Recently, the method for estimation of marine gross production from measurements of 17O/16O and 18O/16O of dissolved O2 has been improved by obtaining a rigorous equation for evaluation of gross production, and it has been suggested that small errors associated with approximations resulted in significant underestimation of gross production rates. The new equation requires accurate knowledge of the isotopic composition of photosynthetic and atmospheric oxygen. While the values for the latter have been determined very precisely, those of the former have never been measured. Here we present the first experimentally derived values of δ17Op and δ18Op of marine photosynthetic O2 (−10.126 and −20.014 ‰). Based on these, we show that the suggested underestimation was not the result of approximation errors, rather it was caused by inaccurate estimated values of δ17Op and δ18Op.
The quantity 17Δ denotes an excess of 17O for a given δ18O. It has been defined in several ways [e.g., Kaiser, 2011], but in most studies the logarithmic form has been used:
The main reason for this choice was that in this form, as noted by Miller , the value of λ is independent of the choice of isotopic reference. This facilitates direct comparison of fractionation data and results of different laboratories. In contrast, it is clear from equation (2) that the magnitude of 17Δ depends on the isotopic composition of the standards (e.g., atmospheric O2 or VSMOW – Vienna Standard Mean Ocean Water) and also on the reference slope λ. Therefore, both should be chosen such that they optimally fit the system under consideration. The background theory of three isotope fractionations of O2 for kinetic processes and under steady state conditions and derivation of equation (1) for these cases are discussed by Luz and Barkan [2005, Appendix A].
Prokopenko et al.  have recently published a new, elegantly derived, rigorous equation by which gross production can be calculated directly from measured δ17O and δ18O of dissolved O2. Furthermore, their paper suggests that calculations with equation (1), which is an approximation, leads to underestimation of the ratio of net to gross O2 production (N/G) by up to 50%. Whereas we agree that equation (1) is an approximation, we disagree with their conclusion that the 50% discrepancy results from approximation errors.
2. Proper Choice of δ17O and δ18O of Photosynthetic O2 for Calculating G
 All the parameters in equation (1), with the exception of 17Δbio, are as defined by Luz and Barkan . The parameter 17Δbio (previously denoted by the less intuitive symbol 17Δmax) stands for pure biological O2 having maximum value of 17Δ (vs. atmospheric O2) that is affected by both photosynthesis and respiration in steady state, but not gas exchange.
where δ*O = *Rs/*Rref − 1. The factor 0.518 in equation (3) is a value of parameter λ in Equation 7 of Prokopenko et al. that equals (1 − α17)/(1 − α18). This ratio for respiration, the most relevant process, was accurately determined in direct measurements of Luz and Barkan  and is not based on any assumptions. Importantly, δ17Op and δ18Op stand for photosynthetic O2 that is not affected by respiration at all. In contrast, 17Δbio in equation (1) is derived from δ17Obio and δ18Obio at biological steady state where rates of photosynthesis and respiration are equal. Correct value of 17Δbio can be derived from δ17Op and δ18Op only with proper choice of λ.
 A major advantage of the new equation is that it makes it possible, for the first time, to calculate G/(kOeq) directly from measured δ17O and δ18O of dissolved oxygen and corresponding values of O2 produced by photosynthesis (δ17Op and δ18Op) and in equilibrium with atmospheric O2 (δ17Oeq and δ18Oeq).
 Prokopenko et al. present results based on two values of λ: 0.518 and 0.516. In the first case the differences from calculations with the approximate equation were ∼38% and ∼50% in G and N/G, respectively, but in the second case the differences were much smaller. Evidently, the authors chose to emphasize the first case by giving it strong presence both in their abstract and Table 1. Furthermore, they made a strong statement in the abstract that small errors associated with approximations resulted in large relative errors in G and N/G. We do not agree with this statement because a different choice of λ leads to reasonable agreement between values obtained with the new and old equations. In fact, the authors themselves demonstrated this point.
