We report diurnal variations in 18O discrimination (18Δ) during photosynthesis (18ΔA) and respiration (18ΔR) of Picea sitchensis branches measured in branch chambers in the field. These observations were compared with predicted 18Δ (18Δpred) based on concurrent measurements of branch gas exchange to evaluate steady state and non-steady state (NSS) models of foliage water 18O enrichment for predicting the impact of this ecosystem on the δ18O of atmospheric CO2. The non-steady state approach substantially improved the agreement between 18Δpred and observed 18Δ (18Δobs) compared with the assumption of isotopic steady state (ISS) for the δ18O signature of foliage water. In addition, we found direct observational evidence for NSS effects: extremely high apparent 18Δ values at dusk, dawn and during nocturnal respiration. Our experiments also show the importance of bidirectional foliage gas exchange at night (isotopic equilibration in addition to the net flux). Taken together, neglecting these effects leads to an underestimation of daily net canopy isofluxes from this forest by up to 30%. We expect NSS effects to be most pronounced in species with high specific leaf water content such as conifers and when stomata are open at night or when there is high relative humidity, and we suggest modifications to ecosystem and global models of δ18O of CO2.
The enrichment or depletion of 18O in ambient CO2 during photosynthesis and respiration is dominated by the isotopic exchange of CO2 and water during interconversion between CO2 and H2CO3. The reaction is catalysed by the enzyme carbonic anhydrase and is thus very fast in leaves (Francey & Tans 1987), but slow in environmental water. As CO2 can enter and diffuse back out of the foliage (retrodiffusion) without being assimilated, its δ18O signature will reflect nearly complete isotopic equilibration with the water it contacts in the leaf. The magnitude of 18O fractionation during CO2 exchange between foliage and canopy air thus depends on the isotopic signature of water close to the evaporation sites. Because the δ18O value of foliage water at these evaporating sites, δ18OE, is transferred to atmospheric CO2 (Farquhar et al. 1993), it is important to test how well the existing models predict the 18O enrichment of foliage water under field conditions over the diurnal cycle.
During foliage transpiration, the lighter isotope diffuses faster than the heavier one, thereby enriching the water at the evaporating sites in 18O. The extent of 18O enrichment is controlled by the bidirectional exchange of water vapour between the foliage surface and air, and is thus inversely proportional to relative humidity in the surrounding air. In the steady state description, the value of δ18OE instantly adjusts to any change in environmental conditions. The steady state enrichment model (Craig & Gordon 1965) is widely used to predict δ18OE in modelling studies at ecosystem and global scales (e.g. Farquhar et al. 1993; Ciais et al. 1997a, Ometto et al. 2005). The basic assumption of this model, that foliage water is at isotopic steady state (ISS), however, is not necessarily fulfilled at all times, especially under field conditions (Bariac et al. 1994; Harwood et al. 1998; Cernusak, Pate & Farquhar 2002). Dongmann et al. (1974) developed a non-steady state (NSS) version of the enrichment equation, incorporating the finite turnover of the foliage water pool. In this approach, the time necessary for foliage water to attain an ISS is determined by the rate of bidirectional exchange of water vapour relative to the water content of the foliage. Cernusak et al. (2002) applied such an NSS model to measurements of phloem sap and leaf water δ18O values and found good agreement between the predicted and observed time courses of δ18O signatures. The NSS model was recently modified to additionally include variations in the foliage water volume (Cernusak, Farqukar & Pate 2005; Farquhar & Cernusak 2005).
Here, we explore the effects of the NSS enrichment of evaporating site water on the δ18O signatures of CO2 exchanged between foliage and canopy air under field conditions. We report on diurnal variations in 18O fractionation occuring during photosynthesis and respiration in a stand of Sitka spruce (Picea sitchensis) growing in Central Scotland. The method we applied combines the analysis of air samples collected from branch chambers with measurements of environmental variables and gas exchange in the chambers, representing well-defined, enclosed subspaces within the forest canopy. This approach enabled us to investigate concurrent processes simultaneously. From the chamber measurements of microclimate and gas exchange, we predicted the δ18O signatures of photosynthetic and respiratory CO2 fluxes on the basis of the steady state and NSS assumptions. We then compared these predictions with the observed δ18O signatures of CO2 from the chamber air samples.
MATERIALS AND METHODS
Study site and gas exchange measurements
The study was conducted in Griffin Forest, a plantation of Sitka Spruce (P. sitchensis) located near Aberfeldy, Perthshire, UK (56°37′N, 3°48′W). Wingate (2003) gives a detailed description of the experimental set-up. Branch chambers were installed in the upper canopy at 10.5 and 9.4 m (chambers 1 and 3, respectively), and in the middle canopy at 8.1 m (chamber 4). The latter was used as a control (empty) chamber in July 2001. The chambers were operated on an automated 20 min cycle. Each chamber was open and ventilated for 15 min, after which it was closed for 5 min, and the CO2 mole fraction, relative humidity, photon flux density and temperatures inside were monitored. From these data, rates of net CO2 assimilation (A), stomatal conductance to water vapour (gs), and the leaf surface, intercellular and chloroplast CO2 mole fractions (Cs, Ci, and Cc) were calculated at each 20 min time step for the beginning as well as over the full closure period (see Appendix b for details of the gas exchange calculations).
