) instantaneous measurements of A and gs at saturating Q; (2) foliar carbon isotope discrimination (Δ13C) from sampled shoots; (3) instantaneous measurements of A and gs at ambient Q and D, then weighted by As; and (4) instantaneous measurements of A and gs at ambient Q and D, then analysed using Eqn 6. Data are presented as means ± SE. Four sets of canopy air samples taken during the day were used to construct profiles of ambient CO2 concentration and δ 13Cah at a given height above ground (h). The functions giving the best fit to the data are described by Ca(h) = 369.9 + 65.7e–h (r2 = 0.99) and δ 13Ca(h) = −8.16 − 2.54inline image (r 2 = 0.99).

Table 2.  Values of Ci presented as means for all measurement methods at different spatial and temporal scales.
Scale SpatialTemporalMethodCiµmol mol−195% confidence interval (or range in calculated values)
  1. Estimates of data variability are presented as 95% confidence intervals for shoot, A-weighted upper canopy and Keeling plot methods, and as a range for canopy model and eddy covariance methods. Canopy heights for Ci calculation were: shoot and As-weighted upper canopy (22 m), upper canopy (Leuning model; 18–22 m) and all canopy methods (0–22 m). Superscript letters indicate that means are significantly different using Student's t-test, while superscript letters in parentheses indicate that the range in values overlaps the 95% confidence intervals. The subscript s refers to shoot.

  2. C1, intercellular CO2 concentration; A, rate of net photosynthesis.

ShootMinuteGas exchange at saturating irradiance288ab276–300
ShootYearCarbon isotope discrimination287ab269–305
Upper canopyDayGas exchange at ambient irradiance, weighted by As274a270–278
Upper canopyDayGas exchange and Leuning (1995) model297b293–304
CanopyMonthCanopy multilayer model288ab(242–344)
CanopyMonthEddy covariance and soil surface exchange294ab(276–308)
CanopyDecadeCarbon isotope composition of ecosystem respiration, using Keeling plots236c232–239

Both δ 13CO2 (Fig. 3a) and Ca of canopy air (Fig. 3b) were almost constant during the day at heights > 4 m above the ground indicating good mixing in the canopy space. Nearer to the ground, Ca increased markedly and δ 13Ca became more negative, consistent with the release of depleted CO2 during soil respiration. There was a general decrease in Δ13C for foliage with increasing height in the canopy (Fig. 3c). Mean air temperature during the growing season in 2001–02 was 1 °C warmer than that during 2000–01, but total Q was remarkably similar between years (J. Hunt, unpublished data). Values of isotope discrimination for D. cupressinum were not significantly different (P > 0.05) between years at either 18 m or 22 m above the ground (Fig. 3c) and, because of the near constant values of δ 13Ca and Ca at heights above 4 m, the decrease in Δ13C corresponded with a decrease in Ci (Fig. 3c,d). Mean (± SE) Ci determined from carbon isotope discrimination at the top of the canopy (287 ± 9 µmol mol−1 at 22 m; Fig. 3d and Table 2) was nearly identical to that estimated from instantaneous gas exchange measurements of gs and A at the top of the canopy (288 ± 16 µmol mol−1 at 22 m; Fig. 3d and Table 2). However, mean shoot Ci in the lower canopy (8–16 m above ground) estimated by carbon isotope discrimination was consistently higher (range of 300–332 µmol mol−1) than shoot Ci estimated by gas exchange measurements at saturating Q (289318 µmol mol−1; Fig. 3d), most likely because lower shoots did not often experience saturating Q under ambient environmental conditions (Fig. 2a).

