Measuring primary production rates in the ocean: Enigmatic results between incubation and non-incubation methods at Station ALOHA



[1] Primary production (PP) rates were estimated using concurrent 14C and 18O bottle incubations and a non-incubation oxygen isotope (17Δ) based method during monthly cruises to the time series station ALOHA in the subtropical N. Pacific Ocean between March, 2006 and February, 2008. The mean gross oxygen production (GOP) rate in the photic layer (0–200m) at ALOHA was estimated at 103 ± 43 and 78 ± 17 mmol O2 m−2 d−1 from the 17Δ and 18O methods, respectively. In comparison, the mean 14C-PP rate (daytime incubations) in the photic layer was 42 ± 7 mmol C m−2 d−1 (502 ± 84 mg C m−2 d−1). Seasonal and depth variability (% change) for GOP rate was 2–3 times that for 14C-PP. The non-incubation 17Δ-GOP rates consistently exceeded the incubation 18O-GOP rates by 25–60%, and possible methodological biases were evaluated. A supersaturation of the dissolved O2/Ar gas ratio was measured every month yielding a mean annual value of 101.3 ± 0.1% and indicating a consistent net autotrophic condition in the mixed layer at ALOHA. The mean annual net community production (NCP) rate at ALOHA estimated from dissolved O2/Ar gas ratio was 14 ± 4 mmol O2 m−2 d−1 (120 ± 33 mg C m−2 d−1 or 3.7 ± 1.0 mol C m−2 yr−1) for the mixed layer. A NCP/GOP ratio of 0.19 ± 0.08 determined from 17Δ and O2/Ar measurements indicated that ∼20% of gross photosynthetic production was available for export and harvest.

1. Introduction

[2] Photosynthetic production of organic carbon is the foundation of ocean's food web and biological carbon pump, leading ultimately to fish production, deep sea carbon sequestration and sustained oxygen levels in the atmosphere. Yet accurately measured rates of marine primary production (PP) remain elusive since there is no absolute PP standard against which methodological accuracy can be tested. Historically, the benchmark aquatic PP method has been 14C bottle incubations [Steemann Nielsen, 1952], which have yielded thousands of ocean PP estimates over the last 50 years. Thus our understanding of PP in the ocean is dominated by the 14C incubation method (14C-PP) and yet the method has had acknowledged flaws that persist today [e.g., Peterson, 1980]. The influence of 14C-PP extends to satellite based estimates of PP since most the satellite PP algorithms rely on 14C-PP measurements to estimate the response of PP to light.

[3] One source of ambiguity of the 14C-PP method is whether this incubation method is a measure of gross primary production (GPP), net primary production (NPP) or some value in between [Marra, 2002, 2009]. In addition to the absence of a reliable physiological model, 14C-PP measurements are influenced to some unknown degree by sampling, manipulation and containment effects that impact all incubation methods. The accuracy of the 14C-PP method is obviously very important because many remineralization-intensive marine ecosystems (e.g., tropical and subtropical habitats) are poised near the balance point between GPP and total community respiration (R). The incubation PP method based on 18O2 production from 18O-labeled H2O [e.g., Bender et al., 1987] provides a direct estimate of gross oxygen production (GOP). Concurrent 18O- and 14C- incubation PP measurement have been used to improve the interpretation of 14C-PP measurements [e.g., Bender et al., 1999; Laws et al., 2000a; Marra, 2002]. However, the 18O incubation PP method (18O-GOP) is still susceptible to sampling, manipulation and containment effects.

[4] Over the past decade two other PP methods, based on Fast Repetition Rate Fluorometry (FRRF) and the isotopic composition of dissolved oxygen, have begun to yield non-incubation based estimates of gross primary production (GPP) rates in the ocean [e.g., Kolber et al., 1998; Luz and Barkan, 2000]. These two PP methods have the advantage of avoiding potential methodological flaws inherent to incubation techniques and are capable of yielding GOP estimates over large regions of the ocean via underway sampling. The opportunity to obtain a new look at PP in the ocean using multiple approaches, rather than the reliance on a single method, is bound to provide new insights into PP variability in the ocean and the ecological and biogeochemical implications of this variability.

[5] From the perspective of the ocean's carbon cycle, the most important biological production rate is net community production (NCP), which equals the difference between GPP and R. NCP, at steady state, represents the organic carbon in both dissolved and particulate phase that is available for export or harvest. NCP has been estimated by several methods (e.g., dissolved oxygen and inorganic carbon (DIC) budgets, thorium-234 budgets, sediment traps, bottle O2 incubations, etc.); however, each approach has substantial uncertainty. Unfortunately, there are only a few sites where multiple NCP methods have been compared (e.g., JGOFS study sites, BATS and ALOHA time series sites). Thus, for most of the ocean NCP is poorly constrained. The lack of accurate NCP estimates may be responsible for reports of net ecosystem heterotrophy in the oligotrophic ocean surface layer based on bottle O2 incubations [e.g., Williams et al., 2004] that are in contradiction to reports of net ecosystem autotrophy based on mixed layer O2 budgets at the same location [e.g., Emerson et al., 1997].

[6] We present the results of two years of monthly GOP and NCP rate estimates based on non-incubation methods using the measured isotopic composition of dissolved O2 and the ratio of dissolved O2/Ar gases, respectively, and compare these estimates to concurrent 18O and 14C incubation based estimates of PP. The study site is the Hawaii Ocean Time series station ALOHA in the subtropical N. Pacific Ocean where PP has been measured by the 14C incubation method for 20 years and NCP has been estimated previously by several methods. Since the non-incubation oxygen isotope GOP method has been applied at only a few sites to date, the results discussed here provide an excellent opportunity to compare PP rates measured by non-incubation and incubation methods at a well-studied site.

[7] We conclude that the oxygen isotope method yields a mean GOP rate of 103 ± 43 mmol O2 m−2 d−1 in the photic layer that was ∼25–60% greater than the GOP rate measured using 18O-bottle incubations and 2–3 times the daytime 14C-PP rates at ALOHA. We measured a supersaturation in dissolved O2/Ar gas ratio every month that yielded a mean annual NCP rate of 14 ± 4 mmol O2 m−2 d−1 (120 ± 33 mg C m−2 d−1 or 3.7 ± 1.0 mol C m−2 yr−1) for the mixed layer that is ∼20% of the GOP and ∼35% of daytime 14C-PP. Compared to the historic 14C-PP data set, the much higher non-incubation based estimates of GPP and NCP at ALOHA may require a re-evaluation of carbon and energy flow in this habitat.

2. Background

2.1. Study Site

[8] The study site is the time series station ALOHA (23°N 158°W) in the subtropical N. Pacific Ocean. Certainly a major advantage of working at ALOHA is the wealth of historic data available that provides a context for evaluating data collected in any given year. There have been monthly measurements of PP at ALOHA using the 14C bottle incubations (in situ dawn to dusk incubations) since 1988 for a total of almost 200 individual depth integrated 14C-PP estimates. The mean daytime 14C-PP is 514 ± 135 mg C m−2 d−1 between 1988 and 2008 with a slight seasonality, i.e., in the typical year there is ∼16% difference between 14C-PP measured in winter (Dec–Mar) and summer (Jun–Aug).

