Keeling Plot Analysis Research Articles

Marine N2O cycling from high spatial resolution concentration, stable isotopic and isotopomer measurements along a meridional transect in the eastern Pacific Ocean

Nitrous oxide (N2O) is a potent greenhouse gas and ozone depleting substance, with the ocean accounting for about one third of global emissions. In marine environments, a significant amount of N2O is produced by biological processes in Oxygen Deficient Zones (ODZs). While recent technological advances are making surface N2O concentration more available, high temporal and spatial resolution water-column N2O concentration data are relatively scarce, limiting global N2O ocean models’ predictive capability. We present a N2O concentration, stable isotopic composition and isotopomer dataset of unprecedently large spatial coverage and depth resolution in the broader Pacific, crossing both the eastern tropical South and North Pacific Ocean ODZs collected as part of the GO-SHIP P18 repeat hydrography program in 2016/2017. We complement these data with dissolved gases (nitrogen, oxygen, argon) and nitrate isotope data to investigate the pathways controlling N2O production in relation to apparent oxygen utilization and fixed nitrogen loss. N2O yield significantly increased under low oxygen conditions near the ODZs. Keeling plot analysis revealed different N2O sources above the ODZs under different oxygen regimes. Our stable isotopic data and relationships between the N2O added by microbial processes (ΔN2O) and dissolved inorganic nitrogen (DIN) deficit confirm increased N2O production by denitrification under low oxygen conditions near the oxycline where the largest N2O accumulations were observed. The slope for δ18O-N2O versus site preference (SP, the difference between the central (α) and outer (β) N atoms in the linear N2O molecule) in the eastern tropical North Pacific ODZ was lower than expected for pure N2O reduction, likely because of the observed decrease in δ15Nβ. This trend is consistent with prior ODZ studies and attributed to concurrent production of N2O from nitrite with a low δ15N or denitrification with a SP >0‰. We estimated apparent isotope effects for N2O consumption in the ETNP ODZ of 3.6‰ for 15Nbulk, 9.4‰ for 15Nα, -2.3‰ for 15Nβ, 12.0‰ for 18O, and 11.7‰ for SP. These values were generally within ranges previously reported for previous laboratory and field experiments.

IRIS analyser assessment reveals sub-hourly variability of isotope ratios in carbon dioxide at Baring Head, New Zealand's atmospheric observatory in the Southern Ocean

Abstract. We assess the performance of an isotope ratio infrared spectrometer (IRIS) to measure carbon (δ13C) and oxygen (δ18O) isotope ratios in atmospheric carbon dioxide (CO2) and report observations from a 26 d field deployment trial at Baring Head, New Zealand, NIWA's atmospheric observatory for Southern Ocean baseline air. Our study describes an operational method to improve the performance in comparison to previous publications on this analytical instrument. By using a calibration technique that reflected the principle of identical treatment of sample and reference gases, we achieved a reproducibility of 0.07 ‰ for δ13C-CO2 and 0.06 ‰ for δ18O-CO2 over multiple days. This performance is within the extended compatibility goal of 0.1 ‰ for both δ13C-CO2 and δ18O-CO2, which was recommended by the World Meteorological Organization (WMO). Further improvement in measurement performance is desirable to also meet the WMO network compatibility goals of 0.01 ‰ for δ13C-CO2 and 0.05 ‰ for δ18O-CO2, which is needed to resolve the small variability that is typical for background air observatories such as Baring Head. One objective of this study was to assess the capabilities and limitations of the IRIS analyser to resolve δ13C-CO2 and δ18O-CO2 variations under field conditions. Therefore, we selected multiple events within the 26 d record for Keeling plot analysis. This resolved the isotopic composition of endmembers with an uncertainty of ≤ 1 ‰ when the magnitude of CO2 signals is larger than 10 ppm. The uncertainty of the Keeling plot analysis strongly increased for smaller CO2 events (2–7 ppm), where the instrument performance is the limiting factor and may only allow for the distinction between very different endmembers, such as the role of terrestrial versus oceanic carbon cycle processes.

