We measured gross primary productivity (GPP) in vitro and GPP and net community production (NCP) in situ on four cruises to the Hawaii Ocean Time series (HOT) station ALOHA during 2002–2003. In vitro GPP, determined by 18O labeling, yielded integrated production (0–100 m) that was on average 1.5 times the 14C integrated production. Mean integrated productivity from two winter and two summer cruises was 575 mg C m−2d−1 and 930 mg C m−2d−1, respectively. In situ GPP, determined from the triple-isotope composition of dissolved O2, averaged 910 mg C m−2d−1 in the winter and 1225 mg C m−2d−1 in the summer/fall, with an uncertainty of ±40%. The NCP/GPP ratio, determined using O2/Ar gas ratio and oxygen isotope measurements, was around 0.1 in the summer, close to the canonical f-ratio for the open ocean, indicating station ALOHA is a net autotrophic system during summer months. The consistently higher gross carbon production measured by the in situ method, which integrates production over the ∼2-week residence time of O2 in the mixed layer, suggests that aperiodic bursts of production contribute significantly to time-averaged mean productivity at station ALOHA.
 Accurate and unambiguous measurement of marine primary productivity (PP) is essential to an understanding of global carbon cycling. As the principle source of organic carbon to marine ecosystems, PP, currently estimated at ∼50 Gt C per year [Longhurst et al., 1995; Antoine et al., 1996; Behrenfeld and Falkowski, 1997], influences the magnitude of ecologically and geochemically relevant quantities, such as net community production (NCP), a quantity often equated with carbon export. Uncertainty over the response of PP and NCP to climate-forced changes in the surface ocean (e.g., temperature, pH, mixed layer depth, and nutrient delivery) and the potential for this oceanic response to feedback on future atmospheric CO2 levels [Sarmiento et al., 1998; Sarmiento and Hughes, 1999] make routine and accurate measurement of marine PP essential. PP rates serve as a benchmark property of ecosystems, and variability in these rates on seasonal, annual, decadal, or multidecadal timescales can serve as an indication of ecosystem function, allowing identification of natural modes of variability or long-term trends [Karl et al., 2001; Chavez et al., 2003].
 Unfortunately, most PP measurements to date have been laden with uncertainty. The 14C in vitro method, introduced by Steemann Nielsen , has over the last 50 years become the standard means of determining PP in the field. However, interpretation of 14C PP measurements has been troublesome. Many of the problems stem from the in vitro methodology. Because in vitro methods involve removing the plankton community from the natural environment and confining it in a small volume, it cannot be assumed that the rates observed in vitro equate to natural, uncontained PP rates [e.g., Peterson, 1980; Harrison and Harris, 1986; Marra, 2002]. Also, because in vitro measurements sample small space and timescales, the variability in PP rates inherent in natural communities is rarely observed, and therefore measured rates often do not reflect the time-averaged mean PP rate [Platt et al., 1989; Karl et al., 2003; Williams et al., 2004]. The tendency to extrapolate these measurements to larger space scales and timescales, despite chronic undersampling, compounds this problem. Williams et al.  suggested that the issues inherent to in vitro methodologies may be responsible for the observations of heterotrophy in the oligotrophic ocean [e.g., delGiorgio et al., 1997; Duarte and Agusti, 1998], despite the lack of evidence for a sustainable supply of organic carbon. Similar arguments have been made for the classic mismatch between PP determined in vitro and PP determined in situ using chemical budgets [Schulenberger and Reid, 1981; Jenkins and Goldman, 1985; Platt et al., 1989].
 Other problems arise with the use of the 14C label. Comparisons among PP derived from 14C and other in vitro methods, such as the O2 light/dark bottle method, and incubations with 18O- labeled H2O, a measure of gross PP (GPP), have shown that 14C in most cases yields a quantity that lies ambiguously between NCP and GPP [Peterson, 1980; Bender et al., 1987, 1999]. Recycling of particulate organic 14C (PO14C) over the course of the incubation is one likely cause of the 14C method yielding a rate less than GPP. In addition, if dissolved organic 14C (DO14C) is produced but not measured, the 14C-derived rate will underestimate GPP. The degree of 14C recycling, the source of the respired POC, the ability for respiratory CO2 to be refixed into POC, and the production of DO14C are processes that presumably depend on phytoplankton community structure and metabolic state and, as a result, complicate attempts to interpret the PP rate that 14C measures [Richardson et al., 1984; Peterson, 1980]. Differences in 14C methodology, including differences in incubation location (shipboard incubators versus in situ), filtration and radioactivity counting of samples, and calculation of PP from the resulting data can lead to substantial variability (∼20%) in PP estimates, even when supplied with standardized samples and data sets [Richardson, 1991].
 The uncertainties associated with the 14C method make comparisons with PP estimates determined using alternative methods useful for constraining PP of marine ecosystems. For example, investigations have used 18O-labeled H2O for in vitro GPP determinations [Bender et al., 1987; Grande et al., 1989; Bender et al., 1999]. This method measures the photosynthetic transfer of an 18O label from water to dissolved O2. Because the pool that is labeled, dissolved O2, is much larger than the POC pool (e.g., 200 μmol/L versus 1–2 μmol/L), it is ∼100 times less sensitive to recycling than the 14C label. In comparisons of 18O and 14C-determined PP the 18O method typically yields a PP rate that is 1.5–2 times the 14C PP [Bender et al., 1987; Grande et al., 1989; Bender et al., 1999]. However, the results of 18O-labeling are subject to the same in vitro uncertainties as 14C-labeling, and must be approached with some caution.
 Recently, Luz and Barkan  showed that measurement of the natural abundance of O2 isotopes could be used to calculate GPP in situ. This method relies on the ability to distinguish photosynthetically produced O2 from atmospheric O2 present in the mixed layer due to air-sea exchange by use of the natural abundance of three isotopes of oxygen (16O (99.76%), 17O (0.04%), and 18O (0.20%)). By measuring the 18O/16O and 17O/16O of dissolved O2 and estimating air-sea gas exchange rates, as discussed below, this method provides a measure of GPP. In addition, measurements of the in situ O2/Ar gas ratio of the mixed layer can be combined with the oxygen triple isotope composition to yield rates of NCP [Bender, 2000]. The advantages of this method are clear. First, as no bottle incubations are employed, there are no uncertainties related to containment effects. Second, because the method measures the isotope composition and O2/Ar gas ratio on a bulk property, mixed layer dissolved O2, the measured PP is the time-averaged mixed layer productivity over the residence time of O2 (∼1–2 weeks). As a result, this in situ method is much more likely to capture episodic changes in PP rates missed by bottle incubations. However, because it relies on a relatively uncertain parameterization of the air-sea gas exchange rate, the uncertainty in the estimated GPP is significant (±40%).
