Diurnal variation of Δ13CO2, ΔC18O16O and evaporative site enrichment of δH218O in Piper aduncum under field conditions in Trinidad

Authors


K. G. Harwood Fax: (44) 191 2225229; e-mail: k.g.harwood@newcastle.ac.uk

Abstract

Concurrent measurements of gas exchange, instantaneous isotope discrimination (Δ) against 13CO2 and C18O16O, and extent of 18O enrichment in H2O at the evaporative sites, were followed in a tropical forest pioneer, Piper aduncum, on two different days in Trinidad during February 1995. Δ13CO2 differed from that predicted from measurements of internal:external CO2 concentration (Ci/Ca) and showed a wide range of values which decreased throughout the course of the day. Derivation of Cc (the CO2 concentration at the carboxylation site) was not possible using carbon isotope discrimination under field conditions in situ and was derived assuming a constant value of internal transfer conductance (gw). Under low rates of assimilation the derived Cc/Ca, like Ci/Ca, remained relatively stable over the course of both days and ΔC18O16O followed evaporative demand. Lower values of ΔC18O16O on day 2 occurred in response to the indirect effect of increased leaf-to-air vapour pressure deficits (VPD) and reduced stomatal conductance. For the first time, direct determination of the δH218O of transpired water vapour (δt) allowed derivation of evaporative site enrichment without the prerequisite of isotopic steady state (ISS) defined in the Craig and Gordon model. Generally, δt was less enriched than the source water (δs) in the morning and more enriched in the afternoon, which would be predicted from an increase and decrease in ambient VPD, respectively. On both days, leaves of P. aduncum approached ISS (indicated where δtδs) between 1300 and 1500 h. Evaporative site enrichment was maintained into the late afternoon, despite a decrease in ambient VPD. The data presented provide a greater insight into the natural variation in isotopic discrimination under field conditions, which may help to refine models of terrestrial biome discrimination.

INTRODUCTION

Plant processes serve a unique role as integrators and indicators of global environmental change (Yakir et al. 1993). Isotopic analysis of atmospheric 13CO2 (Mook et al. 1983; Broadmeadow et al. 1992; Ciais et al. 1995), C18O16O (Francey & Tans 1987; Farquhar et al. 1993; Ciais & Meijer 1997; Ciais et al. 1997) and H218O (Salati et al. 1979; Jacob & Sonntag 1991) has suggested that the terrestrial biosphere exerts a substantial influence on global cycles of CO2 and water. Attention has, therefore, focused on the mechanisms controlling fractionation in biological systems. Intuitively, terrestrial photosynthetic gas exchange will provide a large biospheric contribution to global biogeochemical cycles, along with soil respiration and aquatic photosynthesis. In order to assess any contribution by terrestrial photosynthesis, we must first understand the processes governing fractionation and the natural variation in biological discrimination.

During photosynthetic gas exchange, a combination of diffusive and enzymatic discrimination against the heavier 13C leaves plant material depleted in the heavier isotope relative to atmospheric CO2 (Farquhar, Ehleringer & Hubick 1989). Additionally, through rapid exchange between CO2 and water in the chloroplast, the 18O leaf water isotope signal is transferred to atmospheric CO2 (Francey & Tans 1987; Farquhar & Lloyd 1993; Yakir et al. 1994) and provides a non-destructive probe to investigate terrestrial– atmospheric exchanges of CO2 and water vapour (Farquhar et al. 1993; Yakir & Wang 1996). The magnitude of discrimination expressed by the leaf is a result of both leaf water enrichment and the diffusional resistance of CO2 from ambient air to the chloroplast (Farquhar & Lloyd 1993; Flanagan et al. 1994). The latter is represented by the expression Cc/Ca, the concentration of CO2 within the chloroplast of a leaf relative to that in ambient air, and is a key intracellular parameter which represents leaf photosynthetic gas exchange and is important for interpreting oxygen and carbon isotope discrimination of CO2 during photosynthesis. Estimation of Cc is possible by analysing the difference between observed instantaneous carbon isotope discrimination and that predicted from concurrent measurements of gas exchange [ΔiΔobs; Evans et al. (1986); von Caemmerer & Evans (1991)].

Measurements of photosynthetic response and isotope discrimination under steady-state conditions are necessary to gain an understanding of the interaction between physiological and biochemical processes that occur during photosynthetic gas exchange. Because response times of gas exchange vary with light and water stress, the natural variation in environmental conditions under some field situations may mean that photosynthetic steady state is rarely achieved. Stomatal conductances can take 20 min to achieve steady state following changes in irradiance (Barradas & Jones 1996) and have been reported to take between 20 and 40 min for Piper auritum and P. aequale (Tinoco-Ojanguren & Pearcy 1992). However, field data have shown that despite large shifts in A and gs, Ci/Ca can remain relatively constant throughout the day [see Tenhunen et al. (1984), Wise et al. (1991) and references therein] from which simple discrimination models would predict a constant Δ13CO2 (Farquhar et al. 1989). This suggests that steady-state discrimination may occur without the prerequisite of steady-state photosynthesis.

However, initial observations of diurnal changes in instantaneous discrimination for P. aduncum (a tropical pioneering tree) under field conditions in Trinidad revealed large variability in the measured discrimination signal compared with that predicted (M. S. J. Broadmeadow, unpublished field observations; Gillon et al. 1997). Recent laboratory work has suggested that under low rates of assimilation, changes in (photo)respiratory CO2 evolution can affect the on-line signal (Gillon & Griffiths 1997). Models predicting discrimination are essential for understanding both net and gross exchanges of CO2 between the biosphere and the atmosphere. However, whether the observed oxygen and carbon discrimination in CO2 follows that predicted, despite large changes in A and gs under field conditions, has not been extensively investigated for 13CO2 and only under laboratory conditions for C18O16O.

Additionally, estimation of 18O leaf water enrichment at the sites of evaporation (δe) is important in analysing isotopic leaf water budgets. Using the Craig & Gordon (1965) model of evaporative enrichment for hydrological systems, the enrichment of water at the evaporating surfaces of a leaf can be expressed as the proportion of H218O evaporated relative to H216O (White 1988; Flanagan, Comstock & Ehleringer 1991a; see Appendix A).

Whilst the model is useful for estimating the degree of enrichment at the evaporation sites (δe), it is, by derivation, dependent on the leaf being at isotopic steady state (ISS), the point at which transpired water vapour has the same oxygen isotope composition as the stem water. The time for leaf water to attain ISS is proportional to leaf water turnover rate [defined as the ratio of leaf water volume to transpiration rate, V/E: Farris & Strain (1978); Flanagan, Bain & Ehleringer (1991b); Wang & Yakir (1995)] and the magnitude of the change in environmental conditions that the leaf is exposed to (Flanagan et al. 1991b; Yakir et al. 1994). Under transient environmental conditions, we cannot always assume that a leaf is at ISS (Wang & Yakir 1995). Direct measurement of the δ18O of transpired water vapour during photosynthetic gas exchange would, therefore, allow: (i) the identification of proximity to ISS; (ii) an estimation of evaporative site enrichment irrespective of ISS; and (iii) provide a greater insight into the range of δ18O water vapour signal transferred to the surrounding environment, which may have important implications for scaling up of water transfer from leaf to canopy to atmosphere.