[9] Primary production rates at ALOHA have been simulated using a PP model based on measured chlorophyll concentrations, PAR and the empirically derived photosynthesis versus light relationship or quantum yield [Ondrusek et al., 2001]. Although the PP model yielded an annual mean PP rate (460 ± 33 mg C m−2 d−1) within ∼10% of the mean 14C-PP at ALOHA, it significantly underestimated the variability in PP, which may have been due to a species dependence of maximum quantum yield [Ondrusek et al., 2001]. Ondrusek et al. found that satellite based PP at ALOHA underestimated measured 14C-PP by ∼60% and was mainly due to underestimates of chlorophyll and assumed maximum quantum yield.

[10] The importance of stochastic productivity events at ALOHA is somewhat unclear. Karl et al. [2003] made the illustrative case that episodes of NCP occurring at rates 3x the mean rate and 10% of the time are sufficient to yield a net annual autotrophic condition despite a background heterotrophic condition. The lesson being that frequent NCP measurements would be required to accurately estimate the mean annual trophic state. Letelier et al. [2000] detected an eddy at ALOHA that had 3x background chlorophyll concentrations and a nitracline uplifted by ∼100 m, which demonstrated the potential impact of mesoscale features on the ecosystem structure at ALOHA. Emerson et al. [2002] used a continuous record of surface dissolved O2 gas measurements from a moored sensor at ALOHA to demonstrate significant mesoscale variability of O2 saturation levels and that would, at steady state, correspond to similar variability in NCP. On the other hand, the historic 14C-PP data from ALOHA do not detect large PP events. Only two out of 172 times between 1988 and 2006 was the measured 14C-PP greater than twice the long-term mean and never was it ≥ 3x the mean. Thus, either PP events are relatively unimportant at ALOHA or the 14C-PP method doesn't detect them. It is important to point out, however, that episodic events (eddies, storms, fronts, etc.) potentially have a much bigger impact on the rate of NCP than PP. At ALOHA, NCP estimates (∼90 mg C m−2 d−1, see below) represent ∼20% of the average 14C-PP (∼500 mg C m−2 d−1), with most of the PP being supported by recycled nutrients, and thus episodic events that doubled annual NCP would increase 14C-PP by only ∼20%.

[11] Rates of NCP at ALOHA have been estimated using several approaches including budgets of O2, dissolved inorganic carbon (DIC) and 234Th and sediment traps (Table 1). The NCP estimates range from 30 ± 8 mg C m−2 d−1 to 135 ± 62 mg C m−2 d−1 and yield a main rate (excluding sediment traps) of 90 ± 28 mg C m−2 d−1 (∼2.7 ± 0.9 mol C m−2 yr−1).

Table 1. Estimates of Net Community Production Rate (mg C m−2 d−1) From the Photic Layer at ALOHA
MethodRateTime IntervalReference
Sediment traps30 ± 81989–1997Karl [1999]
O2 + Ar budgets89 ± 561992–1995Emerson et al. [1997]
 46 ± 212000–2001Hamme and Emerson [2006]
DIC + DIC13 budgets90 ± 451994–1999Quay and Stutsman [2003]
 92 ± 321988–2002Keeling et al. [2004]
234Th + OC budgets88 ± 311999–2000Benitez-Nelson et al. [2001]
Mooring O2135 ± 622005Emerson et al. [2008]

2.2. Primary Production Rates

2.2.1. Bottle Incubation Methods

[12] 14C bottle incubations were performed from dawn to dusk (daytime) in triplicate at six depths (5, 25, 45, 75, 100, and 125m) to estimate daytime PP during each HOT cruise using procedures described by Karl and Lukas [1996]. Marra [2002] found that the net O2 production rate measured by bottle incubations during JGOFS was similar to the concurrent daily (24 h incubation) 14C-PP rate assuming a PQ of 1.4 and, thus, concluded that daily 14C-PP approximated net autotrophic PP under most conditions. Letelier et al. [1996] estimated that 90% of the 14C-PP between the surface and 200m occurred within the top 100m at ALOHA. Based on a comparison of concurrent daytime and 24 h (daily) 14C incubation PP measurements at ALOHA, Karl et al. [1996] observed that ∼15% of the organic matter fixed during the daytime was respired at night as compared to previous estimates of 20–25% observed during JGOFS [Laws et al., 2000a; Marra and Barber, 2004]. Karl et al. [1998] observed that unmeasured DOC14 production during the dawn to dusk 14C incubations at ALOHA could yield a 30–50% underestimate of PP.

[13] Rates of GOP were measured each month in triplicate at five depths (5, 25, 45, 75 and 100m) using bottle incubations (24 h) of 18O labeled water [Bender et al., 1987]. Specifically, we used acid cleaned quartz bottles and 18O enriched water that was triple distilled to eliminate nutrient and trace metal contamination following the incubation techniques described by Juranek and Quay [2005]. The 18O method clearly measures GOP and since there is insignificant recycling of the 18O labeled O2 produced during the incubation, the measured GOP rate does not depend on incubation duration or diel variations in R [Bender et al., 1999; Laws et al., 2000a]. However, the conversion from GOP to gross carbon production is not straightforward. Recent studies demonstrate an uncoupling of O2 production from carbon fixation during photosynthesis via, for example, the Mehler reaction, chlororespiration and photorespiration, which can vary depending on light level and nutrient availability [e.g., Zehr and Kudela, 2009; Suggett et al., 2009].

[14] Previous comparisons of 18O-GOP versus daily 14C-PP indicated that GOP was 2–3 times the 14C-PP rates (mol O2/mol C) based on measurements in the equatorial Pacific and Arabian Sea during JGOFS [Bender et al., 1999; Laws et al., 2000a]. Marra [2002] compiled the 18O-bottle and 14C-bottle measurements during JGOFS and found an 18O-GOP/14C-PP of ∼2.7 (mol O2/mol C) for daily (24 h) incubations and ∼2.0 for daytime (dawn to dusk) incubations. Juranek and Quay [2005] measured a mean 18O-GOP/14C-PP (daytime) of 1.7 ± 0.4 during four cruises to station ALOHA in 2002–2003. Recently, however, Robinson et al. [2009] reported a significantly higher 18O-GOP/14C-PP of 4.5 ± 1.2 based on concurrent 24 h on-deck incubations in waters from the Celtic Sea where the 14C-PP rate ranged from ∼200 to 5000 mg C m−2 d−1.

[15] A difficult question to answer is whether any incubation (or non-incubation) based measurement of PP is accurate. Incubating seawater enclosed in bottles or bags introduces artifacts potentially affecting the measured PP rate (e.g., changes in the quality and quantity of irradiance, the phytoplankton, zooplankton and microbial communities, nutrient supply, turbulence, etc.). There is also the issue of extrapolating incubation-based PP measurements to longer time scales. Daily variations in cloudiness, water clarity, mixed layer depth and spatial patchiness in PP yield uncertainty in spatial or temporal extrapolations of 12- or 24-h incubation-based PP measurements. If infrequent episodes or patches of high PP are superimposed on a background of lower PP conditions, then sporadic bottle measurements of PP will underestimate the true PP. In short, the question of how the PP rate measured by traditional incubation methods compares to the in situ PP rate has remained unanswered because of the lack of an absolute PP standard and alternative non-incubation PP methods.

