This paper describes the coupling framework for the Úa ice sheet and FESOM ice-ocean models. This coupling is particularly interesting because both models use unstructured meshes and mesh refinement to resolve better local processes in otherwise global or regional setups. The paper is well-structured and written, and I expect it to be of substantial interest to GMD readers. I find the paper quite mature, and I recommend publishing it once the authors have addressed my suggestions and minor concerns.

Sequential coupling: You use a sequential coupling, which essentially doubles the runtime of the whole system, at least at first guess. Yet, there is no discussion of why you don't use a parallel approach. You need to give at least the reason or justification for why you went with the sequential approach. An estimate of the performance benefits of using a parallel coupling would also benefit the paper.

I don't like your names for sections 2.4 and 2.5, as they could be more descriptive; please consider giving these more descriptive names.

L116: The description of the background mesh left me hanging. How do you decide what regions "could possibly unground during the simulation period"? You describe this later, but more details are in order here. Even saying that it's simulation-specific and will be described in more detail later would suffice (if this is the case).

In section names for sections 3 and 3.3, you use "verification" when "evaluation" would be more fitting. You can only verify the model results if you have an analytical solution.

In section 3.1, I would appreciate more justification for choosing the IceOcean1ra experiment. I expect that there are several idealised experiments in the MISOMIP framework, so you should say why you chose this particular one.

L160: "... with the later described pan-Arctic setup" should be "... with the pan-Arctic setup described later".

L161: "summaries" should be "summarised".

In section 3.3, I miss a reference to what results other modelling groups get for the IceOcean1ra setup. Are your results similar to those of others? Is there a large spread in the results in general?

In section 4.4, you say that you use three different machines to run the model. Such a setup is very unusual, and it would be nice to have more details. Is everything automated, or is there some manual work involved? Do all the machines have access to a shared storage area, or do you need to copy data between machines? Why did you choose this setup?

You only mention Greenland in the "summary and conclusions" sections. You should either remove this or also mention it in the main text. As it is, it comes completely out of nowhere.