 In an effort to evaluate 14C-based PP rates, we measured GPP using the 18O in vitro method and the triple isotope in situ method and measured NCP using the O2/Ar gas ratio in situ method on four cruises to the Hawaii Ocean Time series (HOT) oligotrophic ocean site Station ALOHA (22°45′N, 158°00′W) between February 2002 and August 2003. The HOT program and Station ALOHA, located in the subtropical Pacific approximately 100 km north of the Hawaiian Island of Oahu, are ideal candidates for this study for several reasons. First, as part of a long-term time series study, Station ALOHA is well characterized with respect to hydrographic and biogeochemical measurements [Karl, 1999; Karl et al., 2002]. PP has been determined by the 14C incubation method monthly at Station ALOHA since the HOT program's inception in 1988 [Karl and Lukas, 1996]. Prior to the HOT program, 14C-based PP measurements were made in nearby areas semi-frequently during various process studies (e.g., CLIMAX, VERTEX, PRPOOS). Second, several in situ budgets (e.g., O2, 234Th, DIC/DI13C) have been used to estimate carbon export at Station ALOHA [Emerson et al., 1997; Benitez-Nelson et al., 2001; Quay and Stutsman, 2003]. Comparison of these long-term (annual) integration methods with the short-term (daily) 14C and 18O incubations, and intermediate-term (weekly) triple isotope measurements is helpful for understanding PP and NCP variability on various timescales. Finally, as an oligotrophic regime with relatively low 14C-based PP, HOT Station ALOHA has been argued to be a net heterotrophic system [delGiorgio et al., 1997; Duarte and Agusti, 1998; delGiorgio and Duarte, 2002]. We hope to contribute to the discussion of this issue with the help of the 18O in vitro and triple isotope and O2/Ar in situ tracers.
 Historically, definitions of PP have been quite vague [cf. Williams, 1993]. In an effort to avoid confusion, we clarify our terminology. Throughout this paper we equate gross O2 production, whether measured by the 18O in vitro method or the triple isotope in situ method, with GPP. PP as measured by the 14C method will be referred to as 14C PP. Defining these terms operationally as opposed to conceptually makes presentation of the data straightforward. We will discuss the relationship of these terms to parameters of interest to carbon cycle studies (i.e., gross and net autotrophic carbon production) in section 4.3.
3.1. The 18O In Vitro Method: Background
 The 18O in vitro method provides a measure of GPP by isotopic labeling of photosynthetically produced O2. When water is split during the light reactions of photosynthesis, it forms O2 and reducing agents used for electron transport. This photosynthetic O2 has the same 18O/16O as the water from which it was formed [Guy et al., 1993]. Therefore, when a natural plankton community is spiked with H218O, the measured 18O enrichment of dissolved O2 over the incubation interval can be used to determine GPP by mass balance [Bender et al., 1987; Grande et al., 1989; Bender et al., 1999]:
where δ18O is defined in the conventional way,
and δ18O(O2)init, δ18O(O2)final, δ18Owater, and [O2]init refer to the initial and final δ18O of dissolved O2, the δ18O of the incubation water, and the concentration of dissolved O2, respectively, relative to an air standard.
 There are several advantages of this method. First, the labeled dissolved O2 pool is very large (∼200 μmol/L) compared to the biomass pool (∼1–2 μmol/L), and therefore the 18O GPP method is much less sensitive to recycling than the 14C method. That is, respiration may consume some of the labeled O2, but it does not significantly affect the δ18O of the remaining O2 because the fraction of the pool being removed by respiration is small. Also, the label is restricted to dissolved O2, whereas with the 14C PP method the label can be distributed among various pools, including POC and DOC [Bender et al., 1987, 1999]. Additionally, the method does not require the assumption that light respiration is equivalent to dark respiration. This assumption is necessary in the commonly used O2 light/dark bottle method (hereinafter referred to as the ΔO2 method), although studies have shown that light and dark respiration rates are not always equivalent [Grande et al., 1991; Kana, 1992]. The 18O in vitro method therefore measures GPP, provided there is no intracellular recycling of labeled O2; that is, the method assumes that all photosynthetic labeled O2 reaches the external dissolved O2 pool [Bender et al., 1987]. An important caveat is that PP by bacteria capable of photosynthesizing anoxygenically [Karl, 2002] would cause the 18O method to measure something less than GPP. At present the extent to which these bacteria contribute to GPP is not known.
3.2. The 18O In Vitro Sample Collection and Analysis
 We measured GPP via the 18O in vitro method on four cruises to Station ALOHA on the R/V Ka'imikai-O-Kanaloa: HOT 135 (February 2002), HOT 140 (October 2002), HOT 145 (February 2003), and HOT 151 (August 2003). Sample collection and incubation for HOT 135 was slightly different than for the subsequent cruises in order to compare with simultaneous ΔO2 PP measurements. The HOT 135 samples were collected on a separate hydrocast from the 14C PP samples and were incubated in situ on a free drifting array for 24 hours (dawn to dawn). Details on sample collection and incubation for this cruise are described by Williams et al. . Sample collection for HOT 140, 145, and 151 was from the same hydrocast as the HOT program 14C PP samples. These samples were incubated in situ on the 14C free drifting array for ∼12 hours (dawn to dusk) [Karl et al., 1996] (available at http://hahana.soest.hawaii.edu/hot/protocols).
 For all cruises, water samples were collected from and incubated at standard 14C PP depths (5, 25, 45, 75, and 100 m). Before each cruise the incubation bottles (quartz glass with a nominal volume of 120 mL) were soaked in 10% HCl for a week and rinsed with Milli-Q™ water. The H218O spike was triple distilled prior to use with a sub-boiling Teflon still [Hinn and Nelson, 1997]. Concentrations of several metals, including Cu, Co, and Ni, measured on a Perkin-Elmer Inductively Coupled Plasma Mass Spectrometer (ICPMS), yielded concentrations below background seawater for a 400 μL of spike diluted 300-fold (as in our incubations). Similarly, measured NO3− concentrations were below background seawater concentrations (<0.03 μM). Water samples for incubation were collected in the dark, using “clean” techniques as much as possible [Fitzwater et al., 1982]. Powderless polyethylene gloves and acid-washed silicone tubing were used to collect samples. Incubation sample bottles were rinsed with sample water and then filled after overflow of three volumes of sample.
 Determining GPP from equation (1) requires the measurement of four terms (i.e., δ18O(O2)init, δ18O(O2)final, δ18Owater, and [O2]init). During HOT 140, 145, and 151, one initial and three final measurements of δ18O were made on dissolved O2. For HOT 135, two initial and two final δ18O measurements were made. We collected the dissolved gases for measurement of δ18O (O2) using the method of Emerson et al. . Samples were siphoned from the incubation bottles into pre-evacuated, HgCl2 poisoned, 200 mL glass flasks equipped with a 9-mm-bore Louwers-Hapert™ single or double o-ring valve. Samples were kept free of atmospheric O2 contamination by flushing the space before the valve with CO2, establishing a water lock with a small volume of the sample, eliminating all bubbles in the space before the valve, and maintaining the water lock while the valve was open and the water sample was emitted into the flask. The flasks were filled with ∼50 mL of water taken directly from the bottom of the incubation bottle in order to limit atmospheric contamination. The δ18O (O2)init samples were taken immediately after the array was released, and the δ18O(O2)final samples were taken immediately after the array recovery.
 After sampling, the water was equilibrated with the headspace by continuous agitation for 8–10 hours, allowing 97–98% of the dissolved gases to exsolve [Emerson et al., 1999]. The sample water was then removed by vacuum, leaving the gases in the headspace for analysis. After cryogenic removal of water and CO2 each sample was frozen into a stainless steel finger immersed in liquid helium and then attached to a Finnigan MAT 251 Isotope Ratio Mass Spectrometer (IRMS) for δ18O analysis [Quay et al., 1993].
