This manuscript provides valuable insights into the misrepresentation of the Arabian Oxygen minimum zone (ASOMZ) in 10 historical CMIP5 model simulations, and relates these model deficiencies to the analysis of the different water masses that ventilate this OMZ. Overall, I have found the paper very useful in providing metrics to quantify the representation of the OMZ in these models. The approach of relating deficiencies at different depth of the water column to water masses of different origins provides a new and original understanding of the ventilation pathways of this OMZ, that complements nicely a previous study by the same first author in 2020. The main finding is that CMIP5 models tend to underestimate the lower part of the OMZ due to ventilation of highly oxygenated waters from the Southern Ocean.

I am confident that this work should eventually provide a very valuable contribution but I have a major concern, which is that I found that the writing was often clumsy, sometimes to a point where I was not sure I understood the meaning correctly. Although I did find the ideas and general approach of the paper very promissing, reading it was not as pleasant as it could have been and I had to struggle my way through.

Reply to reviewer #1

We would like to thank the reviewer for taking the time and for providing feedback to improve the manuscript.

My major problem was that I could not really understand how water masses were determined based on my reading of 3.3. For example, I did not understand how the formation regions have been localized (lines 5-6-7). I also did not understand how the water mass properties (T-S) were derived from observations. Therefore, it was difficult to follow 4.3 (water representation in models).

We thank you for raising the point, that the determination and localization of the water masses in the models is hard to understand. We will rewrite that part for better understanding. We have included here a preliminary revision of the chapter:

Red Sea Water and Persian Gulf Water (RSW and PGW) are geographically restricted in their formation regions. Figure 1a shows the formation region for RSW and PGW for which temperature and salinity ranges and mean values are determined (Table 2 and associated new figure).

In contrast to that Indian Central Water (ICW) is not geographically restricted in its formation regions. ICW is a mixed water mass and is characterised by a nearly linear temperature and salinity relation that is density-compensated (Tomczak, 1983) and can be identified in T-S diagrams. With this relation, we were able to define upper and lower temperature and salinity limits of ICW in observations and compared those values to respective values from literature values (see Tab. 2). ICW is formed on zonal oriented fronts in the tropical ocean sub-surface layers (Tomczak, 1983). Sprintall and Tomczak (1993) and Schott and McCreary (2001) described the geographical location of the formation region of ICW. Figure 1b shows the grid boxes where these T-S properties are found in the IO in WOA13 observations. These are in line with the description of the formation region as shown by Sprintall and Tomczak (1993) and Schott and McCreary (2001).

To investigate the formation region of ICW in the models, we followed the same procedure as previously described for the observations. The linear temperature/salinity relation as given by the T-S diagrams of the individual models (Fig. S4) sets the upper and lower temperature and salinity limits (see also Table 2).

Different to the observations and the literature, the resulting locations that determine the formation of the simulated ICW are not restricted to the subduction area of ICW.  For consistency reasons, we therefore limit the formation region of ICW in the models to the subduction area of ICW as prescribed by Sprintall and Tomczac (1993) and Schott and McCreary (2001). We exclude grid boxes with similar T-S properties that are found outside the subduction region as well as those within the upper 200 m so that the oxygen content of subducted ICW is not affected by the well ventilated mixed layer. Figure S2 shows the respective area for each model and the deepest depth at each location, where the T-S properties are found.

Indian Ocean Deep Water (IODW) originates in the Southern Ocean, where it is often referred to as Circumpolar Deep Water and Antarctic Bottom Water, before it travels northward into the deep IO and mixes along its way with the surrounding water masses. IODW is thus defined as the densest water mass in the IO north of 60 °S  that is found below 1500 m depth (Talley et al., 2011).

Fig 1c shows the formation region of IODW derived from observations for which temperature and salinity limits are determined. IODW in the models is defined in the similar way as in observations. In the models the derived formation regions of IODW in the Southern Ocean differ from those we find in observations (Fig. S3).

The oxygen content of the water masses as listed in Table 2 (and shown in the corresponding Figure) is calculated, for each model and the observations, by the arithmetic mean of all grid boxes of the corresponding source waters.