The objectives of this investigation were to follow photosynthetic gas exchange, discrimination against 13CO2 and C18O16O along with the extent of evaporative site enrichment in H218O, under field conditions for a deciduous rainforest in Trinidad. In particular, the natural fluctuations throughout the day were examined, irrespective of whether steady-state photosynthetic gas exchange or isotope discrimination were attained. From this we hoped to ascertain whether steady-state photosynthesis was a prerequisite for steady-state isotopic discrimination, as well as evaluating whether discrimination models reliably predict discrimination, and therefore Cc, under natural conditions in the field.

MATERIALS AND METHODS

Experiments were conducted in the field, at the Asa Wright, Simla Research Station, Trinidad [see Borland et al. (1993) for a full site description] on 2 d in February 1995. An elevated bamboo platform was constructed so that the upper leaves of the P. aduncum canopy (those exposed to full sunlight) could be reached. Leaves selected were fully expanded and located at the same height and position within the canopy. To gain a representative sample of gas exchange and discrimination characteristics, random leaves were used on day 1 (8 February 1995). On day 2 (16 February 1995), the same three leaves were measured through the day to account for any effects of leaf-to-leaf variation.

Additionally, three leaves were collected for δ18O leaf water determination, at 0930 and 1230 h on the first day and placed into Exetainers (Europa Scientific, Crewe, UK). Spring water was collected from a ground water spring within 0·5 km of the field site. Rain water was collected in a rain gauge, positioned in a clearing, fitted with a funnel to minimize evaporation and the oxygen isotope composition was analysed for each major precipitation event, throughout the field campaign. Volumes of 1 cm3 of collected rain or spring water were injected into pre-evacuated Exetainers (Europa Scientific, Crewe, UK).

Photon flux and temperature

Instantaneous photon flux density (PFD) was followed throughout the day using the sensor attached to the portable infra red gas analysis system (IRGA) used for gas exchange measurements (see below). Measurements of total daily PFD were also recorded using an integrating light meter (Delta T, Cambridge, UK). Shade air temperature and relative humidity measurements were taken with the dew point humidity sensor (Protimeter, Marlow, UK) throughout the course of both days.

Leaf water potential

Xylem sap pressure was measured using a pressure chamber to indicate leaf water potential (Soil Moisture Equipment Corporation, Santa Barbara, CA, USA) for day 1 only (because of the limited nitrogen supply in the remote field location). Leaves were taken from the same height and position in the canopy over the day (n = 3).

Gas exchange

Gas exchange measurements were conducted by enclosing 10–20 cm2 sections of leaf within a leaf cuvette attached to a CIRAS-1 portable IRGA (PP Systems, Hitchin, UK) on leaves throughout the day, under ambient light, CO2 concentration and temperature. The leaf temperature was calculated using an energy balance calculation (Parkinson 1983) and gas exchange parameters were recalculated for the respective leaf areas. On day 1, a conifer PLC 3 head (ADC, Hoddeston, U.K.) was used, whilst in the hotter conditions on day 2 this was replaced with a broad-leaf PLC (PP Systems, Hitchin, UK) attached with an external cooling fan. The air supply was taken from a 5 dm3 buffering volume ≈ 15 m upwind of the platform to provide a reference supply with a stable CO2 concentration and isotopic signature. The flow of air passed through the cuvette ranged between 350 and 430 cm3 min–1. Gas exchange measurements were taken immediately before and after the 15 min on-line collection periods, and were averaged for the collection time. The P. aduncum leaves were hypostomatous with an average total leaf area of 85 cm2. Mean stomatal density taken from five stomatal epidermal impressions of three leaves was 193 mm–2. Leaf conductance to water vapour was corrected for hypostomy (CIRAS manual, PP systems 1994). Differences between the two types of cuvettes used were accounted for by correcting for boundary layer resistances of 0·45 and 0·28 m2 s–1 mol–1 for days 1 and 2, respectively.

Collection of CO2 and water vapour for isotope analysis

Atmospheric CO2 and water vapour were collected using a modified glass collection line (Griffiths et al. 1990; Broadmeadow et al. 1992) attached to a rotary vacuum pump. The standard collection procedure was modified to enable separation of CO2 from water vapour for C18O16O and H218O analysis. Air from either upstream (reference) or downstream (analysis) of the leaf cuvette was passed at positive pressure to a metering valve (Swagelock, Ohio, USA). The atmospheric CO2 and water vapour were trapped in a 6 mm internal diameter double spiral cold trap, immersed in liquid nitrogen, under a partial vacuum. Reference air was typically collected twice at the beginning and end of the day and on average after every third analysis collection, to account for diurnal variations in the ambient isotopic composition of canopy air, δa[Fig. 1, see also Borland et al. (1993); Harwood (1997)]. Cryogenically trapped CO2 and water vapour were isolated and evacuated down to 10–2 mbar (10–3 kPa).

Figure 1.

. Diurnal variation in (a) δa13CO2, (b) δaC18O16O and (c) δaH218O of cuvette air supply on day 1 (○) and day 2 (○). Best fit polynomial regressions were applied for both days from which the isotopic composition at the time of the analysis (δin) was derived. The fitted trend lines are represented by fine and thick lines for day 1 and day 2, respectively. The CO2 reference at 1014 h on day 1 (arrowed) was excluded from the regression as the internal mass spectrometer precision of the sample was four times greater than that observed for all other samples.

The CO2 was then liberated using acetone cooled to –70 °C (by addition of liquid nitrogen), for at least 2 min and condensed into a Pyrex side arm and sealed using a butane blow torch. As a precautionary measure, the sealed Pyrex tube was immediately stored in a liquid nitrogen dry shipper (Jencons, Leighton Buzzard, UK). Should complete separation of the CO2 and H2O not have been achieved during sampling, cold storage prevented equilibration (Harwood 1997) allowing the C18O16O signal to be preserved.

A new Pyrex side arm was then fitted to the collection line and evacuated. The water was liberated by heating the trap using a butane blow torch and cryodistilled into the freshly evacuated side arm until no more water was observed to condense in the side arm (this was usually in the order of 2–3 min). An aliquot of CO2 was then introduced into the collection line, so that the δH218O could be measured through equilibration with CO2. This was achieved by backfilling a gas tight syringe with CO2, of known isotopic signature from a modified commercial drink carbonation cylinder and injecting 1 cm3, via a septum fitting, into the collection line. The CO2 was then passed into the side arm, frozen above the water sample and sealed in the vial using the blow torch.

Protocol for determination of C18O16O

Samples of CO2 were stored in liquid nitrogen until purification. Each sample was then taken individually, warmed to room temperature, wiped to remove any external moisture and placed in a bottle breaker attached to a sample purification line. The transfer time from liquid nitrogen to breaking the sample tube under vacuum was typically less than 2 min, so as not to allow equilibration with any residual water.