 The δ18Owater was calculated from calibrated volume of each of the incubation bottles (±0.005 mL), the calibrated volume of the Eppendorf™ pipette delivery (±0.8 μL) and the certified δ18O value of the H218O spike (95.1% 18O, Marshall Isotopes, Ltd., Tel Aviv, Israel). The δ18Owater of the H218O spike was verified by measuring δ18Owater values from 20 diluted samples ranging from −10 to +80‰ by automated equilibration with a headspace of CO2 and subsequent determination of the headspace gas isotopic ratios on a Finnigan MAT 251 mass spectrometer. Predicted δ18Owater values agreed with measured ones within 5%. All samples on the HOT cruises were spiked with 400 μL of H218O to bring the δ18Owater to approximately 1500‰. This enrichment in δ18Owater produced typical δ18O(O2) enrichments for surface incubated samples of ∼10‰ and δ18O(O2) enrichments for 100-m incubations of ∼1‰. Since our typical precision for measuring δ18O(O2) is ±0.05‰, this allowed us to measure GPP toward the base of the euphotic zone where low PP rates are much more difficult to measure by other methods (ΔO2).
 The final term required for the calculation of GPP, the dissolved O2 concentration ([O2]init), was measured by Carpenter-modified Winkler titration [Carpenter, 1965] with visual detection of the endpoint. Samples for dissolved O2 determination were collected at the same time as the incubation samples, temporarily stored in the dark, and fixed immediately after the array deployment (typically 3 hours after sample collection).
 The error in calculated GPP resulting from the analytical errors for each of the four terms in equation (1) is small (1–5%) compared to the typically observed variance in GPP among the three incubated samples (∼10–20%). Calculated GPP values are most sensitive to the degree of δ18O(O2) enrichment between initial and final samples. For example, the error in calculated GPP is typically ±1% for surface incubations, where enrichments are on the order of 10‰, and ±5% for 100 m incubations, where enrichments are typically 1‰.
3.3. Triple Oxygen Isotope and O2/Ar In Situ Methods: Background
 The natural abundance of the three stable isotopes of dissolved oxygen (16O, 17O, 18O) can be used to determine GPP in situ because atmospheric O2 has a triple isotope signature that is distinct from photosynthetic O2 [Luz et al., 1999; Luz and Barkan, 2000]. Most processes fractionate isotopes in a mass dependent way. For example, during respiration a 17O16O molecule is discriminated against half as much as an 18O16O molecule relative to a 16O16O molecule. As a result, the remaining pool of O2 becomes more enriched in both 17O and 18O as O2 is removed by respiration, but the ratio of 17O to 18O remains constant at about 0.5. In contrast, photochemical reactions in the stratosphere involving O2, O3, and CO2 fractionate O2 mass independently; these reactions result in CO2 that is anomalously enriched in 17O relative to 18O and O2 that is anomalously depleted in 17O relative to 18O with a ratio of 1.7 instead of 0.5 [Luz et al., 1999; Thiemens et al., 1995; Lämmerzahl et al., 2002]. The anomalous, 17O-depleted O2 mixes into the troposphere, and subsequently into the surface ocean via air-sea gas exchange. Photosynthetic O2, which has the isotopic composition of seawater, adds O2 that is enriched in 17O relative to the 17O-depleted tropospheric O2 derived from air-sea exchange [Luz and Barkan, 2000].
where δ17O is defined analogously to δ18O in equation (2) and Δ17O is given in units of per meg (1000 × per mil). Tropospheric O2, with a δ17O less than 0.52(δ18O) by 0.25‰ (250 per meg), is defined as the standard and thus has a Δ17O = 0 [Luz et al., 1999; Luz and Barkan, 2000]. Defined this way, input of photosynthetic O2 to the mixed layer results in a positive Δ17O. Respiration will not affect Δ17O, as it is assumed to remove O2 with a slope of 0.52 (see equation (3)).
 The Δ17O anomaly, as defined in equation (3), is sensitive to both the isotopic value of the reference standard and the mass-dependent fractionation slope (0.521) [Miller, 2002]. Following the suggestion of Miller , a new definition was proposed by Angert et al.  to remove these sensitivities,
where 17Δ is measured in per meg units and θ is the mass dependent fractionation slope, measured for dark respiration, the dominant form of respiration, at 0.516. When δ17O and δ18O are close to zero (as is true for all of our mixed layer samples) the 17Δ can be accurately approximated as follows:
Luz and Barkan  showed that GPP in the mixed layer could be determined from the measured 17Δ of dissolved O2 (17Δdiss) using a steady state mixed layer oxygen isotope budget similar to those used previously for O2 [Emerson et al., 1997]. The model used by Luz and Barkan  includes three processes: air-sea gas exchange (G), photosynthesis (P), and respiration (R). They empirically determined the 17Δ signal of mixed layer O2 derived from gas exchange as +16 per meg relative to air O2. The +16 per meg offset from air represents the 17Δ of dissolved O2 in equilibrium with air (17Δeq), and indicates that O2 solubility fractionation has a slightly higher θ value than 0.516. The 17Δ of marine photosynthetic O2 was shown to equal the 17Δ of seawater (17Δw), i.e., +249 per meg relative to air, based on culture experiments [Luz and Barkan, 2000]. The final process, respiration, removes O2 with the 17Δ of dissolved O2, but does not change 17Δdiss. The 17Δ of dissolved O2 is therefore controlled by the balance between P, which increases 17Δdiss toward 249 per meg, and G, which lowers the value toward 16 per meg. If the rate of air-sea O2 gas exchange can be quantified by a parameterization based on wind speed [e.g., Liss and Merlivat, 1986; Wanninkhof, 1992; Nightingale et al., 2000], the GPP can be determined as follows:
where K = gas exchange rate (piston velocity) and Co is the O2 concentration in equilibrium with air for the given temperature and salinity, as shown by Luz and Barkan . We used the average 17Δdiss measured in the mixed layer, with the mixed layer depth defined by 0.125 sigma unit difference from the surface. Thus the GPP calculated represents the average gross production in the mixed layer over the lifetime of O2 (typically 2 weeks).
 The simple mixed layer budget approach used by Luz and Barkan  to calculate GPP only accounts for GPP in the mixed layer, and thus underestimates GPP when the euphotic zone is deeper than the mixed layer. Additionally, the budget lacks terms for advection and mixing effects on 17Δdiss. The effect of these simplifications on GPP is small and will be discussed in section 5.1.
Bender  showed that the measured in situ O2/Ar dissolved gas ratio could be combined with the 17Δdiss GPP to provide an estimate of NCP/GPP. This is possible because dissolved Ar gas behaves similarly to dissolved O2 with respect to physical processes, but does not have a biological source or sink like O2 [Emerson et al., 1997]. Therefore the ratio of the measured O2/Ar, referred to as (O2/Ar)meas, to the O2/Ar expected based on solubility equilibrium with air, referred to as (O2/Ar)eq [e.g., Garcia and Gordon,1992; Hamme and Emerson, 2004], indicates the net effect of biological O2 production and consumption, or NCP. For example, when (O2/Ar)meas/(O2/Ar)eq = 1, there is no net biological production or consumption of O2 and NCP = 0. When 17Δdiss is plotted versus (O2/Ar)meas/(O2/Ar)eq, isolines of constant NCP/GPP (similar to an f-ratio) radiate away from the point that marks equilibrium with the atmosphere [Bender, 2000].