Estimates of Ci at the upper canopy scale

The estimates of Ci weighted for daily changes in A, based on the measurements of A and gs at two canopy heights (18 and 22 m above ground) combined over the three-day measurement period, gave mean (± SE) estimates of Ci of 281 ± 4 µmol mol−1 at 18 m (Fig. 3d) and 274 ± 2 µmol  mol−1 at 22 m (Fig. 3d and Table 2). The A-weighted mean value of Ci at 18 m (281 ± 4 µmol mol−1) was not significantly different (< 0.05) from the Ci estimate from instantaneous gas exchange measurements of A and gs at saturating Q (289 ± 12 µmol mol−1), but both were significantly lower (< 0.05) than the estimate from carbon isotope discrimination (315 ± 10 µmol mol−1) at 18 m (Fig. 3d). These differences are likely attributable to rates of photosynthesis at less than saturating Q in shoots at 18 m above ground level in the canopy. The A-weighted mean value of Ci at 22 m (274 ± 2 µmol mol−1) was lower than the mean Ci estimated from gas exchange at saturating Q (288 ± 16 µmol mol−1) and carbon isotope discrimination (287 ± 9 µmol mol−1) at 22 m (Fig. 3d and Table 2), but the differences were not significant (P > 0.05).

Estimating canopy Ci incorporated coupling between A and gs and the response of gs to D when D > 7 mmol mol−1 (Fig. 1a). When data were selected for conditions when Q > 500 µmol m−2 s−1, there was a linear relationship between gs and the right-hand term in Eqn 6 excluding a (Fig. 1b). The mean (± SE) value for upper canopy Ci at 18–22 m calculated from the slope of the line, α, was 297 ± 2 µmol mol−1 (Fig. 3d; Table 2). This value was significantly higher than Ci estimated from A-weighted gas exchange measurements at ambient Q (< 0.05), but not significantly different (< 0.05) from Ci estimated from either gas exchange measurements at saturating Q or from carbon isotope values of shoots at 22 m above ground in the canopy.

Estimates of Ci at the canopy scale

The multilayer canopy model (Whitehead et al. 2004) provided estimates of A and mean canopy Ci on a daily basis. The model was run for a year when mean (± SE) air temperature was 13.9 ± 0.2 °C and mean daily Q was 12.8 ± 0.4 MJ m−2. For the year, mean daily A for the canopy was 0.25 ± 0.01 mol m−2 and mean A-weighted Ci for the whole canopy was 288 µmol mol−1; daily values of Ci ranged from 242 to 344 µmol mol−1 (Table 2). The mean value fell within the 95% confidence intervals of Ci estimated from gas exchange measurements at saturating Q and carbon isotope values at 22 m, but not within the 95% confidence intervals of Ci from the A-weighted gas exchange measurements at ambient Q or from the Leuning model. However, the range in values overlaps all 95% confidence intervals, except Ci estimated from δ 13C of ecosystem respired CO2 (Table 2).

The eddy covariance and soil surface respiration data were used to generate an A-weighted estimate of whole canopy Ci that reflected the dominant contribution of D. cupressinum. During the time periods selected for eddy covariance data analysis, there was a wide range in D from 5.7 to 15.0 mmol mol−1, with a mean value of 9.4 mmol mol−1. Mean (± SE) daily Q was high with little day-to-day variability (23.2 ± 2.4 MJ m−2). Mean (± SE) midday canopy net photosynthesis (i.e. the difference between ecosystem net CO2 exchange and soil surface respiration rate) was 13.6 ± 0.6 µmol m−2 s−1; the proportion of soil surface respiration was ≈ 30% of canopy A. Adopting Eqns 1 and 2, we estimated that the mean A-weighted canopy Ci during the 11-day measurement period in summer was 294 µmol mol−1; and daily mean values of Ci ranged from 276 to 308 µmol mol−1 (Table 2). The mean value fell within the 95% confidence intervals of Ci estimated from gas exchange measurements at saturating Q, carbon isotope values at a height of 22 m and the Leuning model, but not within the 95% confidence intervals of Ci from the A-weighted gas exchange measurements at ambient Q. However, the range in values of Ci overlaps all 95% confidence intervals, except Ci estimated from δ 13C of ecosystem respired CO2 (Table 2).

