Dissolved oxygen/argon (O2/Ar) ratios in the oceanic mixed layer are indicative of net community production (NCP) because O2 and Ar share similar physical solubility properties, but only O2 is biologically produced and consumed. We describe a membrane inlet mass spectrometer (MIMS) that allows continuous high-precision shipboard analysis of O2/Ar ratios and eventually other gases, calibrated with discrete samples analyzed in the laboratory. We also present O2/Ar data from the eastern equatorial Pacific. Short-term reproducibilities of 0.05% were achieved. Meridional gradients and small-scale phenomena were clearly resolved. O2/Ar undersaturations around the equator reflect the interaction of biological and physical forcings. Mixed-layer NCP estimated from wind speed-gas exchange parameterizations was near zero north of 2.75°N, and about 12 mmol m−2 d−1 south of 6.75°S. Ar supersaturations, calculated from MIMS O2/Ar measurements and accompanying O2 concentration measurements, ranged from −0.8 to +3.0%.
 The dissolved oxygen (O2) concentration of seawater varies because of fundamental physical and biological processes. These include photosynthesis and respiration, diffusive and bubble-mediated gas exchange, temperature and pressure changes, lateral mixing and vertical diffusion.
 The marine oxygen cycle is tightly linked to the carbon cycle via photosynthesis and respiration. In the absence of physical effects, dissolved O2 constrains the balance of these two biological processes, i.e., net community production (NCP). O2 can be used as a geochemical tracer that reflects carbon fluxes integrated over characteristic response times. Warming and bubble injection lead to O2 supersaturation, posing a challenge to this approach.
Craig and Hayward  used oxygen/argon (O2/Ar) ratios to separate O2 supersaturations into a biological and a physical component. This method is based on the similar solubility characteristics of these two gases with respect to temperature and pressure changes as well as bubble-mediated gas exchange. One can define an O2/Ar supersaturation, ΔO2/Ar, as:
 ΔO2/Ar essentially records the difference between photosynthetic O2 production and respiration. c is the dissolved gas concentration (in mol m-3) and csat is the saturation concentration. csat is a function of temperature, pressure and salinity. This approach, in which discrete samples are collected at sea, stored, and analyzed in the lab, has been widely used in subsequent work [e.g., Emerson et al., 1995; Hendricks et al., 2004; Spitzer and Jenkins, 1989].
 Here, we describe a new method for continuous underway measurements of O2/Ar by membrane-inlet mass spectrometry (MIMS), extending earlier oceanographic MIMS applications [Kana et al., 1994; Tortell, 2005]. The measured ΔO2/Ar values can be used in conjunction with suitable wind-speed gas-exchange parameterizations to calculate biologically induced air-sea O2 fluxes and, where conditions are appropriate, NCP. The inferred NCP values represent rates integrated over the characteristic mixed layer gas exchange times (ratio of mixed layer thickness and piston velocity), typically between 10 and 30 days.
 In this paper, we show results of continuous O2/Ar measurements in the eastern equatorial Pacific (EEP). The equatorial Pacific is a region of considerable interest because of its high-nutrient low chlorophyll character [Coale et al., 1996], its role as oceanic CO2 source [Takahashi et al., 2002], and its strong coupling of biogeochemical variability to physical forcing. Here, we address a question related to this topic: How does net community production vary in the EEP between the nutrient-poor region north of the equatorial upwelling zone and the nutrient-rich region in the south? We present O2 and O2/Ar data in the region between 8°S and 12°N, 110°W and 95°W. The results exemplify the use of MIMS for oceanic productivity studies and its potential to provide large-scale constraints for satellite-based productivity estimates [e.g., Behrenfeld et al., 2005].