2.2.2. Oxygen Isotope Method

[16] The rate of aquatic PP has been estimated using the natural isotopic composition of dissolved O2 [Luz and Barkan, 2000]. This method relies on an observed anomalously low 17O/16O for O2 in the atmosphere caused by a mass independent fractionation during reactions between ozone, O2, and CO2 in the stratosphere [Thiemens et al., 1995]. The basis of the oxygen isotope PP method is that air-sea O2 gas exchange drives the 17O/16O and 18O/16O of dissolved O2 in the surface ocean toward that of atmospheric O2 whereas photosynthesis produces O2 with a mass dependent (typical) distribution of O isotopes that has a higher 17O/16O relative to O2 in air. As a result, the measurement of the 17O/16O and 18O/16O of dissolved O2 in the surface ocean yields a direct estimate of the fractions of O2 from air and photosynthesis. Therefore, once the rate of air-sea O2 gas exchange is estimated, one can calculate the rate of GOP from the measured 17O/16O and 18O/16O of O2 dissolved in seawater.

[17] Because the 17O/16O difference (relative to 18O/16O) between O2 in air and O2 produced during photosynthesis is small an isotopic notation (17Δ) expressed in parts per million (per meg), rather than in parts per thousand (per mil) traditionally used in isotopic notation, was adopted [Luz and Barkan, 2000; Angert et al., 2003]. The 17Δ represents the 17O/16O anomaly relative to 18O/16O and is expressed as:

equation image

where δ18O (‰) = [(18O/16O)s/ (18O/16O)std − 1]•1000, s = sample and std = standard (and similarly for δ17O). The expression for 17Δ implies that a process that produces or consumes oxygen with a 17O/16O to 18O/16O reaction rate ratio of ∼0.518 will not change 17Δ. This is important since respiration in plankton and bacteria measured in lab experiments yields a mean 17O/16O to 18O/16O reaction rate ratio of ∼0.518 [Luz and Barkan, 2005] and thus has an insignificant effect on the 17Δ of dissolved O2.

[18] Luz and Barkan [2000] chose O2 in air as the standard for 17Δ and thus, by definition, atmospheric O2 has a 17Δ = 0 per meg. The 17Δ of dissolved O2 in seawater equilibrated with air is 17 ± 4 per meg at ∼25°C [Sarma et al., 2006; Luz and Barkan, 2009]. Based on marine plankton culture studies, the O2 produced during photosynthesis has a 17Δphoto = 249 ± 15 per meg [Luz and Barkan, 2000]. Thus 17Δ of the dissolved O2 in the warm surface ocean ranges between extremes of ∼17 and ∼249 per meg, depending on the relative rates of air-sea O2 exchange and gross photosynthesis.

[19] If the mixed layer budget for O2 and O2 isotopes approaches steady state (i.e., the time rate of change of O2 concentration is small compared to the O2 source and sink rates), then the sources of O2 resulting from air to sea O2 gas invasion and GOP equal the O2 losses resulting from sea to air gas evasion and community respiration (assuming mixing is negligible). Luz and Barkan [2000] showed that the steady state rate of GOP can be expressed in terms of only one field measurement, i.e., the 17Δ of dissolved O2 (17Δdiss), and estimates of the O2 concentration in equilibrium with air (O2eq) and the air-sea O2 gas transfer rate (kas).

equation image

[20] Notably, respiration is not a term in equation (2) because it does not affect 17Δ, as discussed above. Here GOP represents the average rate integrated over the depth of the mixed layer. In practice, GOP rates were calculated from 17Δ measurements using the more accurate version of equation (2) from Hendricks et al. [2004], which yielded GOP rates that were ∼10% higher than rates calculated from equation (2).

[21] Although GOP estimated from equation (2) ignores the effects of mixing and advection, this assumption is often reasonable because the net advective and diffusive O2 fluxes are small compared to the gross O2 flux represented by GOP. However, during periods of mixed layer entrainment (fall and winter), strong upwelling or vertical mixing, GOP can be significantly overestimated by equation (2), as discussed below.

[22] Estimating GOP using the oxygen isotope method (referred to as 17Δ-GOP) has the advantages of eliminating incubation effects and integrating GOP rates over the residence time of O2 in mixed layer (typically 1–3 weeks) and thus potentially capturing PP events likely missed by 12- or 24-h incubations. Furthermore, since it does not require an incubation, the 17Δ-GOP method can be applied anywhere a surface seawater sample can be collected. The disadvantages of the 17Δ-GOP method are the significant uncertainty in the method (∼±40%), the extrapolation from mixed layer to total photic layer, possible biases in the GOP estimates due to mixed layer entrainment, upwelling and vertical mixing and the uncertainty in converting from oxygen production to carbon fixation.

[23] Although the caveats with the 17Δ-GOP method are significant, they are different from the caveats pertaining to 14C-PP incubation method and thus we gain new information about PP in the ocean. However, the lack of an absolute PP standard prevents a determination of accuracy for any PP method. Thus, we must look for consistency (or inconsistency) between PP methods and, in addition, account for the different metrics of PP (e.g., carbon fixation, O2 production, electron transport, fluorescence, etc.) used by individual PP methods.

[24] The isotopic composition of dissolved O2 was measured using the analytical procedures described by Juranek and Quay [2005]. Duplicate samples were collected at 9 depths (5, 25, 45, 75, 100, 125, 150, 200, and 300m) on the HOT cruises using the methods described by Emerson et al. [1999]. The typical precision of the repeated δ18O and δ17O measurements averaged 0.05 and 0.01‰ yielding a standard error in the mean (SEM, where SEM = SD/√n) of ±5 per meg after 75 individual isotope ratio determinations. The mean standard deviation of 17Δ for paired duplicate samples was ±8 per meg. In the mixed layer, where typically 4 to 8 samples were collected each month, the average SEM of the mean 17Δ was ±4 per meg. We estimate the depth of the mixed layer from CTD profiles using a potential density (σθ) change of 0.125 kg m−3 from the surface as the criterion.

[25] The initial application of the 17Δ-GOP method was at the time series station BATS in the subtropical N. Atlantic in 1998–99 by Luz and Barkan [2000]. They measured a 17Δ range of 30 to 47 per meg (mean = 38 ± 8) during six cruises from which they estimated mean 17Δ-GOP rate of 70 ± 35 mmol O2 m−2 d−1 that was 3x the long-term average 14C-PP at BATS. (All uncertainties represent ± 1 standard deviation (SD) unless stated otherwise.) A more detailed presentation of these results by Luz and Barkan [2009] indicated that the 17Δ-GOP rates varied from 29 to 107 mmol O2 m−2 d−1 in the mixed layer between May and October 2000 and were 3.5–8x the concurrently measured 14C-PP rate. Juranek and Quay [2005] measured a 17Δ range of 31–39 per meg in the mixed layer at ALOHA during four cruises in 2002–2003 that yielded a 17Δ-GOP range of 70–185 mmol O2 m−2 d−1 that was 2–3x daytime 14C-PP rates and 1.8 ± 0.8x 18O-GOP rates concurrently measured during these cruises. In the eastern equatorial Pacific (95°W–110°W), Hendricks et al. [2005] measured a 17Δ range of 40 to 100 per meg in surface waters that yielded a mean GOP rate of 102 ± 65 mmol O2 m−2 d−1 for stations poleward of 2°N and 5°S. In the Southern Ocean (40°S to 70°S), Reuer et al. [2007] measured a 17Δ range of ∼10 to 50 per meg in surface waters that yielded GOP range of ∼50 to 500 mmol O2 m−2 d−1 (mean = 166 ± 121 mmol O2 m−2 d−1). In Sagami Bay on the coast of Japan, Sarma et al. [2005] measured a 17Δ range of ∼50 to 110 per meg in surface waters that yielded GOP rates of ∼90 to 350 mmol O2 m−2 d−1.