3.4. Triple Isotope and O2/Ar Sample Collection and Analysis
 We measured 17Δdiss profiles at Station ALOHA on the same four cruises where we measured 18O in vitro GPP, i.e., HOT 135, 140, 145, and 151. Duplicate samples for 17Δdiss were collected at standard depths from 10 to 200 m (10, 20, 40, 60, 80, 100, 120, 140, 200 m) using up to four separate water casts. For HOT 145 and HOT 151 the depth profiles were extended to 300 m. Each depth was sampled for 17Δdiss immediately after recovery, following the sampling protocols of Emerson et al. . Because a larger O2 signal was needed for mass spectrometer analysis, we used 500-mL glass flasks.
 Removal of N2 from the gas mixture is necessary for high-precision triple isotope measurements because differences in the N2/O2 ratio of the gas sample relative to the working reference gas cause interference in the measurement of δ17O and δ18O [Barkan and Luz, 2003]. To remove N2, gas samples were manually prepared following the procedures of Barkan and Luz  and Blunier et al. . First, samples were purified of water and CO2 by passing each through a glass U-tube held at −196°C. The remaining gases, O2, Ar, and N2, were collected on coarse 5-Å molecular sieve held at −196°C. The gases were then warmed to +80°C and eluted with a high purity He carrier gas at 10 cc/min through a chromatographic column (4 m × 1/8 inch stainless steel tube packed with 45/60 mesh 5Å molecular sieve held at 25°C) equipped with a thermal conductivity detector for gas peak detection. O2 and Ar eluted first and were trapped on coarse molecular sieve held at −196°C. After complete elution of O2 and Ar the carrier flow was diverted, separating the N2 from the sample. Excess He was then removed by vacuum while the collected O2 and Ar were held at −196°C. Finally, the O2/Ar gas mixture was warmed to +80°C and transferred into a stainless steel tube immersed in liquid helium, which was later attached to a Finnigan MAT 251 IRMS for measurement of δ17O, δ18O, and O2/Ar.
 Precise determination of 17Δdiss on each sample was attained from 75 paired measurements of δ17O and δ18O against an O2 standard. The isotope ratios were measured by simultaneous collection of mass 32, 33, and 34 in the dry gas mixture [Quay et al., 1993]. The typical precision of the repeated δ17O and δ18O measurements averaged 0.05 and 0.02‰, respectively. The standard error of 17Δdiss, based on 75 individual determinations from each paired δ17O and δ18O, averaged ±5 per meg. Each 17Δdiss value was corrected for sensitivity of the ionization efficiency of the individual oxygen species, 16O16O, 17O16O, and 18O16O, to differences in the O2/Ar ratio. The 17Δdiss values were then referenced to air based on the 17Δ value of an air standard passed through the manual gas separation line each day. The accuracy of this procedure was confirmed by measurement of the 17Δdiss of deionized tap water equilibrated with air (17Δeq = 18 ± 3 per meg (n = 4), which agrees with the +16 per meg value reported by Luz and Barkan ). Duplicate samples collected from the same Niskin had an average standard deviation of ±5 per meg.
 The O2/Ar gas ratio was determined by mass spectrometric measurement of mass peaks 32 and 40 using a single collector following the 75 determinations of δ17O and δ18O. An estimate of the internal instrument precision of the O2/Ar measurement is not possible as the ratio is calculated from a single determination of the gas percentages of O2 and Ar. However, O2/Ar values measured this way agreed with O2/Ar values from duplicate samples determined using the method of Emerson et al. , in which the O2/Ar ratio is determined by the average of three determinations, to within 0.4%. The standard deviation of O2/Ar gas ratios for duplicate samples averaged ±0.3%. The (O2/Ar)eq for each sample was determined using measured sea surface temperature and salinity and the solubilities of O2 and Ar reported by Garcia and Gordon  and Hamme and Emerson , respectively.
 The air-sea gas exchange rate was estimated from the wind speed relationship of Nightingale et al. . We chose this gas exchange parameterization because it is intermediate between the Liss and Merlivat  and the Wanninkhof  relationships. For all cruises except HOT 145 we used wind speed data normalized to a speed 10 m above the sea surface (u10) from National Data Buoy Center (NDBC) buoy 51001, located 170 nautical miles west of the Hawaiian island of Kauai (http://www.ndbc.noaa.gov). Since these data were unavailable for February 2003, we used wind speed data from NDBC buoy 51004, located 185 nautical miles SE of Hilo, Hawaii, for HOT 145. The residence time of O2 in the mixed layer, the time interval over which winds were averaged, was determined based on mixed layer depth and wind speeds, and ranged from 6 to 19 days. Because winds were often variable over this interval, we calculated the gas transfer rate using an average u10 wind speed calculated from the root mean square of hourly wind data.
4. Results and Discussion: 18O GPP
4.1. Comparison of In Vitro Rates: 18O GPP Versus 14C PP
 Profiles of volumetric 18O GPP rates are shown with 14C PP rates determined by the HOT program in Figure 1. A conservative photosynthetic quotient (PQ) of 1.25 (ΔO2/−ΔCO2) was used to convert 18O GPP rates to equivalent carbon production for comparison with 14C PP rates. This PQ value reflects production based on an equal balance of nitrate and ammonia [Laws, 1991] and is similar to those previously used to convert between O2 and carbon production rates in the same study region [Grande et al., 1989; Emerson et al., 1997]. In the oligotrophic ocean the percentage of production supported by ammonia is probably much higher than 50%, and the actual PQ probably closer to 1.1, except when occasional wind-driven mixing events inject nitrate into the mixed layer in winter and spring. We used a PQ of 1.25 for consistency with previous studies and a conservative approximation of carbon production for purposes of the 18O GPP and 14C PP comparison. A more rigorous examination of the carbon equivalents of both 18O GPP and 14C PP will be presented in section 4.3.
 As expected, both labeling methods yield in vitro measurements of PP that decrease with depth. The change in PP with depth and the variability of the PP measurements at a given depth were similar for 14C and 18O within each cruise (Figure 1), indicating that analytical and procedural errors in the individual methods were small with respect to the biological signal and its variability. Additionally, the 18O GPP rates agree well in magnitude with values previously reported from a nearby study site (28°N, 155°W) [Grande et al., 1989].
 Overall, volumetric PP rates were higher when measured by the 18O method. The difference between the two methods for each cruise was greatest at the surface, where 18O GPP was typically 1.5 to 2 times the 14C PP (with the exception of HOT 135), and least toward the base of the euphotic zone, where 14C and 18O rates were comparable. There were also seasonal trends in the surface 18O GPP rates, with lower values (0.7 mmol C m−3d−1) in winter (HOT 135, 145) and higher values (1–1.2 mmol C m−3d−1) in summer and early fall (HOT 140, 151). In contrast, surface PP rates measured by the 14C method were roughly constant throughout the year (∼0.5 to 0.6 mmol C m−3d−1), with only a slight increase during summer and early fall. The ratio of 18O GPP:14C PP therefore varied seasonally, with maximum values in the summer and minimum values in the winter.
