Please can you include a bit more context about this project. This might be included in the Introduction and/or at the start of section 2. I am wondering how the project came about. Did you just happen to find 4 groups that happen to have 2 resolutions of their models lying about? Might there be other groups that could have participated? Are all the models here also contributing to official OMIP-2? How do the resolutions of both the low and high resolution models compare to those used commonly in coupled models, and how they compare to the highest resolution state of the art ocean-only models? It was reviewer 2's comment about the vertical resolutions that brought this to my attention - and your reply that the models were all developed for different purposes. I am wondering what those different purposes are. Also, since you all used the same atmospheric forcing, it seems clear that the model runs were performed especially for this piece of work (so it is not a completely ad-hoc model ensemble). I still don't quite understand, therefore, why the initial conditions vary (or do not use the same protocol). Anyway, it just needs a few sentences of background and context to aid the general GMD reader, or the specialist ocean modeller of the future.

While both reviewers suggested minor revision, one gave only a "fair" mark for reproducibility, but did not leave an indication in their review as to why they marked this aspect down. GMD is very keen on scientific reproducibility, so I dug a bit deeper. While I can see that actually running the models would be difficult, the focus of this paper is the analysis of the results and you have tried quite hard to make model outputs and analysis scripts available. The one thing I noticed that falls below the GMD ideal is that it looks like you are using IDL, a proprietary software package. Can you confirm that your scripts will also run in GDL?

At the early stage of a MIP it is quite common for parts of the protocol to not be defined, or for ad-hoc ensembles to be formed from existing runs or already published runs. This is actually an important thing as it enables experimentation with what protocols work well, which could then be more clearly defined (or not!) in further comparisons. Your MIP is a mixture of all these things. This was explained in your response to me, but you need to add it to the manuscript so that readers not involved in the MIP can also understand the context of the paper.

This is the response I am referring to:
"Ocean-only models are used either to develop the ocean component of a specific climate model or to address specific ocean processes. They are configured using best practices, but each modeling group was empowered to choose what they think is best and that includes initial conditions. The high-resolution experiments are computationally expensive and, when the call for comparison was made, each group leveraged known and proven configurations to perform the requested experiments. Furthermore, some groups had already completed the JRA55-do simulations at high-resolution before this intercomparison was conceived. Given the large computational resources involved, rerunning those experiments to conform to a standard protocol was not an option. "

The points I would like to see in emphasised in the paper at the place where the ensemble is introduced are: that some runs were pre-existing; that modellers were free to choose own boundary conditions; and that the runs are presently very computationally expensive.

Please do switch over to one of the many opensource packages for analysis of model output so that other researchers can get closer to reproducing your published results using the code that you have made the effort to make available. This is not (yet) a hard requirement at GMD, and I appreciate it takes a little bit of effort to make the switch, so on this occasion I am happy to accept the paper as-is.