Estimate of canopy Ci from ecosystem respired CO2

The three Keeling plots exhibited highly significant positive relationships (r2 = 0.99) between δ 13Ca and 1/Ca and the standard errors of the intercepts for individual plots were low (range of 0.14–0.16). The value for δ 13CR determined from Keeling plots was consistent over the three sampling nights. When all samples were included in a single regression analysis, the mean (± SE) value for δ 13CR was −26.3 ± 0.1‰. Using Eqs 4 and 5, and assuming δ 13C for the bulk atmosphere to be −8.0‰ (calculated from the Keeling plot when Ca = 370 µmol mol−1), the ecosystem respiratory value of Ci was estimated to be 236 ± 2 µmol mol−1 (Table 2). This estimate assumed no shifts in isotope composition between photosynthesis and respiration, which may not be realistic (Duranceau et al. 1999; Henn & Chapela 2001; Xu et al. 2004; Klump et al. 2005; Tcherkez et al. 2004).

CO2 respired by ecosystem components varied considerably in stable carbon isotope composition. The most depleted isotopic composition (mean ± SE) was from soil respiration at −28.8 ± 0.4‰, and the least depleted from roots at −25.3 ± 0.7‰, while litter and fresh foliage gave similar results (−27.7 ± 1.4 and −28.1 ± 0.7‰, respectively). Partitioning total ecosystem respiration into contributing components using isotope ratios and a mixing model (Phillips & Gregg 2003) showed the range in values of the proportion of total ecosystem respiratory flux to be 0–0.31 for soil, 0.55–0.74 for roots, 0–0.45 for litter and 0–0.39 for foliage. Clearly, below-ground components dominated ecosystem respiration in this forest.


Canopy gradients in shoot Ci

Shoot Ci increased from the top to the bottom of the canopy in D. cupressinum, as has been shown for many tree canopies (Le Roux et al. 2001; Dawson et al. 2004; Koch et al. 2004; Niinemets et al. 2004). The canopy gradient in shoot Ci is often attributable to lower gs, perhaps in response to higher D at the top of the canopy, and/or higher A due to higher Q in the upper canopy, resulting in reductions in shoot Ci in the upper canopy relative to that in the lower canopy. In Juglans regia, Le Roux et al. (2001) suggested that the increase in shoot Ci in the lower canopy was entirely due to reductions in Q with depth in the canopy. In contrast, water stress reduced Ci in the upper canopy of Populus tremula to a greater degree than canopy Q gradients and that the interaction between Q and water stress was important in determining shoot Ci (Niinemets et al. 2004). In Sequoia sempervirens, a large reduction in shoot Ci was observed at the top of the canopies of very tall trees (85–112 m) compared with the lower canopies, due to very low xylem water potential and reduced gs (Koch et al. 2004). In D. cupressinum, higher shoot Ci with depth in the canopy was associated with a concomitant reduction in Q and to a lesser degree in relation to D. In general, D was constant at heights above 9 m, while most of the variation in shoot Ci occurred at heights above 12 m. In D. cupressinum, gs was sensitive to D and therefore, relatively constant from 9 m to the top of the canopy. Given the variation in shoot Ci above 12 m and the apparent lack of concurrent variation in gs, we conclude that the increase in shoot Ci with increasing depth in the canopy was not due to changes in gs. The gradual, rather than abrupt, increase in shoot Ci with depth in the canopy was more likely due to decreased A at lower Q, as shoots were more shaded lower in the canopy. Therefore, in this rain forest, the canopy gradient in shoot Ci was more likely affected by Q than by D.