Collection and analysis of water samples
Needle and non-green twig samples were collected from the same or adjacent Sitka spruce trees. The samples were taken at the same heights and close (≈ 2 m) to the chamber locations. Soil samples were collected in the vicinity of the trees from the top 5 cm of the soil below the litter layer. In the field, all samples were placed in a cooler in sealed glass containers. These were then transferred to a freezer in the lab and stored until further processing. Water was extracted from the samples cryogenically under vacuum and was collected in small glass vials to which a known amount of CO2 was added. Vials were then left for a minimum of 3 d to allow CO2–H2O equilibration. The δ18O of the equilibrated CO2 was measured on a dual inlet isotope ratio mass spectrometer. The overall precision for water δ18O data was 0.4‰. All water δ18O values are reported with respect to Vienna Standard Mean Ocean Water (VSMOW). Extractions and analyses were performed at the University of Cambridge, UK and at the Weizman Institute of Science, Rehovot, Israel.
Collection and analysis of flask air samples
Pairs of air samples from branch chambers were collected at intervals of approximately 3 h over 24 h in spring (18/19 May) and summer (20 July) 2001. Air was circulated from the branch chambers through sampling lines into a flask sampling system (separate from that used for gas exchange measurements) and back into the chamber. Within the sampling system, the air stream was passed through a magnesium perchlorate cylinder to remove water vapour and was pumped through two flasks in series. We used 1 dm3 glass flasks with a valve (Glass Expansion, Melbourne, Australia) on each end, sealed with Teflon® perfluoroalkoxy (PFA) O-rings (Du Pont Fluoroproducts, Wilmington, DE, USA) and 1.3 dm3 flasks with two valves (Louwers, Hapert, Netherlands) on the same end. Flask samples were taken at two points in the opening and closing sequence of branch chambers. The first flask was collected 3–4 min before closure (‘open sample’) and reflects the CO2 mole fraction and isotopic signature of ambient canopy air. The second flask (‘closed sample’) was collected shortly before reopening of the chamber.
All flasks were analysed at the Max Planck Institut (MPI) für Biogeochemie in Jena, Germany. The CO2 mole fractions in the air samples were determined using a gas chromatograph (HP 6890, Hewlett Packard, CA, USA) linked to a methanizer and flame ionization detector. The CO2 in the dry air samples was then extracted cryogenically (‘BGC-AirTrap’, Werner, Rothe & Brand 2001), and its isotope ratio was determined on a dual inlet isotope ratio mass spectrometer (Delta+XL, Finnigan MAT, Bremen, Germany). The analytical precision was in the order of 0.08 µmol mol−1 for CO2 mole fraction and 0.02‰ for δ18O, reported with respect to Vienna-Pee Dee Belemnite (V-PDB)-CO2. The uncertainties of flask data were estimated from the SDs of the laboratory analysis of duplicate flasks, 0.13 µmol mol−1 for CO2 mole fraction and 0.03‰ for δ18O. Additional uncertainties arising from the sampling procedure, applying to samples collected at the end of closure periods, were estimated from control measurements in an empty chamber. They were 1.3 µmol mol−1 for CO2 mole fraction and 0.2‰ for δ18O.
Calculation of δ18O signatures of CO2 exchange from flask data
In analogy to a Rayleigh process, the observed values of 18Δ during photosynthesis [18ΔA,obs (‰)] and nocturnal respiration [18ΔR,obs (‰)] of foliage in the closed branch chamber were determined after Guy et al. (1989):
where Co and Ce are the mole fractions (µmol mol−1), and Ro and Re are the 18O/16O ratios of CO2 at the beginning and at the end of the closure periods, respectively. These correspond to the samples collected from the open and closed chambers for flask measurements, and to the starting time values and those integrated over the closure periods for gas exchange data (Appendix B). The above definition relates isotopic exchanges to the net exchange of CO2, but note that a large part of the apparent discrimination is associated with isotopic equilibration of CO2 which can occur in the absence of net exchange. This equation can therefore yield large apparent observed 18Δ (18Δobs) at times of small net CO2 fluxes. Estimated uncertainties for 18Δobs were calculated using Gaussian error propagation. They were inversely related to net flux rates, usually < 2‰, but > 10‰ at night, dawn and dusk. Discrimination values were also very large at these times.
