Intermediate water masses, a major supplier of oxygen for the eastern tropical Pacific ocean

It is well known that Intermediate Water Masses (IWM) are sinking in high latitudes and ventilate the lower thermocline (500 – 1500 m depth). We here highlight how the IWM oxygen content and the IWM pathway along the Equatorial Intermediate Current System (EICS) towards the eastern tropical Pacific ocean are essential for the supply of oxygen to the lower thermocline and the Oxygen Minimum Zones (OMZs). To this end, we assess here a heterogeneous subset of ocean models characterized by a horizontal resolution ranging from 0.1° to 2.8°. Subtropical oxygen levels in the lower thermocline, i.e., IWM are statistically correlated with tropical oxygen levels and OMZs. Sensitivity simulations suggest that the oxygen biases of the subtropical IWM oxygen levels contribute to oxygen biases of the tropical thermocline : an increase of the IWM oxygen by 60 mmol.m results in a 10 mmol.m increase in the tropical ocean in a timescale of 50 years. In the equatorial regions, the IWM recirculates into the Equatorial Intermediate Current System (EICS). By comparing tracer and particle release simulations, we show that a developed EICS increases eastern tropical ventilation by 30 %. Typical climate models lack in representing crucial aspects of this supply: biases in IWM properties are prominent across climate models and the EICS is basically absent in models with typical resolutions of ~1°. We emphasize that these biases need to be reduced in global climate models to allow reliable projections of OMZs in a changing climate.


Conceptual reasoning
We compare the oxygen levels in a set of models characterized by different resolutions, integration time scale, forcings, etc.. Despite all these differences, we found common behaviours (part 3.1): the properties of the intermediate waters are poorly represented in all simulations that we analyzed and we found a correlation between oxygen levels in intermediate waters and oxygen levels in tropical regions (part 3.1 of the ms).
It suggests that intermediate waters affect oxygen levels and OMZ volume in tropical regions. We test this hypothesis using a "what if ?" experiment : "If the oxygen levels are realistic south of 30°S and/or below 1500m does it have an impact on OMZs ?". These sensitivity simulations are performed using a single model framework: same resolution, same forcings, same integration time.
(part 3.2) Another second hypothesis that we investigate is "do the intermediate circulation and associated jets play a large role in setting oxygen levels in the equator region ?". To reply to this question, we performed a set of sensitivity simulations using again a single model framework: same integration time, same forcings, but different spatial resolution. (part 4.2).
In addition (part 4.3) we compare the oxygen levels in a climate model suite: similar model framework, same integration time, different ocean resolution.
In summary, we investigate the mechanisms impacting tropical oxygen levels at intermediate depths in a very heterogeneous set of models, by performing dedicated sensitivity simulations that are easy to interpret.
2. Reviewer comment on the heterogeneity of the models and model set-ups that makes it difficult to pinpoint causes for differences of the simulations.

-Atmospheric forcing
We agree that the atmospheric forcing data play a large role in setting ocean properties.   To investigate this aspect, we performed two additional sensitivity simulations using the UVIC model: (i) using the CORE Normal Year Forcing wind stress and (ii) applying the NCEP wind stress data. Both simulations have been integrated for 10000 years. While the oxygen levels show significant differences, the general shape of the OMZ (oxygen lower than 20 mmol.m-3) is similar in both simulations (see Figure 3 below).

Coupled ocean atmosphere experiments
Coupled ocean-atmosphere experiments introduce further discrepancies compared to the use of realistic atmospheric forcings. However, the mean surface velocity is similar in the suite of GFDL models (especially GFDL01 and GFDL025) that we analyzed, suggesting that the effect of atmospheric forcing is likely not dominant when comparing this subset of models (part 4.3).