Technique development showed that small volumes of CO2 inferred incomplete separation, or collection, of the CO2 and water vapour which lead to large outliers. Such samples were, therefore, rejected. This was validated in laboratory trials leading to precision for separated CO2 of ±0·08‰ for δ13CO2 and ± 0·07‰ for δC18O16O (standard deviations, n=8) where isotope compositions are expressed as δ using the per mille (‰) notation given by δ (‰)=[(Rsample/Rstandard)–1]×1000. For 13CO2, R=13C/12C and the standard is Pee Dee Belemnite (PDB) and for C18O16O, R=18O/16O and the standard is CO2 derived from PDB (PDB–CO2).

Protocol for determination of δH218O of water vapour

Samples containing water and equilibration CO2 were left for a minimum of 3 d to equilibrate, after which the sampled was purified in accordance with the standard procedure. Corrections were made for equilibration temperature and respective volume of gas and liquid according to Scrimgeour (1995).

The volume of water vapour present in the side arm (involved in equilibration) was calculated from the flow rate of air passing through the collection line (the rate set by the needle valve), the duration of the collection, and the water vapour concentration of the air supply as measured by the CIRAS water vapour IRGAs attached to the collection line. The average volume of water collected over 15 min generally varied between 0·05 and 0·07 cm3, depending on the ambient vapour pressure. Reproducibility of water vapour samples equilibrated with CO2 were typically ± 0·09‰ (n = 8) and were expressed in the δ notation relative to the water standard, standard mean ocean water (SMOW).

Purification and mass spectrometric analysis of on-line samples

All atmospheric CO2 and water equilibrated CO2 samples were purified by cryodistillation under vacuum on a purification line by passing the gas through two acetone traps at dry ice temperature (– 70 °C, as described in Griffiths et al. (1990) to remove any water and non-condensable gases. Samples were analysed using a dual inlet mass spectrometer modified to a VG 903 triple collector specification by Provac Services (Crewe, UK) and run using the Europa IRMS (Crewe, UK) data collection software, against a working standard of 99·995% CO2 (BDH High Purity Gases, Poole, UK), with a δ13CO2 of – 43·18‰ and a δC18O16O of – 28·72‰ versus PDB–CO2.

Samples of CO2 were corrected for the presence of N2O (as N2O has the same molecular mass, 44, as CO2). Mass spectrometer sensitivity to N2O was determined using the method described in Freidli & Siegenthaler (1988), whereby pure CO2 and N2O are mixed using mass flow controllers to produce different CO2–N2O gas mixtures (where ρ represents the ratio of the volume of N2O to CO2) and admitted into the mass spectrometer. The corrections derived for the ratio of N2O:CO2 representing ambient air (ρ = 0·00088, 310 p.p.b.:350 p.p.m., respectively) of + 0·24‰ for δ13C and + 0·31‰ for δ18O were in agreement with those observed in other investigations (Freidli & Siegenthaler 1988; Flanagan & Varney 1995). Diurnal variation in ambient N2O was considered as negligible and, thus, corrections to CO2 were made using an N2O concentration of 310 p.p.b. and the respective CO2 concentration using the equations of Freidli & Siegenthaler (1988).

Source and rain water

Rain and source water were analysed using a direct equilibration method (Scrimgeour 1995), whereby the water is equilibrated with CO2 of known volume and isotopic composition. Samples were left for at least 3 d to equilibrate, after which they were run on the ANCA CF-IRMS (Europa Scientific, Crewe, UK). Working standards also underwent the same equilibration process as the samples, to account for any variation in sample preparation. After mass spectrometer analysis, samples were corrected for their respective volumes, δ18O of the equilibration CO2 and equilibration temperature. Reproducibility of repeated runs of the same sample generally yielded a standard deviation of ± 0·07‰ (n = 8). The oxygen isotope composition of all source, rain and leaf water (see below) samples were expressed in the δ notation relative to SMOW.

δH218O Leaf water

The δH218O of the leaf water samples were kindly analysed by Prof. D. Yakir and colleagues at the Weizmann Institute, Rehovot, Israel. Leaf water was extracted via vacuum distillation for analysis as described in Wang & Yakir (1995). External precision for the water samples was approximately ± 0·2‰.

On-line discrimination of 13CO2 and C18O16O

Measured (observed) discrimination of 13CO2 (Δobs) was calculated as described by Evans et al. (1986) below, whereby:

inline image

where ξ = ce/(ceco) and δin, ce and δout, co are the isotopic composition and CO2 concentration entering and leaving the cuvette, respectively. Values of co were corrected to the vapour pressure of ce to account for any dilution of CO2 leaving the cuvette from increased water vapour concentration. Background ambient CO2 typically varied between –9·5 and –8·5‰δ13CO2 (Fig. 1a) whilst δC18O16O ranged from –0·5 to 0·5‰ (Fig. 1b). To ensure consistency in the selection of δin for measurements of concurrent discrimination of 13CO2, C18O16O and H218O, values were interpolated from temporal polynomial regressions of the reference air isotope composition (δa) for both days at the time of any given analysis (Fig. 1). The more conventional method, expressing the isotope composition of ambient air as a function of the inverse of the concentration, could not be applied as no relationship held with either δaC18O16O and 1/[CO2] or δaH218O and 1/[H2O], consistent with other observations within this canopy (Harwood 1997).

Instantaneous ‘on-line’ discrimination of C18O16O was derived using the same equation as for Δ13CO2 (Farquhar & Lloyd 1993) by simply substituting the oxygen isotope composition of CO2 for δin and δout in Eqn 1.

Modelled discrimination of Δ13CO2 and ΔC18O16O

Modelled discrimination against 13CO2 (Δi) was derived from the following equation (Evans et al. 1986):

inline image

where Ca and Ci refer to the concentration of CO2 in the air and intercellular leaf spaces, respectively (where Ca represents the CO2 concentration exposed to the leaf and is represented by the CO2 concentration leaving the cuvette), b′ is an assumed value of net discrimination of Rubisco (here taken as 29‰), a is the fractionation due to diffusion through the stomata (4·4‰); Γ* is the CO2 compensation point in the absence of dark respiration (calculated from the regression equations listed by Brooks & Farquhar (1985) and f represents the fractionations associated with photorespiration [here taken as 7‰; Rooney (1988); Gillon & Griffiths (1997)].

The offset between the modelled 13CO2 discrimination, Δi (Eqn 2) and that measured, Δobs (Eqn 1), expressed as ΔiΔobs has traditionally been attributed to additional drawdown imposed by the internal resistance to CO2 diffusion from the intercellular air space CO2 concentration (Ci) to that at the sites of carboxylation in the chloroplast [Cc; Evans et al. (1986); von Caemmerer & Evans (1991)].

