• Angert, A., , Rachmilevitch S. , , Barkan E. , , and Luz B. , 2003: Effects of photorespiration, the cytochrome pathway, and the alternative pathway on the triple isotopic composition of atmospheric O2. Global Biogeochem. Cycles, 17, 1030, doi:10.1029/2002GB001933.

    • Search Google Scholar
    • Export Citation
  • Barkan, E., , and Luz B. , 2005: High precision measurements of 17O/16O and 18O/16O ratios in H2O. Rapid Commun. Mass Spectrom., 19, 37373742.

    • Search Google Scholar
    • Export Citation
  • Bender, M. L., 1990: The δ18O of dissolved O2 in seawater: A unique tracer of circulation and respiration in the deep sea. J. Geophys. Res., 95 (C12), 22 24322 252.

    • Search Google Scholar
    • Export Citation
  • Bender, M. L., , and Grande K. D. , 1987: Production, respiration, and the isotope geochemistry of O2 in the upper water column. Global Biogeochem. Cycles, 1, 4959.

    • Search Google Scholar
    • Export Citation
  • Benson, B. B., , and Krause D. Jr., 1980: The concentration and isotopic fractionation of gases dissolved in freshwater in equilibrium with the atmosphere. 1. Oxygen. Limnol. Oceanogr., 25, 662671.

    • Search Google Scholar
    • Export Citation
  • Cassar, N., , Barnett B. A. , , Bender M. L. , , Kaiser J. , , Hamme R. C. , , and Tilbrook B. , 2009: Continuous high-frequency dissolved O2/Ar measurements by equilibrator inlet mass spectrometry. Anal. Chem., 81, 18551861.

    • Search Google Scholar
    • Export Citation
  • Copin-Montegut, C., 1985: A method for the continuous determination of the partial pressure of carbon dioxide in the upper ocean. Mar. Chem., 17, 1321.

    • Search Google Scholar
    • Export Citation
  • Emerson, S., , Stump C. , , Wilbur D. , , and Quay P. , 1999: Accurate measurement of O2, N2 and Ar gases in water and the solubility of N2. Mar. Chem., 64, 337347.

    • Search Google Scholar
    • Export Citation
  • Frankignoulle, M., , Borges A. , , and Biondo R. , 2001: A new design of equilibrator to monitor carbon dioxide in highly dynamic and turbid environments. Water Res., 35, 13441347.

    • Search Google Scholar
    • Export Citation
  • Garcia, H. E., , and Gordon L. I. , 1992: Oxygen solubility in seawater: Better fitting equations. Limnol. Oceanogr., 37, 13071312.

  • Guéguen, C., , and Tortell P. D. , 2008: High-resolution measurement of Southern Ocean CO2 and O2/Ar by membrane inlet mass spectrometry. Mar. Chem., 108, 184194.

    • Search Google Scholar
    • Export Citation
  • Hendricks, M. B., , Bender M. L. , , and Barnett B. A. , 2004: Net and gross O2 production in the Southern Ocean from measurements of biological O2 saturation and its triple isotope composition. Deep-Sea Res. I, 51, 15411561.

    • Search Google Scholar
    • Export Citation
  • Johnson, J. E., 1999: Evaluation of a seawater equilibrator for shipboard analysis of dissolved oceanic trace gases. Anal. Chim. Acta, 395, 119132.

    • Search Google Scholar
    • Export Citation
  • Juranek, L. W., , and Quay P. D. , 2005: In vitro and in situ gross primary and net community production in the North Pacific Subtropical Gyre using labeled and natural abundance isotopes of dissolved O2. Global Biogeochem. Cycles, 19, GB3009, doi:10.1029/2004GB002384.

    • Search Google Scholar
    • Export Citation
  • Kaiser, J., , Reuer M. K. , , Barnett B. , , and Bender M. L. , 2005: Marine productivity estimates from continuous O2/Ar ratio measurements by membrane inlet mass spectrometry. Geophys. Res. Lett., 32, L19605, doi:10.1029/2005GL023459.

    • Search Google Scholar
    • Export Citation
  • Keeling, R. F., , Blaine T. , , Paplawsky B. , , Katz L. , , Atwood C. , , and Brockwell T. , 2004: Measurement of changes in atmospheric Ar/N2 ratio using a rapid-switching, single-capillary mass spectrometer system. Tellus, 56B, 322338.

    • Search Google Scholar
    • Export Citation
  • Knox, M., , Quay P. D. , , and Wilbur D. , 1992: Kinetic isotopic fractionation during air–water gas transfer of O2, N2, CH4, and H2. J. Geophys. Res., 97 (C12), 20 33520 343.

    • Search Google Scholar
    • Export Citation
  • Körtzinger, A., , Thomas H. , , Schneider B. , , Gronau N. , , Mintrop L. , , and Duinker J. C. , 1996: At-sea intercomparison of two newly designed underway pCO2 systems—Encouraging results. Mar. Chem., 52, 133145.

    • Search Google Scholar
    • Export Citation
  • Levine, N. M., , Bender M. L. , , and Doney S. C. , 2009: The δ18O of dissolved O2 as a tracer of mixing and respiration in the mesopelagic ocean. Global Biogeochem. Cycles, 23, GB1006, doi:10.1029/2007GB003162.

    • Search Google Scholar
    • Export Citation
  • Luz, B., , and Barkan E. , 2000: Assessment of oceanic productivity with the triple-isotope composition of dissolved oxygen. Science, 288, 20282031.

    • Search Google Scholar
    • Export Citation
  • McNeil, C., , Katz D. , , Wanninkhof R. , , and Johnson B. , 2005: Continuous shipboard sampling of gas tension, oxygen and nitrogen. Deep-Sea Res. I, 52, 17671785.

    • Search Google Scholar
    • Export Citation
  • McNeil, C., , D’Asaro E. , , Johnson B. , , and Horn M. , 2006: A gas tension device with response times of minutes. J. Atmos. Oceanic Technol., 23, 15391558.

    • Search Google Scholar
    • Export Citation
  • Murphy, P. P., , Nojiri Y. , , Fujinuma Y. , , Wong C. S. , , Zeng J. , , Kimoto T. , , and Kimoto H. , 2001: Measurements of surface seawater fCO2 from volunteer commercial ships: Techniques and experiences from Skaugran. J. Atmos. Oceanic Technol., 18, 17191734.

    • Search Google Scholar
    • Export Citation
  • Quay, P. D., , Emerson S. , , Wilbur D. O. , , and Stump C. , 1993: The δ18O of dissolved O2 in the surface waters of the subarctic Pacific: A tracer of biological productivity. J. Geophys. Res., 98 (C5), 84478458.

    • Search Google Scholar
    • Export Citation
  • Quiñones-Rivera, Z. J., , Wissel B. , , Justić D. , , and Fry B. , 2007: Partitioning oxygen sources and sinks in a stratified, eutrophic coastal ecosystem using stable oxygen isotopes. Mar. Ecol. Prog. Ser., 342, 6983.

    • Search Google Scholar
    • Export Citation
  • Reuer, M. K., , Barnett B. A. , , Bender M. L. , , Falkowski P. G. , , and Hendricks M. B. , 2007: New estimates of Southern Ocean biological production rates from O2/Ar ratios and the triple isotope composition of O2. Deep-Sea Res. I, 54, 951974.

    • Search Google Scholar
    • Export Citation
  • Sarma, V. V. S. S., , Abe O. , , Hashimoto S. , , Hinuma A. , , and Saino T. , 2005: Seasonal variations in triple oxygen isotopes and gross oxygen production in the Sagami Bay, central Japan. Limnol. Oceanogr., 50, 544552.

    • Search Google Scholar
    • Export Citation
  • Schneider, B., , Kremling K. , , and Duinker J. C. , 1992: CO2 partial pressure in northeast Atlantic and adjacent shelf water: Processes and seasonal variability. J. Mar. Syst., 3, 453463.

    • Search Google Scholar
    • Export Citation
  • Schneider, B., , Sadkowiak B. , , and Wachholz F. , 2007: A new method for continuous measurements of O2 in surface water in combination with pCO2 measurements: Implications for gas phase equilibration. Mar. Chem., 103, 163171.

    • Search Google Scholar
    • Export Citation
  • Stanley, R. H. R., , Kirkpatrick J. B. , , Cassar N. , , Barnett B. A. , , and Bender M. L. , 2010: Net community production and gross primary production rates in the western equatorial Pacific. Global Biogeochem. Cycles, 24, GB4001, doi:10.1029/2009GB003651.

    • Search Google Scholar
    • Export Citation
  • Thiemens, M. H., , Jackson T. , , Zipf E. C. , , Erdman P. W. , , and van Egmond C. , 1995: Carbon dioxide and oxygen isotope anomalies in the mesosphere and stratosphere. Science, 270, 969972.

    • Search Google Scholar
    • Export Citation
  • Tortell, P. D., 2005: Dissolved gas measurements in oceanic waters made by membrane inlet mass spectrometry. Limnol. Oceanogr. Methods, 3, 2437.

    • Search Google Scholar
    • Export Citation
  • Wanninkhof, R., 1992: Relationship between wind speed and gas exchange over the ocean. J. Geophys. Res., 97 (C5), 73737382.

  • Weiss, R. F., , and Price B. A. , 1980: Nitrous oxide solubility in water and seawater. Mar. Chem., 8, 347359.