Conclusion
We agree with the reviewer, the differences induced by the different forcings and integration time have (not surprisingly) an impact on water masses and oxygen levels. Despite the heterogeneity of our simulations, our results nevertheless suggest a strong coupling between subtropical and tropical oxygen content and justify our questioning and the experiments performed in the part 3 and 4 of this study (see 1. Conceptual reasoning) [2] Regarding to sensitivity of tropical IWM oxygen to subtropical and deep dissolved oxygen levels, the authors refer AAIW, NPIW (and the upper part of the PDW) as IWM in this study. I was wondering what will be the relative contributions of each water masses to dissolved oxygen supply, ventilation in the eastern tropical Pacific ocean (particularly North (NPIW) vs South (AAIW)). My impression is that AAIW could be more dominant (e.g. Talley [3] The core of the study is based on a suite of sensitivity simulations from NEMO(NEMO2). In the first reading, I struggled a bit on connecting aim and each sensitivity experiments. The dissolved oxygen restoring simulations aim on investigating sensitivity of tropical IWM oxygen to subtropical and deep dissolved oxygen levels (as stated in section 3.2) and the conservative tracer release simulations are more dedicated to investigate spreading of tracers towards the eastern tropical Pacific (transport by the EICS, as stated in section 4.1). While the standard structure of the manuscript is to introduce overall data and methods in the beginning, (section 2), I suggest to move some of the objective and details of sensitivity experiments to each corresponding sections (referring to sections 3.2 and 4.1) so it is much easier to follow the aim bridging to sensitivity experiments (I think it is still fine to keep brief general descriptions in section 2 including Table 1).
Alternatively, the methods section could be revised to include additional descriptions connecting to corresponding result sections. I will leave this decision to the authors regarding to the structure of the paper but I think the flow could be improved.
We agree and improved the flow of the experiment in the final version of the manuscript.
[4] Another major issue is the figures. Similar issues for multiple maps (such as Fig.5), it will be reader friendly to label maps with "zonal advection", "meridional advection" etc.
The transport terms (Fig 4)  I put few more specific suggestions below and hope this helps to point out the difficulties I am referring to.
Thanks to the reviewer for these suggestions. We have rechecked all captions to make sure that they are correctly describing the panels.
[4.1] Fig.1caption, (L762−763) oxygen levels (mean 500 -1500m) at 160W, I think color shading in b) is not vertical mean (because it is depth-latitude section). Also, is dissolved oxygen in Fig.1from observations such as World Ocean Atlas?
The new legend of the Figure 1 is reproduced at the end of this reply

Minor Comments
[1] I am curious whether CORE v2 climatological forcing (used for NEMO) and NCEP/NCAR climatological forcing (wind stress, used for UVIC) makes a difference in paper spinup states. As far as I know, CORE v2 forcing is based on NCEP/NCAR reanalysis but it has several corrections and adjustments in the forcing and difference between the two could lead to different results, particularly after long-term spinup. Do authors think this is a minor thing ?
The different climatological forcings have indeed a significant impact (see Figure 3 of our response). However we think that differences in resolution play a larger role by resolving additional processes (in particular deep equatorial jets) [2] Are all the GFDL model simulations integrated for the same period following high-resolution [4] Regarding to dissolved oxygen restoring, are the boundaries (and depth inter-face at 1500m) all in the Pacific ocean only (e.g. thinking of for example, 30•N and 30•S zonal walls and 1500m layer in the entire Pacific ocean) or globally ? Also, how strong (i.e. timescale) is the restoring in these simulations ?
The term "restoring" is maybe inadequate and has been replaced by "forcing to the observations" in the manuscript as the oxygen levels are forced to the WOA monthly climatology. The latitude where the forcing is applied has been set globally (however as it is a "forcing", it does not make any difference if it were applied solely in the Pacific Ocean).
[5] Regarding to the respiration rate (in L144), did you set all the simulations respiration rate [6] I am a bit confused by the locations of particle release and IETP/IWTP regions you were referring to (L363−383, Fig.7 and 8). While the the locations of particle release is in sections (shown as black bold lines (or dot) in Fig.7), I thought the IETP/IWTP are basins in specific rectangles and this is different from the locations of particle release (it contains of course) if I understand correctly. If that is the case, I suggest to revise the main text and Figure to include these information more explicitly (I think adding boxes in Fig.7 could help and you can refer to that interpreting Fig.8).
A new Figure 8a has been added, which shows the IETP/IWTP boxes and the release locations.
[7] Just for clarification: do ocean stand-alone simulations (i.e. NEMO and UVIC) paper also use preindustrial pCO2for spinup (related to mean state diagnostics)?
Preindustrial pCO2 is used. This is now stated in the text.
[8] In section 2.1, Table 1, and part of the main text: The author mix use the NEMO and NEMO2 through the manuscript and I have got a bit confused. Since all the simulations use NEMO2, you should make the terminology consistent through the text after introducing (or just NEMO, I will leave this to the authors).
Three versions of NEMO are used : NEMO2 (with biogeochemical cycles), NEMO05, NEMO01 (physics only). We now refer specifically to these versions in the text.
[9] For Table 1, I would suggest to include model integration time information.
The model integration time has been added in the Table 1 (see last section  This is corrected in the final version of the ms [L124]more than 50 years: suggest to change to "60 years" (the same as the statement in latter section, L160). This is corrected in the final version of the ms [L167]5 daily means: I think "5-day mean" is more common.
This is corrected in the final version of the ms [L262−263]Where is the information (figure) of total advective term? Fig. 4g is the vertical advection term difference and I could not find specific information on total term in the figure (although it is possible to infer from all the terms).
The objective of the Figure 4 is to better explain the differences between the model experiments ( Fig 3g). As the patterns are mostly zonal, we did not show in Fig 4 the total term (the zonal mean of the total term is already displayed in Fig. 3g).
[L301]Tsuchuya jets: should be "Tsuchiya jets". This is corrected in the final version of the ms       Table 1) are displayed in contour