2.3. NCP Rates Estimated From O2/Ar

[26] The net imbalance between GOP and R can be estimated from a mixed layer budget for dissolved oxygen. In its simplest form, the budget assumes that the net rate of O2 gas evasion to the atmosphere is balanced by net biological O2 production (i.e., GOP - R). Emerson et al. [1997] utilized the similar temperature dependence of gas solubility in seawater for Argon (Ar) and O2 to determine the biological component of the O2 saturation. In this way, the O2/Ar saturation state, i.e., where (O2/Ar)sat equals the O2/Ar measured divided by the O2/Ar expected in equilibrium with air, yielded the portion of the net sea to air O2 flux that was balanced by net biological production of O2 (or NCP). Thus NCP in the mixed layer is determined from measuring the ratio of dissolved O2 and Ar gases, calculating (O2/Ar)sat and estimating kas, as follows:

equation image

[27] This approach has been applied at ALOHA previously and yielded NCP rates of 10.4 ± 6.6 mmol O2 m−2 d−1 in the 1990s [Emerson et al., 1997], 3–5 mmol O2 m−2 d−1 in 2000–02 [Hamme and Emerson, 2006] and 9–17 mmol O2 m−2 d−1 during summer 2002–2003 [Juranek and Quay, 2005]. At BATS, Luz and Barkan [2009] used O2/Ar measurements to estimate NCP rates of 6–8 mmols O2 m−2 d−1 for the mixed layer. In the eastern equatorial Pacific, Hendricks et al. [2005] used surface O2/Ar measurements to estimate a mean NCP rate of 10 ± 9 mmol O2 m−2 d−1. In the Southern Ocean (45°–65°S), Reuer et al. [2007] used surface O2/Ar measurements to estimate NCP rates of 22–50 mmol O2 m−2 d−1. Kaiser et al. [2005] used continuous underway O2/Ar measurements during a cruise in the eastern equatorial Pacific to estimate NCP rates of ∼0–18 mmol O2 m−2 d−1. For all the above examples, kas was estimated from observed wind speeds and an empirical relationship between kas and wind speed [e.g., Wanninkhof, 1992; Nightingale et al., 2000].

[28] The ratio NCP/GOP in the mixed layer can be estimated by simultaneous 17Δ and O2/Ar measurements (combining equations (2) and (3)) as follows:

equation image

Note that NCP/GOP is independent of kas, which means that the calculated NCP/GOP has substantially less uncertainty than either GOP or NCP (except when (O2/Ar)sat approaches 1) and that the numerator and denominator are measured in the same units (O2 production). Previous NCP/GOP estimates based on 17Δ and O2/Ar measurements averaged 0.13 ± 0.05 during summer cruises at ALOHA [Juranek and Quay, 2005], 0.06 ± 0.05 in the eastern equatorial Pacific [Hendricks et al., 2005], 0.13 ± 0.06 for the Southern Ocean [Reuer et al., 2007] and 0.13 ± 0.05 at BATS [Luz and Barkan, 2009].

3. Results

3.1. Incubation-Based Estimates of PP

[29] Depth profiles of 14C-PP and 18O-GOP rates decreased with depth, as expected, through the photic layer (Figure 1). However, 18O-GOP decreased twice as sharply as 14C-PP with depth, i.e., the 18O-GOP at 100m equaled 15 ± 2% (±SEM) of the rate at 5m whereas the 14C-PP at 100m equaled 31 ± 4% (±SEM) of the rate at 5m. The ratio of 18O-GOP to 14C-PP (mol O2/mol C) decreased with depth from a mean (±SEM) of 2.4 ± 0.3 at 5 m, 1.1 ± 0.05 at 100 m (Figure 2). During the summer (May–Sep), with a mean mixed layer depth (Zml) of 45m, the depth-integrated mixed layer 18O-GOP was 64% of the 18O-GPP integrated to 100 m, whereas in winter (Nov–Mar) when the mean Zml was 75 m, the integrated 18O-GPP in the mixed layer was 90% of 18O-GOP to 100m. For 14C-PP, similar ratios of 60% and 86% were determined for summer and winter, respectively.

Figure 1.

Two representative depth profiles of primary production rates at ALOHA measured by the 18O-GOP (mmol O2 m−2 d−1) and 14C-PP (mmol C m−2 d−1) incubation methods in June and December 2006.

Figure 2.

The depth distribution of the 18O-GOP/14C-PP ratio (mol O2/mol C) measured monthly from March 2006 to February 2008 (HOT cruises 179 to 200). The filled circles represent the mean values at each depth and error bars are ±SEM.

[30] The annual mean 18O-GOP and 14C-PP values were 71 ± 16 mmol O2 m−2 d−1 and 38 ± 6 mmol C m−2 d−1 (455 ± 75 mg C m−2 d−1), respectively, integrated over the top 100m. In comparison, the climatological mean 14C-PP at ALOHA (1988–2008) was 514 ± 136 mg C m−2 d−1. Seasonally, 18O-GOP and 14C-PP in summer (May–Sep) exceeded winter (Nov–Mar) rates by 32% and 18%, respectively (Figure 3).

Figure 3.

Monthly rates of GOP estimated from the 17Δ-GOP and 18O-GOP (mmol O2 m−2 d−1) and 14C-PP (mmol C m−2 d−1) methods depth-integrated to 100 m.

3.2. Estimates of GOP Based on 17Δ

[31] In the mixed layer, where 4–8 individual 17Δ values were measured, the monthly 17Δ ranged from 23 to 45 per meg (n = 131) with a typical SEM of ±4 per meg. The annual mean 17Δ of 33 ± 2 per meg (±SEM) implied that 7% of the dissolved O2 in the mixed layer at ALOHA is from photosynthesis and 93% from air. The 17Δ increased with depth reaching a subsurface maximum at 75–100m that peaked (∼140 per meg) during the late summer (Figure 4). This sub-surface 17Δ build up was a result of photosynthesis occurring within the photic layer at depths isolated from air-sea O2 gas exchange. Similar 17Δ depth profiles and mixed layer values have been observed at BATS [Luz and Barkan, 2009].

Figure 4.

Depth profiles of 17Δ (per meg) measured during four HOT cruises in 2007.

[32] The monthly GOP rates in the mixed layer calculated from 17Δ, using the rigorous version of equation (2), ranged from 40 to 242 mmol O2 m−2 d−1, with an annual mean of 100 ± 51 mmol O2 m−2 d−1. The two highest mixed layer 17Δ-GOP estimates of 210 and 241 mmol O2 m−2 d−1 occurred in December 2006 and February 2007 when the mixed layer was deepest at 90–110m (Figure 3). There is substantially more variability in the monthly 17Δ-GOP estimates compared to 18O-GOP and 14C-PP and no significant annual trend (Figure 3).

