Dear Author,

Both referees now consider that your manuscript could be published after some minor revisions. I agree with all minor revisions asked by the referees. Please provide a new version of your manuscript taking into account those requests for minor revisions, clearly identifying the modifications you made compared to the current version of the manuscript.
With very best regards,
Sophie Valcke

Dear Author,

Thank you very much for your revisions and update version of your manuscript that answers the remarks done by the referees in the second round of reviews. However, I have the following remarks that I would like you to consider before the publication, in order of importance:
- Figure 1: I don't understand the correspondence between c) and a) and b). Figure 1 a) represents a component with 64 grid points in total decomposed over 8 processes with the first process holding (in the global index space) grid points 0,1,8,9,16,17,24,25, the second process holding grid points 2,3,10,11,18,19,26,27, and so on. Figure c) on the left represents (for me) a component with 16 grids points in total decomposed over 8 processes with the first process S0 holding (in the global index space) grid points 0 and 1, the second process S1 holding grid points 2 and 3, the third process S2 holding grid points 4 and 5, and so on. So a) does not correspond to the left of figure c, and b) does not correspond to the right of figure c. Why do you associate them in Figure 1? Please rectify or explain.
- p.4, l.111-115: I really don't understand the sentence: “Existing global implementations depend on global CLGMTs because the distribution of the CLGMT entries is determined by a model, and thus a coupler generally has to view any distribution as random.” , could you clarify?
- Figure 2 captions: What does “and a regular 1-D parallel decomposition is designed for the data interpolation.” mean? Please clarify.
- p.3 l.81-85: please consider changing this sentence for : “... similar complexities to C-Coupler, even if a compressed global index description is, in the case of regular parallel decompositions, used to reduce the memory and time required to detect common grid cells (the compressed description may not work for irregular, such as round-robin, parallel decompositions)
- p.3, l.76: you assume that both src and dst components use the same number of processes; this is OK but I would then change “Given that ...” by “Assuming that ...”
- In the Abstract, l.19, maybe you should write C-Coupler1 to contrast with C-Coupler2 into which the new DiRong scheme has been implemented, as stated later in the paragraph.
- Figure 3 captions: I think the comma should be removed between “entries” and “larger”

Thank you for considering those remarks,

With best regards,
Sophie Valcke

Dear Author,

Thank you very much for your revised manuscript that fully answer my latest remarks. Please consider the following very minor points for the final version of the manuscript:
- p.4, l.109, please change "a model but not a coupler" for “the model but not the coupler”
- p.3, l.83, please add "the" before "time"

Thank you again and very best regards,
Sophie Valcke