Continuous CH<sub>4</sub> and <i>δ</i><sup>13</sup>CH<sub>4</sub> measurements in London demonstrate under-reported natural gas leakage

Abstract. Top-down greenhouse gas measurements can be used to independently assess the accuracy of bottom-up emission estimates. We report atmospheric methane (CH4) mole fractions and δ13CH4 measurements from Imperial College London from early 2018 onwards using a Picarro G2201-i analyser. Measurements from March 2018 to October 2020 were compared to simulations of CH4 mole fractions and δ13CH4 produced using the NAME (Numerical Atmospheric-dispersion Modelling Environment) dispersion model coupled with the UK National Atmospheric Emissions Inventory, UK NAEI, and a global inventory, the Emissions Database for Global Atmospheric Research (EDGAR), with model spatial resolutions of ∼ 2, ∼ 10, and ∼ 25 km. Simulation–measurement comparisons are used to evaluate London emissions and the source apportionment in the global (EDGAR) and UK national (NAEI) emission inventories. Observed mole fractions were underestimated by 30 %–35 % in the NAEI simulations. In contrast, a good correspondence between observations and EDGAR simulations was seen. There was no correlation between the measured and simulated δ13CH4 values for either NAEI or EDGAR, however, suggesting the inventories' sectoral attributions are incorrect. On average, natural gas sources accounted for 20 %–28 % of the above background CH4 in the NAEI simulations and only 6 %–9 % in the EDGAR simulations. In contrast, nearly 84 % of isotopic source values calculated by Keeling plot analysis (using measurement data from the afternoon) of individual pollution events were higher than −45 ‰, suggesting the primary CH4 sources in London are actually natural gas leaks. The simulation–observation comparison of CH4 mole fractions suggests that total emissions in London are much higher than the NAEI estimate (0.04 Tg CH4 yr−1) but close to, or slightly lower than, the EDGAR estimate (0.10 Tg CH4 yr−1). However, the simulation–observation comparison of δ13CH4 and the Keeling plot results indicate that emissions due to natural gas leaks in London are being underestimated in both the UK NAEI and EDGAR.

Impact on air quality of carbon and sulfur volatile compounds emitted from hydrothermal discharges: The case study of Pisciarelli (Campi Flegrei, South Italy)

Volcanoes are currently to be regarded as natural sources of air pollutants. Climatic and environmental forcing of large volcanic eruptions are well known, although gases emitted through passive degassing during periods of quiescence or hydrothermal activity can also be highly dangerous for the environment and public health. Based on compositional and isotopic data, a survey on the spatial distribution in air of the main volatile compounds of carbon (CO2 and CH4) and sulfur (H2S and SO2) emitted from the fumarolic field of Pisciarelli (Campi Flegrei, Pozzuoli, Naples), a hydrothermal area where degassing activity has visibly increased since 2009, was carried out. The main goals of this study were (i) to evaluate the impact on air quality of these natural manifestations and (ii) inquire into the behavior of the selected chemical species once released in air, and their possible use as tracers to distinguish natural and anthropogenic sources. Keeling plot analysis of CO2 and CH4 isotopes revealed that the hydrothermal area acts as a net source of CO2 in air, whilst CH4 originated mainly from anthropogenic sources. Approaching the urban area, anthropogenic sources of CO2 increased and, at distances greater than 800 m from the Pisciarelli field, they prevailed over the hydrothermal signal. While hydrothermal CO2 simply mixed with that in the atmospheric background, H2S was possibly affected by oxidation processes. Therefore, SO2 measured in the air near the hydrothermal emissions had a secondary origin, i.e. generated by oxidation of hydrothermal H2S. Anthropogenic SO2 was recognized only in the furthest measurement site from Pisciarelli. Finally, in the proximity of a geothermal well, whose drilling was in progress during our field campaign, the H2S concentrations have reached values up to 3 orders of magnitude higher than the urban background, claiming the attention of the local authorities.

Quantifying Nitrous Oxide Cycling Regimes in the Eastern Tropical North Pacific Ocean With Isotopomer Analysis