Major understanding issues also involved how Figure 1 was generated, how IODW, ICW, RSW/PGW were identified from Figure 2 (where do the ovals come from?).

The revision of section 3.3 also includes a more detailed description of the generation of Figure 1. The ovals in Figure 2 sketch the limits of the source water mass properties. We now include that in the caption: 'TS diagram of the Indian Ocean from observational data (WOA13) color coded by depth. The source water masses for the water mass mixing analysis are Indian Ocean Deep Water (IODW), Indian Central Water (ICW) and Red Sea and Persian Gulf Water (RSW/PGW). The ovals indicate the water masses providing an overview in the overall picture. Exact values of the water mass properties can be taken from Fig. 3 and Table 2.'

I really liked Figure 4, which is a very nice and synthetic way of representing the OMZ, but I add difficulties because of too many lines on the same plot. I would suggest to have more panels, for instance to group them by set of clusters instead of showing all models together, with WOA in all of them (which would make 4 clusters x 3 panels= 12 panels).

Thank you for raising this point. However, at this stage of the manuscript the clusters have not been introduced. The key massage of that figure is to get a quick overview of all models and to show the overall wide range in the oxygen representation. Thus, we prefer to keep the figure as it is.  

I have the same comment for Figure 5.

However, for Figure 5 we really like your suggestion. In the revised manuscript the panels are divided into the individual clusters. This makes it easier for the reader to capture the differences in the individual clusters.

Also I think it would be easier if the information contained in Table 2 was somehow shown in a set of figures, that would ease the presentation of results and the discussion.

Thank you for your suggestion. In addition to Table 2, we will include a Figure visualizing the information on the water mass properties given in the Table in the revised manuscript.

Information about the age tracer should also be shown in a synthetic figure.

We agree with the reviewer and will add a figure to the supplement visualizing the information about the age tracer that is given in the text.

In the end, because of my misunderstanding, my review is rather limited in terms of how I am able to evaluate the methodology and conclusions, and I believe that the presentation issues that I've raised must be fixed before a full assement of the content can be provided.

General Comments:

The authors present an interesting attempt to answer some of the uncertainties in the OMZ representation using the 10 CMIP5 model outputs. This is an important issue as the future prediction of OMZ shows a large inter-model spread which is a reflection of the current status of model OMZs. The models are first clustered into four bands based on vertical profiles of oxygen and then the models exhibiting most similarity with are observations are compared with other models in their water mass formation process — a representative of ventilation. The authors primarily attribute the uncertainties to a higher oxygen content in the southern ocean and the coarser model vertical resolution. However, quantitative assessment of the model discrepancies is not possible by just looking at model outputs, but I understand this is beyond the scope of the study.

The authors have communicated the scientific evidences with good clarity — outlining the scientific methods and results clearly — which is appreciable. The discussion section can be written a bit more orderly. Scientifically, there is very little discussion of the respiration part other than in the introduction section. Also, the methodology part could be a bit more explanatory to ensure transparency. Overall, the manuscript discusses a very important problem seen in nearly all the earth system models and is promising. I therefore, recommend the study alongside stating my serious concerns in detail below.

Reply to reviewer #2

We would like to thank the reviewer for taking the time and for providing constructive and very specific comments, which will help to improve the manuscript considerably. We will add more details to the methodologies and restructure the discussion part for more transparency. Also, we have carefully addressed the comments. The point-by-point responses to the specific comments follow below.

Specific Comments:

1. One cause of deoxygenation is solubility which is highly sensitive to the temperature of the oceans — warmer the surface ocean, lesser the solubility of oxygen into the ocean. The representation of ASOMZ in the CMIP5 models will also be a function of ocean temperature in the surface ocean waters. The authors discuss ventilation and respiration to some extent — which I agree are the major causes of deoxygenation. However, I feel the solubility factor might also account to some of the OMZ difference seen among the models. How would the authors justify not looking at the solubility parameter while accounting for the uncertainty in CMIP5 models.

That is a good point that the reviewer mentions. We had a look at the temperature in the upper layers of the Arabian Sea and find slightly lower temperatures there in the models compared to the observations. We compute oxygen solubilities and analyse corresponding model-data differences and will add these findings to the revised manuscript.