[33] The 17Δ-GOP values calculated using equation (2) are overestimated during fall and winter, when subsurface water with elevated 17Δ (Figure 4) is entrained into the mixed layer. To avoid this bias in the 17Δ method, GOP rates were calculated from 17Δ using a time dependent depth-integration of 17Δ. This approach calculated the time rate of change of the monthly mean 17Δ integrated to a constant depth, i.e., a depth below the base of the photic layer at which 17Δ varies little during the year (e.g., mean 17Δ = 85 ± 3 (SEM) per meg at 200m) (Figure 4). This estimate of 17Δ-GOPint (integrated to Zint) is expressed as follows:

equation image

In essence, a time rate of change term is added to the numerator of equation (2). This approach avoids an overestimation of GOP caused by entrainment and vertical mixing by accounting for the observed depth-integrated 17Δ increase (oxygen concentration weighted) during the summer and decrease during winter (Figure 5).

Figure 5.

The monthly measurements of the 17Δ (per meg) in the mixed layer and depth-integrated to 200m and depth of mixed layer (m) at ALOHA from Mar 2006 to Feb 2008.

[34] The average 17Δ-GOPint rate (integrated to 200m) was 123 ± 44 mmol O2 m−2 d−1 for summer (May–Sep) and 83 ± 42 mmol O2 m−2 d−1 for winter (Nov–Mar) yielding an annual average rate of 103 ± 43 mmol O2 m−2 d−1 (Table 2). (The mean annual 17Δ-GOPint rate integrated to 150m was 5% lower). The 17Δ-GOPint rates were higher by ∼50% in summer than winter, in marked contrast to the 17Δ-GOP rates calculated using equation (2) which were about equal in winter than summer (Table 2).

3.3. Estimates of NCP Based on O2/Ar

[35] The measured mixed layer O2/Ar exceeded that expected at equilibrium with air every month (Figure 6) yielding an annual mean O2/Ar saturation of 101.3 ± 0.1% (SEM). In summer, the O2/Ar saturation at 101.4 ± 0.1% (SEM) was slightly greater than in winter at 101.0 ± 0.1% (SEM) and was inversely correlated with mixed layer depth and, in 2007, correlated with 18O-GOP (Figure 6). Vertical mixing had a small effect on O2/Ar in the mixed layer, i.e., a vertical mixing rate (Kz) of 10−4 m2 s−1 decreased the annual average mixed layer (O2/Ar)sat by 0.1%, implying that the measured O2/Ar is predominantly controlled by air-sea O2 gas exchange and NCP. An annual mean rate of NCP of 14 ± 4 mmol O2 m−2 d−1 for the mixed layer was calculated using equation (3), monthly measured (O2/Ar)sat and estimated air-sea O2 gas transfer rates [Nightingale et al., 2000], which converts to 120 ± 35 mg C m−2 d−1 (3.7 ± 1.0 mol C m−2 yr−1) assuming a PQ of 1.4.

Figure 6.

The supersaturation (%) state of the dissolved O2/Ar gas ratio (i.e., [(O2/Ar)meas/ (O2/Ar)eq − 1] •100) in the mixed layer, 18O-GOP (mmol O2 m−2 d−1) and mixed layer depth (m) measured monthly at ALOHA.

3.4. Error Analysis

[36] The uncertainty in the GOP, NCP and NCP/GOP estimated from 17Δ and O2/Ar measurements was determined using a Monte Carlo approach. An error was assigned for each term in the equations for GOP, NCP and NCP/GOP. A value for each term in these equations was randomly selected assuming a normal distribution based on its mean value and uncertainty (±1 SD). For example, a value of GOP was calculated for a given set of values for each term in equation (2) and this procedure was repeated 3000 times. The variability (±1 SD) of the calculated mean GOP was determined from the 3000 individual estimates of GOP.

[37] The following errors in the terms were used: ±25% for kas representing the range between Liss and Merlivat [1986] and Wanninkhof [1992], ± 4 per meg for the mean monthly 17Δdiss in the mixed layer representing the typical SEM, ±15 per meg for 17Δphoto and ±3 per meg for 17Δeq [Luz and Barkan, 2009], ±0.2% for O2eq, ± 0.1% for (O2/Ar)sat representing the observed SEM for mean monthly O2/Ar in the mixed layer. For the calculation of GOP using equation (5), the error in the time rate of change term (d17Δ/dt) was the error in the slope of 17Δ versus time regression. This analysis yielded errors of ±40% for GOP (equation (2)), ±35% (summer) and ±50% (winter) for GOPint (equation (5)), ±25% for NCP (equation (3)) and ±0.1 for NCP/GOP (equation (4)).

4. Discussion

4.1. Estimates of PP Based on Incubations

[38] The overall ratio of 18O-GOP to 14C-PP (daytime incubation) at ALOHA is 1.9 ± 0.1 (mol O2/mol C) based on a regression of individual bottle measurements made at the same depths on each HOT cruise. This value is slightly higher than the 1.4 ± 0.2 value previously observed during a single day incubation experiment at a nearby PRPOOS study site by Grande et al. [1989] but similar to the 2.0 value based on a compilation of JGOFS data [Marra, 2002]. Assuming 15% of organic carbon produced during the daytime is respired at night at ALOHA, as estimated by Karl et al. [1996], then the ratio of 18O-GOP to daily (24-h) 14C-PP would have been 2.2 ± 0.1 and slightly lower than the 2.7 value observed during JGOFS [Bender et al., 1999; Marra, 2002].

[39] In the deeper portion of the photic layer where growth rates are light limited and slower, however, the 14C-PP rate approaches the 18O-GOP rate (Figure 2) as previously observed [Grande et al., 1989; Bender et al., 1999; Juranek and Quay, 2005]. There are several possible explanations of these observations. There could have been a decrease with depth in the photic layer of the O2 production to carbon fixation ratio, the autotrophic respiration to photosynthesis ratio, or the proportions of recycled 14CO2 or unmeasured DOC14 production.

[40] There was a stronger seasonality in 18O-GOP, with summertime (May–Sep) rates exceeding wintertime (Nov–Mar) rates by 32%, compared to 14C-PP that varied by 18%, between seasons.

4.2. Estimates of GOP Based on 17Δ

[41] The ALOHA site is a tough test for the 17Δ-GOP method because the PP rates are low and have low variability (based on historic 14C-PP). The annual mean mixed layer 17Δ of 32 ± 2 (SEM) per meg is close to the atmospheric equilibrium value (17Δeq) of 17 ± 3 per meg. Since 17Δ-GOP is proportional to the difference between the measured 17Δ and 17Δeq (see equation (2)), an SEM of ±4 per meg for each monthly 17Δ measured in the mixed layer typically yields ∼±25% uncertainty in GOP. Adding the ±25% error in air-sea gas transfer rates increases the overall uncertainty of the monthly GPP estimates to ∼±40%, as discussed above. If the seasonality of GOP at ALOHA was ∼30%, as indicated by the bottle 18O-GPP measurements, then the uncertainty in an individual 17Δ-GOP determination precludes detection of monthly variations in GOP. Nonetheless, the 17Δ-GOP method yields useful estimates of seasonal and annual mean GOP rates, as discussed below.