  • View in gallery

    Flow diagram for the mass spectrometer inlet. Pressures in the sample and reference lines and the line going to the mass spectrometer are controlled by MKS type 250 controllers. The changeover valve selects between the sample and reference gases every 10 s. A fraction of the selected gas goes through a capillary into the mass spectrometer for sampling. The rest of the selected gas and the unselected gas are wasted to a vacuum pump.

  • View in gallery

    (a) Equilibrator used to measure dissolved oxygen. The main equilibration tube is on the left in the photo, and is filled with glass Raschig rings to increase the surface area for gas exchange and increase the travel time of gas and water. The water outflow tube is on the right in the photo. Stoppers are used to set the water outflow height. (b) A line drawing of the equilibrator showing the air- and water flows. Air bubbles up from the bottom (dashed arrows) and water falls from a showerhead. A liquid water sensor above the equilibrator prevents water from entering the air line.

  • View in gallery

    Airflow in the pressure-controlled equilibrator loop. Dashed region and line show the setup for N2 gas additions to determine the equilibration time, as described in section 4c.

  • View in gallery

    Schematic of the gas tension device. A peristaltic pump pushes water through a MicroModule Membrane Contactor (arrows show water flow). A needle valve keeps the hydrostatic pressure high in the gas tension device. A differential pressure transducer is connected to the air volume in the membrane contactor.

  • View in gallery

    Field setup at the SIO pier. Arrows show the water flow. Water is pumped from a depth of 2 m below MLLW. Most of the water goes into a tank, which houses the optode and gas tension device. The rest of the water is filtered with a 130-μm strainer. Some of this water flows into the equilibrator, which is also shown in Fig. 2. The rest of the filtered water goes into a bag filter, from which water is pumped through the gas tension device.

  • View in gallery

    Model of gas transfer between water and headspace: Q is the water flow rate, Pi is the headspace pressure of gas i, Ci is the concentration of gas i in the water entering the equilibrator, and k is the gas exchange coefficient.

  • View in gallery

    Example of the gas tension device response after its air volume was vented to the atmosphere. The line shows the fit to a model with a linear trend and an exponential decay of a pulse. Water was pumped through the GTD at 100 mL min−1. The e-folding time is 5.35 ± 0.59 min.

  • View in gallery

    (a) Gas tension measured from the gas tension device, and calculated from the optode and mass spectrometer using Eqs. (6) and (20)(25). (b) The calculated gas tension subtracted from the measured gas tension. The measured gas tension is on average 4 torr higher than expected from the optode and mass spectrometer measurements.

  • View in gallery

    Response in δ(Ar/N2) to N2 gas addition during one test. The line shows the fit to a model with a linear trend and an exponential decay of a pulse. Data between 6 and 13 min were removed for the fit. The weighted average of τ based on four tests is 7.36 ± 0.74 min.

  • View in gallery

    Modeled relationship between the time constant τ for δ(Ar/N2) and at a water flow rate of 5.6 L min−1. At large , τ becomes insensitive to . Horizontal dashed lines show the mean and 1σ uncertainty of τ determined experimentally (7.36 ± 0.74 min). Vertical dotted lines show the values of corresponding to the measured τ (7.3, 9.1, and 12.8 L min−1).

  • View in gallery

    Model prediction of the sensitivity of (a) δ(O2/N2), (b) δ(O2/Ar), (c) δ18O, and (d) Δ17O to equilibrator pressure and makeup gas composition, for a gas tension of 790 torr, a makeup gas flow rate of 5 mL min−1, a Q of 5.6 L min−1, and a of 7.3 L min−1 (black lines) or a approaching infinity (gray lines). Here, ΔP is the headspace pressure minus the gas tension. The y axis shows the difference between the steady-state delta value of the headspace (i.e., the values that would be observed) and the delta value of the makeup gas. The contour values show the correction, defined in Eq. (27), for a of 7.3 L min−1, in units of per mil. Some contour values are also shown for a approaching infinity.

  • View in gallery

    Hourly mean percent oxygen saturation measured with the optode and the mass spectrometer. Gray triangles show points where the oxygen percent saturation was within 5% for the last 5 h. The black text gives the slope of all the data, and the gray text gives the slope of the more stable data. Looking only at periods where the oxygen saturation was stable, the slope of the data comes closer to the ideal slope of 1.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 14 14 2
PDF Downloads 11 11 5

An Equilibrator System to Measure Dissolved Oxygen and Its Isotopes

View More View Less
  • 1 Scripps Institution of Oceanography, University of California, San Diego, La Jolla, California
© Get Permissions
Full access

Abstract

An equilibrator is presented that is designed to have a sufficient equilibration time even for insoluble gases, and to minimize artifacts associated with not equilibrating to the total gas tension. A gas tension device was used to balance the pressure inside the equilibrator with the total gas tension. The equilibrator has an e-folding time of 7.36 ± 0.74 min for oxygen and oxygen isotopes, allowing changes on hourly time scales to be easily resolved. The equilibrator delivers “equilibrated” air at a flow rate of 3 mL min−1 to an isotope ratio mass spectrometer. The high gas sampling flow rate would allow the equilibrator to be interfaced with many potential devices, but further development may be required for use at sea. This system was tested at the Scripps Institution of Oceanography pier, in La Jolla, California. A mathematical model validated with performance tests was used to assess the sensitivity of the equilibrated air composition to headspace pressure and makeup gas composition. Parameters in this model can be quantified to establish corrections under different operating conditions. For typical observed values, under the operating conditions presented here, the uncertainty in the measurement due to the equilibrator system is 2.2 per mil for δ(O2/N2), 1.5 per mil for δ(O2/Ar), 0.059 per mil for δ18O, and 0.0030 per mil for Δ17O.

Corresponding author address: Lauren Rafelski, Scripps Institution of Oceanography, University of California, San Diego, 9500 Gilman Dr., La Jolla, CA 92093-0244. E-mail: lauren.rafelski@gmail.com

Abstract

An equilibrator is presented that is designed to have a sufficient equilibration time even for insoluble gases, and to minimize artifacts associated with not equilibrating to the total gas tension. A gas tension device was used to balance the pressure inside the equilibrator with the total gas tension. The equilibrator has an e-folding time of 7.36 ± 0.74 min for oxygen and oxygen isotopes, allowing changes on hourly time scales to be easily resolved. The equilibrator delivers “equilibrated” air at a flow rate of 3 mL min−1 to an isotope ratio mass spectrometer. The high gas sampling flow rate would allow the equilibrator to be interfaced with many potential devices, but further development may be required for use at sea. This system was tested at the Scripps Institution of Oceanography pier, in La Jolla, California. A mathematical model validated with performance tests was used to assess the sensitivity of the equilibrated air composition to headspace pressure and makeup gas composition. Parameters in this model can be quantified to establish corrections under different operating conditions. For typical observed values, under the operating conditions presented here, the uncertainty in the measurement due to the equilibrator system is 2.2 per mil for δ(O2/N2), 1.5 per mil for δ(O2/Ar), 0.059 per mil for δ18O, and 0.0030 per mil for Δ17O.

Corresponding author address: Lauren Rafelski, Scripps Institution of Oceanography, University of California, San Diego, 9500 Gilman Dr., La Jolla, CA 92093-0244. E-mail: lauren.rafelski@gmail.com

1. Introduction

Measurements of dissolved gases in the ocean can give information about air–sea gas exchange, oceanic carbon uptake, and other biological and physical processes. Because biological oxygen production and consumption are closely tied to carbon consumption and production, measurements of processes affecting oxygen can be used to determine gross and net biological carbon uptake in the ocean. In particular, dissolved oxygen isotopes can be used together with dissolved oxygen concentration to determine rates of gross primary production and net community production. Here, we show that accurate, continuous measurements of dissolved oxygen isotopes can be made from a stationary platform using a novel equilibrator design.

As defined by
e1
δ18O has been used with oxygen measurements as a tracer of water masses (Bender 1990; Levine et al. 2009), as well as a measure of the ratio of gross primary production to respiration (Bender and Grande 1987; Quiñones-Rivera et al. 2007). These computations rely on the fact that oxygen produced during photosynthesis is isotopically lighter than oxygen in air (i.e., less 18O) and respiration preferentially consumes lighter oxygen isotopes. For example, oxygen in air has a δ18O of 23.88 per mil on the Standard Mean Ocean Water (SMOW) scale (Barkan and Luz 2005), whereas photosynthetic oxygen has a δ18O close to zero. However, these calculations are sensitive to uncertainties in the fractionation factor of respiration (Bender and Grande 1987; Quay et al. 1993).

The triple isotopic composition of dissolved O2 (isotopes 16, 17, and 18) can be used to measure gross primary production without making assumptions about the respiration fractionation factor (Luz and Barkan 2000). Using the triple isotopic composition of oxygen to measure gross primary production is possible because, in the stratosphere, photochemical reactions related to ozone production cause mass-independent fractionation of oxygen isotopes, with the heavier 17O and 18O preferentially ending up in CO2 (Thiemens et al. 1995). This is in contrast to fractionation processes such as respiration, where the degree of fractionation is roughly proportional to the mass difference between the major isotope and the minor isotope (“mass-dependent fractionation”).

Dissolved O2 in the ocean can come from air–sea gas exchange or photosynthesis. Because of the mass-independent fractionation of oxygen in the stratosphere, photosynthetic O2 has an excess of 17O relative to atmospheric O2. This excess is defined as
e2
where λ is the relationship between δ17O and δ18O due to mass-dependent fractionation [approximately 0.52, Luz and Barkan (2000)]. Equation (2) is an approximation of
e3
where 17R or 18R is the ratio of the oxygen isotope to 16O2 (e.g., 17O16O/16O2) and Rreference is the reference ratio (Angert et al. 2003). Photosynthetic O2 has a typical Δ17O of 249 per meg relative to atmospheric O2 (Luz and Barkan 2000). This makes it possible to separate these two sources of oxygen in the ocean.