1. Discussion, Ln 15-24: The authors conclude that in their study there is no definite linkage found between the model resolution and the representation of OMZ — which to me is surprising. Fundamentally, the ASOMZ is located in what we call the shadow zone where ventilation occurs through mixing processes mainly caused by mesoscale eddies [Resplandy et al., 2012, Lachkar et al., 2016]. Increasing the model horizontal resolution should result in more mesoscale eddy activity hence allowing more ventilation (due to eddy mixing) and hence changing the OMZ. An absence of a linkage between the OMZ and model resolution highlights serious issues in the parameterization of subgrid scale processes in the models or the possibility that the increased model ventilation is being balanced by an increase in model respiration. I am not sure if you can conclude that increasing the horizontal resolution has no effect on the OMZ. Please explain your disagreement to my explanation, if any, and state your reasons for making such a conclusion.

Thanks for raising this point. In principle we agree with the reviewer. However, in all models considered in our study the sub grid processes are parameterized. Even those models that have a higher resolution are only eddy permitting. The comparison of these models shows that there is no improvement with resolution of the OMZ representation as long as all models are non eddy resolving. We agree with the reviewer that regional models eddy resolving resolutions definitely improve the representation of the OMZ. We will clarify this point in the discussion to avoid misunderstanding of our conclusions:

‘Among the models considered here, we confirm the lack of an apparent correlation between model resolution and better representation of the OMZ in the IO, because we cannot establish a relationship between the oxygen clusters and the respective resolution of the models (Tab. 1 & 2). This is contradicting to what regional model studies show, that ventilation of the ASOMZ occurs through mixing processes mainly caused by mesoscale eddies (e.g. Resplandy et al., 2012; Lachkar et al., 2016). An increased horizontal resolution of the model should therefore lead to more mesoscale eddy activity, which allows for more ventilation and thus a change in the ASOMZ. However, we must take into account that all CMIP5 models are far from eddy resolving. As long as sub grid processes are parameterized an increase of the resolution seems to have no impact on the representation of the OMZ. Inclusion of mesoscale processes in the CMIP6 models resulted in moderate improvements in subsurface oxygen representation (Kwiatkowski et al., 2020).’

1. Statement: “There is some evidence that these model flaws are related to a deficient representation of ventilation pathways in models. On this basis, it is hardly possible to say whether the models’ biogeochemistry does have deficiencies that are associated with the oxygen representation”

Comment: OMZ is shaped majorly by respiration and ventilation. The study highlights the difference in model mixing of different water masses in the OMZ. However, arising from the fact that OMZ are located in the world’s major upwelling zones — which depicts the importance of respiration in shaping OMZ — it is very possible that the model’s biogeochemical component is highly (if not equally) responsible for the OMZ volume simulated in the models. Previous studies have addressed the weak representation of biophysical processes in the model, which lays a strong possibility of the deficiencies in the biogeochemical component thus, shaping the model’s OMZ. Most of the burden here is placed on the physical parameterizations, whereas the biogeochemical part is a little under looked. Please justify.

You are right. The OMZ volume simulated in the models depends strongly on the models’ biogeochemistry as well as the representation of circulation.

The focus of this paper, however, is placed on the physical processes. The underlying physical circulation has a large impact on the biogeochemical model components. Deficiencies of the physical circulation model can be compensated by over-tuning the biogeochemical model. This can be illustrated with the given example: If the upwelling strength in the model is deviating from the observations, this would strongly affect the nutrient supply in the AS and thus the phytoplankton growth that influences the respiration.

We have realised that there is an imbalance between the introduction and the discussion in the manuscript related to the consideration of biogeochemical processes and model uncertainties. We will clarify in the introduction that the focus lies on the physical model components. In addition, the discussion will focus more on the fact that the biogeochemical models also have uncertainties that can influence the OMZ representation.

1. If you notice the core region of ASOMZ, then you would find almost all the models largely overestimating primary production in the region. This to me, suggests that large respiration should be occurring in the models. I suggest to check whether respiration is well represented in the models. This will confirm that the problem lies largely with the physical processes or biological processes.