[42] During the summer, under well-stratified conditions, the mean mixed layer 17Δ-GOP was 81 ± 39 mmol O2 m−2 d−1. This mixed layer 17Δ-GOP value was extrapolated to 200 m by assuming that the proportion of 17Δ-GOP integrated to 100 m that occurred in the mixed layer equaled the measured proportion of 18O-GOP (integrated to 100 m) that occurred in the mixed layer and that 10% of GOP occurred between 100 and 200m as observed for 14C-PP by Letelier et al. [1996]. This extrapolated summertime 17Δ-GOP estimate (0–200m) of 136 ± 51 mmol O2 m−2 d−1 exceeded by only 10% the summertime 17Δ-GOPint value of 123 ± 44 mmol O2 m−2 d−1 calculated using the time dependent and depth-integrated (0–200 m) version of the 17Δ-GOP method (equation (5)); see Table 2. Given the uncertainties in each GOP estimate, this difference is insignificant (although most of the errors in the two estimates are correlated except for the time rate of change term) and suggests that the 17Δ-GOP method as typically applied to the mixed layer (equation (2)) under well stratified conditions may be only slightly overestimated due to mixing of subsurface waters. Similarly, Sarma et al. [2005] estimated that mixing with subsurface waters caused a slight (<14%) overestimate in mixed layer 17Δ-GOP rates during summer in Sagami Bay.

[43] The situation during winter at ALOHA is very different when the mean 17Δ-GOP of 144 ± 66 mmol O2 m−2 d−1 (equation (2) and extrapolated to 200 m, as described above) exceeded by ∼75% the mean 17Δ-GOPint of 83 ± 42 mmol O2 m−2 d−1 (0–200 m); see Table 2. Subtropical locations like ALOHA (and BATS) are particularly sensitive to entrainment induced biases in mixed layer 17Δ-GOP estimates (equation (2)) because GOP rates are low and elevated subsurface 17Δ levels exist during early fall at the onset of the mixed layer deepening (Figure 4). The bias in 17Δ-GOP caused by entrainment explains why mixed layer 17Δ-GOP rates (equation (2)) were higher in winter than in summer, opposite to the measured seasonal trends in 18O-GOP, 14C-PP and PAR (Figure 3). In contrast, the mean depth-integrated 17Δ-GOPint (0–200m) in summer exceeded winter 17Δ-GOPint by ∼50%, which was similar in magnitude to the 40% difference in PAR between summer and winter.

Table 2. Rates of Primary Production in the Photic Layer Based on 17Δ-GOP, 18O-GOP (mmol O2 m−2 d−1), and 14C-PP (mmol C m−2 d−1) for Summer (Jun–Sep), Winter (Dec–Mar), and Annually at ALOHA Between March 2006 and February 2008
17Δ-GOPa17Δ-GOPintb18O-GOPc14C-PPcequation imageequation image
  • a

    Estimated from rigorous version of equation (2), assuming steady-state mixed layer 17Δ conditions and extrapolated to 200m assuming the relative depth decrease of 18O-GPP rates measured to 100m applied to 17Δ-GPP rates and 10% of PP occurred between 100 and 200m based on comparison of 14C-PP incubations [Letelier et al., 1996].

  • b

    Depth-integrated to 200m using time dependent method (equation (5)).

  • c

    Depth integrated to 100 m using bottle measurements at five depths and extrapolated to 200 m, as described above.

136 ± 51123 ± 4489 ± 1845 ± 81.4 ± 0.62.0 ± 0.5
144 ± 6683 ± 4267 ± 1338 ± 61.2 ± 0.71.8 ± 0.4
140 ± 58103 ± 4378 ± 1742 ± 71.3 ± 0.61.9 ± 0.5

[44] One test of the validity of the 17Δ-GOPint approach is whether the depth-integrated 17Δ increase observed during the summer was balanced by the 17Δ decrease observed during the winter so that the depth profile of 17Δ is reset after an annual cycle. This was the situation at ALOHA where the wintertime depth-integrated 17Δ decrease to 200 m was within 2% of the summertime depth-integrated 17Δ increase. This balance is demonstrated by the observation that the annual mean 17Δ-GOPint of 103 ± 43 mmol O2 m−2 d−1 integrated over summer and winter equaled the mean annual mixed layer 17Δ-GOP of 100 ± 51 mmol O2 m−2 d−1 (equation (2)) and occurs because equation (5) reduces to equation (2) when the time rate of change term approaches zero. One benefit of this observation is that the calculated standard error in the mean (SEM) for the annual mixed layer 17Δ-GOP at ±17 mmol O2 m−2 d−1 is significantly smaller than the error in the 17Δ-GOPint method at ±43 mmols O2 m−2 d−1, which will help in the comparison between 17Δ-GOP and 18O-GOP estimates discussed below.

4.3. Incubation Versus Non-Incubation Estimates of GOP

[45] The concurrent monthly measurements of 17Δ and 18O-bottle incubations at ALOHA for two years provided an unprecedented opportunity to compare incubation and non-incubation estimates of GOP (Table 2). We find that 17Δ-GOP always exceeds 18O-GOP. During the well-stratified summer season (May–Sep), the mean mixed layer 17Δ-GOP (rigorous version of equation (2)) of 81 ± 39 mmol O2 m−2 d−1 (SEM = 12, based on the eleven individual monthly 17Δ-GOP estimates that comprised the summer mean value) exceeded by 60 ± 12% (based on SEMs) the mixed layer 18O-GOP of 51 ± 14 mmol O2 m−2 d−1 (SEM = 4). Similarly, but with greater uncertainty, the summertime 17Δ-GOPint (integrated to 200 m) of 123 ± 44 mmol O2 m−2 d−1 was 38 ± 57% higher than the average 18O-GOP of 89 ± 18 mmol O2 m−2 d−1 and the wintertime 17Δ-GOPint of 83 ± 42 mmol O2 m−2 d−1 was 25 ± 66% higher than the mean wintertime 18O-GOP of 67 ± 13 mmol O2 m−2 d−1 (where 18O-GOP was integrated to 100 m and extrapolated to 200 m, as described above). Thus 17Δ-GOP consistently exceeded 18O-GOP at ALOHA, yet there is substantial uncertainty in whether this difference is significant. If the difference (∼25–60%) between 17Δ-GOP and 18O-GOP estimates at ALOHA is accurate, then either the 18O method underestimated GOP, most likely because of bottle incubation effects or missed stochastic PP events, and/or the 17Δ-method overestimated GOP most likely because of an overestimated air-sea O2 gas exchange rate.

[46] Let's first consider the air-sea O2 gas transfer rate used in the 17Δ-GOP method. Estimates of kas at ALOHA were derived from wind speeds measured at a nearby National Data Buoy Center buoy site 51001 (23.5°N, 162.2°W) weighted following Reuer et al. [2007] and the empirical relationship between kas and wind speed reported by Nightingale et al. [2000]. At ALOHA, kas estimates would have been ∼20% higher and lower based on the work of Wanninkhof [1992] and Liss and Merlivat [1986], respectively. The kas values used here are in the middle of the possible range, agree with recently determined kas versus wind speed relationships [Ho et al., 2006; Sweeney et al., 2007] and likely overestimated by no more than 20%. In short, the 25–60% difference between 17Δ-GOP and 18O-GOP is unlikely to be solely a result of overestimated 17Δ-GOP.