To measure the isotopic composition of dissolved oxygen, a mass spectrometer must be used. To make measurements of dissolved gases, the gases must first be extracted from a solution, because only the gases themselves can be measured on the mass spectrometer. One common method of extracting gases to measure oxygen isotopes is to take bottle samples of water and allow the dissolved gases to exchange with gases in the headspace. After an equilibration time of 8 (Emerson et al. 1999) to 24 h (Sarma et al. 2005), the headspace gas can be sampled on a mass spectrometer. This technique has been used to measure Δ17O around the world (e.g., Hendricks et al. 2004; Juranek and Quay 2005; Sarma et al. 2005; Reuer et al. 2007; Stanley et al. 2010). A major limitation of this method is that it is time consuming, which limits the number of samples that can be measured and the time scales of the processes that can be observed.

A technique used to take continuous measurements of dissolved gases is membrane inlet mass spectrometry (MIMS). In this method, the mass spectrometer has a semipermeable membrane at its inlet. Dissolved gases pass through the membrane and are measured in the mass spectrometer. This technique has been used to measure O2, N2, Ar, CO2, dimethyl sulfide (DMS), and H2S (Tortell 2005; Kaiser et al. 2005; Guéguen and Tortell 2008), but has not been used to measure oxygen isotopes.

Another method of taking rapid measurements of dissolved gases is to use an equilibrator. In an equilibrator, a large surface area of contact between headspace gas and water allows gases in the water to exchange rapidly with gases in the headspace. The headspace gas is then sampled in real time, allowing for continuous measurements of the gas of interest. A large surface area at the air–water interface can be achieved through the use of a showerhead, where water “rains” down into a headspace, and gases can exchange on the surfaces of water droplets (e.g., Johnson 1999); a bubbling system, where gases can exchange on bubble surfaces (e.g., Copin-Montegut 1985; Schneider et al. 1992); or other surfaces inside the equilibrator, such as marbles (e.g., Frankignoulle et al. 2001). These equilibrators have been used very successfully for measurements of carbon dioxide, because carbon dioxide is very soluble in water.

Using an equilibrator to measure O2 is more challenging than measuring CO2. Compared to CO2, O2 is much less soluble, which means the gas flux of O2 is lower, and the equilibration time is longer. For example, an equilibrator described by Schneider et al. (2007) has a time scale for equilibration of around 1 h for O2, compared with around 1 min for CO2 (Körtzinger et al. 1996). Faster equilibration of O2 can be achieved by reducing the headspace volume or increasing the surface area of the air–water interface. For example, Schneider et al. (2007) describe a bubble-type equilibrator used to measure O2 that has a much smaller headspace volume than a similar equilibrator used for CO2, an increased airflow, and a “frit” to increase bubble formation. In another study, Cassar et al. (2009) use a very small device (Liqui-Cel MicroModule Membrane Contactor) as an equilibrator for O2. This device contains porous membranes that create a large, compact surface area that separates water from a very small headspace volume. This device was used to measure O2 and Ar, and had a time constant of 7–8 min.

The techniques described above for measuring dissolved oxygen could also be used to measure dissolved oxygen isotopes if the equilibrating instrument were interfaced with an appropriate mass spectrometer. The isotope ratio mass spectrometer that we use requires a high flow rate of gas to stabilize the pressure in the mass spectrometer as the changeover valve switches, which allows for faster switching between the sample and the reference gases (Keeling et al. 2004). Additionally, a high flow rate allows water vapor to be removed from the sample using a cold trap. For these reasons, a device such as the MicroModule Membrane Contactor employed by Cassar et al. (2009) would not be suitable because it provides a sampling flow rate that is much too low. In this paper, we describe an equilibrator that was designed for rapid equilibration of oxygen and allows for a sampling flow rate of 3 mL min−1. It allows gas exchange on the surfaces of water droplets from a showerhead, bubbles, and Raschig rings, and has a headspace volume of approximately 930 mL.

The headspace pressure inside the equilibrator is set to the total gas tension using gas from a cylinder (the “makeup” gas). This is different than many other equilibrators, where a vent keeps the headspace at atmospheric pressure (e.g., Johnson 1999). The vent flow is typically inward because of sampling, and the mixture of outside air with equilibrated air causes a scale contraction in the measurement of the gases [the “disequilibration factor” in Johnson (1999)]. In addition, an equilibrator that does not equilibrate to the total gas tension cannot reach true equilibrium; it only approaches a steady state that is offset from true equilibrium, because different gases equilibrate at different rates. By setting the headspace pressure to the total gas tension, we minimize the pressure effect and better control the venting effect, because the composition of the makeup gas is controlled and constant.

We also present a model for describing the equilibration of major gases (O2, N2, and Ar), as well as oxygen isotopes, which differs from models for CO2 equilibration because of the need to account for the dilution of major gases by other major gases. The equilibrator, when interfaced with an IsoPrime isotope ratio mass spectrometer, allows for continuous measurements of dissolved oxygen and its isotopes, and has the potential to aid in our understanding of dissolved gas processes.

2. Methods

a. Mass spectrometer

An IsoPrime isotope ratio mass spectrometer was used to measure dissolved gas composition. This mass spectrometer has a flight tube, source, and detector integrated into a single stainless steel enclosure, and has 10 collectors that simultaneously measure 10 m/z ratios: 28, 29, 30, 32, 33, 34, 36, 40, 44, and 45. This allows for the measurement of δ(O2/N2), δ(Ar/N2), and δ(O2/Ar), as well as δ18O and δ17O, where δ(a/b) is defined similarly to Eq. (1):
e4
The difference in ratios between the sample and the reference gases is very small, so the delta values are multiplied by 103 to give units of per mil.

The mass spectrometer employs an inlet system (Fig. 1) designed following Keeling et al. (2004) to allow for continuous sampling, but the scheme modified to run at a lower flow rate. Specifically, a Licor nondispersive infrared (NDIR) CO2 analyzer was removed and the lines were changed to -in. stainless steel to reduce the volume of the inlet system. Sample and reference gases flow through precision flow control valves at 3 mL min−1 and pressures of 500 torr. The pressure is controlled by MKS type 250 controllers and 10-torr differential pressure transducers, which are referenced to a volume with a pressure of 500 torr. Gases then flow through a changeover valve, which determines whether the sample gas or the reference gas enters the mass spectrometer. The pressure in the line to the mass spectrometer is controlled at 300 torr to minimize pressure fluctuations as the changeover valve switches. The selected line flows through a pickoff “T,” where some gas flows through a capillary tube into the mass spectrometer, and most of the gas is wasted to a vacuum. Gas in the unselected line is also wasted to maintain flow. The changeover valve switches every 10 s, which allows for the calculation of delta values every 20 s. Upstream of the mass flowmeter in Fig. 1, sample gas flows through a −55°C trap made of a 16.5-cm-long piece of ¼-in. outer diameter [0.18-in. inner diameter (I.D.)] stainless steel tubing filled with 3-mm glass beads to reduce its volume. This trap removes water vapor from the sample, and lasts for 24–48 h before ice has to be removed from the trap.

Fig. 1.
Fig. 1.

Flow diagram for the mass spectrometer inlet. Pressures in the sample and reference lines and the line going to the mass spectrometer are controlled by MKS type 250 controllers. The changeover valve selects between the sample and reference gases every 10 s. A fraction of the selected gas goes through a capillary into the mass spectrometer for sampling. The rest of the selected gas and the unselected gas are wasted to a vacuum pump.

Citation: Journal of Atmospheric and Oceanic Technology 30, 2; 10.1175/JTECH-D-12-00074.1

The mass spectrometer is calibrated to account for scale contraction due to incomplete sample/reference replacement during switching (“crossover contamination”), as in Keeling et al. (2004). This calibration is of the form
e5
where δ(32/28) is the measured value from the mass spectrometer and c is the correction factor needed to convert the measurement to δ(O2/N2).

To find c, a gravimetric method was used in which a cylinder of air was prepared to be highly enriched in O2, by adding pure O2 to an evacuated cylinder, and then filling the remainder of the cylinder with air. The cylinder was weighed after evacuation, after O2 addition, and after it was filled. From these weights, the final composition of the gas in the cylinder was determined. The gas composition was then measured on the mass spectrometer. The ratio between the δ(O2/N2) of the gas in the cylinder, calculated from the weights of the added gases, and the δ(32/28) measured on the mass spectrometer gives the correction factor c. For this mass spectrometer, c was determined to be 1.2. This high correction factor is likely due to the low flow rate of the gases, which would increase the crossover contamination.

b. Equilibrator

The equilibrator is a “counterflow”-type equilibrator, where gas and water flow in opposite directions. It uses water droplets, bubbling, and Raschig rings to create the surface area for gas exchange. The equilibrator is made of clear, 5.2-cm I.D. PVC pipe that is 53.3 cm long (Fig. 2). Water showers down from the top of the equilibrator and gas bubbles up from the bottom. The equilibrator is filled with 12-mm glass Raschig rings, which are small hollow cylinders that provide a large surface area for gas exchange between the air and water, and increase the time it takes for water and gas to travel through the equilibrator. The air pressure in the equilibrator is controlled using a pressure-control loop, as described below.

Fig. 2.
Fig. 2.