 Similar trends with season and/or depth have been observed in previous comparisons between 14C PP and 18O GPP [Grande et al., 1989; Bender et al., 1999] and between 14C PP and ΔO2 GPP [e.g., Williams et al., 2004]. Williams et al.  suggested that changes in in situ irradiance may be the cause for both seasonal and depth trends in the ΔO2 GPP:14C PP, arguing that at saturating irradiances, CO2 in intracellular pools may become limiting and internal 12C-rich CO2, such as respiratory CO2, may be utilized, causing 14C PP to measure closer to net rather than gross production. Conversely, in light-limiting conditions, 14C measures closer to gross production.
 Differential rates of DO14C production with depth and season may provide another possible explanation for the observed seasonal and depth trends in the 18O GPP:14C PP ratio. Two lines of evidence from the work of Karl et al.  support this claim. First, production of DO14C, which is not fully accounted for in the standard 14C PP method used at ALOHA, may contribute substantially (30–50% of PO14C production) to total production. Second, profiles of DO14C and PO14C production measured in July 1996 at Station ALOHA indicate that DO14C production might account for a larger percentage of the total production (PO14C + DO14C) in the surface and a smaller percentage at depth. Seasonality in DOC production at ALOHA might be expected in response to changes in temperature, irradiance, or nutrient availability and the effects these environmental cues have on structuring the phytoplankton community. For example, there is growing evidence from several data sets, including measurements of N:P stoichiometry, δ15N of sediment trap material, and N2-fixing organism population abundances, that cyanobacterial N2 fixation may support a substantial proportion (>75%) of primary production during the summer and early fall when surface ocean mixing is reduced at Station ALOHA [Karl et al., 1998]. In contrast, Karl et al.  estimate that winter N2-fixation supports a much smaller proportion (∼20%) of the total production. This seasonality in community composition affects nutrient cycling and export rates during this time [Karl et al., 1998], and might conceivably be linked to enhanced DOC production. If DO14C is produced in greater quantites in the surface relative to depth and in the summer relative to the winter, this would cause the 18O GPP:14C PP ratio to change with depth and with season, as observed. However, there is no direct evidence of enhanced DOC production in communities supported by N2-fixation.
 The 18O GPP rates measured during HOT 135 were anomalous, as they were similar to 14C PP rates for the whole water column. The collection and incubation methods of HOT 135 samples were slightly different from subsequent cruises because the 18O GPP were collected from a hydrocast that occurred 3 hours prior to the 14C PP hydrocast and samples were incubated for 24 hours (instead of 12 hours). It is unlikely that the length of incubation influenced results because the 18O GPP method is not affected by respiration [Bender et al., 1987]. Water mass heterogeneities and the 3-hour time lag between collection of 18O GPP samples and 14C PP samples (see section 3.2), however, could complicate the comparison. Consistent with the 18O GPP estimates, GPP determined for HOT 135 by the ΔO2 method on samples collected from the same cast as our 18O GPP samples are also anomalously low compared with other winter data [see Williams et al., 2004]. The fact that GPP profiles determined by both methods are atypically comparable to 14C PP profiles suggests that either some methodological feature common to both oxygen methods was in error, or that heterogeneity in the environment caused both methods to give anomalously low GPP estimates.
4.2. The 18O GPP and 14C PP Integrated Production
 Depth-integrated 14C PP and 18O GPP rates were determined using a trapezoidal approximation for the upper part of the euphotic zone (0–100 m) and for the mixed layer (Table 1). We were unable to integrate to the base of the euphotic zone (∼175 m [Letelier et al., 1996]) because our 18O GPP measurements only extended to 100 m. However, production below 100 m, based on historic 14C PP measurements at ALOHA, is typically only 10% of total euphotic zone production. Production rates integrated to 100 m ranged from 480 to 1040 mg C m−2d−1 for 18O GPP and 470 to 600 mg C m−2d−1 for 14C PP. Overall, the seasonal trends in 18O GPP and 14C PP integrated to 100 m were similar to the trend observed in the volumetric rates. Seasonality was much stronger in the 18O GPP integrated rates, with maximum values in the late summer/early fall (820–1040 mg C m−2d−1) that were 1.6 times higher than winter values (475–675 mg C m−2d−1). In contrast, integrated 14C PP rates increased only slightly (∼20%) in the summer and early fall (560–600 mg C m−2d−1) compared to winter rates (470–520 mg C m−2d−1), and agreed well with magnitudes of 14C integrated PP observations in the last 10 years at Station ALOHA [see Karl et al., 2001, 2002]. Production rates integrated to the depth of the mixed layer (85, 70, 80, and 45 m for HOT 135, 140, 145, and 151, respectively) ranged from 440 to 710 mg C m−2d−1 for 18O and 350 to 460 m−2d−1 for 14C.
Table 1. In Vitro Productivity at Station ALOHA Determined by 18O and 14Ca
Productivity integrated to 100 m and to the base of the mixed layer by trapezoidal approximation.
14C PP (measured by the HOT program). Uncertainty in these estimates comes from the variability in triplicate samples at each depth.
18O GPP rates scaled to units of carbon using a PQ of 1.25. Uncertainty as above.
18O GCP rates scaled to units of carbon using the 18O GPP rates, a PQ of 1.25, and a 15% correction for photorespiration and Mehler activity.
18O NACP, estimated from 18O GCP assuming a 17% autotrophic respiratory loss over the course of 12 hours (see text).
MLD,as determined by a density criterion of 0.125 sigma unit difference from the surface.
HOT 135/Feb 2002
520 ± 40
475 ± 35
470 ± 40
440 ± 40
HOT 140/Oct 2002
560 ± 40
820 ± 30
460 ± 30
710 ± 30
HOT 145/Feb 2003
470 ± 20
675 ± 20
430 ± 20
620 ± 20
HOT 151/Aug 2003
600 ± 30
1040 ± 25
350 ± 20
620 ± 20
 Integrated GPP measured in this study by the 18O method is up to 40% higher than the canonical values reported for Station ALOHA based on 14C, with summer GPP values reaching a rate of 1 g C m−2d−1. This supports the claim of Karl et al. , based on their DO14C production measurements, that 14C is underestimating productivity at Station ALOHA by 30–50%. The significance of DOC export has also been indicated in the surface layer carbon budgets at HOT reported by Emerson et al.  and Benitez-Nelson et al. . For example, downward diffusion of DOC and accumulation of DOC within the mixed layer accounted for approximately 25% of the overall carbon sink terms in the budget of Benitez-Nelson et al. . Emerson et al.  calculate that DOC export accounts for 30–50% of the organic carbon export, and is thus comparable in magnitude to the POC export. If DO14C production is primarily responsible for the difference between the in vitro 18O and 14C primary production estimates, models of organic carbon at Station ALOHA need to be revised to better represent the importance of DOC cycling.
4.3. Estimating Carbon Production From 18O GPP
 Net autotrophic carbon production (NACP, here we use the specific term instead of net primary productivity (NPP), which can be in units of O2 or C) and gross carbon production (GCP) are terms of fundamental interest in carbon cycle studies. GCP represents total carbon fixation by primary producers. NACP represents carbon available to the heterotrophic community. The response of the GCP/NACP ratio to environmental cues helps identify changes in algal physiology that can affect carbon supply to the surface community. However, measurement of these rates is not typically straightforward. GCP can be approximated by measuring the gross evolution of O2, but there are some caveats in the conversion of O2 to carbon. NACP is impossible to observe directly, since autotrophic respiration cannot be isolated from community respiration. Quantification of NACP therefore requires highly uncertain estimates of autotrophic respiration [e.g., Langdon, 1993].