All calculations were performed in Interactive Data Language (IDL version 6.1, Research Systems Inc., Boulder, CO, USA). Correlation parameters between predicted 18Δ (18Δpred) and 18Δobs were obtained from least absolute deviation regressions using the flask observations as independent variable. Night-time data points were excluded because of the limited number of conductance measurements at night. Four day-time data points were also excluded because of lack of reliable gas exchange measurements [18 May, 0710, infrared gas analyser (IRGA) tubing detached; 0740, relative humidity constant at 50%; 1740, sensor artefacts from direct sunlight; 20 July, 0340, air saturated].
where Ca, δ18Oa and Cc, δ18Oc–eq are the mole fractions and isotopic compositions of CO2 in ambient air and at the site of isotopic equilibration, respectively. The offset between the equilibrated CO2 (δ18Oc–eq) and foliage water (δ18Ocw) depends on leaf temperature T (K) as ɛc–eq = 17 604/T − 17.93 (Brenninkmeijer, Kraft & Mook 1983). In the following, we assumed that the isotopic exchange is limited to the chloroplasts (Cc), that the δ18OE of evaporating site water is a good approximation for δ18Ocw (Farquhar & Lloyd 1993; Cernusak et al. 2004) and that the isotopic exchange is complete (i.e. θ = 1, Gillon & Yakir 2000). The fractionation coefficient for diffusion of CO2 to the sites of isotopic exchange, ā, expresses the mean of the successive diffusion steps through leaf boundary layer (5.8‰), stomata (8.8‰) and the liquid phase (0.8‰), weighted by the CO2 drawdown associated with each step (Farquhar & Lloyd 1993). The boundary layer conductance was estimated from temperature, light and water flux data (1.5 ± 0.5 mol m−2 s−1). The mesophyll conductance required for ā and Cc was estimated from concurrent 13Δobs data (0.16 mol m−2 s−1, see Seibt 2003; Wingate 2003). The calculation of 18Δ from chamber data during flask sampling periods is detailed in Appendix B.
Equations 1 and 2 relate the 18Δ signature to net CO2 exchange. If isotopic exchange occurs without any net CO2 flux, neither of the two descriptions can be applied. For comparison, the isotopic signatures of nocturnal foliage respiration using the (now obsolete) net-flux description (δ18OR,net), corresponding to Eqn 2 for the case where Cc >> Ca, were also expressed as 18Δ values, 18ΔR,net (‰):
The δ18O signature of water at the evaporating sites of foliage
where wa/wi is the ratio of ambient to leaf intercellular vapour mole fraction, and δ18OS and δ18OV are the δ18O signatures of source water and canopy water vapour. The equilibrium fractionation for the liquid to vapour phase transition was calculated from leaf temperature T (K) as Îµeq = exp (1137/T2 − 0.4156/T − 0.0020667) (Majoube 1971). The kinetic fractionation, ɛk, was derived from the isotope effects during water vapour diffusion through the stomata and boundary layer (Farquhar et al. 1993) as Îµk = (32rs + 21rb)/(rs + rb), where rs and rb are the stomatal and boundary layer resistances to diffusion of water vapour, and 32 and 21‰ are their respective fractionation coefficients (Cappa et al. 2003). The simplified expression in Eqn 4 underestimates δ18OISS by ≈ 0.1‰, usually negligible compared with experimental uncertainties (Farquhar & Lloyd 1993; Cernusak, Wong & Farquhar 2003).
The expression for the NSS isotopic composition of evaporating site water, δ18ONSS, follows the time course of the leaf water isotopic composition towards an ISS (Dongmann et al. 1974):
where δ18OISS(t) is the enrichment at time t in an ISS with environmental conditions (Eqn 4), δ18ONSS(t − Δt) is the NSS enrichment at the previous time step, and Δt is the interval between time steps (usually 20 min). NSS calculations were initialized with δ18OISS values on the preceding day at 1600 when NSS and ISS predictions were usually closest. The leaf water turnover time is defined as: τ = V/(gwi), with leaf water volume (V, mol m−2) and conductance to water vapour (g, mol m−2 s−1) combining boundary layer and stomatal components. We estimated V (11.8 ± 0.7 mol m−2, no clear diurnal cycle) from the difference between fresh and dry weight (FW and DW, respectively) of 18 needle samples with respect to the projected leaf areas. Equation 5 is an analytical solution of the differential equation given by Cernusak et al. (2002) assuming constant V (derivation given in Appendix A). Under this condition, the equation is easier to apply because it does not require iterative solution.
Diurnal patterns of branch gas exchange
During 18/19 May and 20 July 2001, we measured photon flux density, relative humidity, temperatures and CO2 mole fraction in branch chambers containing Sitka spruce branches. The chambers were closed for 5 min each on an automated 20 min cycle. From the changes in CO2 and water mole fractions monitored every 5 s during chamber closure, we calculated gs (Fig. 1b) for each closure period and A (Fig. 1a), and all other variables (Cc, ā, etc.) for each 5 s time step within a closure period. When direct gs data was not available because of high humidity at night, the average observed nocturnal gs (0.022 mol m−2 s−1) was used in the calculations. Further diurnal environmental and gas exchange data are presented in Wingate (2003).
