Dear Dr. Hioki,

Thank you for submitting your manuscript entitled "Degree of ice particle surface roughness inferred from polarimetric observations" to ACP. I have received 2 reviews of the revised manuscript, which can be found in the MS record. While reviewer 1 is happy with your response and revision (but please include your response to “aerosol above cloud” in the revision), reviewer 2 feels that his/her comments are not fully addressed.

I think that the proposed method is exciting and I appreciate this opportunity to publish this manuscript in ACP. However, I agree with reviewer 2 that we need to understand a bit more about why/when the proposed method doesn’t lead to successful retrievals. If we look at Figure 13, the proportion of “unphysical” roughness values is very large, suggesting that perhaps additional constraints should have been incorporated in the proposed method. This paper would be very useful for the community if more detailed discussions and directions for improvement are provided.

Specifically, regarding your response to reviewer 2:

1) Comment 1: I would think this could be answered by looking some cases and perhaps design some experiments to test where the inconsistency comes from.

2) Comment 2: I think that reviewer 2’s statement is fair and I don’t quite understand your explanations. Could you please elaborate on this?

3) Comment 4: I am bothered by the extrapolations of the linear regression relationship between EOF1 score and roughness, because when a regression is performed, the relationship should be valid only in the range used in the regression process. As pointed out by reviewer 2, we don’t really know if this relationship remains the same outside the range. However, I agree with you that it seems less useful if any roughness outside the range is unphysical anyway. If the wrong choice of ice particle model could be the main reason for the unphysical EOF1 score/roughness, could you please at least show one example, demonstrating how sensitive EOF1 score is to the choice of the ice particle model? Since it may be inevitable to apply an inappropriate ice particle model for actual observations, what would be appropriate criteria to remove unphysical retrieval? In short, a couple of paragraphs that detail when/what factors will lead to unphysical EOF1 score/roughness and how sensitive retrievals are to those factors will be helpful for readers to apply the proposed method and to understand what those retrievals represent.

4) Comment 5–6: More detailed analysis and supporting evidence are needed.

If you need more time for revision, please feel free to let us know.

I look forward to receiving your revised manuscript.

Sincerely,

Christine

Dear Souichiro et al.,

Thanks very much for submitting the revised manuscript. I am generally happy with your revision.

I am not sure if you can read Referee #1's comments that were submitted on 31 Mar 2016, suggesting that you repeat your response to aerosol above cloud in the revision. Could you please incorporate your response into the manuscript? I understand that you have mentioned it in the last paragraph of Section 3.2, but I thought information like 22%of pixels that are possibly contaminated is useful to be included.

Thanks again for submitting your manuscript to ACP. I hope you can figure out why you got so many unphysical roughness retrievals soon.

Kind Regards,
Christine