(a) Equilibrator used to measure dissolved oxygen. The main equilibration tube is on the left in the photo, and is filled with glass Raschig rings to increase the surface area for gas exchange and increase the travel time of gas and water. The water outflow tube is on the right in the photo. Stoppers are used to set the water outflow height. (b) A line drawing of the equilibrator showing the air- and water flows. Air bubbles up from the bottom (dashed arrows) and water falls from a showerhead. A liquid water sensor above the equilibrator prevents water from entering the air line.

Citation: Journal of Atmospheric and Oceanic Technology 30, 2; 10.1175/JTECH-D-12-00074.1

Water falls from a showerhead at the top of the equilibrator and exits the equilibrator through a parallel 5.2-cm I.D. PVC pipe, with 2.5-cm-diameter holes drilled every 7.6 cm. Rubber stoppers can be placed inside the holes to change the height of the water outlet. A balance between the water outlet height and the air pressure inside the equilibrator determines the water height inside the equilibrator. If the water level of the outlet is too low relative to the pressure inside the equilibrator, the water level inside the equilibrator will be too low, and large bubbles will escape from the equilibrator, making full equilibration of the headspace gas impossible. If the water level of the outlet is too high, water will rise inside the equilibrator and eventually be pulled into the air lines. The adjustable water outlet height allows the equilibrator to be used under a wide range of pressures without the water level inside the equilibrator getting too low or too high. The water level in the equilibrator also determines how much of the gas exchange occurs on bubble surfaces, and how much of the exchange occurs on Raschig rings or water droplets. A clear ½-in.-diameter tube runs parallel to the main equilibrator tube, and is connected at the bottom and top of the equilibrator. This tube fills with water to a height equal to the mean water level in the equilibrator and acts as a “sight glass” to observe the water level. It is difficult to observe the mean water level in the equilibrator itself because there are many large air pockets as the water and airflow around the Raschig rings.

Gas enters the equilibrator through a ¼-in. stainless steel tube in the bottom of the equilibrator, and exits through the top of the equilibrator, via an opening next to the showerhead. An absolute pressure transducer is connected to the top of the equilibrator to measure the headspace pressure. A KNF type N05 diaphragm pump circulates the air in the equilibrator at 0.5–1 L min−1 through a loop made of ⅛-in. stainless steel tubing (Fig. 3). The total volume of the gas and water in the loop and equilibrator is approximately 1200 mL. From the loop, 3 mL min−1 of gas (STP) is continuously drawn into the mass spectrometer, via a “T” in the loop that connects to -in. stainless steel tubing, which connects to the cold trap, and then to the mass spectrometer inlet system.

Fig. 3.
Fig. 3.

Airflow in the pressure-controlled equilibrator loop. Dashed region and line show the setup for N2 gas additions to determine the equilibration time, as described in section 4c.

Citation: Journal of Atmospheric and Oceanic Technology 30, 2; 10.1175/JTECH-D-12-00074.1

In most equilibrators, the headspace pressure is kept at atmospheric pressure, using a vent to the surrounding air. The headspace pressure in the equilibrator presented here can be actively set. The headspace pressure is matched to the total gas tension so that full equilibration can be reached. To keep the loop pressure at the setpoint, the gas lost to the mass spectrometer or to bubbles escaping from the equilibrator is replaced by gas from a makeup cylinder; without the addition of makeup gas, active pressure control would be problematic. The flow rate of gas from the makeup cylinder to the loop is measured by an electronic bidirectional mass flowmeter (Honeywell, AMW2150V) with a range of ±30 mL min−1. The amount of gas that enters from the makeup cylinder is controlled using an absolute pressure transducer, a precision flow control valve, and an MKS type 250 controller. If the pressure in the loop is too low, the valve opens and gas from the makeup cylinder enters the loop. If the pressure is too high, the valve closes and air is pulled out of the loop through a continuously running 300-torr vacuum until the pressure in the loop drops to the setpoint. This system keeps the equilibrator at the setpoint pressure.

One of the biggest risks of running equilibrated gas into a mass spectrometer is the possibility of seawater entering the mass spectrometer. This risk is in part avoided by having a cold trap upstream of the mass spectrometer; small amounts of water will freeze out of the sample before going into the mass spectrometer. However, a large pressure spike in the loop could draw more water into the sample line than can be removed by the cold trap. To eliminate the possibility of seawater getting into the mass spectrometer, a liquid water sensor was installed in a clear, 2.0-cm I.D. PVC pipe in the air line at the top of the equilibrator. If water from the equilibrator gets into the air line, a four-port, two-position valve that is interfaced to the sensor will isolate the equilibrator from the sampling loop.

c. Gas tension device

For the setpoint pressure inside the equilibrator, we chose to target the total gas tension of water, which was independently measured. Gas tension is defined as the sum of the partial pressures of gases dissolved in water, including water vapor. It can be approximated as
e6
where PT is the total gas tension and Px is the partial pressure of gas x.

Commercial devices exist to measure gas tension (McNeil et al. 2005, 2006). However, these devices either have a long time constant [e.g., 11 min; McNeil et al. (2005)] or need to be a few meters underwater to work properly (e.g., McNeil et al. 2006). We built a fast-responding gas tension device (GTD) using a Liqui-Cel 0.75 × 1 MicroModule Membrane Contactor (Membrana; Fig. 4). The MicroModule Membrane Contactor is a polycarbonate housing containing a porous polypropylene membrane shaped into hollow fibers, with a total surface area of 392 cm2. Inside the hollow fibers, there is an air volume of 3.4 mL. Because the membrane is hydrophobic, liquid water cannot pass through the pores unless a pressure threshold is exceeded. Water vapor, however, will pass through the pores. At equilibrium, the pressure of the gas in the air volume equals the total gas tension.

Fig. 4.
Fig. 4.

Schematic of the gas tension device. A peristaltic pump pushes water through a MicroModule Membrane Contactor (arrows show water flow). A needle valve keeps the hydrostatic pressure high in the gas tension device. A differential pressure transducer is connected to the air volume in the membrane contactor.

Citation: Journal of Atmospheric and Oceanic Technology 30, 2; 10.1175/JTECH-D-12-00074.1

The contactor housing has two tubes for water flow, and one tube that connects to the air volume. Water is pumped over the outside of the fibers, allowing gases to exchange between the water and the air volume. A needle valve downstream of the GTD keeps the water pressure inside the GTD constant and high (around 25 kPa) so that gas tension measurements are not affected by changes in the pump pressure. A differential pressure transducer (All-Sensors Corp., range of 56 torr) is attached to the air volume via a 68.6-cm-long, -in.-diameter stainless steel tube. A barometer is used to convert the differential pressure to absolute pressure.

d. Field setup

The equilibrator, gas tension device, and mass spectrometer were used at the end of the pier at the Scripps Institution of Oceanography (SIO), in La Jolla, California (32°52.52′N, 117°15.30′W), from 22 June to 1 August 2011. The pier extends 330.4 m from shore and is 10.2 m above the mean lower low water (MLLW) level. The instruments were housed in an air-conditioned room with a hole in the floor that allowed for water sampling. A “filtered effluent pump” (Little Giant) was lowered through the hole to around 2 m below MLLW. It pumped water up through an 18-m-long, ¾-in.-diameter rubber hose at a flow rate of 40 L min−1 into a line that filtered and diverted the water to different water streams (Fig. 5). The high flow rate kept the residence time in the hose short (8 s), which helped prevent warming of the water before it went through the equilibrator.

Fig. 5.
Fig. 5.

Field setup at the SIO pier. Arrows show the water flow. Water is pumped from a depth of 2 m below MLLW. Most of the water goes into a tank, which houses the optode and gas tension device. The rest of the water is filtered with a 130-μm strainer. Some of this water flows into the equilibrator, which is also shown in Fig. 2. The rest of the filtered water goes into a bag filter, from which water is pumped through the gas tension device.

Citation: Journal of Atmospheric and Oceanic Technology 30, 2; 10.1175/JTECH-D-12-00074.1

Most of the water was diverted into a 95-L insulated tank (Igloo Products Corp.), in which the GTD and an Aanderaa optode O2 sensor were submerged. The rest of the water flowed through a 130-μm Amiad strainer that was modified to be self-cleaning, as explained below. The water flow then split, with 5.6–5.8 L min−1 flowing to the equilibrator, measured using a paddlewheel-type water flow sensor (Gems Sensors), and the rest of the water flowing into a 2-L metal beaker lined with nested filter bags (50, 25, and 10 μm). The filtered water was pumped from this beaker through the GTD at 100 mL min−1 using a peristaltic pump. Water exited the 95-L tank through a pipe inserted into a 5.1-cm-diameter hole cut into the side of the tank. The height of the pipe inlet inside the tank set the water level in the tank. Temperature sensors were placed in the line to the equilibrator and in the tank. A pressure transducer was placed at the downstream end of the hose to monitor the pump head pressure.

In the nearshore environment, there is so much algae and sand in the water that the strainer would clog quickly. The strainer had a removable valve at the bottom, which could be opened so that instead of water flowing through the strainer to the rest of the water line, water would flow inside the strainer and out the bottom, clearing out filtered particles that had collected inside the strainer. However, the strainer clogged so quickly (often in less than 1 h) that manually clearing the strainer was not practical for continuous use. Furthermore, the filtered algae caused the particles to be too sticky to be flushed away from the strainer when the valve was opened. To correct this issue, the valve was removed and a bottle brush that just fit into the strainer cylinder was inserted through a T into the bottom of the strainer. The end of the brush was coupled to a rotating motor. At the arm of the T, a solenoid valve was attached. When the flow rate to the equilibrator dropped below 5.3 L min−1, signaling that the strainer was clogging, the motor automatically started rotating, and the valve opened. The rotating brush swept away sand and organic matter from the strainer, while the diverted flow washed the filtered sediment out of the bottom of the strainer. This stopped the flow to the equilibrator for 5–10 s at a time. When the water flow to the equilibrator stopped and started again, the pressure would change in the equilibrator, and the loop would have to adjust back to the pressure setpoint. The self-cleaning strainer allowed the water to flow continuously without manual intervention.