We determined a total of 29 values of 18ΔA,obs and 18ΔR,obs from flask sample pairs (Eqn 1). The observed values showed pronounced diurnal variability (Fig. 2). The extremely high dawn and dusk 18ΔA,obs values (≈ 100‰) were in sharp contrast to those during the morning (≈ 10‰) and afternoon (20–60‰). The highest day-time value (126‰) was measured in the mid canopy at noon in May. This and the high dusk and dawn values coincided with low net fluxes (Fig. 1). Extremely large 18ΔR,obs values were also observed at night (up to −260‰). Note that negative 18ΔR,obs values mean that the nocturnal foliage isoflux has a positive sign, i.e. the nocturnal gas exchange results in the 18O enrichment of canopy CO2 like the day-time gas exchange.
The δ18O signatures of bulk needle, twig and soil water
Direct sampling of leaf water within the chambers would have been destructive and was not considered an option in our experiments. Instead, we collected needle samples from adjacent trees during the field campaigns (Fig. 3). The average bulk water δ18O from twig samples (Table 1) were used to define the source water composition δ18OS (Eqn 4). Efforts to measure water vapour δ18O proved unsuccessful for this study. The water vapour composition δ18OV was assumed to reflect isotopic equilibrium with that of precipitation from recent rain events, captured in water samples from the soil surface, δ18OSW (Table 1). The δ18O signatures of bulk needle water had only small diurnal variations (5–8‰). The difference between needle and twig water δ18O was large and surprisingly stable (≈ 10‰), even at night. These differences were established over small distances as twigs and needles were usually sampled together, i.e. the attached needles were separated from the respective twigs.
Table 1. Average values of parameters used in the simulation of δ18OE and 18O discrimination (18Δ) for 18/19 May and 20 July 2001: day-time canopy air temperatures, the δ18O values of bulk twig and soil water samples, the δ18O composition of canopy water vapour estimated from soil water δ18O, the isotopic composition of canopy CO2, δ18Oa and the flux-weighted diffusional fractionation, ā, for the upper (u) and middle (m) branch chambers
Daily average values of:
18/19 May 2001
20 July 2001
Day-time canopy air temperature
Twig bulk water δ18O
−7.9 ± 1.6‰
−6.9 ± 0.4‰
Soil bulk water δ18OSW
−6.9 ± 1.5‰
−8.5 ± 0.7‰
Water vapour δ18O, δ18OV
Day-time δ18Oa of canopy CO2
1.3 ± 0.1‰
0.9 ± 0.1‰
7.3‰ (u), 6.2‰ (m)
Predictions of the δ18O signature of evaporating site water
Compared with controlled laboratory experiments, plants in the field usually experience fluctuating environmental conditions. Thus, their foliage water may rarely reach ISS, the fundamental assumption of the Craig & Gordon (1965) model normally used to predict the δ18O composition of leaf water. To examine the role of NSS effects, the δ18O signature of evaporating site foliage water was calculated on the basis of the chamber measurements (gs, etc.) and water δ18O data (Table 1) for each closure period assuming steady state (δ18OISS, Eqn 4) and NSS (δ18ONSS, Eqn 5) conditions. Typical maximum δ18ONSS values were ≈ 5‰ lower and shifted towards the afternoon compared with those of δ18OISS (Fig. 3). In July, changes in environmental conditions were gradual, with a clear maximum of predicted δ18OISS at noon. The May sampling day (not shown) had more fluctuations in δ18OISS on timescales of around 1 h as a result of rapid changes in cloud cover. These were effectively smoothed in δ18ONSS because of the dependency of NSS enrichment on leaf water turnover. In both months, predicted δ18OISS values increased rapidly by 5–10‰ within 1–2 h after dawn, whereas increases in δ18ONSS were much smaller as a result of the low transpiration rates usually found in the morning. On the other hand, δ18ONSS stayed at a higher level of enrichment in the evening and during the night when δ18OISS decreased in the absence of evaporative enrichment. Thus, predicted δ18ONSS started rising from already enriched signatures compared with δ18OISS at dawn.
Predictions of 18Δ during foliage gas exchange
With the chamber gas exchange data (A, Cc, etc.) and δ18OISS or δ18ONSS, we predicted the foliage 18ΔA (Eqn 2) and 18ΔR,bi (Eqn 2), and the (now obsolete) 18ΔR,net (Eqn 3) for each 5 s time step as described in Appendix B. We also calculated the isotopic signature of chamber air (δ18Oa) on the same time step. The δ18Oa values were initialized with constant δ18Oa corresponding to their averages from open branch chamber data (Table 1). For the flask sampling periods, we used the δ18Oa observed in the open branch chambers. We produced two sets of data: One consisted of data at the beginning of closure periods, thus corresponding more closely to the conditions of branches in the absence of chambers. These starting time values of 18Δ (lines in Fig. 4) were used for scaling branch data to the ecosystem (Tables 1 & 2), but they are not exactly comparable to 18Δ measured using the flask samples of chamber air (symbols in Figs 3 & 4). Therefore, we calculated a second set of integrated 18Δ by simulating the flask filling during the chamber closure periods (Appendix B). By incorporating the feedbacks of changing chamber conditions on isotopic gas exchange as captured in the flask samples, this set of 18Δ predicted from gas exchange measurements could then be quantitatively compared with flask 18Δobs data (Fig. 5). To illustrate the importance of integrating values over the chamber closure periods (Appendix B), the integrated 18ΔR,bi values for 20 July, 2001 were 50 and −65‰ (ISS and NSS, respectively), less extreme than the starting time values of 69 and −88‰ (ISS and NSS, respectively).