We do not find any confirmation in the literature that almost all the models largely overestimate primary production in the ASOMZ region. Based on the historical data of the CMIP5 models, Bopp et al. (2013) shows that the multi model mean underestimates the NPP in the AS in the upper 600 m. In detail, Roxy et al. (2016) looked at the chlorophyll bloom. According to them, only three out of the here considered 10 models overestimate the chlorophyll bloom in the AS. Two of these models are from cluster HIGH and one belongs to cluster MEDIUM. Thus, we cannot confirm where the uncertainties come from.

1. What clustering method is performed to identify the clusters. Please add some details about the clustering technique in the methods section of the manuscript.

As written in the methods section 3.2 we used the Hierarchical Agglomerative Cluster Analysis that was introduced by Johnson (1967). With this method we clustered the correlation between the vertical profiles in the Arabian Sea for oxygen and for salinity separately. We will add some more details in the revised manuscript:

‘To reduce the high amount of data of the model output and detect similarities between the models and observations we grouped them with the Hierarchical Agglomerative Cluster Analysis (Johnson, 1967). Here, the correlation between the vertical oxygen profiles was used as the distance measure for the clusters. This means that profiles that are more similar to each other than to others are grouped together in a cluster.’

1. There are 10 CMIP5 models used in the study. However, there are ~15-16 ESM models which participated in CMIP5. Please state your choice to choose these 10 models and leave the others.

We chose the 10 models from the CMIP5 models that provided oxygen data for the historical period. We will clarify that in the revised manuscript: “In this study we included all ESMs from the CMIP5 project (Taylor et al., 2012), where output of dissolved oxygen was available. The suit of ten model simulations includes …”

1. What is the reason to choose the oxygen threshold value of 60 mircomol/litre?

Unfortunately this is a misunderstanding. We did not choose an oxygen threshold for the analysis but used averaged oxygen profiles in order to be able to compare the OMZs in a way that is as generally valid as possible. The thresholds that are mentioned in the text are used to make different statements, as the behaviour among the models show systematic differences when accounting a specific threshold. We will clarify this in the revised manuscript.

In addition, we add two vertical lines to Figure 4a for more clarity. With this modification, all thesholds that are mentioned in the text are included in Fig 4a.

1. Use of T-S diagrams to resolve the water mass characteristics are not quite effective near the shelves. In such a case, how reliable are the estimates taken for the RSW and PGW water masses. Can this be the reason for the models showing large deficiencies in the RSW and PGW masses. What is the author’s opinion?

This is a good point. As the formation regions are clearly defined by the geographical location, the T-S properties given here are exactly as simulated by the models. This does not mean that they are close to the observations. Coastal regions and shelf areas are not well resolved in the coarse resolution models. This might be one reason for the models showing deficiencies in the RSW and PGW water masses. We will include this point to the manuscript: …‘A possible reason for this offset in RSW and PGW could be the poor resolution of coastal regions and shelf areas in the coarse resolution models, which includes the shallow marginal seas.’

1. Figure 4a: The authors have shown a vertical line of 50 micromol/litre. Why is this value pointed out when the threshold for hypoxia is considered as 60 in the rest of the study.

As mentioned in point 7 above, our analysis does not rely on a single threshold for oxygen. To avoid this misunderstanding in the revised manuscript, we clarified this in the text, as well as we modified Fig. 4. For details please see Point 7.

1. It is advised to shorten the discussion section slightly. Nonetheless, it can be organized a bit more clearly.

Thank you for your advice, we will revise the discussion section.

1. Summary, Ln 14-15: The authors suggest improved parameterizations of Persian gulf and Red sea water masses in the models. However, I am not confident if improvement in the parameterization of these water mass overflows into the OMZ region would significantly improve the OMZ. Instead, a more local process improvement would be inclusion of eddies into the parameterizations which would affect the ventilation. Please provide strong evidence supporting your solution put forward to the problem under question.

We agree with the reviewer that eddies might have a large impact on the OMZ. This important aspect will be added to the revised manuscript. However, upon the mismatches we find are in addition deficiencies in the representation of the RSW and PGW. To give a complete picture of potential error sources this needs to be mentioned as well.