3. Mathematical model of equilibrator

Understanding the equilibration time in this equilibrator is complex for two reasons. First, we are equilibrating major gases, and not just the minor species. Therefore, changes in the mole fraction of one gas (such as oxygen) will cause changes in the mole fraction of another gas (such as nitrogen) because of dilution. Second, because our equilibrator is kept at constant pressure, gas must enter from a cylinder or leave the equilibrator in order to keep the pressure constant. To understand the behavior of our equilibrator, we use a forward model. Following Johnson (1999), we first present a closed model, where headspace gas is only influenced by exchange with gas dissolved in water, and then expand to a more realistic model where gas is removed from the headspace for sampling, and replaced with gas from a makeup cylinder.

For a closed model, the rate of change in the number of moles in the headspace can be defined as
e7
where Ni is the number of moles of gas i in the headspace and Fi,water is the flow rate of gas i entering the headspace from the water (mol min−1).
The value of Fi,water is determined using an idealized case where water flows through a box without mixing and exchanges with gas in a headspace (Fig. 6). For this case, Fi,water can be derived to be
e8
where Q is the water flow rate (L min−1), S is the Henry’s law solubility constant (mol L−1 atm−1), and k is the gas exchange coefficient (L min−1). In addition, Ci is the concentration of gas i dissolved in the water entering the equilibrator (mol L−1), and is assumed to be constant, while Pi is the pressure of gas i in the headspace (atm), and is assumed to be uniform in the headspace, and constant for the time it takes a water parcel to flow through the equilibrator. All of these parameters can be measured directly except k. This equation is applied to N2, O2, and Ar. The gas exchange coefficient k is set for O2, and then scaled to N2 and Ar using a square root dependency on the ratio of the Schmidt numbers (Wanninkhof 1992).
Fig. 6.
Fig. 6.

Model of gas transfer between water and headspace: Q is the water flow rate, Pi is the headspace pressure of gas i, Ci is the concentration of gas i in the water entering the equilibrator, and k is the gas exchange coefficient.

Citation: Journal of Atmospheric and Oceanic Technology 30, 2; 10.1175/JTECH-D-12-00074.1

Here, Pi is related to Ni using Pi = (PdNi)/Nd, where Pd is the total headspace pressure of dry air, defined as , and Nd is the total number of moles of dry air. Through this relationship, Eq. (7) can be integrated to find Ni as a function of time, which shows which parameters determine the time constant of equilibration:
e9
where N0 is the initial value of Ni, and the time constant for the closed system τc is given by
e10
This model differs from the model described in Johnson (1999) in that the gas exchange is explicitly computed using a gas exchange coefficient k rather than assuming that a constant fraction of the total possible equilibration occurs as a water parcel flows through the equilibrator. Our model allows gas exchange to be limited by either Q or k, as shown in Eq. (8). In the limit where the gas exchange coefficient is much higher than the water flow rate (k/Q ≫ 1), the flux of gas becomes
e11
and is limited by the flow rate, Q. In the limit where the water flow rate is much higher than the gas exchange coefficient (k/Q ≪ 1), the flux of gas becomes
e12
and is limited by the gas exchange coefficient k.
If we consider the limits above (k/Q ≫ 1 or k/Q ≪ 1), the time constant [Eq. (10)] simplifies to
e13
or
e14
Oxygen isotopes are also modeled to determine δ18O and Δ17O. The flux of 18O16O or 17O16O can be expressed as
e15
where x is 18O16O or 17O16O, Rx is the ratio of the oxygen isotope to 16O2 in the air, αeq,x is the equilibrium fractionation factor for oxygen dissolution into water, and αk,x is the kinetic fractionation factor for oxygen dissolution into water. The quantity αeq,18 is 1.000 73 at a water temperature of 19°C (Benson and Krause 1980) and αk,18 is 0.9972 (Knox et al. 1992). Both αeq,17 and αk,17 are determined by assuming that the fractionations are mass dependent with a relationship of 0.52, giving values of 1.000 38 and 0.9985, respectively. The modeled 18O16O and 17O16O values can be used to determine δ18O and Δ17O.
We now explore a more realistic model where, instead of a closed system, gas is removed for sampling and added from a cylinder to keep the pressure constant at a setpoint. In this case, the rate of change in the number of moles of gas i in the headspace becomes
e16
where Xi,cyl is the mole fraction of gas i in the cylinder, Fcyl is the flow rate of gas entering the headspace from the cylinder (mol min−1), Xi,h is the mole fraction of gas i in the headspace, and Fexit is the flow rate of gas leaving the headspace (mol min−1), from bubble loss and sampling. Here, the mole fractions and flow rates are for water-free headspace air. This model does not explicitly account for gas loss due to bubbles escaping from the equilibrator.

The pressure in the headspace, and therefore the total number of moles, remains constant. This means that the sum of Eq. (16) for O2, N2, and Ar is defined as zero, which allows for the calculation of Fcyl, since Fexit is prescribed by the mass spectrometer sampling rate and bubble loss. Because these fluxes are small, bubbles entering the equilibrator from seawater could potentially be an important gas source. However, in our system, transparent tubing at the equilibrator inlet showed that there was no bubble flux into the headspace from the seawater supply. Through the calculation of Fcyl, the equilibration of one major gas is linked to the equilibration of other major gases. This differentiates our model from models of trace gas equilibration, including that in Johnson (1999).

When the headspace pressure is close to the total gas tension, the net flux of gas into the headspace will be close to zero. In this typical case, Fexit will be greater than the net flux of gas from the water, and gas will need to be added from the cylinder. If Fwater is the sum of the fluxes of all the gases (Fwater = ΣFi,water), then, in the typical case where Fexit > Fwater,
e17

If the setpoint pressure in the headspace were much lower than the gas tension, gases would be stripped out of the water. In this case, Fwater could be greater than Fexit, and no gas would be added to the loop from the cylinder. Instead, gas would be lost from the loop to the vacuum, through the bidirectional flowmeter shown in Fig. 3, to maintain the setpoint pressure.

The model described in Eq. (16), where gas is sampled from the loop and added to the loop from a cylinder, introduces a new time constant due to sampling τs:
e18
This time constant reduces the total adjustment time of the system, as also described in Johnson (1999). Furthermore, when gas is constantly added, the headspace gas composition can never reach complete equilibrium with gas in the water. The time constants can be combined to get the overall time constant, as in Johnson (1999):
e19

As mentioned above, every parameter in this model can be measured experimentally except for k. Also, although the flow to the mass spectrometer is known, Fexit is not, because gas loss due to bubbles contributes to Fexit. However, Fcyl can be directly measured and is equal to Fexit in steady state. The model is run with forward time stepping of Eq. (16) to compute the gas composition of the headspace as equilibration occurs. Below, we use experiments and model runs to determine the value of k, and explore the equilibration time and the measurement errors introduced by the equilibrator.

4. Results

a. Equilibration time of the GTD

Since the equilibrator pressure is set to match the total gas tension, it is useful to know how quickly the gas tension device can respond to changes in the gas tension. The e-folding time of the GTD can be measured by perturbing the pressure. This was achieved by opening a vent in the -in. tubing that connects the GTD air volume to a pressure transducer. When the vent was opened, the pressure dropped to 0 torr relative to the atmospheric pressure. When the vent was closed, the pressure gradually returned to the gas tension as water was pumped through the GTD at 100 mL min−1. This test was repeated 6 times. For each test, in order to determine the time constant, we fit the data to a model that consisted of a linear trend, which allows for slow changes in ambient seawater gas tension, and an exponential decay of a pulse, which gives the time constant, as shown in Fig. 7. These results were averaged using 1/σ2 as the weight. The weighted mean e-folding time of the gas tension device is 5.35 ± 0.59 min.

Fig. 7.
Fig. 7.

Example of the gas tension device response after its air volume was vented to the atmosphere. The line shows the fit to a model with a linear trend and an exponential decay of a pulse. Water was pumped through the GTD at 100 mL min−1. The e-folding time is 5.35 ± 0.59 min.

Citation: Journal of Atmospheric and Oceanic Technology 30, 2; 10.1175/JTECH-D-12-00074.1

b. Accuracy of GTD

The accuracy of the GTD can be determined using percent O2 saturation from the optode, and δ(O2/N2) and δ(Ar/N2) from the mass spectrometer, which together provide an independent measure of the total gas tension. For a 3-week period, these measurements were taken at the SIO pier.

The optode measures O2 concentration and water temperature, and uses the solubility relationship of Garcia and Gordon (1992) to compute percent oxygen saturation. The optode was initially calibrated with a two-point calibration using air-saturated water and a zero-oxygen solution, as described in the Aanderaa manual. This gives an internal calibration to define 100% oxygen saturation and 0% oxygen saturation. During the optode deployment at the SIO pier, Winkler titrations were taken periodically to provide an independent calibration of the optode. The optode readings were on average 2.9 ± 1.1 μM below the Winkler results, a mean difference of around 1.2%. The variability in ambient dissolved O2 is small enough that correcting by applying an additive offset is roughly equivalent to multiplying by a constant factor. We multiplied the optode data by 1.012 so that the optode data matched the Winkler titrations.