AbstractNitrous oxide (N2O), a potent greenhouse gas, is produced disproportionately in marine oxygen deficient zones (ODZs). To quantify spatiotemporal variation in N2O cycling in an ODZ, we analyzed N2O concentration and isotopologues along a transect through the eastern tropical North Pacific (ETNP). At several stations along this transect, N2O concentrations reached a near surface maximum that exceeded prior measurements in this region, of up to 226.1 ± 20.5 nM at the coast. Above the σθ = 25.0 kg/m3 isopycnal, Keeling plot analysis revealed two sources to the near‐surface N2O maximum, with different δ15N2Oα and δ15N2Oβ values, but each with a site preference (SP) of 6‰–8‰. Given expected SPs for nitrification and denitrification, each of these sources could be comprised of 17%–26% nitrification (bacterial or archeal), and 74%–83% denitrification (or nitrifier‐denitrification). Below the σθ = 25.0 kg/m3 isopycnal, box model analysis indicated that the observed 46‰–50‰ SPs in the anoxic core of the ODZ cannot be reproduced in a steady state context without an SP for N2O production by denitrification, and may indicate instead a transient net consumption of N2O. Furthermore, time‐dependent model results indicated that while δ15N2Oα and δ18O‐N2O reflect both N2O production and consumption in the anoxic core of the ODZ, δ15N2Oβ predominantly reflects N2O sources. Finally, we infer that the high (N2O) observed at some stations derive from a set of conditions supporting high rates of N2O production that have not been previously encountered in this region.

Seasonal and diurnal variations of greenhouse gases in Florence (Italy): Inferring sources and sinks from carbon isotopic ratios

In this study, the results of a continuous monitoring of (i) CO2 fluxes, and (ii) CO2 and CH4 concentrations and carbon isotopic ratios (δ13C-CO2 and δ13C-CH4) in air, carried out from 7 to 21 July 2017 and from October 10 to December 15, 2017 in the city centre of Florence, are presented. The measurements were performed from the roof of the historical building of the Ximenes Observatory. CO2 flux data revealed that the metropolitan area acted as a net source of CO2 during the whole observation period. According to the Keeling plot analysis, anthropogenic contributions to atmospheric CO2 were mainly represented by vehicular traffic (about 30%) and natural gas combustion (about 70%), the latter contributing 7 times more in December than in July. Moreover, the measured CO2 fluxes were about 80% higher in fall than in summer, confirming that domestic heating based on natural gas is the dominant CO2 emitting source in the municipality of Florence. Even though the continuous monitoring revealed a shift in the δ13C-CO2 values related to photosynthetic uptake of atmospheric CO2, the isotopic effect induced by plant activity was restricted to few hours in October and, to a lesser extent, in November. This suggests that urban planning policies should be devoted to massively increase green infrastructures in the metropolitan area in order to counterbalance anthropogenic emissions. During fall, the atmospheric CH4 concentrations were sensibly higher with respect to those recorded in summer, whilst the δ13C-CH4 values shifted towards heavier values. The Keeling plot analysis suggested that urban CH4 emissions were largely related to fugitive emissions from the natural gas distribution pipeline network. On the other hand, δ13C-CH4 monitoring allowed to recognize vehicular traffic as a minor CH4 emitting source.

Nitrous oxide cycling in the Eastern Tropical South Pacific as inferred from isotopic and isotopomeric data

The ocean accounts for up to 25% of global emissions of nitrous oxide (N2O), a potent greenhouse gas. Much of this N2O flux occurs in upwelling regions near the ocean's oxygen deficient zones (ODZs), areas known for intense N2O cycling. The Eastern Tropical South Pacific (ETSP) ODZ is one such area, and large uncertainties surround the balance of processes regulating N2O production and emission in this region. Here we examined the distributions of dissolved N2O concentration and stable isotopic composition, in concert with nitrate (NO3-) and nitrite (NO2-) isotopic ratios, to understand the mechanisms that drive N2O production, consumption, and emission from the ETSP ODZ. Keeling plot analysis identified N2O production from both nitrification and denitrification (or nitrifier-denitrification) in the near-surface and in the oxycline, where the largest accumulations of N2O were found. In the N2O concentration maximum that occurs below the ODZ, a higher 15N site preference (SP) indicated nitrification was more prominent. Within the ODZ, significant enrichments were apparent in δ15Nbulk (14–22‰), δ18ON2O (68–100‰) and SP (39–60‰), implying active N2O consumption. Further scrutiny of N2O isotope data in the ODZ highlights a deviation from the relative increases in δ18ON2O and SP expected for bacterial denitrification. At high levels of N2O consumption, SP increased more than expected for the increase in δ18ON2O. This appeared to be due, at least in part, to a decrease in δ15Nβ driven by N2O production in the ODZ, rather than further increases in δ15Nα. Isotopic analysis of co-occurring NO3- and NO2- suggests that NO3- may be the dominant source of N2O in the offshore ETSP ODZ.