1. The authors discuss a very important problem using the available model outputs. However, using the model outputs it is very difficult to quantitatively separate the discrepancies in the physical and biogeochemical processes. It would be interesting to have a quantitative estimation of the model discrepancies in between the individual processes using some model experiments. I understand this is beyond the scope of this paper, but this can be mentioned as a future scope of the work undertaken.

Thank you for mentioning this point. This is indeed a future scope and we will mention it in the revised manuscript: …‘Therefore, an important next step would be a quantitative estimate of the model discrepancies between the individual processes based on model sensitivity experiments.’

1. The authors start the discussion section mentioning that all the models underestimate ASOMZ [as seen in Fig4a]. However, if we look at the vertical profile of oxygen in core region of OMZ [fig 5] we see that almost all the models have higher concentrations of oxygen. Is this possible that the models are overestimating oxygen in core region of OMZ? Please clarify.

You are absolutely right. Underestimating the oxygen minimum zone and higher than observed oxygen concentrations in the core region of the OMZ do not contradict each other. Looking at the OMZ volume or expansion as shown in Fig. 4 always needs a predefined oxygen threshold to define the OMZ. An underestimated OMZ thus means that the volume of water containing less oxygen than the threshold is smaller than in observations. Therefore, averaged profiles (Fig. 5) show higher oxygen values.

1. It is advised to include how the water masses are calculated in the methods section.

We apologise that the description of the water mass calculation was not clear. We will rewrite and rephrase section 3.3 ‘Determination of water masses in models’ to make our method comprehensible. We have included a preliminary revision of the chapter:

Red Sea Water and Persian Gulf Water (RSW and PGW) are geographically restricted in their formation regions. Figure 1a shows the formation region for RSW and PGW for which temperature and salinity ranges and mean values are determined (Table 2 and associated new figure).

In contrast to that Indian Central Water (ICW) is not geographically restricted in its formation regions. ICW is a mixed water mass and is characterised by a nearly linear temperature and salinity relation that is density-compensated (Tomczak, 1983) and can be identified in T-S diagrams. With this relation, we were able to define upper and lower temperature and salinity limits of ICW in observations and compared those values to respective values from literature values (see Tab. 2). ICW is formed on zonal oriented fronts in the tropical ocean sub-surface layers (Tomczak, 1983). Sprintall and Tomczak (1993) and Schott and McCreary (2001) described the geographical location of the formation region of ICW. Figure 1b shows the grid boxes where these T-S properties are found in the IO in WOA13 observations. These are in line with the description of the formation region as shown by Sprintall and Tomczak (1993) and Schott and McCreary (2001).nTo investigate the formation region of ICW in the models, we followed the same procedure as previously described for the observations. The linear temperature/salinity relation as given by the T-S diagrams of the individual models (Fig. S4) sets the upper and lower temperature and salinity limits (see also Table 2). Different to the observations and the literature, the resulting locations that determine the formation of the simulated ICW are not restricted to the subduction area of ICW.  For consistency reasons, we therefore limit the formation region of ICW in the models to the subduction area of ICW as prescribed by Sprintall and Tomczac (1993) and Schott and McCreary (2001). We exclude grid boxes with similar T-S properties that are found outside the subduction region as well as those within the upper 200 m so that the oxygen content of subducted ICW is not affected by the well ventilated mixed layer. Figure S2 shows the respective area for each model and the deepest depth at each location, where the T-S properties are found.

Indian Ocean Deep Water (IODW) originates in the Southern Ocean, where it is often referred to as Circumpolar Deep Water and Antarctic Bottom Water, before it travels northward into the deep IO and mixes along its way with the surrounding water masses. IODW is thus defined as the densest water mass in the IO north of 60 °S  that is found below 1500 m depth (Talley et al., 2011).

Fig 1c shows the formation region of IODW derived from observations for which temperature and salinity limits are determined. IODW in the models is defined in the similar way as in observations. In the models the derived formation regions of IODW in the Southern Ocean differ from those we find in observations (Fig. S3).

The oxygen content of the water masses as listed in Table 2 (and shown in the corresponding Figure) is calculated, for each model and the observations, by the arithmetic mean of all grid boxes of the corresponding source waters.