Shoot Ci was generally high in D. cupressinum, ranging from 287 µmol mol−1 at the top of the canopy to 332 µmol mol−1 at the bottom of the canopy, due to low A (c.µmol m−2 s−1 at saturating Q for all canopy positions; Tissue et al. 2005) in this nutrient-limited forest (Whitehead et al. 2004). In comparison, shoot Ci in Acer negundo, determined from gas exchange and stable isotope measurements, was lower in well-watered (236–269 µmol mol−1) and water-stressed (221–281 µmol mol−1) conditions than in D. cupressinum, in part because A was five to eight times larger (Dawson et al. 2004). Using values of Δ13C reported elsewhere, and assuming Ca to be 370 µmol mol−1, we used Eqn 5 to calculate shoot Ci at other sites. Shoot Ci at the top of the canopies of Quercus alba, Quercus prinus and Acer rubrum ranged from 260 to 272 µmol mol−1 (Baldocchi & Bowling 2003), while much lower values of shoot Ci were observed in the top canopy of Pinus taeda (238 µmol mol−1; Katul et al. 2000), Picea mariana (229 µmol mol−1), Pinus banksiana (222 µmol mol−1), and Populus tremuloides (252 µmol mol−1; Flanagan et al. 1997). The lowest recorded value of shoot Ci in trees was observed at the top of two 112-m-tall trees of S. sempervirens, based on estimates derived from carbon isotope discrimination (160 µmol mol−1) and instantaneous gas exchange (170 µmol mol−1; Koch et al. 2004). We conclude that high shoot Ci in D. cupressinum reflects both low A and minimal stomatal limitation in this rain forest.

Scaling Ci from shoots to the canopy

Estimates of Ci at the shoot and canopy scales, and across temporal scales from minutes to years, were remarkably similar at our site. The small range in mean Ci observed in shoots at the top of the canopy, and in the upper and whole canopy, indicates that shoot Ci at the top of the canopy was generally representative of Ci integrated for the whole canopy. The foliage area density in D. cupressinum is low in the upper canopy, and gradually increases with canopy depth to reach a maximum value in the mid-canopy and, in midsummer conditions, half of the total canopy A in D. cupressinum occurs in the mid-canopy (Whitehead et al. 2004). Transmittance of Q into the mid- and lower canopies is relatively high, with only 26% of Q intercepted in the upper canopy. Thus, differences in As, gs and foliage nitrogen concentration with depth in the canopy are very small (Tissue et al. 2005). The good agreement between estimates of Ci made at different spatial scales in the canopy must be at least partly attributable to the lack of marked gradients in photosynthetic properties with depth in the canopy (Tissue et al. 2005). Lower shoots spend less time in high Q conditions, but when shoots are fully sunlit Ci will be similar in the upper and lower parts of the canopy.

Temporal scaling of Ci in the canopy was also possible because of the relatively small variation in environmental conditions at the site throughout the year. Temperature variation is low and seasonal changes in volumetric root-zone water content are small and restricted to short periods because of the high rainfall that is evenly distributed throughout the year. Further, daily variation in canopy conductance to water vapour (maximum mean daily value of 0.11 mol m−2 s−1) is small because D is frequently low (Whitehead et al. 2004). While instantaneous Ci at the shoot scale reflects A, gs and environmental conditions immediately prior to the measurements, estimates of shoot Ci from measurements of carbon isotope discrimination gives values integrated across the life of the foliage. This is likely to be up to several years in D. cupressinum because of very slow rates of foliage growth and high foliage longevity (Whitehead et al. 2002). Similarly, Dawson et al. (2004) demonstrated that shoot Ci in A. negundo estimated from instantaneous gas exchange measurements and long-term integrated analysis of foliage carbon isotopes were similar in well-watered conditions. However, during dry periods, estimates of Ci for the Acer trees using the two methods differed, reflecting temporal effects on A. Similarly, differences in the Ci/Ca ratio estimated using short-term gas exchange and long-term carbon isotopes in three boreal tree species also reflected temporal variation in A due to variation in resource availability (Flanagan et al. 1997). Clearly, the low variability in environmental conditions was an important factor contributing to the efficacy of scaling Ci across temporal scales at our site.