 Continuous O2/Ar ratio measurements were made by MIMS on board NOAA R/V Ron Brown during a cruise to service TAO (Tropical Atmosphere Ocean Project) [McPhaden et al., 1998; McPhaden, 1995] buoys in the eastern equatorial Pacific in October/November 2003 (RB-03-09-TAO). Meridional and zonal transects were performed in the following order: 12°N to 8°S along 95°W, 95°W to 110°W along 8°S, 8°S to 8°N along 110°W, 110°W to 95°W along 8°N, 8°N to equator along 95°W. The ship's underway sampling system was used to pump water through a Teflon AF membrane (Random Technologies) mounted in a stainless steel vacuum manifold, which was connected to a quadrupole mass spectrometer (Pfeiffer Vacuum Prisma). The intake of the underwater sampling system is located at the bow at a nominal depth of 5 m. The inner diameter of the membrane of 600 μm and the head pressure of the ship's continuous seawater pump (3 bar) determined the volumetric flow rate of 12 cm3/min through the 10 cm-long membrane tubing. In order to reduce O2/Ar variations due to temperature and water vapor pressure effects and to avoid degassing, the vacuum manifold with the membrane was held at a constant temperature of 20°C (≥2°C below the surface seawater temperature, SST) and the flight tube was in a thermally insulated box maintained at 40°C.
 We calibrated the O2/Ar ratio measurements with discrete water samples taken from the same seawater outlet as used for the MIMS measurements. 300 cm3 samples were drawn into pre-evacuated glass flasks poisoned with 7 mg HgCl2 [Emerson et al., 1999]. These samples were later analyzed with an isotope ratio mass spectrometer (IRMS, Thermo Finnigan) for their dissolved O2/Ar ratios and the oxygen triple isotope composition relative to air [Hendricks et al., 2004]. The precision of the IRMS O2/Ar ratios was about 0.1%, based on analyses of duplicate samples. MIMS O2/Ar ion current ratio measurements were made every 10 to 30 s and had a short-term stability of 0.05% (standard deviation s of 10 sample-mean; error estimates in the following are ±1s). The raw O2/Ar ion current ratios were calibrated by linear regression against the IRMS-measured dissolved O2/Ar ratios. The mean residual was (0.0 ± 0.3)%.
 In addition to the O2/Ar ratios, O2 concentrations were measured continuously with an optode (Aanderaa), calibrated by automatic Winkler titration of discrete water samples with potentiometric endpoint detection. The analytical precision of the Winkler method was better than 0.1%. Short-term (60 s) precision of the optode measurements was 0.03%. Calibration was achieved by regression of the temperature-corrected optode readings against the Winkler results. The mean difference between calibrated optode and Winkler measurements of the O2 supersaturation (ΔO2) was (0.0 ± 0.3)%.
 The spatial sampling resolution at a cruise speed of 12 kn is 60–180 m for the MIMS and about 40 m for the optode measurements. Response times (1/e-fold) are about 60 s for the MIMS and 30 s for the optode.
 Dissolved O2 was also measured in surface water samples from Niskin bottles in order to assess whether any gas losses occurred from the water pumped from the seawater intake to the laboratory due to warming and potential outgassing or O2 loss to the pipe walls. A mean ΔO2 decrease of (0.5 ± 0.3)% was recorded and corrected for in the results presented here. About one third of the ΔO2 change can be explained with outgassing due to warming by about 0.08°C between seawater intake and outlet in the lab. We did not measure O2/Ar ratios in Niskin samples on this cruise. On a 2004 cruise of R/V Ka'Imimoana, however, we found that the mean difference between Niskin and continuous measurements was negligible, with a similar change in ΔO2 as observed for the RB-03-09-TAO cruise. We therefore do not apply any corrections to the ΔO2/Ar values.
 ΔO2 and ΔO2/Ar values were calculated from the measured O2 concentrations, the O2/Ar ratios, the observed SST and salinities (SSS) from the ship's thermosalinograph and the local barometric pressure. For O2, the solubility parameterization of García and Gordon  based on data from Benson and Krause  is used. For Ar, we use the parameterization of Hamme and Emerson . SST measured during the CTD cast was within (0.00 ± 0.04)°C of the temperature at the seawater intake, but SSS had to be corrected upward by 0.20 psu, due to calibration errors of the thermosalinograph. The argon supersaturation (ΔAr) can be inferred from ΔO2 and ΔO2/Ar as ΔAr = (ΔO2 − ΔO2/Ar)/(1 + ΔO2/Ar).