[47] Next, let's consider the likely impact of episodic PP events on 17Δ-GOP and 18O-GOP due to the different integration times of the two methods. Infrequent PP events (either spatially or temporally) that contribute to the overall productivity of a region would cause periodic (monthly) single day incubations to underestimate the average PP [Karl et al., 2003]. In contrast, the 17Δ-GOP method integrates GOP over the residence time of O2 in the mixed layer (typically 1–2 weeks at ALOHA) and thus would more often capture PP events and could yield a mean 17Δ-GOP rate higher than the mean incubation-based 18O-GOP rate. However, there is no evidence from the 20 year record of monthly 14C-PP measurements at ALOHA that pulses of enhanced PP are important. To illustrate this point, let's assume that a 40% higher rate of (non-incubation) 17Δ-GOP compared to (incubation) 18O-GOP was the result of episodic PP events. This difference would result from PP events that were 5x stronger than the background PP rate occurring 10% of the time. In contrast, the historic 14C-PP data at ALOHA indicate that only two out of 172 14C-PP measurements (1%) exceeded the mean 14C-PP rate by >2x and none of the 172 14C-PP measurements exceeded the mean by 3x. Thus there is no evidence from the historic 14C-PP data that PP events occur with sufficient frequency or magnitude to explain the 25–60% difference between non-incubation 17Δ-GOP and incubation-based 18O-GOP. Although storm events or eddies can have a significant impact on annual rates of NCP at ALOHA [e.g., Emerson et al., 2002] such events will have a much smaller impact on rates of PP. At ALOHA, estimates of NCP at ∼90 mg C m−2 d−1 (Table 1) are ∼20% of the mean 14C-PP (∼500 mg C m−2 d−1), thus a hypothetical doubling of the annual mean NCP due to episodic events would only increase PP by ∼20%.

[48] Finally, let's consider possible biases in incubation-based PP methods. Once seawater is enclosed in bottles and incubated at fixed depths, the characteristics of the environment change (e.g., light, biological community composition, nutrients, excretion products, grazing, turbulence, etc.), which in turn potentially impacts the PP rate. However, neither the magnitude nor even direction of these incubation effects is well known. If the 25–60% difference between 17Δ-GOP and 18O-GOP estimates observed at ALOHA is too large to be explained by uncertainty of the 17Δ-GOP method or differences in integrating episodic PP events, as discussed above, then one should consider whether the difference may be caused by a bias in the incubation method. If so, the environmental changes inherent to isolating seawater in bottles (or bags) could cause biases in all rate measurement based on incubation methods (e.g., 18O, 14C, and O2).

4.4. Estimates of NCP and NCP/GOP Based on O2/Ar and 17Δ

[49] A mean NCP rate of 14 ± 4 mmol O2 m−2 d−1 (120 ± 35 mg C m−2 d−1 or 3.7 ± 1.0 mol C m−2 yr−1, assuming a PQ = 1.4) for the mixed layer at ALOHA for 2006–2008 was estimated from the measured monthly saturation state of the dissolved O2/Ar gases in the surface layer and estimates of kas (equation (3)). Previously, O2/Ar measurements at ALOHA yielded NCP estimates for the mixed layer of 10.4 ± 6.6 mmol O2 m−2 d−1 in the 1990s [Emerson et al., 1997], 3–4.7 mmol O2 m−2 d−1 in 2000 [Hamme and Emerson, 2006] and 9–17 mmol O2 m−2 d−1 in summer 2004 [Juranek and Quay, 2005]. Emerson et al. [2008] used a continuous record of dissolved O2 in 2005 from a moored O2 sensor at ALOHA to estimate a NCP rate of 13.1 ± 7.2 mmol O2 m−2 d−1. The NCP rate below the mixed layer has been estimated at 6.0 ± 0.8 mmol O2 m−2 d−1 using the annual cycle in dissolved O2 measured by a profiling float near ALOHA [Riser and Johnson, 2008] and 3.5 ± 0.4 mmol O2 m−2 d−1 using the subsurface O2 field at ALOHA measured by a SeaGlider survey [Nicholson et al., 2008]. Combining these sub mixed layer NCP estimates with our O2/Ar derived NCP estimate for the mixed layer yields an NCP of ∼19 ± 4 mmol O2 m−2 d−1 for the photic layer, which corresponds to NCP of 5 ± 1 mols C m−2 yr−1 (for a PQ = 1.4).

[50] The ratio of net community oxygen production to gross oxygen production (NCP/GOP) in the mixed layer was estimated from measured 17Δ and O2/Ar using equation (4). During the summer, the mean NCP/GOP was 0.22 ± 0.08. During winter the mean NCP/GOP was lower at 0.12 ± 0.05 but likely underestimated because entrainment elevated mixed layer 17Δ values significantly, as discussed above. The summertime NCP/GOP estimate is slightly higher than previous range of estimates of NCP/GOP (0.08 ± 0.05 to 0.13 ± 0.05) based on 17Δ and O2/Ar measurements at ALOHA, BATS, equatorial Pacific and Southern Ocean, as discussed above. An annual mean NCP/GOP of 0.19 ± 0.08 for the photic layer at ALOHA was estimated by dividing the mean NCP of 19 ± 4 mmol O2 m−2 d−1 for the photic layer, as discussed above, by the mean 17Δ-GOPint (0–200m) of 103 ± 43 mmol O2 m−2 d−1.

[51] Prior to 17Δ and O2/Ar measurements, the ratio of new or export production to primary productivity was typically estimated from bottle incubation based measurements of new production (e.g., 15NO3 and/or 15NH4 uptake) or estimates of organic carbon export (e.g., 15NO3 uptake, sediment traps, 234Th budgets, O2 budgets, etc.) compared to measured 14C-PP. Thus, most previous estimates e- and f-ratios potentially suffer from incubation artifacts. In contrast, the NCP/GOP estimated from 17Δ and O2/Ar measurements is independent of incubations, has oxygen production as a common currency and is independent of the air-sea gas exchange rate estimate with its substantial uncertainty. Thus NCP/GOP estimates based on 17Δ and O2/Ar are likely to be more accurate than previous estimates of e- and f-ratios when measured O2/Ar is significantly greater than 100% saturation. Converting the 17Δ and O2/Ar based NCP/GOP to an e-ratio of NCP/14C-PP (mol C/mol C), assuming a PQ of 1.4 and the observed 17Δ-GOP/14C-PP of ∼2.5, yields a NCP/14C-PP of 0.34 ± 0.14 for the mixed layer that is about double a previous e-ratio estimate of ∼0.15 at ALOHA derived from an O2 budget-based estimate of NCP and measured 14C-PP [Laws et al., 2000b].

5. Oceanic PP: Where Do We Stand?

[52] The oceanographic community has relied overwhelmingly on an incubation-based methodology to obtain PP rates. The recent introduction of non-incubation based 17Δ and FRRF PP methods provide independent means to evaluate incubation-based PP methods. However, since an absolute standard for PP does not exist the accuracy of any method cannot be verified. An additional complication is that the methods measure different metrics of PP (e.g., C fixation, O2 production, fluorescence, etc.). The best we can do at this time is to compare multiple incubation and non-incubation (including satellite) based PP methods systematically and determine whether consistent relationships exist between methods.

[53] At ALOHA, we found that the non-incubation 17Δ method yielded gross oxygen production (GOP) rates that were consistently ∼25–60% higher than GOP rates measured by 18O incubation method. Yet, the uncertainties in the individual GOP estimates is comparable to the difference between the estimates. Is the non-incubation or incubation GOP method more accurate? Each method has potential biases. High 17Δ-GOP rates could be a result of an overestimated air-sea gas transfer rate or a result of mixed layer entrainment of subsurface water. Low 18O-GOP rates could reflect missed PP events or incubation artifacts. However, the chosen gas transfer rate parameterization was in the middle of the possible range with uncertainties that could explain only up to 20% of the difference between the methods. The depth-integrated 17Δ-GOPint method eliminated the entrainment bias. Incubation methods could underestimate mean PP because of missed PP events, however, two decades of 14C-PP measurements at ALOHA indicate that PP events occur too infrequently to be an important factor. Thus the consistent difference between 17Δ-GOP and 18O-GOP estimates at ALOHA, despite substantial uncertainty, raises the question whether incubation methods underestimate PP rates in the ocean.