Technical Corrections:

1. Summary, Ln 10: “overestimate oxygen concentrations….” —> I think it should be underestimate.

Overestimate is right in this sentence. We say that the models overestimate the overall oxygen concentration in the Arabian Sea, which is right for our study (Fig. 5) as well as for the CMIP6 models (Seferian et al., 2020).

1. Introduction, Ln 16-17: Provide references.

We will add the reference: ‘The strong influence of the semi-annually changing monsoon winds on the circulation and resulting upwelling and subduction in the AS shapes the OMZ (Schott & McCreary, 2001; Schmidt et al., 2020).

1. Discussion, Ln 16: “Recent studies analyzing….” —> Provide the references.

We will add the reference: ‘Recent studies by Seferian et al. (2020) and Kwiatkowski et al. (2020) analysing CMIP5 and CMIP6 model data show that increasing the horizontal resolution does not overcome the major problems with respect to simulating oxygen in the open ocean.’

1. It is advised to rephrase a few sentences in the discussion section as they are confusing.

We will go through the discussion section again and rephrase sentences that are hard to understand.

1. Summary: Rephrase the first two sentences

We will rephrase the sentences: ‘In this paper we compared 10 ESMs from the CMIP5 historical experiment and analysed their representation of the modelled OMZs in the AS. We systematically grouped the models with a cluster analysis to recognize their similarities. By comparing the representation of water masses and mixing in the models with observations, we identified systematic weaknesses in the ESMs that lead to deficient oxygen concentrations in the AS in the northern IO.’

General comments:

This paper aims to assess the representation of the Arabian Oxygen minimum zone

(ASOMZ) in 10 CMIP5 model historical simulations and relates the error to water mass properties. The topic is interesting and important. However, there are several major issues that should be addressed. The authors stated that none of the selected CMIP5 ESMs reproduces the observed oxygen distribution. It would be interesting to examine these aspects in their upgraded CMIP6 versions to check if they had substantially improved or worsen in representing OMZ and water mass properties.

The water mass properties over the north Indian Ocean region and their implications on ocean biogeochemistry in CMIP models have not been studied extensively. This paper presents important and fresh perspectives through the use clustering and quantification of uncertainties in water mass mixing ratios. However, the paper has not been written clearly. This manuscript was not structured well, especially the introduction and results sections. Introduction needs to be organised. Re-structuring of the manuscript can be done to make it easily readable and highlight the novelty of the study. I would recommend a major revision.

Reply to reviewer #3

We would like to thank the reviewer for taking the time and for providing constructive and very specific comments, which will help to improve the manuscript considerably.

We agree with the reviewer that it would be interesting to examine these aspects also for the upgraded CMIP6 models. However, we do not find it advisable to include the analysis of the CMIP6 models in this study. First of all, for a meaningful discussion of our results the previous work done on CMIP5 models by other scientists was important. This enables us to set our results into perspective and to draw conclusions. However, for the relatively new set of CMIP6 models, such a base is so far not available. Second, the protocols for the new CMIP6 models differ from the older CMIP5 ones. With this, they form a new and independent set of experiments and cannot be treated identically to CMIP5. Therefore, a comprehensive discussion and interpretation of the results is only possible to a very limited extent. However, we are aware of the importance to fully investigate the CMIP6 models, therefore we included all relevant available information (and the respective references) in the discussion.

We will revise the introduction and make sure that the novelty of the study is clear throughout the manuscript.

We have carefully addressed all the comments. The point-by-point responses to the specific comments follow below.

Specific comments:

1. Why did the authors choose 50 threshold to define OMZ? Please clarify in the methodology section.

Unfortunately, this is a misunderstanding. We did not choose an oxygen threshold for the analysis but used averaged oxygen profiles in order to be able to compare the OMZs in a way that is as generally valid as possible. The thresholds that are mentioned in the text are used to make different statements, as the behaviour among the models show systematic differences when accounting a specific threshold. We will clarify this in the revised manuscript.

In the methods section page 5, line 11-13 we explained the choice of this threshold for the plot. To prevent misunderstandings, we will rewrite the sentence: “For a first spatial comparison, we chose our threshold to be 50 μmol l−1 to make it comparable to previous studies on CMIP5 oxygen distribution (e.g. Cabré et al., 2015; Cocco et al., 2013) and looked at the horizontal extension of the OMZ dependent on depth and the actual location of these areas in a map.”