Ecosystem respiration estimates of canopy Ci

There was a large difference in the estimate of canopy Ci calculated from δ 13CR (236 µmol mol−1) compared with estimates at the shoot (273–332 µmol mol−1 for the range of mean values at all canopy heights) and canopy (242–344 µmol mol−1) scales. This confirms that A and gs during the measurement period do not alone determine δ 13CR. In a number of ecosystems, large discrepancies have been noted in δ 13C of respired CO2 (i.e. the Keeling plot method we used to estimate canopy Ci) and δ 13C of carbon sources (i.e. the method we used to determine shoot Ci integrated over the lifetime of the shoots). This may be explained by time lags between A and respiration and by shifts in isotopic composition prior to respiration (Pataki et al. 2003; McDowell et al. 2004).

Partitioning of ecosystem respiration into four components (soil, roots, litter and foliage) indicated that below-ground respiration, and in particular root respiration, was the dominant contributor. Root respiration was estimated to contribute between 55 and 74% of the respiration for the ecosystem, and therefore to our δ13CR-generated estimate of canopy Ci. Recently fixed carbon is the most likely source for respiration by roots (Ekblad & Högberg 2001; Knohl et al. 2005), but carbon supplied by shoots to the roots may display shifts in isotope ratio prior to respiration (Schmidt & Gleixner 1998; Duranceau et al. 1999; Klump et al. 2005). Further, the non-statistical distribution of 13C within respiratory substrates (i.e. non-random patterns of 13C enrichment within carbohydrates; Rossman, Butzenlechner & Schmidt 1991) may result in release of 13C-enriched CO2, that is fragmentation fractionation (Tcherkez et al. 2004). While shifts in isotopic composition during transport of carbohydrates and fragmentation fractionation both act to enrich the respired CO2 (i.e. estimates of Ci from δ 13CR would be closer to estimates from other techniques if fractionations were taken into account), the extent of these fractionations is currently poorly understood.

Our estimate of canopy Ci from δ 13CR must relate to true canopy Ci to some extent, as Ci determines δ 13C of carbon fixed by the canopy. Subsequently, after days, weeks or years, this canopy carbon is respired by the ecosystem (Bowling et al. 2002; McDowell et al. 2004; Scartazza et al. 2004; Barbour et al. 2005a). However, variable time lags and shifts in isotopic composition between photosynthesis and respiration make the direct estimation of canopy Ci from δ 13CR difficult. Therefore, δ 13CR may be more appropriate for assessing variation in canopy Ci between sites (Fessenden & Ehleringer 2002; Mortazavi et al. 2005), or between seasons at a single site (Bowling et al. 2002; Scartazza et al. 2004), rather than providing an absolute estimate of canopy Ci at one site.


Gradients in shoot Ci through the canopy have been variously attributed to gradients in Q or D (Le Roux et al. 2001; Niinemets et al. 2004). In D. cupressinum, the gradual increase in shoot Ci with depth in the canopy was more likely attributable to decreases in A resulting from lower Q than to increases in gs or changes in D in the lower canopy. The lack of marked vertical gradients in A and gs at saturating Q through the canopy (Tissue et al. 2005), and the low seasonal variability in environmental conditions, contributed to the efficacy of scaling Ci. Estimates of shoot and canopy Ci across temporal scales from minutes to years were remarkably similar (range of 274–294 µmol mol−1) suggesting much smaller seasonal variation in A and gs in D. cupressinum than is commonly observed in other forests (Schleser 1990; Le Roux et al. 2001; Niinemets et al. 2004). However, an estimate of canopy Ci calculated from δ 13CR (236 µmol mol−1) was much lower than other estimates of canopy Ci. Partitioning δ 13CR into four components (soil, roots, litter and foliage) indicated root respiration was the dominant (> 50%) contributor to δ 13CR. Variable time lags and shifts in isotopic composition between A and respiration make the direct estimation of canopy Ci from δ 13CR problematic and reveals the complexities of scaling Ci from individual dominant trees to the canopy.


