Dear Juliette Mignot and co-authors,

Thank you for your response and revised manuscript.

I am convinced that the paper should be published by GMD, but I require the further minor changes described below, which I will review on resubmission.

Kind regards,

Paul Halloran (Topical Editor)

I concur with the reviewers inference that it is still ambiguous as to whether this paper is focused on a model subsetting method with the Senegalo-Mauritanian upwelling system is a case study, or whether it is a validation paper on the Senegalo-Mauritanian upwelling system performance in CMIP5 which happens to use a certain methodology. It is clear from your response (and it was my initial interpretation) that you consider it to be the former. To make this distinction clearer, I would suggest a more substantial reorganisation of the manuscript, beginning with the discussion about model skill/subsetting starting on P2 L10, then once all of this discussion is out of the way, introduce the upwelling system specifically as a case study.

Your response to Reviewer 1’s comments ‘Give a brief summary of the…’ and ‘More detail is needed on the ERSST_v3b…’ are very light. They do not do as the reviewer asks, which may be appropriate (for example if you are trying to move away from too much focus on the case study, and stick to the technique), but you need to justify this. For example, in your response ‘Yet, sensitivity of the method to the target field should definitely be addressed in the future’ why not here? Please explain why it is not important, or do it.

I am convinced by your response to Rev2’s comments ‘A core problem that needs to be addressed in the paper can be illustrated by considering the section starting on line…’ and ‘I also cannot see an adequate description for what the…’ but please integrate these responses fully into the manuscript.

P3 L3 ‘Nevertheless, these models constitute a fully independent ensemble’.
It is not correct to consider the CMIP models independent in a generic sense.

As was alluded to by one of the reviewers, the language use is on occasion not succinct and precise or contains mistakes. I highlight a small number of examples of this below, but the manuscript needs to go through a comprehensive copy editing process. GMD does offer this: https://publications.copernicus.org/services/copy_editing_for_english.html but please don’t feel obliged to use this service if you co do this yourself or by different means.
. P1 L17 ‘an efficient multi-model’ What does efficient mean in this context?
. P1 L20 ‘combination of 12’ implies that the models within those 12 offers something unique or are ordered in some way. This perhaps plays into the reviewer's view that you were trying to weight models rather than subset models. It maybe that you are suggesting they offer something unique, but my interpretation here is that you mean ‘subset’.
. P2 L19 ‘for a large variety of variables the multi-model average mostly agrees better with observations of present day climate than any single model, and that the average also consistently scores higher in almost all diagnostics’ - the second part of this sentence is redundant.
. P2L30 ‘each other in the averaging procedure used to build the multi-model’ - missing ‘mean’.

Dear Juliette Mignot and co-authors,

Thank you for your revised manuscript. I'm please to accept it, but just spotted two very minor and specific things to address first while reading through the corrections.

1st, the sentence 'The present paper is dedicated to the elaboration of an objective method...' is I'm sure aiming to add the caveats asked for by the reviewer about the fact that any method like this is not completely objective, but I think it could well be read as stating that the method is objective. Perhaps something like the way to describe this in the title, e.g. 'The present paper works towards an objective method...' would be clearer?

Secondly, I noted the typo 'On the contrary, the Senegalo-Mauritanian upwelling presents on the contrary...'.

Please can you address these two items, then we can move to publication.

Kind regards,
Paul Halloran (Topical Editor)