 This cruise was the first deployment of our newly-developed instrument. Due to experimental problems with corrosion and clogged membranes, O2/Ar ratio coverage is not continuous. We obtained reliable O2/Ar data from the zonal transects and from two partial meridional transects between 8°S and the equator at 110°W and between 8°N and the equator at 95°W.
 Continuous ΔO2 records based on optode data are generally similar between 2°N and 8°N for the 95°W and 110°W transects (Figure 1). Clear differences between 95°W and 110°W are seen in the upwelling region near the equator and smaller differences exist south of the equator, although the ΔO2 variations at 95°W and 110°W are qualitatively similar between 2°S and 8°S. The surface water ΔO2 and SST variations reflect the physical transport mechanisms that prevail in the equatorial Pacific. Foremost, the upwelling of cold, O2-undersaturated waters is evident between the equator and 2°S at 95°W and between just south of 2°N and 1.5°S at 110°W. The northern transition corresponds to the boundary between the westward flowing South Equatorial Current (SEC) and the eastward flowing North Equatorial Countercurrent (NECC). It is particularly well-articulated at 2°N near 110°W and occurs within less than half a nautical mile. South of the equator, waters are re-saturated at about 2°S at 95°W. However, at 110°W O2 is again undersaturated between 2°S and 4°S. This could be a remnant due to lateral advection of low O2 waters from the coastal upwelling region west of Peru [Toggweiler et al., 1991]. Another upwelling region is located between 9°N and 10°N near the boundary between NECC and the westward flowing North Equatorial Current.
 Outside of the regions influenced by upwelled waters, O2 is at or above its saturation concentration, except for a region between 5°N and 1°N sampled during the second 95°W transect. A possible explanation for this undersaturation is to be found in the equally undersaturated Ar concentrations with ΔAr values down to −0.8%. There are two mechanisms that could explain its origin: a recent increase in barometric pressure (which contributed about −0.2%) or a decrease in temperature (which was not observed for the time between the first and second 8°N to 0°-transect at 95°W). Sampling and analysis errors may also contribute. This is suggested by a duplicate discrete ΔO2/Ar measurement at 2°N, which is 0.5% higher than the ΔO2/Ar values measured by MIMS. South of the equator, ΔAr values range from 1 to 3% (Figure 2), once again confirming the importance of correcting for the physical supersaturation of O2.
Figure 2 shows the ΔO2/Ar data for the northern part of the first 95°W transect and the southern part of the 110°W transect, in addition to ΔO2 optode measurements, sea surface temperatures and inferred ΔAr values. Both the MIMS and optode give excellent spatiotemporal resolution.
 As discussed above, ΔO2/Ar can be used to estimate the contribution of biology to the observed O2 supersaturation. Neglecting vertical mixing, a mixed-layer O2 balance can be written as
where zmix is mixed layer depth, G is gross production, R is respiration, k(O2) is the O2 gas-exchange coefficient (or piston velocity), csat(O2) is the O2 saturation concentration, and Finj/Fexch are air-sea fluxes due to bubble injection and bubble exchange, with the dimensionless parameters being χ(O2) – O2 mixing ratio in air, Sc(O2) – Schmidt number of O2, α(O2) – Ostwald solubility coefficient ([Hamme and Emerson, 2002] replace the diffusion coefficient of Hamme and Emerson's equation (4) by the Schmidt number for consistency). Taking the time-derivative of (1) and assuming ∂2c/∂t2 = 0 and constant zmix, G, R, k(O2), and Fbubble, we obtain
ψ represents the influence of temperature, salinity and pressure changes on csat. A similar equation applies to Ar, with G = R = 0. Assuming the same relative response of Ar and O2 to variations in ψi (true to within 1% for seawater) yields Fbio = G − R =
Fbio denotes biologically induced sea-air O2 fluxes, which can be equated with mixed-layer NCP in areas where the above assumptions are valid and there is negligible vertical mixing across the base of the mixed layer. Fbio, G, R, kcsat(O2), and Fbubble χ(O2) are O2 fluxes in units of mmol m-2 d-1. To calculate k, 24 h-average winds at the 10 m-height level (NCEP Reanalysis Project, http://www.cdc.noaa.gov/) are used in combination with the quadratic wind-speed gas-exchange relationship and Schmidt numbers of Wanninkhof . The first bracketed term in (3) is equal to ΔO2/Ar + (ΔO2/Ar − 0.01)ΔAr and can be approximated by ΔO2/Ar. The second/third bracketed terms are about 0.1/−0.01, and will be neglected, because even if the entire observed Ar supersaturation of mostly <2% were due to bubbles, the O2/Ar supersaturation would be affected by <0.2%. Fbio is therefore approximated by k(O2) csat(O2) ΔO2/Ar. Values for k are calculated iteratively from the fraction of the mixed layer ventilated on a particular day before sampling, for 120 days backwards in time (M. K. Reuer et al., New estimates of Southern Ocean biological production rates from O2/Ar ratios and the triple oxygen isotope composition of O2, submitted to Deep-Sea Research, Part I, 2005). Mixed layer depths, taken here as the depth where θ is 0.5°C less than at the surface, vary from 10 m in upwelling regions near the equator at 10°N to 93 m at 2.5°S on the 110°W line. Mixed layer depths only enter implicitly into the calculation.