This work was funded by the Foundation for Research, Science and Technology and the Marsden Fund (New Zealand), Andrew W. Mellon Foundation (USA), Centre for Environmental Research and Conservation of Columbia University (USA), National Science Foundation International Program (USA) and Black Rock Forest Consortium (USA). We thank Timberlands West Coast Ltd and the Department of Conservation for access to the forest site and logistical support in construction of the canopy access tower. John Byers, Tony McSeveny and Graeme Rogers contributed by designing and building a canopy access tower, providing us with access to the trees. Landcare Research provided David Tissue the resources and an environment conducive to completing this manuscript during his sabbatical visit to New Zealand.


The full derivation of discrimination against 13C during C3 photosynthesis is given by (Farquhar, O’Leary & Berry 1982):

image((A1) )

where Ca, C′, Ci and Cc are the CO2 concentrations in the atmosphere, at the foliage surface, in the intercellular air spaces, and at the sites of carboxylation, respectively. The parameters ab, a, es, a1, b, e and f refer to discrimination associated with diffusion through the boundary layer (2.9‰; Farquhar 1980), diffusion through the stomata (4.4‰; Craig 1954), dissolution of CO2 (1.1‰ at 25 °C; Vogel 1980), diffusion in the liquid phase (0.7‰; O’Leary 1984), carboxylation, dark respiration (Rd) and photorespiration, respectively. k is the carboxylation efficiency and Γ* is the CO2 compensation point in the absence of dark respiration (Brooks & Farquhar 1985). Discrimination by ribulose 1.5-bisphosphate carboxylase/oxygenase has been shown to be about 30‰, but net discrimination during carboxylation must include discrimination during carboxylation by phosphoenol pyruvate carboxylase, so that the true value of b lies between 28.2 and 30‰ (Brugnoli & Farquhar 2000). Most researchers apply the simplification:

image((A2) )

which uses a lower value of b (usually 27‰, rather than 30‰) to include isotopic effects of gm, dissolution into and diffusion through water, and assumes that the effects of e and f are negligible. This simplification also allows interpretation of variation in Δ when gm are significant, but when variation in Δ is driven by variation in stomatal conductance (Brugnoli & Farquhar 2000).

In the strictest sense, Δ-values cannot be used to back-calculate Ci, as it is more closely related to Cc. However, inspection of the relative importance of the various discriminating effects on Δ reveals that when a value of 27‰ is used in Eqn A2, Ci values are almost identical to those estimated from Eqn A1 for the studied trees.

At the photosynthesis-weighted average Ci calculated from gas exchange data (274 µmol mol−1 in the upper canopy), the photosynthetic rate was 3.8 µmol m−2 s−1 (from raw gas exchange data), and using estimates of gm from field-grown conifers (gm = 0.20 mol m−2 s−1 for Pseudotsuga menziesii in the upper canopy; Warren et al. 2003) Cc may be estimated to be 255 µmol mol−1. Furthermore, assuming Ca is 370 µmol mol−1, Cs = Ca for the small, scale-like foliage of Dacrydium cupressinum, and using values from a recent review (Brugnoli & Farquhar 2000) to parameterize Eqn A1[a = 4.4‰, a1 = 0.7‰, es = 1.1‰, b = 30‰, f = 7‰, e = 4‰ (Xu et al. 2004), Γ * = 3.2 (DeLucia et al. 2003b), Rd = 1.0 (DeLucia et al. 2003b), k = 0.27 (Farquhar et al. 1980)], Δ may be calculated to be 21.7‰. The measured value of Δ in the upper canopy is 21.9‰, that is within 0.2‰ of the value estimated from Eqn A1. Estimating Ci from Eqn A2 rather than from the full derivation translates to an error of just 1 µmol mol−1, well within measurement uncertainty. Even if the true value of gm were at the lower end of those observed by Warren et al. (2003) at 0.14 mol m−2 s−1, Δ calculated from Eqn A1 would be 21.1‰, resulting in Eqn A2 underestimating Ci by just 8 µmol mol−1.