The difference between measured and modelled discrimination (ΔiΔobs) has, therefore, been used to estimate Cc by rearranging the following equation (von Caemmerer & Evans 1991):

inline image

Additionally, if values of gw are known then from Ficks’ law of diffusion, estimates of Cc can also be derived using the following expression (von Caemmerer & Evans 1991; Harley et al. 1992):

inline image

where A is assimilation in μmol m–2 s–1, P is atmospheric pressure and gw is the internal transfer conductance for CO2 (in mol m–2 s–1 bar –1). Whilst conductance for CO2 (gs) between Ca and Ci can vary, the internal transfer conductance (gw) from Ci to Cc within the same leaf can be considered as constant.

Assuming there is complete equilibration between CO2 and water within the chloroplast, the on-line discrimination of C18O16O can be predicted using the following equation (Farquhar & Lloyd 1993):

inline image

where δ is the oxygen isotope composition of CO2, C is the concentration of CO2 and subscripts c and a refer to the chloroplast and ambient air, respectively, and a¯ is the weighted mean of discrimination occurring during the diffusion from ambient air to the sites of carboxylation within the chloroplast, here taken as 7·4‰ (Farquhar et al. 1993).

Determination of the δH218O of transpired water vapour (δt)

The δH218O of transpired water vapour was derived by mass balance from measuring the water vapour concentration and isotope composition before and after a leaf cuvette, as follows (see Appendix A):

inline image

where ξL= eout/(eoutein) and δin, ein and δout, eout are the oxygen isotope composition and vapour pressure (in mbar) of the water vapour entering and leaving the cuvette, respectively. This was substituted for Rs in Eqn A3 and solved for Re, to provide an estimate for enrichment at the evaporative sites of the P. aduncum leaves, without the prerequisite of ISS. The oxygen isotope composition of the reference water vapour, δa H218O, varied between –7·5 and –10·0‰ on day 1 and between –9·0 and –10·9‰ on day 2 (Fig. 1c). Variation in the flux and isotope composition of transpired water vapour under the high and variable leaf-to-air VPDs may have accounted for the differences in δaH218O between days 1 and 2. As with CO2, the value of H2O δin at the time of the analysis was derived from a temporal polynomial regression of the δ18O of water vapour entering the cuvette (Fig. 1c).

RESULTS

Photon flux, temperature and leaf water status

The average PFD on day 1 was lower and more intermittent than on day 2, due to greater cloud cover (Fig. 2a & b). The accumulated PFD recorded reached 26·7 and 37·2 mol m–2 d–1, for days 1 and 2, respectively. In general, day 1 experienced substantially lower temperatures and higher relative humidities throughout the photoperiod than day 2 (data not shown). These brighter, drier conditions on day 2 resulted in higher average leaf temperatures of 37·3 °C between 1000 and 1400 h on day 2, compared with a mean of 31·6 °C for the same period on day 1 (Fig. 2c).

Figure 2.

. Diurnal variation in (a, b) above canopy photon flux density (PFD) and (c) leaf temperature for day 1 (○) and day 2 (○). (d) Variation in leaf water potential (expressed by xylem sap pressure) for the average of three Piper aduncum leaves over day 1 (n = 3, error bars show standard error).

The leaf water potential (Ψ, measured on day 1 only) decreased from early morning, reaching a minimum of –1·1 MPa at 1300 h and then increased in the afternoon (Fig. 2d). The leaf water potential had fully recovered to the predawn value of –0·21 MPa, by 1800 h. In general, the diurnal transition in leaf water potential was gradual, reflecting the overall diel progression in evaporative demand, and contrasted with the variability seen in leaf gas exchange which was influenced by individual leaf stomatal responses (Fig. 3a & b).

Figure 3.

. Diurnal variation in (a) assimilation (A), (b) stomatal conductance (gs), (c) Ci/Ca and (d) leaf-to-air vapour pressure deficits (VPD), for different Piper aduncum leaves on day 1 (○) and three separate leaves on day 2 (▵ leaf 1, ○ leaf 2, □ leaf 3). Points are the average of two readings.

Gas exchange

On day 1, gas exchange measurements were made continuously throughout the day on a range of leaves. Assimilation was greatest in the morning around 0800 h and declined as the day progressed (open symbols, Fig. 3a), whilst the general diurnal trend in stomatal conductance tended to follow leaf-to-air VPD (Fig. 3b & d). Under cloud cover at midday (Fig. 2a), a decrease in leaf-to-air VPD (Fig. 3d) gave rise to a transient increase in average A and gs from 4·2 to 5·8 μmol m–2 s–1 (Fig. 3a) and 120–185 mmol m–2 s–1 (Fig. 3b), respectively. The concentration of substomatal CO2, expressed as Ci/Ca, remained relatively stable throughout the day (Fig. 3c), with a mean value of 0·78 (standard error = 0·02, n = 12), with the exception of the early morning, between 0800 and 0900 h, where high rates of assimilation increased the drawdown from Ca to Ci reducing Ci/Ca to 0·65.

For day 2, the same three leaves were followed throughout the day (closed symbols, Fig. 3). Higher PFD and temperatures (Fig. 2) induced greater leaf-to-air VPDs (Fig. 3d) reaching a maximum of 42 mbar (4·2 kPa) at midday. Assimilation and stomatal conductance were greatest in the morning before 1000 h and declined over the course of the day. Both A and gs on day 2 were slightly reduced in the drier conditions, compared with day 1. For all leaves, Ci/Ca slightly increased over the day (Fig. 3c), due to the drop in assimilation, but on average was significantly lower (P = 0·00026, students t-test) than day 1 at 0·68 (standard error = 0·02, n = 11).

Instantaneous isotope discrimination

On day 1, measured Δ13CO2 declined gradually over the day, from above 30‰ in the morning to around 10‰ in the late afternoon (open symbols, Fig. 4a). Whilst the sampling procedure showed that individual leaves can have distinct responses, as demonstrated by the regular decline over the course of day 2 in all three leaves (closed symbols, Fig. 4a), variation in measured discrimination was also systematically related to changing environmental conditions. The first measurements of both days, taken as soon as rates of photosynthesis for isotopic discrimination were measurable (the time at which solar elevation exceeded the surrounding canopy) produced unusually high discrimination signals (circled points, Fig. 4a) compared with that modelled (Fig. 4b). This ‘postdawn’ response was observed not only on both days in 1995, but also on the same P. aduncum canopy 3 years earlier (M. S. J. Broadmeadow, unpublished field observations 1992; Gillon et al. 1997).

Figure 4.

. Diurnal variation in (a) instantaneous observed Δ13CO2 for different Piper aduncum leaves on day 1 (○) and three separate leaves on day 2 (▵ leaf 1, ○ leaf 2, □ leaf 3), (b) discrimination from predicted Ci/Ca for days 1 and 2 (symbols as above). Relationship between the offset in predicted and measured Δ13CO2 (ΔiΔobs) with (c) the quotient A/Ca and (d) the expression of the amount of CO2 fixed relative to the concentration in the atmosphere ξ. The shaded triangle in (d) represents the range of values associated with a mass spectrometric precision of ± 0·1‰. High values of Δ13CO2 observed at the beginning of both days are circled in (a), (b) and (d) to ease interpretation.