In addition, we add two vertical lines to Figure 4a for more clarity. With this modification, all thesholds that are mentioned in the text are included in Fig 4a.

1. Are there any criteria adopted in selecting the specific ESMs? Are they good at representing the Arabian Sea mean state? Provide references if available.

No, there are no criteria for the selection of the models. We chose the 10 models from the CMIP5 models that provided oxygen data for the historical period. We will clarify that in the revised manuscript: “In this study we included all ESMs from the CMIP5 project (Taylor et al., 2012), where output of dissolved oxygen was available. The suit of ten model simulations includes …”

As we focus on oxygen, we give an overview of the oxygen mean state in these models (Fig. 4) and see that they are not that good in representing it. Other variables and processes connected to the representation of the OMZ in the Arabian Sea that were already analysed for the CMIP5 models were referenced in the discussion.

1. Description of OMZ along west coast of India can be included in the introduction section.

Thank you for your suggestion. We are aware of the coastal OMZ and the complex dynamics right off the west coast of India. However, the resolution of the ESMs used in this study is too coarse, so coastal processes are not fully resolved and the model bias in these areas is expected to be large. We therefore excluded the coastal areas for the determination of the clusters and focus on the open ocean OMZ.

We will briefly discuss this point in the introduction and emphasis the central Arbian Sea as the focus area of this study.

1. The description of mixing ratio coefficients is not clear. Please elaborate. Define in terms of their corresponding water mass.

We will specify the description of the mixing ratio coefficient in the revised manuscript, and we will explicitly mention the corresponding water masses used in this context:

… ‘The three main source water masses in the AS are IODW, RSW/PGW and ICW (Fig. 2). We used a linear mixing approach and restricted the input to physical water mass properties from observational data. By considering potential temperature (θ), salinity (S) and mass conservation this yielded the possibility to resolve the mixing ratio of the three main source water masses in the AS. The set of linear equations was:

θ=αθ(IODW) +βθ(ICW) +γθ(RSW/PGW) (1)  

S = αS(IODW) + βS(ICW) + γS(RSW/PGW) (2)

1=α+β+γ (3)

α, β and γ were the mixing ratio coefficients for IODW, ICW and RSW/PGW, respectively.’

1. Apart from the errors associated with ventilation, it would be interesting to describe the static stability and solubility parameter in these models. Stratification of upper layers associated with warming and weakened surface winds restrict mixing oxygen-rich surface waters to intermediate depths, leading to oxygen depletion. Please clarify.

That is a good point that was mentioned as well by reviewer #2. We will compute oxygen solubilities and analyze corresponding model-data differences and will add these findings to the revised manuscript. We will also compare the static stability in the upper layers and discuss the findings in the revised manuscript.

1. Page 5, line 10: “We chose our threshold to be 50 ”. But a threshold of 60 is referred to state the general underestimation of OMZ volume (e.g.: Abstract section). Please clarify.    

As explained above (see point 1), we did not choose a single oxygen threshold for the analysis. In the discussion we state that “All ten models underestimate the ASOMZ volume when we consider oxygen thresholds of 60 μmol l−1 or higher (Fig. 4a).” This is just the threshold that fits the statement and is not related to the Figures 4b & c.

To avoid further misunderstanding, we also added the two thresholds of 20 and 60 μmol l−1 to Fig. 4a.

1. Page 16, line 5: “…….physical model components show no obvious deficiencies in circulation and mixing”. The analysis presented in this paper is not sufficient to conclude this. Please clarify.

We would like to apologise, as this sentence was misleading. We wanted to say that the physical models show deficiencies, but that these are not large enough to adequately explain the deviations in oxygen. We will rephrase it in the revised manuscript.

Technical corrections:

Page 4, line 20: “……..OMZ between 200 and 1800m”. Provide references.

The concrete depth of the OMZ depends on the oxygen threshold and varies among the models. Thus, there are various depth ranges related to the OMZ. We neglected that fact while writing such a general statement that refers to the observations and the threshold of 50 μmol l−1. We apologize for that and will rewrite this sentence:

‘Averaging also neglects the seasonal cycle. The seasonal oxygen cycle is weak in the upper layers of the AS and not noticeable at greater depth (Schmidt et al., 2020).  Thus, averaging is a reasonable approach for a uniform process analysis over large parts of the water column.’