Table 2. Daily foliage isofluxes (= 18Δ × flux, per m2 ground area) estimated from branch chamber gas exchange data (time step, 20 min) integrated over 24 h for 19 May (upper rows) and 20 July (lower rows) 2001. The δ18O signatures of evaporating site foliage water were calculated in isotopic steady state (ISS) and non-steady state (NSS) versions. Net canopy isofluxes (ΔNFN, foliage only) reflect combinations of day-time photosynthetic (ΔAFA) and nocturnal respiratory (ΔRFR) isofluxes. Net canopy and nocturnal respiration isofluxes are based on either the net-flux (ΔR,net, Eqn 3) or bidirectional (ΔR,bi, Eqn 2) approaches to nocturnal foliage gas exchange
The diurnal curves of 18Δ predictions using δ18ONSS (18ΔNSS) closely followed that of the ratio of retroflux to net uptake of CO2[Cc/(Ca − Cc), Fig. 4a], whereas the diurnal 18ΔISS curves show the opposite pattern because of their negative values in the morning, evening and at night (Fig. 4b). Note that the change in sign between day and night 18ΔNSS is a result of the change in sign of the net CO2 flux. During the majority of the day (≈ 0600–1500), the differences between 18ΔA,ISS and 18ΔA,NSS were small (1–2‰). In contrast, they differed by 40–200‰ at dawn, dusk and during the night.
Comparison of simulations and observations
The bulk water δ18O data were in better agreement with the δ18ONSS values than with those assuming steady state, δ18OISS, especially in the evenings and at night (Fig. 3). The δ18ONSS values had small diurnal variations (≈ 10‰) similar to those of bulk water δ18O data (5–8‰), whereas the steady state predicted δ18OISS had much larger diurnal variations (≈ 20‰). Bulk water δ18O data were generally lower than evaporating site water δ18O values, as expected because of the Péclet effect (Farquhar & Lloyd 1993; Barbour & Farquhar 2003). A more quantitative comparison between bulk and evaporating site δ18O values was not possible as the needle samples for bulk water analyses were collected outside the chambers.
The prediction, 18ΔNSS, were generally in better agreement with 18Δobs than those assuming an ISS, 18ΔISS (Fig. 4b). During the day, this was most obvious at dusk, dawn and in the late afternoon, when only the 18ΔNSS predictions accounted for both the magnitude and sign of 18ΔA,obs. The consequences of NSS effects were even more pronounced at night. Clearly, the NSS approach combined with the bidirectional formulation of nocturnal foliage gas exchange (18ΔR,bi, Eqn 2) was required to account for the large negative 18ΔR,obs values found at night. In most cases, the steady state assumption did not even predict the sign of the observed 18ΔR,obs correctly. For example, the integrated values of 50 and −65‰ (ISS and NSS, respectively) for 18ΔR,bi can be compared with the 18ΔR,obs value of −40‰ (20 July, 2001). At the same time, the previously used net flux description (18ΔR,net, Eqn 3) predicted much smaller values, 12 and 1‰ (ISS and NSS, respectively), and thus failed to account for the magnitude of the observed values.
The combined observations from all sampling campaigns were compared with the 18Δpred in Fig. 5. Dawn and dusk samples (18ΔA,obs of 92–126‰) were representative for less than 10% of the photosynthetic period but constituted almost one-third of all samples because of our sampling strategy. To reduce any bias introduced by this, the correlation parameters and SEs of 18Δpred versus 18Δobs were obtained using the least absolute deviations method. We found the 18Δpred assuming NSS enrichment of foliage water at the evaporating sites in better agreement with the observations (18ΔNSS, slope: 0.81 ± 0.13, intercept: 3.6 ± 9.5, R2: 0.74) than predictions assuming an ISS (18ΔISS, slope: 0.26 ± 0.15, intercept: 12.0 ± 7.8, R2: 0.45).
Daily foliage isofluxes
The daily canopy isofluxes (= 18Δ × flux, foliage only) were estimated by integrating chamber flux measurements and their calculated 18Δ values over 24 h (Table 2). The canopy isoflux (ΔNFN) combines those of day-time photosynthesis (ΔAFA) and nocturnal respiration (ΔRFR, ΔR,bi/net from Eqn 2 or 3). For our field days, assuming ISS would result in underestimation of net canopy isofluxes by ≈ 30% when combined with the bidirectional version of nocturnal foliage gas exchange. Approximately half of that occurs due to NSS-related changes in δ18OE, i.e. the difference between negative δ18OISS and positive δ18ONSS values, particularly at night (Fig. 3); the other half is caused by bidirectional exchange-related amplification of 18ΔR, i.e. the difference between Eqns 2 and 3.