Plume mapping and isotopic characterisation of anthropogenic methane sources

Methane stable isotope analysis, coupled with mole fraction measurement, has been used to link isotopic signature to methane emissions from landfill sites, coal mines and gas leaks in the United Kingdom. A mobile Picarro G2301 CRDS (Cavity Ring-Down Spectroscopy) analyser was installed on a vehicle, together with an anemometer and GPS receiver, to measure atmospheric methane mole fractions and their relative location while driving at speeds up to 80 kph. In targeted areas, when the methane plume was intercepted, air samples were collected in Tedlar bags, for δ13C–CH4 isotopic analysis by CF-GC-IRMS (Continuous Flow Gas Chromatography-Isotope Ratio Mass Spectrometry). This method provides high precision isotopic values, determining δ13C–CH4 to ±0.05 per mil. The bulk signature of the methane plume into the atmosphere from the whole source area was obtained by Keeling plot analysis, and a δ13C–CH4 signature, with the relative uncertainty, allocated to each methane source investigated. Both landfill and natural gas emissions in SE England have tightly constrained isotopic signatures. The averaged δ13C–CH4 for landfill sites is −58 ± 3‰. The δ13C–CH4 signature for gas leaks is also fairly constant around −36 ± 2‰, a value characteristic of homogenised North Sea supply. In contrast, signatures for coal mines in N. England and Wales fall in a range of −51.2 ± 0.3‰ to −30.9 ± 1.4‰, but can be tightly constrained by region. The study demonstrates that CRDS-based mobile methane measurement coupled with off-line high precision isotopic analysis of plume samples is an efficient way of characterising methane sources. It shows that isotopic measurements allow type identification, and possible location of previously unknown methane sources. In modelling studies this measurement provides an independent constraint to determine the contributions of different sources to the regional methane budget and in the verification of inventory source distribution.

The stable isotopic signature of biologically produced molecular hydrogen (H<sub>2</sub>)

Abstract. Biologically produced molecular hydrogen (H2) is characterised by a very strong depletion in deuterium. Although the biological source to the atmosphere is small compared to photochemical or combustion sources, it makes an important contribution to the global isotope budget of H2. Large uncertainties exist in the quantification of the individual production and degradation processes that contribute to the atmospheric budget, and isotope measurements are a tool to distinguish the contributions from the different sources. Measurements of δ D from the various H2 sources are scarce and for biologically produced H2 only very few measurements exist. Here the first systematic study of the isotopic composition of biologically produced H2 is presented. In a first set of experiments, we investigated δ D of H2 produced in a biogas plant, covering different treatments of biogas production. In a second set of experiments, we investigated pure cultures of several H2 producing microorganisms such as bacteria or green algae. A Keeling plot analysis provides a robust overall source signature of δ D = −712‰ (±13‰) for the samples from the biogas reactor (at 38 °C, δ DH2O= +73.4‰), with a fractionation constant ϵH2-H2O of −689‰ (±20‰) between H2 and the water. The five experiments using pure culture samples from different microorganisms give a mean source signature of δ D = −728‰ (±28‰), and a fractionation constant ϵH2-H2O of −711‰ (±34‰) between H2 and the water. The results confirm the massive deuterium depletion of biologically produced H2 as was predicted by the calculation of the thermodynamic fractionation factors for hydrogen exchange between H2 and water vapour. Systematic errors in the isotope scale are difficult to assess in the absence of international standards for δ D of H2. As expected for a thermodynamic equilibrium, the fractionation factor is temperature dependent, but largely independent of the substrates used and the H2 production conditions. The equilibrium fractionation coefficient is positively correlated with temperature and we measured a rate of change of 2.3‰ / °C between 45 °C and 60 °C, which is in general agreement with the theoretical prediction of 1.4‰ / °C. Our best experimental estimate for ϵH2-H2O at a temperature of 20 °C is −731‰ (±20‰) for biologically produced H2. This value is close to the predicted value of −722‰, and we suggest using these values in future global H2 isotope budget calculations and models with adjusting to regional temperatures for calculating δ D values.