 In the preceding discussion we assumed that measurements of O2 production (as determined by 18O) could be related to GCP by a PQ. However, the 18O GPP method measures total O2 production, regardless of the fate of the O2 produced, and therefore yields an overestimate of GCP. Gross O2 production measured by the 18O labeling method is greater than GCP because not all reducing agents generated in photosystem II when O2 is formed are used for carboxylation. In the Mehler reaction O2 is reduced in photosystem I without any associated carboxylation; this registers as gross O2 production because a molecule of 18O16O is produced while an ambient 16O16O molecule is consumed [Bender et al., 1999]. Photorespiration, a process that involves the oxidation rather than carboxylation of ribulose 1,5-bis-phosphate (RuBP), results in conversion of newly fixed carbon to either glycolate or CO2. If glycolate is formed and excreted, then photorespiration is linked to DOC production. If CO2 is produced, then photorespiration, like the Mehler reaction, results in 18O16O production and 16O16O consumption without any associated carboxylation [Bender et al., 1999].
 Since the Mehler reaction and photorespiration only occur in the light, quantitative estimates of their importance can be inferred from observations of light-enhanced O2 consumption [e.g., Grande et al., 1991; Kana, 1992]. These estimates might be flawed, however, since “dark respiration” (mitochondrial respiration, or the prokaryotic equivalent) rates may also be higher in the light [e.g., Weger et al., 1989]. Nonetheless, Bender et al.  and Laws et al.  used measurements of light-enhanced O2 consumption in culture experiments to estimate that both the Mehler reaction and photorespiration together accounted for 15–20% of gross O2 production as measured by 18O in the Equatorial Pacific and Arabian Sea, respectively. Using this estimate, we calculated GCP from our 18O productivity data by multiplying the gross O2 production by 0.85 to account for Mehler and photorespiration and scaling with a PQ of 1.25 to convert to units of carbon. Laws et al.  argue that the uncertainty associated with this correction is probably no more than ±5%.
 NACP is much more difficult to estimate than GCP. If autotrophic respiration rates are known, NACP can be calculated from GCP. However, estimates of autotrophic respiration are highly uncertain. On the basis of models of autotrophic respiration by Langdon , Bender et al.  suggested that plant respiration accounts for ∼35% of GCP over 24 hours. According to Langdon  this estimate pertains to cyanobacteria-dominated populations, and therefore seems appropriate for application to the autotrophic community at Station ALOHA [see Karl, 1999]. Since we only measured O2 production for 12 hours, and we have no way of estimating the proportion of plant respiration that occurs only in the day, we will assume that plant respiration in our study was approximately half of the 24-hour value, 17% (e.g., NACP = 0.83(GCP) for 12-hour incubations). Again, this estimate is based on a theoretical model and is highly uncertain.
 How do productivities measured by 14C relate to net and gross carbon production? A comparison of 14C PP with GCP and NACP, calculated with correction factors assumed above, is shown in Figure 2. Water column integrated GCP and NACP are reported in Table 1. Although the estimates of NACP are uncertain, our results indicate that 14C PP is equivalent to NACP, except in the upper euphotic zone during summer, when 14C PP underestimates NACP by ∼20%. Agreement between 14C PP and NACP has been noted previously. For example, Bender et al.  found that when gross O2 production rates from the equatorial Pacific were corrected for Mehler activity and photorespiration (15%), autotrophic respiration (35%), and their estimate of DOC production (6%), the remainder (45%) was consistent with measured PO14C production over 24 hours. Marra  used comparisons of net O2 production and 14C PP from the North Atlantic Bloom Experiment and theoretical arguments regarding the observed linear uptake of 14C during incubations to suggest that 14C measures “net production.” The ambiguous “net production” term was used because agreement between net O2 production and 14C PP suggests that 14C approximates NCP while the model results suggest that, to first approximation, 14C estimates NACP. Marra  argues that both the inability to determine the PQ with sufficient accuracy and the variability in the field data may complicate attempts to observe heterotrophic respiration, and therefore to distinguish between NACP and NCP in vitro.
 The agreement between our estimates of NACP and 14C PP, with the exception of the summer/early fall surface estimates, suggests that either heterotrophic respiration of new photosynthate over the 12-hour incubation is negligible or it is within the variability of the data [Marra, 2002] and the uncertainty of the NACP estimate. The 20% offset between 14C PP and NACP estimates observed in the upper euphotic zone during summer is also consistent with a seasonally variable DO14C production (lower in winter and higher in summer, as discussed above). Arguably, the error in the estimation of NACP makes these conclusions tentative. However, the similar findings by Bender et al.  and Marra  lend some weight. If 14C PP underestimates NACP during periods of high productivity, as our results suggest, than estimates of annual production based on 14C need to be revised upward. In addition, our comparison of in vitro 18O and 14C PP rates at ALOHA underscore the need to evaluate what each tracer measures in order to apply productivity rate estimates to carbon cycle studies.
5. Results and Discussion: In Situ 17Δ-Based GPP and O2/Ar–Based NCP
5.1. Triple Isotope GPP
 Depth profiles of 17Δdiss were measured in February 2002, August 2002, February 2003, and October 2003 (HOT 135, 140, 145, and 151, respectively) (Figure 3). The 17Δdiss profiles vary with depth and season. These variations can be explained by changes in the relative rates of air-sea O2 gas exchange and PP, the two processes that affect 17Δdiss. In the mixed layer, photosynthesis increases 17Δdiss toward a maximum value of + 249 per meg while gas exchange decreases 17Δdiss toward the air-sea equilibrium value of +16 per meg [Luz and Barkan, 2000]. Measured 17Δdiss values in the mixed layer of 20–40 per meg are much closer to the atmospheric O2 value, indicating that most of the dissolved O2 in the surface ocean is from the atmosphere. Lower 17Δdiss values were measured in the winter and higher values were measured in the summer and early fall. A consistent, slight depth gradient in mixed layer 17Δdiss exists, with values increasing toward the base of the mixed layer, indicating that the influence of gas exchange on 17Δdiss decreases with depth within the surface mixed layer.
 The higher subsurface 17Δdiss values can be explained by the decreasing importance of gas exchange with depth in the water column. Between the base of the mixed layer and the base of the euphotic zone (∼175 m), photosynthesis occurs in the absence of gas exchange, allowing 17Δdiss values to reach much higher values than are found at the surface. Below the base of the mixed layer, 17Δdiss values range from 40 to 150 per meg, with maximum values in the summer and early fall. The peak in 17Δdiss during the summer and early fall is roughly coincident in depth and timing with the subsurface O2 saturation maximum that forms every year at Station ALOHA by a combination of radiative heating and biological O2 production [Craig and Hayward, 1987]. Below the base of the euphotic zone the 17Δdiss values are moderate (60–100 per meg) and less seasonally variable. Since respiration does not affect the 17Δdiss signal (assuming the mass dependence factor is 0.516 in equation (4)), these sub-euphotic layer 17Δdiss values reflect the 17Δdiss values that this water mass had when it was at its surface outcrop location plus any photosynthetically induced changes that accumulated in the euphotic zone as it moved along on its isopycnal path to ALOHA. Variability in 17Δdiss at these sub-euphotic layer depths therefore reflect differences in the 17Δdiss values of surface source waters and time spent in the euphotic zone while in transit to ALOHA.