 We emphasize that a properly chosen value of λ for calculating δ17Op must be smaller than 0.518. Indeed, for steady state between photosynthesis and respiration for which Prokopenko et al. derived their equation, λ should be obtained from equation 17 of Angert et al. . This equation was obtained with the specific aim to explaining photosynthesis-respiration steady-state. Depending on the known sizes of respiratory fractionations (about −30 to −14 ‰), λ can be calculated to range between 0.5142 and 0.5162. This point is discussed in considerable detail by Nicholson . Yet, realizing the confusion associated with proper choice of λ, we present here direct determination of δ17Op and δ18Op from experimental data of Helman et al. , Eisenstadt et al.  and Barkan and Luz .
 In Figure 1 we show δ17Op and δ18Op values calculated from data in the above studies. As can be seen, they are always larger than the corresponding values of VSMOW. Furthermore, in the ln(δ18Op + 1) vs. ln(δ17Op + 1) plot there is very tight correlation (R2 = 0.99999) with regression slope of 0.5237. This slope is significantly larger than the 0.518 ratio of ordinary respiration or cyanide resistant respiration [Helman et al., 2005; Luz and Barkan, 2005]. Because the experiments were run at zero O2 in the medium, we can rule out respiration due to lack of substrate. It was suggested that the 17O and 18O enrichments were caused by rapid recycling of O2 by chlororespiration or PTOX in very close proximity to the production site in photosystem-2 [Eisenstadt et al., 2010].
 The new values of δ17Op and δ18Op are given in Table 1 along with corresponding G/(kOeq) and N/G calculated with equation (3), and relative deviations calculated with respect to G/(kOeq) and N/G obtained with equation (1). In addition, similar deviations were calculated for δ17Op and δ18Op from Prokopenko et al. As can be seen, the deviations for all marine organisms, except coccolithophores are less than ∼13% in both G/(kOeq) and N/G, and are within the range of estimation errors with either equation (1) or equation (3). We also present results of calculations made by Prokopenko et al. for λ = 0.518 and λ = 0.516. Clearly, while in the first case the differences are large, in the second case they are only ∼5%, the same order as for the studied phytoplankton. This independently confirms our point of view that for proper calculation of G/(kOeq), λ = 0.516 is a better choice.
 The obtained average phytoplankton values of δ17Op and δ18Op are at present the most accurate ones, and thus, we recommend using them for calculations of marine gross O2 production. We note that for the purpose of improving G/(kOeq) accuracy, it will be worthwhile studying more marine organisms and also the potential gross production rate of each group of phytoplankton.
 From the discussion above, it is begging asking the question of whether calculations with the new rigorous equation of Prokopenko et al.  and direct measurements of δ17Op and δ18Op is an improvement over the previous approach of Luz and Barkan  with 17Δbio. The answer is clearly positive. Mainly so because as done by Luz and Barkan  the value of 17Δbio = 249 per meg was derived with relatively low precision (±15 per meg) from difficult experiments with corals and green algae, one of these does not represent open ocean phytoplankton. In contrast, the current average values of δ17Op and δ18Op are based on four groups of phytoplankton, three of which (cyanobacteria, diatoms and coccolithophores) are very important primary producers in the world's oceans.
 Finally, Prokopenko et al.  pointed out that the N/G ratios obtained with the approximate equation were erroneously high and exceeded biologically feasible value of 0.4 to 0.5. We strongly disagree with this statement and point out that there is no basis for the claim that N/G values larger than 0.5 are not biologically feasible. For example, in the seminal paper on gross and net O2 production of Bender et al. , the authors conducted careful laboratory experiments and determined N/G values in the range of 0.79 to 0.96 [see Bender et al., 1987, Tables 5 and 7]. Yet, such high N/G ratios are exceptional as are the 0.4 and 0.55 values of Halsey et al.  and Bender et al. . Much smaller ratios (0 to 0.3) are more common over wide oceanic areas [e.g., Reuer et al., 2007].
 In conclusion, the new rigorous equation of Prokopenko et al.  is a very welcome addition to studies of gross O2 production in the ocean surface and, in our opinion, should be adopted and used from now on by marine biogeochemists. However, we emphasize that estimates done previously with the approximate equation yielded results that are close to those that can now be calculated with the rigorous one.
 This research has been supported by grants from the Israel Science Foundation.
 The Editor thanks Maria Prokopenko and an anonymous reviewer for their assistance in evaluating this paper.