This study aims to provide an evaluation of the steady state and NSS models of foliage water 18O enrichment under field conditions. We assumed that the discrimination theory (Farquhar & Lloyd 1993) would capture the variability of photosynthetic and respiratory 18Δ values in the field, given measurements of the CO2 and water mole fractions and flux rates and estimates of the 18O enrichment of water at the evaporating sites of foliage. By comparing 18Δ predictions based on the two enrichment models with flask observed values of 18Δ, we tested how important NSS effects on foliage water δ18O are in shaping the diurnal patterns of δ18O signatures of foliage CO2 fluxes.
We found that foliage 18Δ was determined by the interplay of three related factors: the ratio of retrodiffusion to net flux of CO2[Cc/(Ca − Cc)], the NSS effects of foliage water turnover on the 18O enrichment of evaporating site water, and the effects of nocturnal gs on the isotopic exchange of CO2 and water between the foliage and canopy air at night. The main results of this study were (1) the direct observational evidence for NSS effects; (2) the extremely high apparent 18Δ values in the morning, evening and at night; and (3) the persistently enriched δ18O values of bulk leaf water at night. We found that the NSS approach (Dongmann et al. 1974) substantially improved the agreement between 18Δpred and 18Δobs (Fig. 5) compared with the ISS assumption (Craig & Gordon 1965) for the δ18O signature of evaporating site water.
The relative humidity of canopy air controls 18Δ through both the steady state level of foliage water enrichment and gs, which limits the rate of CO2 retrodiffusion and the approach to steady state (Eqn 5). While steady state foliage water enrichment could be calculated without gs data, measurements of gs during the day and at night now turn out to be crucial in estimating NSS foliage water 18O enrichment and thus the isotopic impact of foliage gas exchange on atmospheric CO2. On the diurnal timescale, the retrodiffusion ratio was more important in determining the shape of the 18Δ curve (Fig. 4) than variations in leaf water enrichment. Thus, diurnal variations in the retrodiffusion ratio need to be considered, for example, in ecosystem scale studies. On the longer timescales relevant for global scale studies, leaf water enrichment played a bigger role. For example, the daily mean photosynthetic 18ΔA values (Table 3) were 12‰ smaller in July, when both CO2 and water fluxes were lower, compared with May, with the effects of less enriched leaf water outweighing those of lower internal CO2 gradients. For both months, our estimates of daily mean 18ΔA were 2–3‰ higher when taking NSS effects into account (Table 3) because of the higher levels of enrichment at the evaporating sites already present in the mornings. Firstly, this indicates that the isotopic exchange between enriched leaf water and depleted canopy water vapour, scaled by gs, can be slow enough to cause needle water δ18O values to stay enriched until morning. Consistent with the nocturnal stomatal exchange, Grace, Malcolm & Bradbury (1975) reported that the stomata of Sitka spruce needles were open in the dark at high relative humidity. Secondly, this and the large difference between needle and twig water δ18O at night also indicate that the bulk needle water does not return to the depleted δ18OS at night, consistent with a large longitudinal Péclet number (Farquhar & Gan 2003).
Table 3. Daily flux weighted values of photosynthetic 18O discrimination (18Δ) (Eqn 2) for 19 May and 20 July 2001 assuming isotopic steady state (18ΔISS) or non-steady state (18ΔNSS) for the enrichment of water at the evaporating sites of foliage
Mean 18ΔA (‰)
19 May 2001
20 July 2001
The biggest uncertainty in the simulations of δ18OE and 18Δ was probably introduced by the lack of canopy water vapour δ18O measurements because of technical difficulties during the collection of water vapour samples. To illustrate this, we calculated that a 1‰ shift in δ18OV would lead to errors of ≈ 0.4‰ in δ18OE, and ≈ 0.9‰ at dawn and dusk. Diurnal variations in δ18OV have been mainly attributed to transpiration fluxes (Harwood et al. 1999) which were small in this study. Nevertheless, δ18OV is an important variable and attempts to collect δ18OV data should be made in future studies. Gradients in leaf water δ18OE along the needles should not affect the (whole-leaf integrated) 18Δ under steady and approximately uniform gas exchange conditions (Farquhar & Gan 2003). It has also been shown that δ18ONSS calculations are not sensitive to variations in V (Farquhar & Cernusak 2005). Control measurements using an empty chamber confirmed that the isotopic exchange between CO2 and water was not affected by condensation in the chambers or sampling lines.
Other, minor caveats include the contributions of woody tissue to the net and isotopic gas exchange of CO2 and water. We estimated typical bark fluxes in the order of 1–5% of foliage fluxes at low light, and even less during the rest of the day. Lastly, if the carbonic anhydrase catalysed isotopic exchange between CO2 and leaf water is incomplete, using Eqn 2 leads to an overestimation of 18Δ (Farquhar & Lloyd 1993). For coniferous trees, isotopic exchange has been assumed very close to complete (Wang et al. 1998). To illustrate how a potentially incomplete isotopic exchange would affect 18Δ, we calculate that applying a coefficient (θ) of 0.9 would decrease 18Δ by 1–2‰ (Gillon & Yakir 2000) compared with full isotopic equilibrium of CO2.
