Dear author,

Thank you very much for your revised manuscript and the details with which you answer the two referees' concerns. I consider you globally satisfied their remarks besides few points that I list here and this is why I am asking for an « editor review ». So thank you for addressing the issues below that I think are important before considering publication of the manuscript.
(Please note that when I cite pages and lines, I refer to your gmd-2017-87-author_response-version3.pdf document).

Major comments

1) English : I suspect that the text has not been written by an english native writer. In many places the text is heavy and not easy to read. Please have the english revised by an english native speaker. For example:

-The sentence “The question, if the aggregation of the subgrid-scale information provided by the smaller scale model to the larger scale model constitutes an added value for the larger scale model is still under debate and might be answered with the help of 2-way coupled applications.” could be changed for “The fact that whether or not the aggregation … is still under debate and this question might be answered with the help of 2-way coupled applications”

-p. 19, the sentence “the two slices of the boundary fields, for which COSMO performs a linear time interpolation, are filled with the data of the actual time step” could be changed for “ the two time slices of the boundary fields between which COSMO usually performs a linear time interpolation are now filled with the data of the actual time step


2) Regarding the section on model performances (comments c by referee #1 and R13 by referee #2 R13), you added a new section, but only the computing time for data transformation is shown while a more interesting measure would be the elapse time of the whole coupled system and an analysis on how the two-way coupling impacts this balance with respect to the one-way coupling. Is it possible to have few numbers on comparative elapse time and load balance of the two-way coupled system wrt the one-way coupled system?

3) Following comments by referee #1 and #2 on the meaning of “Ideally” in section 3, I consider that your introduction on grid transformation algorithm is way too general and too long. This whole section should be revised to clearly describe on what GRID really provides and not on what an ideal library would provide.
One specific comment: given your description of grids, I don't understand what is the difference between “curvi-linear” and “non-rectangular strcutured” grids; can you be more precise or give an example (but only if if these grids are relevant for GRID).
Finally, the second sentence of first paragraph of 3.1 “Three different grid types ...” is completely redundant with the introduction of section 3 (lines 20-25)

Minor comments:

1) The modifications done on what is now p.8 regarding the use of “0” and “1” should also be done on what is now p.9 for the use of “0”, “1” and “2”

2) Regarding your Figure 2, you write that you want to add it to the supplement and make a reference to the supplement in the text. First, I don't see where that reference is done. Second, on the figure, I don't see any black line (triangles) so I think it is not appropriate to refer to these in the captions.

3) I think that “optimisation” in” without any resolution dependent optimisation of the emission scheme” could be changed for “tuning”.

4) R3 and A3: Regarding the vertical grid transformations, I think your text is more precise now but you still should mention the inconsistencies linked to the different methods used for the parent-to-child and for the child-to-parent vertical transformations

5) R4 and A4: Regarding the use of INT2LM, I think you should clarify in the text why you did not exchange the horizontal and vertical interpolation routines as you explain in your author response.

6) R15: I still don't understand what is the “patch approach”. You write: The user can specify the number of patches” but patches of what? Please explain better as it is not evident at all for non regional modellers.

7) p.5, you write “GRID can be used for all grid mapping operations ...”; I suppose this is not true as in practice it cannot be used for data going into COSMO.

8) p.10, l.12: change “serveral” for “several”

9) p.12, l.19: please correct sentence “GRID_TRAFO comprises the in EMAC ...”

10) p.13, 1st parag of 3.1.2; please rephrase “It is for the rediscretisation ...”; maybe: “It is used for the rediscretisation ...”

11) p.19, l.6: change “aver” of “over”

Dear author,

Thank you very much for your revised manuscript that answers basically all my concerns. I just have the few minor points that I think you should address before publication.

-Figure S1: I think it would be better to remove the sentences “Black line: initial profile in the parent model” and “Note: the blue line is always on top of the black line” and replace “blue line (circles): initial profile in the child model” by “blue line (circles): initial profile in the child model and in the parent model”.
-p1, l17: correct "specificially"
-p3, l3: change "patches approach" for "patch approach"
-p5, l28: correct "tranfer"
-p11, l23: I don't think ICON can be qualified as a structured grid, cf https://www.dkrz.de/SciVis/hd-cp-2/en-icon_grid.pdf
-Fig 5 & 6: Use lower case to refer to panel; for d) put (a - c)/a * 100.
-Fig 6, I don’t see that data has been masked, as indicated in the captions
-p19, l15: I am not sure "exemplarily" is an English word?

With best regards,
Sophie Valcke