 GPP, calculated using the method of Luz and Barkan  (see equation (6)), are given along with gas exchange coefficients (piston velocities), average mixed layer 17Δdiss, and Co for each cruise in Table 2. Mixed layer GPP estimates ranged from 70 to 185 mmol O2 m−2d−1, corresponding to carbon production rates of 670 to 1780 mg C m−2d−1 or 560 to 1470 mg C m−2d−1, depending on whether or not gross O2 production is corrected for Mehler reaction and photorespiration as described in section 4.3. The high GPP value observed for HOT 151 (185 mmol O2 m−2d−1) occurred during a period of anomalously high winds, despite a typical 17Δdiss signal. Wind speeds 15 days prior to these observations averaged ∼10 m s−1 at a height of 5 m above the sea surface versus a range of 6 to 8 m s−1 observed during the other three cruises.
Table 2. In Situ Productivity at Station ALOHA Determined from 17Δdiss and O2/Ar
Measured O2/Ar gas ratios divided by calculated solubility equilibrium.
Average lifetime of O2 in the mixed layer.
GPP, calculated from gas exchange rate (K), equilibrium O2 concentration (Co), and average 17Δdiss in the mixed layer using the method of Luz and Barkan  (equation (6)). Uncertainty estimates come from a Monte Carlo analysis and include uncertainties in gas exchange and measurement of 17Δdiss.
GPP calculated as above, but in carbon equivalents determined using a PQ of 1.25.
Same as footnote above, but with a 15% correction for photorespiration and Mehler activity.
NCP/GPP ratios, determined from 17Δdiss and O2/Ar measurements. Uncertainty is from variance within mixed layer measurements for each cruise.
NCP, as determined from (O2/Ar)meas/sat. Uncertainty includes both variability in mixed layer (O2/Ar)meas/sat and uncertainty in gas transfer rate.
HOT 135/Feb 2002
0.991 ± 0.010
100 ± 50
960 ± 480
−0.13 ± 0.10
−120 ± 150
HOT 140/Oct 2002
1.011 ± 0.001
70 ± 30
670 ± 290
+0.13 ± 0.07
90 ± 30
HOT 145/Feb 2003
1.001 ± 0.002
90 ± 40
860 ± 380
0.00 ± 0.05
0 ± 30
HOT 151/Aug 2003
1.010 ± 0.001
185 ± 70
1780 ± 670
+0.10 ± 0.03
180 ± 50
 On average, in situ mixed layer GPP values (Table 2) were much higher than in vitro mixed layer 18O GPP (Table 1). Several factors could cause in situ productivity to be higher than in vitro productivity. Harris and Piccinin  estimated that in vitro productivity could be as low as 20–80% of in situ productivity, since naturally circulating phytoplankton communities are not exposed to continuous high light as are in vitro communities. They argued that holding incubation bottles at stationary positions in the water column could induce light stress, and hence, photoinhibition and photorespiration. A second possible reason for the disagreement between the two methods is the difference in the time and space scale over which each method integrates. The in vitro 18O GPP method measured production over the course of 1 day within a small volume. The 17Δ method integrates production for longer time periods (6 to 18 days), depending on wind speed and mixed layer depth. On the basis of data from mooring-based O2 sensors, Karl et al.  suggested that autotrophic production is intermittent at Station ALOHA, occurring in bursts at rates much higher than the background rate. They concluded that the current monthly sampling frequencies of ocean observatories like the HOT program are not sufficient to reflect the time-averaged productivity, since these infrequent episodes of productivity may contribute substantially to the annual metabolic balance of the ecosystem. The 17Δdiss in situ method should yield higher GPP rates than in vitro GPP, since the longer time period of integration (days to weeks) will capture more of the episodic productivity events in this system.
 The 17Δ method does have complications. First, GPP is calculated from a simple model of the surface ocean that neglects effects of advection and vertical mixing. In the oligotrophic ocean horizontal gradients of most tracer properties are typically not large, and therefore advection effects are likely to be small. More specifically for O2, Emerson et al.  reported that horizontal O2 advection was less than one tenth the value of other terms in their mixed layer O2 budget. Vertical mixing can affect the mixed layer 17Δdiss by the injection of relatively high 17Δ O2 from subsurface waters into the mixed layer. However, this effect is also likely to be small. Using O2 concentration and 17Δdiss profiles from HOT 151 in August 2003, when gradients of O2 and 17Δdiss below the mixed layer are at a maximum, and an eddy diffusivity of 1.0 cm2s−1, the upward O2 flux is only ∼3% of the gross O2 production rate. The effect this upward flux of O2 would have on 17Δdiss is small, and any correction to GPP is therefore negligible. Two other complications of the 17Δ method do contribute significantly to the error in calculated GPP. These are the uncertainties in the measurement of 17Δdiss, and the estimated air-sea gas exchange rate, which is approximated from empirical relationships based on wind speed [e.g., Liss and Merlivat, 1986; Wanninkhof, 1992; Nightingale et al., 2000].
 We evaluated the sensitivity of our calculated GPP to uncertainties in gas exchange rate and measurement of 17Δdiss using a Monte Carlo approach, in which the input values of equation (6) were chosen randomly from a data field distributed in a Gaussian way about the mean, with standard deviations equal to the uncertainty estimates [Quay et al., 1993; Emerson et al., 1997]. Mean values for gas transfer rate and 17Δdiss were 6 m d−1 and 35 per meg, i.e., average values for our data set. Uncertainties were ±30% for the gas exchange rate and ±5 per meg for 17Δdiss. On the basis of 1000 randomly chosen budget scenarios, the uncertainty in GPP due to uncertainties in gas exchange rate alone, 17Δdiss alone, and both gas exchange and 17Δdiss were 30%, 28%, and 41%, respectively. The uncertainty in calculated GPP due to 17Δdiss measurement error gets progressively lower as the magnitude of 17Δdiss increases, i.e., as the difference between 17Δdiss and 17Δeq increases in the numerator in equation (6). For example, increasing 17Δdiss from 35 to 60 per meg results in an overall uncertainty of 32% and an uncertainty due to measurement of 17Δdiss alone of 14%. This indicates the method is more robust in high productivity systems. GPP uncertainty calculated for each cruise by the Monte Carlo approach is reported in Table 2.
 A final concern is the value of θ in equation (4). The 17Δ method relies on the fact that the 17Δdiss used in equation (6) reflects only air-sea gas transfer and PP, and that respiration does not affect the 17Δ signal. This is only true if the slope used to calculate 17Δ reflects the 17O/18O discrimination of community respiration. Since different respiratory pathways fractionate 17O and 18O slightly differently, values of θ can range from 0.506 (photorespiration) to 0.516 (mitochondrial respiration) [Angert et al., 2003]. We examined the sensitivity of our results to the respiration slope in equation (4) by varying θ between the range of known values, 0.506 and 0.516. Lowering the slope to 0.506 had virtually no affect on the 17Δdiss of mixed layer values, increasing them only slightly (1–2 per meg). Since photorespiration is specific to autotrophs, the contribution of this respiratory pathway is expected to be a minor component of total (heterotrophic + autotrophic) community respiration in the open ocean, and the actual value of θ is expected to be within a much narrower range, 0.514 (the θ for the alternative oxidation pathway) to 0.516 (mitochondrial respiration). The error in GPP due to variations in θ is therefore expected to be less than 3%.