With Ci/Ca relatively constant over both days (Fig. 3c), modelled Δ13CO2 (Δi) also remained constant (Fig. 4b), so that the observed discrimination (Δobs) could be either greater or lower than that modelled (Δi) from Ci/Ca (Fig. 4c). The offset between observed and modelled discrimination (ΔiΔobs), when compared with the quotient of assimilation and CO2 concentration, A/Ca, can usually be used to derive the internal resistance to CO2, from ambient air to the sites of carboxylation in the chloroplast, Cc (Evans et al. 1986; von Caemmerer & Evans 1991; Evans & von Caemmerer 1996). However, a general decrease in ΔiΔobs with A/Ca was observed (Fig. 4c) and contrasted with the increase expected in order to calculate gw.

The systematic shift in Δ13CO2 was not a function of the gas exchange system: calculation of Δobs is partly dependent on ξ, the fractional uptake of CO2 (Eqn 1), where large values of ξ (low CO2 drawdown) may reduce the accuracy of Δobs (von Caemmerer & Evans 1991). However, the relationship between ξ and ΔiΔobs for P. aduncum revealed that scatter occurred over the full range of ξ (8–45), not just at large values (Fig. 4d). In addition, the majority of the data points fell outside the range of ΔiΔobs associated with a mass spectrometric precision of ± 0·1‰ (shaded triangle, Fig. 4d).

On day 1, measured ΔC18O16O generally increased from 10‰ in the morning, to 37‰ at 1330 h, and declined again in the afternoon (open symbols, Fig. 5a). As with Δ13CO2, variations from this trend were related to both distinct leaf responses (demonstrated by the three individually measured leaves on day 2, closed symbols, Fig. 5a) and changes in environmental conditions induced by intermittent cloud cover (Fig. 2a & b). For day 2, leaves 2 and 3 followed a similar diurnal pattern, but had maximum midday enrichment of 21 and 23‰, respectively, somewhat lower than that measured on day 1. In contrast, ΔC18O16O of leaf 1 remained fairly constant between 8 and 12‰, over the three time intervals measured (also Fig. 5a).

Figure 5.

. Diurnal variation in (a) instantaneous ΔC18O16O for different Piper aduncum leaves on day 1 (○) and three separate leaves on day 2 (▵ leaf 1, ○ leaf 2, □ leaf 3). (b) Relationship between observed ΔC18O16O and Cc/Ca (derived from Eqn 4) with that predicted (Eqn 5) from Cc/Ca with δ18O chloroplast water in equilibration with source water (–3·2‰) and evaporative site water at 2, 10 and 20‰ (from the observed range of δe, Fig. 6b).

Because of the large variation in Δobs relative to Δi for 13CO2 discrimination (Fig. 4c), the concentration of CO2 in the chloroplast (Cc) for P. aduncum was estimated using Eqn 4. Internal CO2 transfer conductance (gw) was derived from taking the assimilation of 16 μmol m–2 s–1 attained under 1000 μmol m–2 s–1 PFD for the pioneering species P. auritum (Tinoco-Ojanguren & Pearcy 1992) and the relationship between assimilation under high light and gw in the literature (von Caemmerer & Evans 1991; Lloyd et al. 1992; Loreto et al. 1992; Evans et al. 1994; Epron et al. 1995). Assuming a constant Rubisco sink strength, the internal CO2 transfer conductance (gw) for P. aduncum (a hypostomatous tree) was estimated as 0·25 mol m–2 s–1 bar–1.

Assuming constant gw and under low rates of assimilation, by definition, the offset between Ci and Cc, and, hence, Cc/Ca, in P. aduncum will have remained relatively stable over the course of the day. A reduced chloroplast CO2 concentration on day 2 (closed symbols, Fig. 5a) is, therefore, suggested to partially account for the lower observed values of ΔC18O16O. Furthermore, the low C18O16O discrimination observed for leaf 1 on the second day corresponded with lower values of Cc/Ca (closed triangles Fig. 5a & b). Measured values of ΔC18O16O generally lay between that predicted (Eqn 5, dashed lines, Fig. 5b) for chloroplast water in equilibration with that at the evaporative sites of between 2 and 10‰. This agreed well with the evaporative enrichment derived from the δ18O of transpired water vapour (see Fig. 6b) for day 1, but did not reflect the high evaporative enrichment (20‰) in the late afternoon of day 2.

Figure 6.

. Diurnal variation of (a) the δ18O of transpired water vapour for different Piper aduncum leaves on day 1 (○) and three separate leaves on day 2 (▵ leaf 1, ○ leaf 2, □ leaf 3) with respect to source water at –3·2‰ (dashed line), indicating the vapour pressure deficit (VPD) status and general trend over the day (solid line). The rain water signal generally varied between + 1·0 and –1·0‰ throughout the field campaign, whilst the ground water had a δ18O of –3·2‰. With high evaporative demand and infrequent and low intensity rain showers (data not shown) it was likely that the water supply for the Piper originated from ground water, rather than precipitation-fed soil water. Isotopic composition of ground water has been shown to remain constant despite input from precipitation with variable δ18O (Neal & Rosier 1990), so the δ18O of spring water collected on 20 February 1995 was assumed to be the source water signal for the 2 d investigated. (b) Evaporative site enrichment (δe) for different leaves on day 1 and three separate leaves on day 2 (symbols as above). Solid lines represent diurnal trends in δe for day 1 and day 2 calculated using δH218O of transpired water vapour. Dashed lines represent the trendline fitted to the same data points derived assuming isotopic steady state (ISS) held throughout the day. Crosses indicate two average measurements of bulk leaf water δH218O (n = 3) taken on day 1.

On both days, the δH218O signal of transpired water was generally more depleted than stem water in the morning and more enriched in the afternoon (Fig. 6a). The sole exception to this trend was at 0740 h on day 1, which approximated that of source water and may have occurred due to low and stable VPD conditions early in the morning. An increase in the δH218O of transpired water vapour over the day, would be expected with an approach to steady state, under increasing and decreasing VPD, respectively. The transpired water vapour had, on average, approximately the same signal as the stem water (–3·2‰) between 1300 and 1500 h on both days indicating that the leaves were probably close to or at ISS.

Leaf water enrichment at the evaporative sites (δe) derived from the measurements of transpired water vapour, was tightly related for all leaves over both sampling days, but consistently higher on day 2 (Fig. 6b). Again, there was no systematic relationship with the fractional proportion of water lost (ξL), as for both days 17 of the 23 data points ranged between 3 and 7 (data not shown). The largest remaining values of ξL recorded occurred shortly after dawn and approaching dusk and ranged from 8 to 15.