Page 4, line 25: “…….depth levels ranges from 31 to 63”. Please rewrite this sentence. What are the numbers 31 and 63?

The numbers are the numbers of resolved depth levels in the models. We rewrote the sentence to make that clear: ‘The horizontal resolution ranges from 2° x 2° to 0.4° x 0.4° and the vertical resolution varies between 31 and 63 resolved depth levels.’

Page 5, line 10: “We thus compare the volume of the OMZ for a wide range of thresholds.” Please provide the values.         

We included the values and the new sentence is: “We thus compare the volume of the OMZ for a range of thresholds from 0 to 100 μmol l−1.”

Page 5, line 25: “…….Oxygen profiles in the AS for all models and the observations.” All models or selected ESMs?

With all models we mean all the 10 models considered for this study. We changed the sentence to avoid misunderstandings: “We performed the cluster analysis for oxygen profiles in the AS for all 10 models considered in this study and the observations.”

Page 5, line 30: Is that the area shown in Fig. 4c? The location of the central AS can be better shown on a map.      

Yes, that is the area marked in Figure 4c. We will include the information to the text and reference the Figure accordingly: ‘For this analysis we chose to exclude coastal areas and focus on the open ocean core of the ASOMZ in the central AS between 16 and 22 °N, 61 and 67 °E and from 10 to 1800 m depth and analysed averaged profiles in this region, that is marked in Figure 4c.’

Page 6, line 30: “…… three different source water masses”. Please mention three source water masses.      

We will mention them here: ‘The three main source water masses in the AS are IODW, RSW/PGW and ICW (Fig. 2). We used a linear mixing approach and restricted the input to physical water mass properties from observational data. By considering potential temperature (θ), salinity (S) and mass conservation this yielded the possibility to resolve the mixing ratio of the three main source water masses in the AS.’

Page 7, line 5:” ……IODW, RSW and PGW and ICW”. Please rewrite this sentence. Should it be like………PGW/RSW? Please provide proper references to the methods described to determine the source water masses (Page 7, line 5-15).

We apologize for the deficient description how we determined the source water masses. We will restructure and rewrite parts of that section to make our methods more comprehensible (see also point 4):

… ‘were the mixing ratio coefficients for IODW, ICW and RSW/PGW, respectively. The equations were solved at each data grid point.

To solve the equations the temperature and salinity input of the source water masses has to be determined first. Therefore, we used values from the literature that are based on observations (Table 2, Fig. 3a) and solved the equations for each observational WOA13 data grid point in the box in the ASOMZ (Fig. 3b). Figure 3b shows now mixing ratios for the upper 200 m in the AS. This is due to the limitations of the analysis method: It is not possible to mix the source water masses in a realistic way and get a higher/lower temperature and salinity than the highest/lowest temperature and salinity input of the source water masses. With the input of the described three source water masses this limits our analysis results to the central AS and thus the core region of the ASOMZ, which is of the main interest of this study. In the models the source water properties deviate from the observations. To obtain an uncertainty range of the water mass analysis that can be related to such a change of the input, we solved the equations again for each observational WOA13 data grid point in the box in the ASOMZ. This time we used arithmetic temperature and salinity mean of the WOA13 data in the IO, following the calculations described in section 3.3 for oxygen (Fig. 3c & d). This allows us to draw conclusions on the uncertainties in the mixing in those models.’

Provide references or describe the method to obtain the age of water masses in selected models (Page 9, line 30).

We will include the method how we obtained the age of the water masses in the models: ‘To find out more about the differences between clusters in the oxygen consumption of IODW on the way to the AS, we looked at the age since surface contact of two models. Age since surface contact is an ideal age tracer that is included only in two of the considered ten model data sets. We obtained the age of IODW in the Southern Ocean by the arithmetic mean of all grid boxes of the formation region of the source water mass, similar to the calculation of the oxygen content (section 3.3). In the deep AS the age is calculated by the mean within the averaging box of the profiles (Fig. 5) below 1800 m depth.’