A cross-correlation analysis revealed that the mass spectrometer measurements lagged the optode measurements by 8.3 min. The travel time from the equilibrator to the mass spectrometer is roughly 2.5 min, leaving a 5.8-min lag due to the time constant of the equilibrator. In addition to a time offset, the optode can measure variability at a higher frequency than can the equilibrator. To account for these effects, a 10-min moving boxcar average was applied to the optode data, as well as a lag of 8.3 min.

In addition to these corrections to the optode data, some of the optode and mass spectrometer data were masked out based on the makeup flow rate. There were many occasions where the makeup flow rate, measured by the bidirectional flowmeter, became very high (>60 mL min−1) or very low (<−20 mL min−1, indicating that gas was leaving the loop). As mentioned in section 2d, some of these fluctuations were due to pressure changes as the strainer was cleaned out, which stopped and restarted the water flow. At flows greater than 60 mL min−1, so much gas would be added to the loop that the mass spectrometer measurements would be highly influenced by the makeup gas. We defined a “spike” in the makeup flow rate as flows greater than 50 mL min−1 or less than −10 mL min−1. All data within 30 min after a spike in the makeup flow rate were removed, for both the optode and mass spectrometer data.

As shown in Eq. (6), the primary contributors to the gas tension are O2, N2, Ar, and water vapor. Water vapor pressure can be determined from temperature using (Weiss and Price 1980)
e20
where T is the temperature in kelvins and S is the salinity in parts per thousand.
Ratios of the partial pressures of O2, Ar, and N2 can be computed from δ(O2/N2) and δ(Ar/N2) using
e21
where δ(O2/N2) is in units of per mil and (0.209 46/0.780 84) is the ratio of O2 to N2 in the atmosphere, and
e22
where δ(Ar/N2) is in units of per mil and (0.009 34/0.780 84) is the ratio of Ar to N2 in the atmosphere.
The value of was calculated using the optode percent O2 saturation:
e23
where 159.1896 torr is the partial pressure of oxygen at a total air pressure of 760 torr. We calculated and PAr using
e24
and
e25

The water temperature inside the GTD was 0.36 ± 0.12°C higher than the water temperature inside the equilibrator. The amount of warming increased when the difference between the room temperature and water temperature increased, suggesting that the warming was due to the peristaltic pump, and to heat transfer from the room air to the water through the Tygon tubing connecting the pump to the GTD. For a water temperature of 19°C (the mean measured water temperature), the warming of the water in the GTD would cause the measured gas tension to be around 100.4%–100.8% of the actual gas tension.

The sum of the partial pressures of water vapor, O2, N2, and Ar are compared with the measured total gas tension in Fig. 8. On average, the measured gas tension shows variability that is similar in timing but lower in magnitude than the calculated gas tension. The measured gas tension is higher than the calculated gas tension by around 4 torr. This is around the uncertainty, ±3 torr, based on the uncertainty in the optode calibration.

Fig. 8.
Fig. 8.

(a) Gas tension measured from the gas tension device, and calculated from the optode and mass spectrometer using Eqs. (6) and (20)(25). (b) The calculated gas tension subtracted from the measured gas tension. The measured gas tension is on average 4 torr higher than expected from the optode and mass spectrometer measurements.

Citation: Journal of Atmospheric and Oceanic Technology 30, 2; 10.1175/JTECH-D-12-00074.1

c. Equilibration time of the equilibrator and determination of the gas exchange coefficient

The equilibration time of the equilibrator determines its time resolution relative to ambient changes. Although this is a useful metric of performance on its own, the equilibration time can also be used to determine the gas exchange coefficient k, the one parameter in the model that cannot be measured directly. The model described in Eq. (16) can be used to model the relationship between the time constant, τo, and k. Then, k can be determined using the measured τo. The determination of k allows the behavior of gases in the equilibrator to be explored using the model.

We measured the equilibration time by running seawater through the equilibrator as explained in section 2d, with gas circulating through the equilibrator at a fixed pressure. A T was put into the makeup cylinder line, and attached to a valve and a cylinder filled with nitrogen gas (dashed box in Fig. 3). After the headspace gas composition stabilized, the valve connected to the N2 cylinder was opened to add a small amount of N2 gas into the makeup gas stream, perturbing the headspace gas composition. Because the natural variability of δ(Ar/N2) is lower than δ(O2/N2), we looked at the response in δ(Ar/N2) as the headspace returned to the gas composition that it had before the N2 gas addition.

This test was repeated 4 times. For each test, the time constant was determined using the model fit described in section 4a (e.g., Fig. 9). The resulting values of τ were then averaged using a weight of 1/σ2. The weighted average and standard deviation of τ is 7.36 ± 0.74 min.

Fig. 9.
Fig. 9.

Response in δ(Ar/N2) to N2 gas addition during one test. The line shows the fit to a model with a linear trend and an exponential decay of a pulse. Data between 6 and 13 min were removed for the fit. The weighted average of τ based on four tests is 7.36 ± 0.74 min.

Citation: Journal of Atmospheric and Oceanic Technology 30, 2; 10.1175/JTECH-D-12-00074.1

To determine k, the model described in Eq. (16) was used to relate τo to k. The model was initialized with the headspace gas composition simulating an N2 spike; the initial N2 concentration was high, and the O2 and Ar concentrations were diluted proportionally. Parameters were set to reflect values measured during sampling at the SIO pier. We used a water flow rate of 5.6 L min−1 (the mean water flow during sampling), a headspace volume of 930 mL (the measured total equilibrator loop volume minus the estimated water volume in the equilibrator, based on the water height), and an Fcyl of 5 mL min−1 (measured using the bidirectional flowmeter). Equation (16) was run with forward time stepping to model the response of the headspace gas as water with a different dissolved gas composition flowed through the equilibrator. The model was run until δ(Ar/N2) of the headspace was stable. An equation of the form
e26
was fit to the model output of δ(Ar/N2) versus time, where τ is the e-folding time of equilibration. This equation allows for an exponential approach back to equilibrium. This was repeated for 18 values of between 5 and 75 L min−1, where is the gas exchange coefficient of O2. The best-fitting τ for each value of is shown in Fig. 10.
Fig. 10.
Fig. 10.

Modeled relationship between the time constant τ for δ(Ar/N2) and at a water flow rate of 5.6 L min−1. At large , τ becomes insensitive to . Horizontal dashed lines show the mean and 1σ uncertainty of τ determined experimentally (7.36 ± 0.74 min). Vertical dotted lines show the values of corresponding to the measured τ (7.3, 9.1, and 12.8 L min−1).

Citation: Journal of Atmospheric and Oceanic Technology 30, 2; 10.1175/JTECH-D-12-00074.1

Although k is a fixed property of the equilibrator, the model results show that, for a given k, τ varies depending on the dissolved oxygen concentration of the water (not shown). However, this variability is small relative to the uncertainty in the measured τ. We show results of the relationship between τ and k for a dissolved gas composition in the middle of the range of measured O2 concentration.

As shown in Fig. 10, τ approaches an asymptote at large values of . Using Eqs. (13), (18), and (19), τo can be computed for the limit where k/Q approaches infinity. We find that τo is 5.14 min for O2 and 10.74 min for N2. The forward model asymptotes to a τ of 6.0 min for δ(O2/N2) at large values of k. This value falls between the values of τo for O2 and N2.

Based on the model results shown in Fig. 10, at the mean measured τ is 9.1 L min−1. The range in based on the standard deviation in the measured τ is 7.3–12.8 L min−1. Our system is approaching the high k/Q limit where τ is insensitive to k.

d. Corrections for incomplete equilibration

The observed steady-state headspace composition can differ from what is expected from true equilibration due to the influence of the makeup gas, as well as the differences between the actual gas tension and the headspace pressure. The model described in section 3 can be used to determine the corrections needed to the measurements to account for these effects. The overall difference can be expressed as an additive correction x. For example, for δ(O2/N2),
e27
where δ(O2/N2)observed is the observed steady-state headspace composition and δ(O2/N2)water is the composition in equilibrium with the water. The correction takes into account the extent of equilibration, the sensitivity to pressure, and the sensitivity to the difference between the measured and makeup gas compositions.

To determine how the correction varies with headspace pressure and dissolved gas composition, we ran the model with a range of headspace pressures and dissolved gas compositions using conditions similar to our typical operating conditions, with a total gas tension of 790 torr and a Q of 5.6 L min−1. The correction also varies with Fcyl, for which we used the makeup flow rate, measured by the bidirectional flowmeter. Here, we only show results using an Fcyl of 5 mL min−1. The model was run for two values of to illustrate the sensitivity to . We show results using a of 7.3 L min−1 and a approaching infinity. As shown in Fig. 11, the results using these values of are similar.

Fig. 11.
Fig. 11.

Model prediction of the sensitivity of (a) δ(O2/N2), (b) δ(O2/Ar), (c) δ18O, and (d) Δ17O to equilibrator pressure and makeup gas composition, for a gas tension of 790 torr, a makeup gas flow rate of 5 mL min−1, a Q of 5.6 L min−1, and a of 7.3 L min−1 (black lines) or a approaching infinity (gray lines). Here, ΔP is the headspace pressure minus the gas tension. The y axis shows the difference between the steady-state delta value of the headspace (i.e., the values that would be observed) and the delta value of the makeup gas. The contour values show the correction, defined in Eq. (27), for a of 7.3 L min−1, in units of per mil. Some contour values are also shown for a approaching infinity.