Winter respiration of allochthonous and autochthonous organic carbon in a subarctic clear‐water lake

We studied a small subarctic lake to assess the magnitude of winter respiration and the organic carbon (OC) source for this respiration. The concentration and stable isotopic composition (δ13C) of dissolved inorganic carbon (DIC) accumulating in the lake water under ice was analyzed over one winter (7 months). The DIC concentration increased and the δ13C of DIC decreased over time, with the greatest changes at the lake bottom. Winter respiration was 26% of annual respiration in the lake. Keeling plot analysis demonstrated that the δ13C of respired DIC varied spatially, high δ13C values occurring at shallow (2.5 m, 221.7‰%) compared with intermediate (4 m, 225.1‰) and deep (6 m, 227.8‰) locations in the lake. The variation in the δ13C of respired DIC was related to the variation in the δ13C of the sediments between locations, suggesting that sediment OC supported much of the winter respiration and that the dominant OC source for respiration was OC from benthic algae at shallow locations and settled OC, of predominately terrestrial origin, at deep locations. The respiration of OC from benthic algae constituted 55% of the winter respiration, equaling 54% of the primary production by benthic algae the previous summer. The study indicates the importance of temporal and spatial variation in respiration for the metabolism and net DIC production in unproductive high‐latitude lakes; both allochthonous and autochthonous carbon can contribute to winter DIC accumulation and, consequently, to spring CO2 emissions from lakes.

Cold air drainage in a forested valley: Investigating the feasibility of monitoring ecosystem metabolism

Our objectives were to: characterize spatial and temporal variation in wind speed, direction and air temperature within a steeply sloping 96 ha forested watershed in the Oregon Cascade Mountains; assess the area contributing to advection in cold air drainage; identify appropriate conditions for sampling advected gases representative of the entire watershed; estimate ecosystem respiration from mass balance estimates. The flow dynamics could be categorized into four periods: daytime flows, evening transition, nighttime conditions (formation of a cold air pool) and morning transition. On summer days, the wind above the canopy flowed up-valley whereas the wind direction below the canopy was often down-valley. During the evening transition, the below-canopy cold air drainage became well mixed after the vertical temperature profile at the tower became isothermal, and sodar and tethersonde data revealed that a second drainage flow developed above the canopy. Between 20:00 and 24:00 h (PST), a cold air pool formed within the valley; cold air drainage continued, but wind speed decreased as the night progressed. After sunrise, CO 2 in air passing the tower remained well mixed, but its concentration decreased. After the base of the watershed began to receive direct solar radiation, the in-canopy vertical CO 2 concentration profile became stratified and the cold air drainage ceased or was confined below the canopy. After the cold air pool formed, the entire watershed supplied the respired CO 2 that advected past the tower. During this period, the potential temperature profile within the watershed indicated a strong inversion within the watershed. Because the air passing the tower was well mixed and the cold air pool encompassed the entire watershed, it is probable that air samples collected during this period could provide the best estimate of watershed-scale respiration and carbon isotope composition of respired CO 2 (Keeling plot analysis). However, sampling prior to the formation of the cold air pool may confound interpretations as the area contributing to the tower is likely changing. Mass balance calculations using an inert tracer (SF 6) released below the canopy demonstrated that data from a single vertical profile could provide plausible estimates of ecosystem respiration at the watershed scale.

Respiration of allochthonous organic carbon in unproductive forest lakes determined by the Keeling plot method

We carried out short‐term (2 d) experiments in nine unproductive lakes in northern Sweden in order to investigate organic carbon sources supporting lake water respiration. Surface water was incubated in gas‐tight bottles in the dark, and the concentration and isotopic composition (δ13C) of dissolved inorganic carbon (DIC) were measured at the start and end of the incubations. Keeling plot analyses revealed that the δ13C of the respired carbon was between ‐28.4% and ‐30.6% in the lakes and that the respired carbon was mainly of allochthonous organic carbon (AlloOC) origin. The respiration of AlloOC corresponded well with metabolic imbalances indicated by negative net ecosystem production (NEP) values in the lake waters. Keeling plot analysis of DIC accumulating in the hypolimnion of two lakes during summer stratification showed δ13C values of around ‐26.6% for excess DIC, implying that the accumulation of DIC was mainly derived from respiration of AlloOC. Our data provide direct evidence that net heterotrophy of these lakes is caused by input and respiration of AlloOC. We conclude that the Keeling plot method is a powerful technique that enables characterization and quantification of the organic carbon sources contributing to respiration in aquatic systems.