[54] Utilizing multiple methods to estimate PP potentially yields new insights into the ecosystem function as illustrated by the difference in seasonal and depth variability of 17Δ-GOP, 18O-GOP and 14C-PP at ALOHA. During summer the mean 17Δ-GOPint and 18O-GOP were 48% and 32% higher, respectively, than winter rates, whereas for 14C-PP the difference was only 18%. Similarly, 18O-GOP decreased on average by sixfold between the surface and 100m, whereas 14C-PP decreased by only threefold. Possible explanations of these observations could involve variations in autotrophic respiration to photosynthesis ratio or the O2 production to carbon fixation ratio. If true, we need to better understand the physiological reasons for these variations. If on the other hand, the explanation is a result of methodology (e.g., incubation/bottle effects, variations in CO2 recycling or DOC excretion, etc.), this needs to be determined experimentally.

[55] There are only a few comparisons of concurrent incubation (18O, 14C and O2) and non-incubation (FRRF and 17Δ methods) based estimates of PP with which to compare our observations. At ALOHA, Corno et al. [2005] found that yearlong FRRF-based estimates of GOP were 1.9 ± 0.2 to 2.9 ± 0.2 times concurrently measured daytime 14C-PP (where the range depended on specific volume, light or chlorophyll normalizations), which was similar to the 17Δ-GOPint/14C-PP of 2.7 (summer) and 2.2 (winter) reported here. At BATS, Luz and Barkan [2009] determined mixed layer 17Δ-GOP rates during four monthly cruises between May and October 2000 that were much higher (4–8x) than concurrent 14C-PP rate estimates. In Sagami Bay, off the coast of Japan, Sarma et al. [2005] found that 17Δ-GOP rates were, on average, 1.8 ± 0.5x GOP rates estimated from bottle O2 incubations and 1.6 ± 0.3x FRRF-GOP rates, where GOP ranged from ∼100–350 mmol O2 m−2 d−1. In the Celtic Sea, Robinson et al. [2009] measured mean 18O-GOP rates that were 4.5 ± 1.2x concurrent 14C-PP rates based on on-deck incubations and measured FRRF-GOP rates that were on average only 14% and 40% of GOP rates estimated by 18O and O2 bottle incubation methods, respectively.

[56] A potential methodological bias inherent to incubation-based NCP methods would affect our assessment of the metabolic balance of the ocean. At ALOHA, the consistently supersaturated O2/Ar levels measured every month for two years indicated an autotrophic condition with an annual net O2 production rate of 14 ± 4 mmol O2 m−2 d−1. A similar consistent monthly O2/Ar supersaturation condition was measured at ALOHA for a one year interval by Hamme and Emerson [2006]. Yet, incubation-based net O2 production rates measured monthly at ALOHA over an annual cycle indicated an overall heterotrophic condition with an annual net O2 consumption of 25 ± 4 mmol O2 m−2 d−1 [Williams et al., 2004]. Although Williams et al. acknowledge that a net heterotrophic condition at ALOHA is unlikely, they attributed the underestimation of O2 production by the bottle incubation method to missed productivity events. However, the consistently supersaturated O2/Ar levels measured at ALOHA (24 consecutive times over two years) and the magnitude of the difference between the incubation and O2/Ar based NCP estimates make it unlikely that episodic NCP events could explain the difference. Thus, one needs to consider an alternate explanation that O2 incubation methods underestimate photosynthetic O2 production and/or overestimate respiratory O2 consumption rates compared to in situ conditions. Such a bias in the O2 incubation method would explain the contradiction between the net heterotrophic conditions (∼15 mmol O2 m−2 d−1) measured by Williams et al. every month for one year by the O2 incubations in the lower half of the photic layer at ALOHA (75–150 m) and net autotrophic conditions (∼1.5 mmol O2 m−2 d−1) for this same depth interval determined by Riser and Johnson [2008] based on the annual cycle in O2 measured by an ARGO float near station ALOHA. Until possible incubation biases are resolved, conclusions of net heterotrophic conditions in the photic layer of the oligotrophic ocean based on the O2 incubation method [e.g., del Giorgio and Duarte, 2002; Robinson et al., 2002; Williams et al., 2004; Gist et al., 2009] need to be viewed cautiously.

[57] The non-incubation based estimate of an e-ratio using coupled 17Δ and O2/Ar measurements provides a much needed alternative to traditional incubation based estimates. Since the e-ratio is a measure of the ocean's biological pump efficiency, it is an important index of ecosystem function and biogeochemical cycling. The 17Δ and O2/Ar based NCP/GOP estimate of 0.19 ± 0.08 at ALOHA indicates that ∼20% of the gross photosynthetic production escapes respiration in the mixed layer (and photic layer) and is available for export or harvest.

[58] Clearly, more systematic comparisons of incubation and non-incubation PP methods are needed over a range of productivity regimes in the ocean. The few available comparisons of simultaneous non-incubation (17Δ, FRRF) and incubation (18O, 14C and O2) based PP methods yield variable results [e.g., Sarma et al., 2005; Corno et al., 2005; Luz and Barkan, 2009; Robinson et al., 2009]. A better understanding about the relationship between photosynthetic O2 production and carbon fixation under varying light intensities, nutrient supply and photosynthesis rates is needed to accurately convert FRRF and 17Δ based GOP rates to carbon fixation rates. More accurate air-sea gas exchange rates are needed to improve the accuracy of the 17Δ-GOP method. The careful elimination of biases in the 17Δ-GOP method resulting from entrainment and mixing is needed. Despite the uncertainties in the non-incubation PP and NCP methods, however, the 17Δ-GOP and O2/Ar-NCP results at ALOHA raise a cautionary flag about incubation-based PP and NCP estimates (i.e., including 14C, 18O and O2 methods) in the open ocean.

[59] Primary production is a fundamental process at the foundation of the ocean's biological pump. Yet, the results of newer non-incubation PP techniques can yield very different estimates of PP rates from traditional incubation methods upon which the oceanographic community has relied heavily. Experiments need to be designed to evaluate multiple PP methods while simultaneously examining the ecosystem composition and plankton physiology in a well constrained physical environment that would be analogous, for example, to the many GasEx experiments designed to better constrain air-sea gas exchange rates. If such PP experiments demonstrate that non-incubation methods yield more accurate PP rates than incubation methods, then our current view of the magnitude and variability of biological productivity (including satellite-based estimates) and the efficiency of organic carbon export in the ocean may need significant reassessment.


[60] We thank the crew and personnel during all the HOT cruises for their assistance with sample collection. Thanks to Mark Haught for helping with the oxygen isotope and O2/Ar measurements and to Laurie Juranek for advice on sample collection and 18O spiking procedures. In particular, we want to acknowledge the financial support by NSF Ocean Sciences under grant OCE 0525843 (P.D.Q.) and OCE 0326616 (D.M.K.) and the Gordon and Betty Moore Foundation (D.M.K.).