Canopy isofluxes and implications for models
Because the steady state equation tends to overpredict the foliage water 18O enrichment in the morning and to underpredict it in the afternoon, it could be argued that the errors associated with NSS effects might cancel over the course of the day. This was not the case in this study. As the higher levels of enrichment reached in the afternoon persisted through the night, the nocturnal foliage gas exchange contributed substantially to the increase in the net isoflux (Table 2). We expect these effects to be large under conditions of high relative humidity, for species with high specific leaf water content, and when stomata are open at night. This might occur in boreal and tropical forests, but also in all other types of ecosystems during times of high relative humidity. Thus, the two effects described here, NSS-related changes in leaf water δ18OE and bidirectional exchange-related amplification of 18ΔR, both have important implications for global modelling (Farquhar et al. 1993; Ciais et al. 1997a, Cuntz et al. 2003a) and for the interpretation of field measurements of the δ18O of CO2 (Bowling et al. 2003; Ometto et al. 2005). In the context of bidirectional exchange, the feedbacks between foliage 18Δ and the δ18O of canopy CO2 (Eqn 2) also become important, particularly at night. We suggest that NSS effects and the explicit description of an interactive canopy air space should be included in ecosystem or global studies using the δ18O signatures of CO2.
We observed pronounced diurnal variations in 18ΔA and 18ΔR by Sitka spruce branches in branch chambers deployed in the field. We used the ISS and NSS models of foliage water 18O enrichment to estimate 18Δ based on the theoretical approach of Farquhar & Lloyd (1993), and compared these estimates with the field observations. Predictions of 18Δ assuming ISS did not agree well with our data at most times, highlighting the limitations of the Craig & Gordon (1965) model widely used in numerical models. Taking the gradual development of 18O enrichment throughout the day into account (Dongmann et al. 1974) substantially improved the agreement between 18Δpred and 18Δobs. The direct observational evidence for NSS effects in this study comprises, firstly, the extremely high apparent signatures of CO2 fluxes (18ΔA/R,obs) in the morning, evening and at night, and secondly, the small diurnal amplitude and enriched night-time values of bulk needle water δ18O. We conclude that NSS effects on foliage water enrichment can play an important role in determining the δ18O signatures of photosynthetic and respiratory CO2 exchange between foliage and canopy air. NSS effects are probably most pronounced in species with low transpiration rates or with high specific leaf water content such as conifers, and when stomata are open at night or when there is high relative humidity. This might apply, for example, to large parts of boreal ecosystems and thus plays a role in shaping the interhemispheric gradient in δ18O of CO2. It is therefore important to include NSS effects in global simulations of the δ18O of atmospheric CO2.
We wish to thank H. Geilmann, A. Jordan, M. Rothe and R. Werner for carrying out the analyses of air and organic samples at the MPI für Biogeochemie in Jena, Germany; N. Betson, G. Lanigan and H. Griffiths for analyses of water samples in Cambridge, UK; and D. Hemming and D. Yakir for extractions and analyses of water samples in Rehovot, Israel. We thank R. Clement, V. Finlayson, S. Patiño, F. Ripullone, J. Schmerler and A. Zerva for help in the field. We are grateful to W. Brand and M. Cuntz for advice, to S.C. Wong and J. Severinghaus for loan of equipment, and to M. Heimann for continued support and discussions. We also thank D. Yakir and the anonymous reviewers for their valuable comments. This work was partly supported by a Marie Curie International Fellowship to US (MOIF-CT-2004-2704) and a Natural Environment Research Council Studentship to LW (GT04/98/93/TS).
Non-steady state (NSS) enrichment for constant leaf water volume (V, mol m−2)
We consider the isotopic ratio of evaporating site foliage water (Re) as described for isotopic steady state (ISS) (Craig & Gordon 1965; Farquhar & Gan 2003) and NSS conditions (Dongmann et al. 1974; Bariac et al. 1994; Farquhar & Cernusak 2005). At constant V (mol m−2), the flux of water into the leaf equals that out of the leaf, i.e. the transpiration rate (E, mol m−2 s−1). The isotopic ratio of the flux of water into the leaf (Rs) is assumed to be constant; that of the transpiration flux out of the leaf (Rt) can vary. The change in Re over time can then be described by (Dongmann et al. 1974)
and defining a turnover time, τ = V(1 − wa/wi)/E[or, since E = g(wi − wa), τ = V/(gwi)], this yields
In the case of ISS, dRe/dt = 0, and Eqn A1.4 is simplified to
i.e. the Craig-Gordon equation (Craig & Gordon 1965). If evaporating site enrichment is not at an ISS, but transpiration rate can be assumed constant, then Eqn A1.4 can be solved to give a time-dependent expression for the enrichment:
or integrated over a time step, Δt, in the terms of Eqn 5 of the main text:
so that for small Δt or large τ, e–Δt/τ approaches 1 and δ18ONSS ≈ δ18Ot−1NSS, i.e. no change in leaf water enrichment, and for large Δt or small τ, e–Δt/τ approaches 0 and δ18ONSS ≈ δ18OISS, i.e. maximum (steady state) leaf water enrichment. Note that permil values such as ɛ, ā, etc. need to be divided by 1000 when they occur in terms like (1 − ɛ) or (1 − ā).