Evaporative site enrichment (δe) generally increased in the morning, peaked between 1300 and 1400 h, and then became slightly reduced in the afternoon, on both days. On day 1, δe reached a maximum enrichment of 12·7‰ at 1334 h and remained high in the afternoon (Fig. 6b). For day 2, the maximum enrichment was higher, reaching 20·9‰ at 1423 h, for both leaves 1 and 2. The transient decrease in δe at 1300 h on day 1 was likely to have been due to a decrease in air temperature and leaf-to-air VPD, as a result of cloud cover. The dashed lines in Fig. 6(b) represent values of δe assuming that ISS held throughout the entire day. Such an assumption would have over-estimated evaporative site enrichment under periods of increasing leaf-to-air VPD and under-estimated δe at times of decreasing leaf-to-air VPD, with an increased effect under high VPDs (up to 4‰, in Fig. 6b).

Bulk δH218O leaf water values for P. aduncum (taken on day 1) were 2·56‰ (± 0·2, n = 3) and 7·01‰ (± 0·3, n = 3) for 0930 and 1230 h, respectively (crosses, Fig. 6b). Both measurements were in close agreement with that calculated for the evaporation sites (δe), including the transient decrease in enrichment around 1300 h on day 1 due to cloud cover.

DISCUSSION

Gas exchange and leaf water status

On both days, maximum assimilation of P. aduncum occurred in the morning, around 0800 h and gradually declined over the day as leaf-to-air VPD increased and water potential decreased (Figs 2d & 3a). This is similar to that found for Jacaranda copaia, in French Guiana (Huc, Fehri & Guehl 1994) and in other field observations of diurnal gas exchange (Wise et al. 1990; Wise et al. 1991; Epron, Dreyer & Breda 1992). The minimum leaf water potential (Ψ) at midday of –1·1 MPa for P. aduncum was similar to that of –1·5 MPa for the pioneering tree species J. copaia in the field in French Guiana (Huc et al. 1994). In the afternoon, photosynthesis and stomatal conductance remained depressed (Fig. 3a & b), despite a recovery in Ψ[Fig. 2d, cf. Quick et al. (1992)].

Higher leaf-to-air VPDs in the brighter conditions of day 2 are likely to have accounted for the reduction in stomatal conductance observed in P. aduncum. Reduced stomatal conductance under high VPDs has been reported for another pioneering Piper species (P. auritum, Tinoco-Ojanguren & Pearcy 1993), maintaining instantaneous water use efficiency, as a decline in gs reduced assimilation and transpiration in tandem. The variability in stomatal conductance on day 1 was due to leaf-to-leaf variation and intermittent light resulting from cloud cover. Concurrent variation in assimilation was minimal and may have been due to the more rapid response of assimilation to changes in PFD (Tinoco-Ojanguren & Pearcy 1992; Barradas & Jones 1996).

Despite variation in A and gs, Ci/Ca remained relatively stable over the day. Observations of relatively stable Ci/Ca have also been reported, despite variation in gs during light flecking in a related species, P. auritum (Tinoco-Ojanguren & Pearcy 1992). Furthermore, due to the low rates of assimilation, the theoretical drawdown between Ci and Cc never exceeded 30 μmol mol–1 (corresponding with a maximum assimilation of 7 μmol m–2 s–1), thus, the diurnal variation in Cc/Ca approximated that of Ci/Ca.

On-line discrimination of 13CO2

Measured Δ13CO2 (Δobs) in P. aduncum was similar to that found for other C3 mesophytes (Evans et al. 1986). However, the variation in Δobs could not be accounted for by Ci/Ca alone (Fig. 4). The observation of a wide range in Δobs compared with that modelled from gas exchange was observed on the same P. aduncum canopy 3 years earlier (M. S. J. Broadmeadow, unpublished field observations 1992; Gillon et al. 1997).

Differences between modelled and observed on-line Δ13C have previously been accounted for by gw, the CO2 transfer conductance from the leaf intercellular spaces to the chloroplast (Evans et al. 1986; von Caemmerer & Evans 1991). The maximum drawdown of 30 μmol mol–1 from Ci to Cc likely to be encountered in P. aduncum, only reduces modelled Δ13C by 2‰ and does not explain the full extent of the shift between modelled and measured Δ13CO2 (ΔiΔobs). Any effect of drawdown to Cc on Δobs should be reflected by a positive correlation between ΔiΔobs and A/Ca (Evans et al. 1986). In contrast, on both days, P. aduncum appeared to exhibit a negative correlation, whereby the greatest values of ΔiΔobs coincided with low assimilation rates, when the drawdown from Ci to Cc should have been minimal. Thus, the data are not explained by conventional theory, as neither Ci or Cc could fully account for the observed range of measured Δ13CO2.

Further analysis also revealed that large variations in Δobs could not be explained by the influence of mass spectrometric precision at high ξ (Fig. 4d) or diurnal variation in δa13CO2 (Fig. 1a). Thus, similarities between data on both days and that recorded 3 years previously on the same P. aduncum stand suggest the range of Δ13C observed here may be a function of plant physiology rather than any technical problems.

Increased stomatal patchiness has been associated with water stress (see Eckstein et al. (1996) for a recent review) and also with concurrent changes in relative humidity and irradiance (Cardon, Mott & Berry 1994). Any existence of patchy stomatal closure may lead to an over-estimation of Ci calculated from measurements of gas exchange (Farquhar 1989; Mott 1995). In this case, actual leaf substomatal CO2 concentration would be lower, and in turn reduce modelled discrimination (Δi). Any shift in Δi would still leave values of Δobs both above and below that modelled and, thus, stomatal patchiness seems unlikely to be wholly attributable for the full range of discrimination values observed. However, investigation of asymmetric stomatal closure was beyond the scope of the field work and more experimental work is required to assess any interaction with discrimination.

The range of A/Ca encountered with P. aduncum in the field, 0·005–0·02 (Fig. 3c), was generally lower than values found in controlled conditions [e.g. 0·04–0·12; Evans et al. (1986)]. Values of ΔiΔobs, increasing at lower assimilation rates, suggest that measured Δ13CO2 is lower than modelled from Ci/Ca. This phenomenon was also observed in Phaseolus vulgaris, where the δ13C of dark respired CO2 was very negative at c.–50‰ when plants were grown air depleted in 13C (Gillon & Griffiths 1997), as well as other gas exchange studies under field conditions [M. S. J Broadmeadow, unpublished field data 1992; K. G. Harwood, unpublished field data on Quercus 1996; J. S. Gillon, unpublished field data on Prosopis 1997; see also Gillon et al. (1997)]. At such low assimilation rates, it was suggested that CO2 from dark respiration in the light can make a significant contribution to net CO2 exchange, whereby respiratory CO2 is more depleted in 13C than the CO2 being assimilated via photosynthesis. Any back diffusion or re-fixation of this depleted CO2 from the leaf reduces the enrichment in 13CO2 observed in air passing over the leaf (i.e. a smaller difference between δin and δout, Eqn 1). Thus, the net discrimination expressed by the leaf is lower than predicted (i.e. reduced Δobs). The data would also suggest that respired CO2 is relatively more depleted in 13C than the initial Rubisco product. Although no fractionation during respiration has been reported for cells in vitro (Lin & Ehleringer 1997), such a situation may arise from short-term differences in the isotopic composition of the respiratory substrates and CO2 being assimilated, due to the slow turnover time of carbohydrate pools, c. 4 h (Parnik, Keerberg & Viil 1972). Similarly, a transient enrichment in the 13C of respired CO2 (relative to that fixed by photosynthesis), possibly as a function of a nocturnal metabolic modification of the carbohydrate pools’ isotope composition, may account for the occurrence of very high Δobs, above that predicted, evident in the few hours after dawn on both days.