Citation: Journal of Atmospheric and Oceanic Technology 30, 2; 10.1175/JTECH-D-12-00074.1

Figure 11a shows results for δ(O2/N2). Corrections at a ΔP of 0 torr show the effect of the makeup gas addition on equilibration. For example, when the observed δ(O2/N2) is 50 per mil greater than the makeup gas, the correction is 2.50 per mil using a of 7.3 L min−1, and 1.86 per mil using a approaching infinity. This is a 3.7%–5.0% correction, which means that the headspace gas is around 95%–96% equilibrated, and around 4%–5% of the sampled gas is makeup gas. For headspace pressures higher than the gas tension (ΔP > 0), the correction becomes more positive, and for headspace pressures lower than the gas tension, the correction becomes more negative. This behavior results because O2 equilibrates faster than N2, so pO2 is more able than pN2 to match the equilibrium partial pressure.

Figures 11b–d show results for δ(O2/Ar), δ18O, and Δ17O. Compared with δ(O2/N2), the relative corrections in δ(O2/Ar), δ18O, and Δ17O are similar for ΔP = 0, but are less sensitive to pressure. This is because the differences among the equilibration rates of Ar, 16O2, 17O16O, and 18O16O are small compared with the difference between O2 and N2, so the influence of pressure is not as strong.

As mentioned in section 4b, warming of water in the GTD causes a higher measured gas tension. Since this gas tension was used as the equilibrator setpoint pressure, the pressure in the equilibrator would be too high. For a gas tension of 790 torr, the pressure offset would be around 5 torr for a warming of 0.36°C. For our setup, the actual pressure in the equilibrator is measured to around ±1 torr, and the offset from the true gas tension of 5 torr is known to ±1.7 torr.

Table 1 shows the corrections needed to typical values of δ(O2/N2), δ(O2/Ar), δ18O, and Δ17O for our typical operating conditions. A Monte Carlo approach was used to determine the uncertainty in the corrections, using representative standard deviations in ΔP of 1.7 torr, in k of 1.8 L min−1, and in Fcyl of 3.5 mL min−1. The uncertainty in the correction is most sensitive to the uncertainty in ΔP for δ(O2/N2), and to the uncertainty in Fcyl for δ(O2/Ar), δ18O, and Δ17O. Using the representative uncertainty estimates, the uncertainty in the correction is 2.2 per mil for δ(O2/N2), 1.5 per mil for δ(O2/Ar), 0.059 per mil for δ18O, and 0.0030 per mil for Δ17O.

Table 1.

Corrections to delta values with typical signals for the typical operating conditions of Fcyl = 5 mL min−1, ΔP = 5 torr, and = 9.1 L min−1. Here, Fcyl is the makeup gas flow rate, ΔP is the difference between the gas tension and the headspace pressure, and is the gas exchange coefficient of O2. Also shown are the uncertainties in the corrections from the uncertainties in Fcyl (±3.5 mL min−1), ΔP (±1.7 torr), and (±1.8 L min−1), as well as the combined uncertainties from these three parameters.

Table 1.

e. Verification of model-based correction

The model-based correction to the data, described in section 4d, can be verified using Eqs. (6) and (20)(22) to determine the partial pressure of O2 from the measured gas tension, δ(O2/N2) and δ(Ar/N2). This can be compared with the optode O2 measurements. If the presence of makeup gas were undercorrected, the apparent changes in O2 (measured by the mass spectrometer) would be smaller than the actual changes (measured by the optode).

Data were corrected as described in section 4b, and averaged into hourly means. The mass spectrometer hourly means were corrected for incomplete equilibration, as described in section 4d, using the mean makeup flow for that hour as Fcyl. The computed partial pressure of O2 was divided by 159.1896 torr (the partial pressure of oxygen at a total air pressure of 760 torr) to get the percent oxygen saturation.

Figure 12 shows the mass spectrometer percent O2 saturation versus the optode percent O2 saturation. The slope of the mass spectrometer versus the optode percent O2 was 0.94 ± 0.02. We also tested this relationship by taking into account the different time constants of the optode and the equilibrator. Because the optode has a faster time constant than the equilibrator, the equilibrator will not respond as quickly to rapid changes in oxygen, and the change in the oxygen saturation measured by the equilibrator will be smaller than the change in the oxygen saturation measured by the optode. To eliminate periods with rapid fluctuations in oxygen, we stipulated that, for each hourly mean data point used, the hourly means of the optode data had to be stable to within 5% oxygen saturation for 5 h before the data point. These results are also shown in Fig. 12. Using this stability criterion achieves a slope of 1.03 ± 0.06. These tests show that our model-based correction to the data is valid to within the uncertainties of the data.

Fig. 12.
Fig. 12.

Hourly mean percent oxygen saturation measured with the optode and the mass spectrometer. Gray triangles show points where the oxygen percent saturation was within 5% for the last 5 h. The black text gives the slope of all the data, and the gray text gives the slope of the more stable data. Looking only at periods where the oxygen saturation was stable, the slope of the data comes closer to the ideal slope of 1.

Citation: Journal of Atmospheric and Oceanic Technology 30, 2; 10.1175/JTECH-D-12-00074.1

5. Discussion and conclusions

The equilibrator presented here has an e-folding time of 6.5–8 min for O2. This is as fast as the equilibration method described by Cassar et al. (2009), and is several times faster than using an equilibrator designed to measure CO2. Based on model results, the e-folding time is limited by the water flow rate, rather than the gas exchange coefficient. With an e-folding time of 6.5–8 min, hourly changes are easily resolvable.

The gas tension device has an e-folding time of around 5–6 min. This is around twice as fast as the gas tension device presented in McNeil et al. (2005), and is within the range of the e-folding time of the gas tension device presented in McNeil et al. (2006). The mean measured gas tension agrees with the mean expected gas tension, which was calculated using simultaneous measurements of percent oxygen saturation, δ(O2/N2), and δ(Ar/N2).

The model analyses show that it is important to have a low flow rate to minimize errors due to makeup gas addition. The introduction of makeup gas to the “equilibrated” airstream causes a scale contraction as the headspace air is brought toward the makeup gas concentration. The more gas added, either because of a high sampling flow rate or gas lost to bubbles escaping from the equilibrator, the further away from true equilibrium the sample will be, and the larger correction that will be needed. For example, for a measured δ(O2/N2) of 50 per mil, with no difference between the gas tension and headspace pressure, a flow rate of 3 mL min−1 will need a correction of 1.5 per mil, and a flow rate of 20 mL min−1 will need a correction of 9.7 per mil. This emphasizes the need for a low sampling rate. Also, to avoid large bubbles escaping from the equilibrator, it is important to balance the pressure inside the equilibrator with the water outflow height, so that the water level inside the equilibrator is not too low.

If there were an offset between the measured gas tension and the true gas tension, then the setpoint pressure inside the equilibrator would not match the true total gas tension. This would have implications for measurements of δ(O2/N2), δ(O2/Ar), and oxygen isotopes, because the steady-state values in the equilibrator are sensitive to differences between the headspace pressure and the total gas tension. In our system, warming of water in the GTD causes a pressure offset, which requires a large correction for δ(O2/N2). The pressure offset is less important for the measurement of δ(O2/Ar), δ18O, and Δ17O, because the correction is less sensitive to pressure (Fig. 11). To reduce the correction needed for δ(O2/N2), the pressure offset could be computed and corrected in real time, and this corrected value could be used to set the equilibrator pressure.

A drawback of bubble-type equilibrators is that they have been shown to cause biases in the equilibrated gas because the pressure in the bubble is higher than the ambient pressure due to hydrostatic pressure and the surface tension of the bubble (Murphy et al. 2001; Schneider et al. 2007). This equilibrator does use bubbles in its equilibration. However, the bubbles in this equilibrator are much larger than bubbles produced with a frit, so the surface tension effect is lower. Also, the circulating air only bubbles through the lower part of the equilibrator. After the air rises through the water, gases still exchange on the surface of the Raschig rings and on the water droplets. If the bubbles did produce a bias in the gas composition, it would likely be largely removed by gas exchange on these other air–water interfaces.

This equilibrator was used at a pier to measure gas composition in seawater. If the equilibrator were used on a ship, the motion of the ship would change the height of the water outlet tube relative to the equilibration tube. This would change the water height inside the equilibrator. Therefore, for measurements on a ship, the water outflow would need to be redesigned so that the motion of the ship would not affect the water height inside the equilibrator.

We have described an equilibrator with a fast time constant for dissolved oxygen that allows for a sampling flow rate of 3 mL min−1. Although we used a mass spectrometer for our analyses, this high sampling flow rate would allow the equilibrator to be interfaced with a variety of instruments. We have also presented a model that describes the equilibration of major gases and shows how changes in pressure, gas composition, and sampling flow rate influence the steady-state gas composition in the equilibrator. Using this model, we have determined the corrections needed to account for these influences. The equilibrator, when interfaced with an IsoPrime isotope ratio mass spectrometer, can provide high-frequency measurements of dissolved O2 and its isotopes.

Acknowledgments

We thank Yann Bozec, for initial work on the mass spectrometer; Adam Cox, for assistance preparing gas cylinders; Todd Martz, for providing the optode; Yui Takeshita, for providing Winkler titration measurements; Roberta Hamme, for suggestions on ways to test the equilibrator; Mark Altabet for helpful discussions about building a gas tension device; and Angeline Ta, for help testing the gas tension device and equilibrator. We also thank three anonymous reviewers for their comments, which helped improve this manuscript. This material is based upon work supported by the National Science Foundation under Grant 0421546.