Analysis of closed-chamber measurements
Gas exchange in closed chambers will lead to transient changes in microenvironmental conditions, air composition and fluxes during the closure period. Flask samples collected at the beginning and at the end of chamber closure periods integrate over such changes. Also, the air flushed through the flasks and back into the chambers was dried, weakening the increase in chamber water vapour content. We take both into account when comparing flask data with 13Δ and 18Δ predictions based on gas exchange data. Here, we demonstrate how chamber measurements (5 s time step) can be integrated during closure periods to obtain 13Δ and 18Δ predictions directly comparable with observations from the flask sampling system. The quantitative descriptions are based on one example, chamber 1 on 20 July 2001, 1440–1445. The data processing described in the following were implemented in an IDL program.
Photon flux density was fairly stable during chamber closure at 127 ± 3 µmol m−2 s−1. Air and leaf temperature were similar at 14 ± 0.7 °C. Stomatal conductance to water vapour (gs) of P. sitchensis has been shown to adjust slowly (up to 45 min) to changes in environmental conditions (Ludlow & Jarvis 1971). Therefore, we assumed that to a good approximation, gs remained constant during chamber closure (5 min). Preliminary analyses highlighted a lack of sensitivity in our set-up to monitor changes in the H2O mole fraction using the infrared gas analyser (IRGA). Instead, relative humidity data (h) was used to calculate the vapour mole fraction of the enclosed air [wa(h), Fig. 6a]. During flask sampling periods, the air vapour mole fraction was corrected for the flow of dry air (3 dm3 min−1) returned from the flask sampling system [wa(corr)]. Assuming saturated air at leaf temperature yielded a leaf vapour mole fraction (wi) of 16.4 mmol mol−1. The total g (0.051 mol m−2 s−1 for this period) was obtained from the increase in air vapour mole fraction [wa(corr) from 9 to 14 mmol mol−1; h from 56 to 78%] by fitting
to leaf and air vapour mole fraction data at time t during the closure periods, where wa0 is the initial air vapour mole fraction at the starting time t0. La and Va are the leaf area and air molar volume enclosed in the chamber. The calculated transpiration rate (not shown) decreased from 0.4 to 0.1 mmol m−2 s−1 with decreasing vapour mole fraction deficit during the closure period. At times of a negative leaf to air mole fraction difference (e.g. at night), the mean value of conductance measured under non-saturated conditions at night was used.
The CO2 mole fraction of chamber air (Ca, Fig. 6b) decreased from 350 to 275 µmol mol−1 over the closure period. We assumed that the concurrent change in the rate of net CO2 assimilation (A, Fig. 6b) was on the linear part of the CO2 response curve (Ludlow & Jarvis 1971). The Ca data were fitted with a quadratic equation, its derivative yielding a linear approximation of A (most R2 > 0.99). The decrease in A obtained from changes in Ca (from 5.1 to 3.4 µmol m−2 s−1) compared well with that estimated from the normalized slope of the linear part of the CO2 response curve (from 5.1 to 3.6 µmol m−2 s−1). Because Ca and A varied concurrently, the change in the intercellular CO2 mole fraction (Ci, Fig. 6b) calculated from A and g incorporating ternary effects (Jarman 1974) was less pronounced than that of Ca, leading to a small increase in Ci/Ca over the closure period (Fig. 6c).
Using the δ18Oa and δ13Ca values observed in the open chambers as starting point values, we then calculated δ18Oa(pred) and δ13Ca(pred) (Eqn 1) from 18Δ (Eqn 2) and 13Δ for each time step. The instantaneous 13Δ value increased from 17 to 18‰, whereas that of 18Δ decreased from 21 to 16‰ (Fig. 6c). The calculated δ13Ca(pred) and δ18Oa(pred) increased from −7 to −2‰ and from 1 to 7‰, respectively (Fig. 6d). Weighted averages of δ13Ca, δ18Oa and Ca corresponding to the mixture of air collected in the ‘closed’ flask sample were obtained using exponentially increasing weights for data points prior to the pressurizing phase and equal weights thereafter, each contributing half of the air sample, consistent with flask filling tests conducted at the Max Planck Institut (MPI) for Biogeochemistry, Jena, Germany. The integrated 13Δ and 18Δ values were then derived from the δ13Ca, δ18Oa and Ca at the start, and their weighted averages at the end of chamber closure periods. This approach is equivalent to the calcuations using flask data, making both methods directly comparable. The integrated values were ≈ 0.3‰ higher than starting point values for 13Δ and ≈ 2.1‰ lower for 18Δ. For the example used here, the integrated 13Δ and 18Δ values of 16.9 and 19.1‰ were in better agreement with the flask observed values of 17.0 and 16.9‰ than their respective starting point estimates of 16.6 and 21.2‰.