Any effects of slow carbohydrate pool turnover on Δobs due to increased rates of dark respiration or re-fixation occurring at high temperatures, may be exacerbated in the tropics, and are unlikely to be encountered in crop species usually studied, with characteristically higher values of A/Ca, under controlled laboratory conditions.

On-line discrimination in C18O16O

Discrimination in C18O16O reflects both the enrichment of 18O in leaf water, with which it equilibrates before diffusing back out of the chloroplast, and also the extent of back diffusion of CO2, which allows the leaf water enrichment to be expressed (Farquhar et al. 1993). With a relatively small variation in Cc/Ca (derived from Ci/Ca) over the course of both days, ΔC18O16O generally reflected leaf water enrichment, increasing in the morning, until just after midday and slightly decreasing in the afternoon (Fig. 5a). Discrimination of C18O16O is strongly influenced by the non-linear relationship with Cc/Ca (Farquhar et al. 1993; Flanagan et al. 1994). This was demonstrated by a general reduction in values of ΔC18O16O on day 2 compared with day 1, when stomatal conductance and Cc/Ca were reduced under high VPDs (Fig. 5b). Using the estimated gw of 0·25 mol m–2 s–1 bar–1, the observed ΔC18O16O generally lay between that predicted for chloroplast enrichment of 2 and 10‰ (assuming complete equilibration between δe and δc, Fig. 4c), lower than concurrent measurements of δe which ranged from 2 to 20‰ (Fig. 5b).

Despite the codependence of Δ13CO2 and ΔC18O16O on the extent of back diffusion of CO2 from the chloroplast (Evans et al. 1986; Farquhar et al. 1993; Williams & Flanagan 1996) the two processes were not necessarily coupled in field conditions (Figs 4a & 5a), largely because Cc/Ca imposes a linear control on Δ13CO2, whereas ΔC18O16O is dependent on the non-linear relationship with Cc/Caand the interaction with VPD-induced evaporative enrichment (Williams, Flanagan & Coleman 1996). Additionally, any contribution from dark respiratory CO2 may alter the signal of on-line Δ13CO2, yet not affect the ΔC18O16O, as the δ18O signal of respiratory CO2 will be lost in the equilibration with leaf water in the chloroplast.

Enrichment of leaf water at the evaporative sites (δH218O/δe)

In an approach to ISS, transpired water vapour was generally less enriched in 18O than source water during periods of increasing VPD (e.g. morning) and more enriched than source water during times of decreasing VPD (e.g. afternoon). A relative stability in VPD combined with a slight lag in leaf water turnover rate, indicated that leaf water enrichment was close to ISS between 1300 and 1500 h (Fig. 6a). This observation of proximity to ISS in the field has important implications in modelling the contribution of transpired water vapour to terrestrial water budgets using the isotopic composition of atmospheric water vapour (Bariac et al. 1989, 1994; Brunel et al. 1992). This supports the assumption of ISS in modelling ΔC18O16O discrimination (Farquhar et al. 1993), because the time when ISS occurs coincides with the most photosynthetically active time of the day, when the greatest fluxes of CO2 and H2O should take place, for many terrestrial plants. However, both this and other field observations have demonstrated that under some conditions, optimum exchange of CO2 and water vapour can occur in the early part of the day, both at the leaf level (Roberts et al. 1980; Leverenz et al. 1982; Roberts, Wallace & Pitman 1984; Briggs, Jurik & Gates 1986; Roberts, Cobral & Ferreira de Aguiar 1990; Wise et al. 1990; Epron et al. 1992; Hinckley et al. 1994) and at the whole canopy level (Grace et al. 1995). In these cases, the associated changes in environmental conditions may yield non-steady-state photosynthesis and transpiration, for which steady-state estimations of both Δ13CO2 and ΔC18O16O, derived largely from laboratory experiments, may be in error.

Analysis of transpired water vapour allowed the direct calculation of 18O enrichment of water at the evaporation sites (δe) over the full diurnal period, without the prerequisite of ISS. For the conditions experienced, a conventional derivation of δe assuming ISS throughout the day would lead to an over-estimation of evaporative site enrichment early in the morning and an under-estimation later in the afternoon. The extent of evaporative site enrichment was largely maintained in the afternoon, despite the decrease in leaf-to-air VPD (Fig. 6b) suggesting that 18O enrichment at the evaporative sites may also be affected by species-dependent rates of leaf water turnover (Wang & Yakir 1995), in addition to evaporative control imposed by ambient VPD. Maintenance of evaporative enrichment of leaf water late into the afternoon has also been noted for cotton (Yakir, DeNiro & Gat 1990), alfalfa (Bariac et al. 1989) and maize (Bariac et al. 1994) and also to some extent in mature oak and beech leaves (Forstel 1978) and barley (Walker & Lance 1991), but contrasts with the relatively stable leaf water enrichment found in sclerophyllous juniper and mistletoe by Flanagan, Marshall & Ehleringer (1993). Two direct measurements of bulk leaf water 18O composition were close to that derived for the evaporative sites and suggest that the evaporative enrichment model provided approximate estimates of bulk leaf water δH218O at the two times investigated (cf. Farris & Strain 1978; Forstel 1978; Walker et al. 1989; Yakir et al. 1990; Flanagan et al. 1991a, b; Flanagan et al. 1993). The diurnal trend in measured ΔC18O16O agreed qualitatively with the modelled evaporative enrichment (δ H218O at the evaporation sites, or δe). However, quantitatively, ΔC18O16O was greater on day 1 and this chloroplast signal was tightly coupled to changes in Cc/Ca, associated with VPD and conductance effects. Meanwhile, there was a much more gradual transition in leaf water enrichment at the evaporation sites (δe, Fig. 6b), similar to overall leaf water status (Fig. 2d) and contrasted with the more variable stomatal influence on ΔC18O16O (Fig. 5a). Values of measured ΔC18O16O were low, particularly for day 2, when compared with that predicted from the Farquhar and Lloyd model using measured values of δe (Eqn 5, Fig. 5b).