REFERENCES

  • Angert, A., , Rachmilevitch S. , , Barkan E. , , and Luz B. , 2003: Effects of photorespiration, the cytochrome pathway, and the alternative pathway on the triple isotopic composition of atmospheric O2. Global Biogeochem. Cycles, 17, 1030, doi:10.1029/2002GB001933.

    • Search Google Scholar
    • Export Citation
  • Barkan, E., , and Luz B. , 2005: High precision measurements of 17O/16O and 18O/16O ratios in H2O. Rapid Commun. Mass Spectrom., 19, 37373742.

    • Search Google Scholar
    • Export Citation
  • Bender, M. L., 1990: The δ18O of dissolved O2 in seawater: A unique tracer of circulation and respiration in the deep sea. J. Geophys. Res., 95 (C12), 22 24322 252.

    • Search Google Scholar
    • Export Citation
  • Bender, M. L., , and Grande K. D. , 1987: Production, respiration, and the isotope geochemistry of O2 in the upper water column. Global Biogeochem. Cycles, 1, 4959.

    • Search Google Scholar
    • Export Citation
  • Benson, B. B., , and Krause D. Jr., 1980: The concentration and isotopic fractionation of gases dissolved in freshwater in equilibrium with the atmosphere. 1. Oxygen. Limnol. Oceanogr., 25, 662671.

    • Search Google Scholar
    • Export Citation
  • Cassar, N., , Barnett B. A. , , Bender M. L. , , Kaiser J. , , Hamme R. C. , , and Tilbrook B. , 2009: Continuous high-frequency dissolved O2/Ar measurements by equilibrator inlet mass spectrometry. Anal. Chem., 81, 18551861.

    • Search Google Scholar
    • Export Citation
  • Copin-Montegut, C., 1985: A method for the continuous determination of the partial pressure of carbon dioxide in the upper ocean. Mar. Chem., 17, 1321.

    • Search Google Scholar
    • Export Citation
  • Emerson, S., , Stump C. , , Wilbur D. , , and Quay P. , 1999: Accurate measurement of O2, N2 and Ar gases in water and the solubility of N2. Mar. Chem., 64, 337347.

    • Search Google Scholar
    • Export Citation
  • Frankignoulle, M., , Borges A. , , and Biondo R. , 2001: A new design of equilibrator to monitor carbon dioxide in highly dynamic and turbid environments. Water Res., 35, 13441347.

    • Search Google Scholar
    • Export Citation
  • Garcia, H. E., , and Gordon L. I. , 1992: Oxygen solubility in seawater: Better fitting equations. Limnol. Oceanogr., 37, 13071312.

  • Guéguen, C., , and Tortell P. D. , 2008: High-resolution measurement of Southern Ocean CO2 and O2/Ar by membrane inlet mass spectrometry. Mar. Chem., 108, 184194.

    • Search Google Scholar
    • Export Citation
  • Hendricks, M. B., , Bender M. L. , , and Barnett B. A. , 2004: Net and gross O2 production in the Southern Ocean from measurements of biological O2 saturation and its triple isotope composition. Deep-Sea Res. I, 51, 15411561.

    • Search Google Scholar
    • Export Citation
  • Johnson, J. E., 1999: Evaluation of a seawater equilibrator for shipboard analysis of dissolved oceanic trace gases. Anal. Chim. Acta, 395, 119132.

    • Search Google Scholar
    • Export Citation
  • Juranek, L. W., , and Quay P. D. , 2005: In vitro and in situ gross primary and net community production in the North Pacific Subtropical Gyre using labeled and natural abundance isotopes of dissolved O2. Global Biogeochem. Cycles, 19, GB3009, doi:10.1029/2004GB002384.

    • Search Google Scholar
    • Export Citation
  • Kaiser, J., , Reuer M. K. , , Barnett B. , , and Bender M. L. , 2005: Marine productivity estimates from continuous O2/Ar ratio measurements by membrane inlet mass spectrometry. Geophys. Res. Lett., 32, L19605, doi:10.1029/2005GL023459.

    • Search Google Scholar
    • Export Citation
  • Keeling, R. F., , Blaine T. , , Paplawsky B. , , Katz L. , , Atwood C. , , and Brockwell T. , 2004: Measurement of changes in atmospheric Ar/N2 ratio using a rapid-switching, single-capillary mass spectrometer system. Tellus, 56B, 322338.

    • Search Google Scholar
    • Export Citation
  • Knox, M., , Quay P. D. , , and Wilbur D. , 1992: Kinetic isotopic fractionation during air–water gas transfer of O2, N2, CH4, and H2. J. Geophys. Res., 97 (C12), 20 33520 343.

    • Search Google Scholar
    • Export Citation
  • Körtzinger, A., , Thomas H. , , Schneider B. , , Gronau N. , , Mintrop L. , , and Duinker J. C. , 1996: At-sea intercomparison of two newly designed underway pCO2 systems—Encouraging results. Mar. Chem., 52, 133145.

    • Search Google Scholar
    • Export Citation
  • Levine, N. M., , Bender M. L. , , and Doney S. C. , 2009: The δ18O of dissolved O2 as a tracer of mixing and respiration in the mesopelagic ocean. Global Biogeochem. Cycles, 23, GB1006, doi:10.1029/2007GB003162.

    • Search Google Scholar
    • Export Citation
  • Luz, B., , and Barkan E. , 2000: Assessment of oceanic productivity with the triple-isotope composition of dissolved oxygen. Science, 288, 20282031.

    • Search Google Scholar
    • Export Citation
  • McNeil, C., , Katz D. , , Wanninkhof R. , , and Johnson B. , 2005: Continuous shipboard sampling of gas tension, oxygen and nitrogen. Deep-Sea Res. I, 52, 17671785.

    • Search Google Scholar
    • Export Citation
  • McNeil, C., , D’Asaro E. , , Johnson B. , , and Horn M. , 2006: A gas tension device with response times of minutes. J. Atmos. Oceanic Technol., 23, 15391558.

    • Search Google Scholar
    • Export Citation
  • Murphy, P. P., , Nojiri Y. , , Fujinuma Y. , , Wong C. S. , , Zeng J. , , Kimoto T. , , and Kimoto H. , 2001: Measurements of surface seawater fCO2 from volunteer commercial ships: Techniques and experiences from Skaugran. J. Atmos. Oceanic Technol., 18, 17191734.

    • Search Google Scholar
    • Export Citation
  • Quay, P. D., , Emerson S. , , Wilbur D. O. , , and Stump C. , 1993: The δ18O of dissolved O2 in the surface waters of the subarctic Pacific: A tracer of biological productivity. J. Geophys. Res., 98 (C5), 84478458.

    • Search Google Scholar
    • Export Citation
  • Quiñones-Rivera, Z. J., , Wissel B. , , Justić D. , , and Fry B. , 2007: Partitioning oxygen sources and sinks in a stratified, eutrophic coastal ecosystem using stable oxygen isotopes. Mar. Ecol. Prog. Ser., 342, 6983.

    • Search Google Scholar
    • Export Citation
  • Reuer, M. K., , Barnett B. A. , , Bender M. L. , , Falkowski P. G. , , and Hendricks M. B. , 2007: New estimates of Southern Ocean biological production rates from O2/Ar ratios and the triple isotope composition of O2. Deep-Sea Res. I, 54, 951974.

    • Search Google Scholar
    • Export Citation
  • Sarma, V. V. S. S., , Abe O. , , Hashimoto S. , , Hinuma A. , , and Saino T. , 2005: Seasonal variations in triple oxygen isotopes and gross oxygen production in the Sagami Bay, central Japan. Limnol. Oceanogr., 50, 544552.

    • Search Google Scholar
    • Export Citation
  • Schneider, B., , Kremling K. , , and Duinker J. C. , 1992: CO2 partial pressure in northeast Atlantic and adjacent shelf water: Processes and seasonal variability. J. Mar. Syst., 3, 453463.

    • Search Google Scholar
    • Export Citation
  • Schneider, B., , Sadkowiak B. , , and Wachholz F. , 2007: A new method for continuous measurements of O2 in surface water in combination with pCO2 measurements: Implications for gas phase equilibration. Mar. Chem., 103, 163171.

    • Search Google Scholar
    • Export Citation
  • Stanley, R. H. R., , Kirkpatrick J. B. , , Cassar N. , , Barnett B. A. , , and Bender M. L. , 2010: Net community production and gross primary production rates in the western equatorial Pacific. Global Biogeochem. Cycles, 24, GB4001, doi:10.1029/2009GB003651.

    • Search Google Scholar
    • Export Citation
  • Thiemens, M. H., , Jackson T. , , Zipf E. C. , , Erdman P. W. , , and van Egmond C. , 1995: Carbon dioxide and oxygen isotope anomalies in the mesosphere and stratosphere. Science, 270, 969972.

    • Search Google Scholar
    • Export Citation
  • Tortell, P. D., 2005: Dissolved gas measurements in oceanic waters made by membrane inlet mass spectrometry. Limnol. Oceanogr. Methods, 3, 2437.

    • Search Google Scholar
    • Export Citation
  • Wanninkhof, R., 1992: Relationship between wind speed and gas exchange over the ocean. J. Geophys. Res., 97 (C5), 73737382.

  • Weiss, R. F., , and Price B. A. , 1980: Nitrous oxide solubility in water and seawater. Mar. Chem., 8, 347359.

Save