On the application and interpretation of Keeling plots in paleo climate research – deciphering δ<sup>13</sup>C of atmospheric CO<sub>2</sub> measured in ice cores

Abstract. The Keeling plot analysis is an interpretation method widely used in terrestrial carbon cycle research to quantify exchange processes of carbon between terrestrial reservoirs and the atmosphere. Here, we analyse measured data sets and artificial time series of the partial pressure of atmospheric carbon dioxide (pCO2) and of δ13C of CO2 over industrial and glacial/interglacial time scales and investigate to what extent the Keeling plot methodology can be applied to longer time scales. The artificial time series are simulation results of the global carbon cycle box model BICYCLE. The signals recorded in ice cores caused by abrupt terrestrial carbon uptake or release loose information due to air mixing in the firn before bubble enclosure and limited sampling frequency. Carbon uptake by the ocean cannot longer be neglected for less abrupt changes as occurring during glacial cycles. We introduce an equation for the calculation of long-term changes in the isotopic signature of atmospheric CO2 caused by an injection of terrestrial carbon to the atmosphere, in which the ocean is introduced as third reservoir. This is a paleo extension of the two reservoir mass balance equations of the Keeling plot approach. It gives an explanation for the bias between the isotopic signature of the terrestrial release and the signature deduced with the Keeling plot approach for long-term processes, in which the oceanic reservoir cannot be neglected. These deduced isotopic signatures are similar (−8.6‰) for steady state analyses of long-term changes in the terrestrial and marine biosphere which both perturb the atmospheric carbon reservoir. They are more positive than the δ13C signals of the sources, e.g. the terrestrial carbon pools themselves (−25‰). A distinction of specific processes acting on the global carbon cycle from the Keeling plot approach is not straightforward. In general, processes related to biogenic fixation or release of carbon have lower y-intercepts in the Keeling plot than changes in physical processes, however in many case they are indistinguishable (e.g. ocean circulation from biogenic carbon fixation).

Seasonal and interannual variations of carbon and oxygen isotopes of respired CO2 in a tallgrass prairie: Measurements and modeling results from 3 years with contrasting water availability

We made weekly measurements of carbon (δ13C) and oxygen (δ18O) isotopes of atmospheric CO2 in a C3/C4 tallgrass prairie during the growing season for 3 years with contrasting soil moisture conditions. Air samples above and within canopies were collected using 100‐ml flasks at night to characterize isotopic composition of ecosystem respiration. We used a two‐source mixing line (Keeling plot) approach to estimate isotope ratios of ecosystem respired CO2 for both carbon (δ13CR) and oxygen (δ18OR). Measured net ecosystem CO2 exchange (NEE) showed the largest net carbon uptake in 2004, followed by 2003 and 2002. This interannual difference in NEE strongly depends on the amount and distribution of precipitation received by this tallgrass prairie. Precipitation also affects the timing of the seasonal transition from C3 dominance in spring to C4 dominance in summer. Variations of δ13CR showed that C4 plants dominated ecosystem respiration in 2003 and 2004, except in early spring when C3 plants were more active. In contrast, contributions of C3 plants were relatively higher for an extended period in the summer of 2002, when a severe drought occurred. Typically, C3 forbs extract water and nutrients from soil layers below that of the C4 grasses and remain photosynthetically active in periods when C4 grasses have water stress that limits photosynthesis. Drought‐reduced C4 grass photosynthesis was lower than temperature‐limited C3 forb growth during this period. We used an integrated isotope land surface model (ISOLSM) to simulate (and compare to measurements) net CO2 fluxes, δ18O values of leaf and soil water, and δ18O values of aboveground and soil respiration. The Keeling plot analysis becomes less reliable for estimating δ18OR values when the surface soil is dry. We suspect this is due to low CO2 production in the soil when water is limiting, in which case the invasion (abiotic) effect is more significant. ISOLSM reasonably captured seasonal variations of measured δ18OR in all 3 years, indicating the model's consistency of predicting δ18OR in different soil water conditions. Model simulations also showed that nighttime δ18O values of aboveground respiration were variable, often becoming very positive in water‐stressed conditions primarily because of the low relative humidity and resultant elevated δ18O values of leaf water.