By rearranging the Farquhar and Lloyd C18O16O discrimination model the measured ΔC18O16O can be used to yield an estimate of isotopic composition of water within the chloroplast (δc), which can then be compared with δH218O at the evaporative sites (δe). Preliminary analysis for P. aduncum reveals that in most cases, δe is greater than derived δc (below the 1:1 line of unity, Fig. 7). The model used to derive δc is partially dependent on the estimated value of gw used for deriving Cc (here taken as 0·25mol m–2 s–1 bar–1). However, estimation of δc using both double and half of the internal conductance value selected of 0·125 and 0·5 mol m–2 s–1 bar–1, respectively [representing the natural range of gw reported in the literature; Loreto et al. (1992)], only results in an average respective shift of 1·5 and 0·7‰ (error bars, Fig. 7). Thus, uncertainty in Cc can still not account for the full offset observed between δc and δe for P. aduncum.

Figure 7.

. Relationship between evaporative site enrichment δH218O (δe) inferred from measurements of transpired water vapour (Eqns A7 and A3) and chloroplast δH218O (δc) reconstructed from on-line measurements of ΔC18O16O (Eqn 1, corrected for temperature-dependent equilibration between CO2 and water) for different Piper aduncum leaves on day 1 (○) and three separate leaves on day 2 (▵ leaf 1, ○ leaf 2, □ leaf 3). Positive and negative error bars denote the potential range of δc using a low (0·125 mol m–2 s–1 bar–1) and high (0·5 mol m–2 s–1 bar–1) estimate of gw, respectively. Dotted line represents the 1:1 line of unity.

Observations of chloroplast water less enriched than that at the evaporative sites agrees with the conventional theory of a continuum of enrichment from the unfractionated source water, to the most fractionated water at the sites of evaporation (Farquhar & Lloyd 1993). Chloroplast water can be depleted by up to 10‰ in δH218O compared with that at the evaporative sites (Yakir et al. 1994) and could partially account for the low values of δc observed in this study. However, since the bulk leaf water values were close to that derived for the evaporative sites (Fig. 6), the lower values of ΔC18O16O (Fig. 5) and, hence, δc (Fig. 7) may have been a result of incomplete isotopic equilibration between CO2 and water vapour within the chloroplast (Farquhar & Lloyd 1993; Flanagan et al. 1994), although this may also be questionable under such low rates of gas exchange. Understanding variations in δc and δe are important for discerning the physiological controls of leaf water enrichment and provides a greater insight into the relationship between leaf water enrichment and the δ18O signal transferred to the leaf cellulose. These observations suggest that further investigation into the variation in δe with derived δc is required under field conditions and in particular for reliable estimates of Cc.

Under high temperatures in Trinidad, measurement of Δ13CO2 in the field deviated from that modelled from gas exchange measurements of Ci/Ca. Simple models applied to calculating discrimination and deriving single point values of gw, under tropical field conditions, should, therefore, be used with caution. This study also suggests discrepancies may still exist between C18O16O and H218O estimation of 18O leaf water enrichment under field conditions. Additionally, the assumption of ISS may not hold under changeable VPDs and a more faithful representation of δe can be derived from measuring the oxygen isotope composition of transpired water vapour. Thus, observed isotopic discrimination under natural field conditions can differ from that predicted under steady-state photosynthesis in controlled laboratory experiments. Whilst the models provide us with an insight into the mechanistics of discriminatory processes, investigation of discrimination under field conditions is necessary to ensure accurate prediction of biome discrimination for global models (Farquhar et al. 1993; Ciais et al. 1995, 1997) and to provide a greater understanding of the factors controlling discrimination under natural, transient, environmental conditions.

Acknowledgements

Thanks must go to Charles McDavid, Department of Plant Science and Biochemistry and Joss Knight, Department of Physics both at the University of the West Indies, Trinidad for the loan of equipment. Also to Dan Yakir and Xue Feng Wang for useful discussion and the laboratory at the Weizmann Institute, Israel for 18O leaf water analysis. The comments of three anonymous reviewers are also acknowledged for substantially improving this manuscript, one of whom provided the simplified expression for Eqn A7. This work was supported by the National Environment Research Council (NERC) under small grant no. GR9/1767 and studentship no. GT4/92/27 to K.G.H.

Appendix

APPENDIX A: ESTIMATION OF EVAPORATIVE SITE ENRICHMENT IN 18O

Flanagan et al. (1991a) modified the Craig & Gordon (1965) equation, whereby the evaporative enrichment within a leaf is derived from the proportion of 18O evaporated relative to 16O (E18/E16). The difference in evaporation of the heavy and light isotope can be seen in the transpired water, and expressed as a function of the isotope composition of water at the evaporation sites, the kinetic and equilibrium isotope effects, the leaf-to-air VPD experienced and the atmospheric water vapour signal, as follows:

inline image

where R is the ratio of heavy to light isotopes, and subscripts t and e represent transpired water vapour and (liquid) water at the evaporation sites, respectively, α* is the liquid–vapour equilibration isotope effect corrected to leaf temperature (Majoube 1971); αk is the kinetic isotope diffusion fractionation factor, through the stomatal pore, which is derived from the relative rates of diffusion of the light to heavy isotope, which is 1·0285 (Merlivat 1978). When using a leaf cuvette to measure photosynthetic gas exchange, ea represents the H2O concentration the leaf is exposed to and is reflected by the H2O concentration leaving the cuvette.

Because at ISS the transpired water vapour has the same signal as source water, Rt can be substituted by Rs, the isotopic composition for stem water, and then solved for Re as follows (White 1988; Flanagan et al. 1991a):

inline image

Since fractionation is different through turbulent and laminar flow, the model was further adapted to take into account leaf boundary layer effects as follows (Flanagan et al. 1991a):

inline image

where es is the vapour pressure at the leaf surface and αkb is the ratio of the diffusion of light to heavy isotope molecules in a boundary layer (1·0189). Vapour pressure at the leaf surface is calculated from the evaporation and boundary layer resistance measured during leaf gas exchange by rearranging the following:

inline image

where E is the evaporation, rb the boundary layer resistance and P atmospheric pressure.

Re can then be expressed in the delta notation with respect to a set standard by δ (‰) = [(Rsample/Rstandard) – 1]×103 where Rstandard for 18O/16O is SMOW. Correct derivation of δe using the above, therefore requires the leaf to be at ISS.

By consideration of mass balance, the vapour pressure of H218O in air as it passes through a leaf chamber can be described by

inline image

where e is the vapour pressure, R is the ratio 18O/16O and the subscripts in, out and diff refer to the water vapour entering the cuvette, leaving the cuvette and that added by evaporation from the leaf. Rt is the oxygen isotope ratio of transpired water vapour. The above expression holds assuming that there is no change in temperature, no condensation within the cuvette, that all the water vapour is collected and the ediff remains constant over the measurement period (the authors acknowledge X. F. Wang for helpful discussion).

Solving for Rt, and expressing isotopic composition in the small delta notation (‰) gives

inline image

Using the term ξL to represent the proportion of H2O transpired relative to that in the ambient air allows simplification of the above expression to one analogous to measured Δ13CO2, as follows:

inline image

where ξL= eout/(eout– ein) and δin, ein and δout, eout are the oxygen isotope composition and vapour pressure (in mbar) of the water vapour entering and leaving the cuvette, respectively.

Ancillary