5.2. In Situ Net Community Productivity
17Δdiss is plotted against the “biological O2 saturation”, (i.e., (O2/Ar)meas/(O2/Ar)eq, discussed in section 3.3) for each cruise (Figure 4), and average mixed layer NCP/GPP ratios are reported for each cruise in Table 2. It is important to note that the NCP/GPP ratio is better constrained than GPP, since it is not dependent on the gas exchange rate [Hendricks et al., 2004],
where is the biological O2 saturation, i.e., (O2/Ar)meas/(O2/Ar)eq. Uncertainty in the NCP/GPP ratio therefore depends on the error in the measurement of (O2/Ar)meas/(O2/Ar)eq and 17Δdiss. As stated previously, the estimated uncertainty in measurement of (O2/Ar)meas/(O2/Ar)eq, based on precision of replicate samples, is 0.3%. Error in measurement of 17Δdiss is 5 per meg. Of the two variables, the error in 17Δdiss has the most effect on the calculated NCP/GPP ratio. For example, varying 17Δdiss by ±5 around its mean value, 35 per meg, results in a 17% uncertainty in the calculated ratio. As the value of 17Δdiss increases the uncertainty decreases (e.g., varying 17Δdiss from 40 to 50 per meg results in a smaller uncertainty, 12%). Varying (O2/Ar)meas/(O2/Ar)eq by 0.3% around its mean value, 1.002, changes the ratio by less than 1%.
 In addition to the measurement error, uncertainty in the NCP/GPP ratio is introduced by the simplification of the mixed layer 17Δdiss and O2/Ar budgets (e.g., lack of a mixing term). We have already discussed how this simplification has little influence on the 17Δdiss and GPP calculation (section 5.1). Similarly, Hamme  showed that an order of magnitude change in the mixing rate had no effect on mixed layer integrated NCP calculated from an O2/Ar budget at station ALOHA.
 In general, winter data indicate either a neutral (HOT 145) or net heterotrophic state (HOT 135), with NCP/GPP ratios of 0.0 and −0.13, respectively. Summer and early fall data indicate a net autotrophic state, with average NCP/GPP ratios of 0.13 and 0.10 for HOT 140 and HOT 151, respectively. Assuming a biological steady state, the NCP/GPP ratio is comparable to an f-ratio, and NCP is comparable to carbon export. Summer NCP/GPP ratios agree surprisingly well with the canonical f-ratio of 0.10 reported by Eppley and Peterson  for the oligotrophic ocean and with f-ratio estimates specific to Station ALOHA of 0.05–0.10 [Karl, 1999].
 Given our small data set, it is difficult to extend our measurements to the scale of an annual budget. However, it is possible to determine if our results are roughly consistent with previously reported annual carbon export rates at ALOHA. If we use the mean GPP and NCP/GPP values estimated for October 2002 and August 2003, when NCP/GPP > 0, an average carbon export of 2.3 mol C m−2 yr−1 is determined. This estimate of NCP at ALOHA compares favorably with carbon export rates of 2.4 to 2.7 determined from annual budgets of O2, DIC, and 234Th deficiency [Emerson et al., 1997; Quay and Stutsman, 2003; Benitez-Nelson et al., 2001]. Higher resolution observations of both NCP and GPP over a complete annual cycle are needed for a more rigorous determination of annual carbon export based on 17Δdiss and O2/Ar.
 The ability to estimate GPP and NCP in situ over a period of days to weeks is one of the fundamental strengths of the 17Δ and O2/Ar methods. Comparisons of in vitro and in situ data have classically disagreed in the literature, with productivity inferred indirectly from in situ data being, in most cases, much greater than in vitro data [e.g., Schulenbergerand Reid, 1981; Jenkins and Goldman, 1985]. It has long been recognized that extrapolation of in vitro rates to annual ecosystem budgets may not be appropriate, and is probably a major cause of the discrepancy between in vitro and in situ data [Platt et al., 1989]. The recent work of Karl et al.  further indicates that extrapolation of in vitro rates to time periods of days to weeks may be inappropriate for open ocean systems. Williamset al.  argue that infrequent sampling typical of in vitro methods can incorrectly skew the ecosystem view for oligotrophic oceans toward net heterotrophy. They suggest this is the reason for the observation of heterotrophy in their study at Station ALOHA and in other open ocean studies [e.g., delGiorgio and Duarte, 2002]. The high rates of mixed layer-integrated GPP and the positive NCP/GPP ratios observed in situ at Station ALOHA by the 17Δ and O2/Ar method support the argument that intermittency in primary production contributes substantially to time-averaged production and to the long-term autotrophic/heterotrophic balance.
6. Summary and Conclusions
 Our measurement of organic carbon production in vitro with the 18O GPP method and in situ with the 17Δ and O2/Ar method at the HOT Station ALOHA study site indicates the following.
 1. In vitro GPP, as measured by the gross evolution of O2 with the 18O label, is greater than the carbon production rate measured by 14C labeling, regardless of whether or not measured O2 evolution is corrected for photorespiration and Mehler activity. On average, GPP measured with 18O is 575 mg C m−2d−1 in the winter and 930 mg C m−2d−1 in the summer and early fall.
 2. In situ GPP, as measured by the triple isotope (17Δ) method, is greater than carbon production measured by 18O in vitro, averaging 910 mg C m−2d−1 in the winter and 1225 mg C m−2d−1 in the summer/fall, although with substantial uncertainty (±40%). The in situ method is expected to yield PP rates that are equal to or greater than PP determined from in vitro methods, since the former method integrates episodic PP events and does not suffer from bottle effects.
 3. NCP/GPP ratios, determined using O2/Ar gas ratios and 17Δdiss, are around 0.1 in the summer, close to reported f-ratios for the open ocean [e.g., Eppley and Peterson, 1979; Karl, 1999]. This indicates Station ALOHA is a net autotrophic system during summer months.
 4. Estimates of NACP calculated from in vitro 18O PP rates, although uncertain, agree with 14C PP from the winter and from below the mixed layer in summer. Summer mixed layer data were approximately 20% higher. The seasonal difference might be explained by seasonal differences in DOC production rates.
 Our observations demonstrate the importance of evaluating carbon cycling in the surface ocean with many geochemical tools, even at locales that are relatively well characterized. They also illustrate the need to understand carbon cycling in the oligotrophic ocean on a range of space scales and timescales. Continued application of many geochemical tools, including the 14C and 18O in vitro methods and the 17Δ and O2/Ar in situ methods, and the development of new tracers adept at characterizing fundamental biological processes, will improve our understanding of mechanisms responsible for carbon export to the ocean interior and our predictions of the response of the ocean biological pump to future climate change.
 We thank D. Wilbur and C. Stump for their technical expertise. We also thank the captain and crew of the R/V Ka'imikai-O-Kanaloa, and Dave Karl and the HOT program staff for their invaluable assistance in all field aspects of this study. We gratefully acknowledge the HOT biogeochemistry group for their collection of core data sets, in particular, the 14C PP. P. Morris, P. J. Williams, and D. Karl offered helpful comments throughout the course of this study. This research was supported by a National Defense Science Engineering Graduate Fellowship (L. W. J.) and by National Science Foundation grant OCE 0095534 (P. D. Q.).