 The derived Fbio values are shown in Figures 3 and 4for the meridional and zonal transects, respectively. Clearly, Fbio is higher at 8°S (5–18 mmol m−2 d−1) than at 8°N (0–4 mmol m−2 d−1). To the extent that Fbio can be related to NCP, this meridional gradient is in semi-quantitative agreement with nitrate-based estimates of new production in the eastern equatorial Pacific [Fiedler et al., 1991], which showed about 10 times higher new production at 8°S than at 8°N. This may be due to higher nutrient concentrations in the south, advected into the region by the SEC from the Peru upwelling area, which is supported by modeling results [Jiang and Chai, 2005]. Nitrate concentrations were indeed between 6 and 9 mmol m−3 south of the equator, and below 0.4 mmol m−3 north of 0.5°N at 95°W and north of 2°N at 110°W, suggesting nitrate limitation in the north. Silicate was shown to regulate new production further west in the equatorial Pacific (140°W) [Dugdale and Wilkerson, 1998], but was not measured during the present cruise. Low iron concentrations, together with data showing that iron amendments enhance growth [Gordon et al., 1998], indicate that the waters are iron-limited [Coale et al., 1996]. However, iron concentrations seem to be lower south of the equator than north of it, which would rather support a productivity trend opposite to the one we observed [Jiang and Chai, 2005].
4. Discussion and Conclusions
 Our simple O2 mass balance approach does not constrain net community production (NCP) where vertical mixing affects the O2 budget, especially in the equatorial upwelling region. Similar to Hendricks et al.  and based on the observed SST and ΔO2 values, we therefore restrict our analysis to north of 2.75°N and south of 6.75°S. North of the equator, mixed-layer NCP is low throughout a reach of 600 km between 8°N to 2.75°N. South of the equator, we cannot say how far the higher productivities extend north from 6.75°S because of the competing influences of productivity, upwelling/advection of low O2 waters and gas exchange. Our NCP estimates are similar to previous values derived from discrete O2/Ar measurements, nitrate budgets and 15N incubations [Hendricks et al., 2005, and references therein].
 O2/Ar ratios, together with wind-speed gas-exchange parameterizations, yield the biologically induced sea-air O2 flux, Fbio. This term integrates NCP over larger reaches of the ocean when vertical mixing and lateral mixing are neglected, but an exhaustive interpretation likely requires an ocean transport model. The potential of MIMS to measure other gases and gas ratios [Tortell, 2005] and to provide absolute gas concentrations in connection with precise O2 concentrations measurements make it a very promising technique, e.g., in order to use of ΔAr values to infer heat fluxes or ΔN2 values to unravel the influence of bubble injection.
 We thank the crew and scientific party of NOAA R/V Ron Brown for help and support, the TAO Project (Mike McPhaden and Kristy McTaggart) for providing access and calibrated CTDO results, Greg Cane and Dennis Graham for nitrate measurements and the Gary Comer Foundation, NSF and Princeton University (Hess Fellowships to J. K. and M. K. R.) for funding.