Dear authors,

First of all, I would like to acknowledge the two reviewers for their constructive feedback on your original manuscript. I would also like to thank you for addressing their comments and for submitting a revised manuscript.

On reading your responses to the reviewers, I feel that you have largely addressed their comments. However, I have the following comments that I would like you to clarify. Once these are addressed, I will be very pleased to recommend that your manuscript be accepted for publication in GMD.

Regards,
Fiona O’Connor

------

Comments
1. Reviewer 1 - My primary criticism of this paper is the lack of detailed description of the GCM model, the downscaling, regional model configuration, and execution. Even though references are given for the CESM modeling and the chemical and aerosol processes are briefly described I would like to see further description of the CESM physics, spin-up, constraints, etc. I do not understand how this represents a climate scenario when it is for past years and is evaluated against observations. What does RCP4.5 for these years represent. Do these runs use observation based SSTs? If these runs were spun-up from pre-industrial times without any observed data constraints, there would be no reason to expect agreement with observations. If bias corrections are made to both the meteorology and chemistry, then how do these runs substantially differ from re-analysis driven runs? Please explain the rationale and expectations of these runs.

There are essentially 2 questions here. The first is related to the model and downscaling descriptions and the second seems to arise from a confusion about the use of RCP4.5 to model the time period 2006-2010. You’ve very clearly set out the rationale for using RCP4.5 although I ask for some further clarification on a related comment by the same reviewer - see below.

In answer to the first question, I’m a firm believer that a paper should stand alone in its own right and that the reader shouldn’t be expected to look for basic information in other papers. However, I think there is also a balance to be struck between the level of detail that one puts in a paper and the extent to which you cite the pre-existing literature. Having said that, I think your response to this comment is appropriate, adding in more detail on the GCM, downscaling etc. (Section 2.2. and Table S1). However, I have two minor comments related to the newly added text that I would like you to address:
(i) Why are the ice and the ocean models initialised using CESM default settings rather than using output from an appropriate simulation of CESM run for CMIP5?
(ii) There’s a bit too much repetition in the added text in Section 2.2, particularly lines 22-24 i.e. The CESM simulation … fully-coupled CESM .. with CESM configuration… fully-coupled CESM …

2. Reviewer 1 - P4lns21-22: This statement about “correcting the roughness length by increasing the friction velocity by 1.5 times when calculating wind speeds in the ACM2 PBL scheme to reduce the overpredictions of wind speeds” needs more explanation. First, if the roughness lengths need correcting why not change them and not the friction velocity. Second, what is the problem with roughness lengths? How are they specified and what are they? Our experience has not shown general overpredictions in windspeed. Windspeed and friction velocity are strongly affected by the LSM and surface layer scheme which are not even mentioned here. Also the LU scheme and data are important. The USGS 24cat data is way out of date especially for China where urbanization has been dramatic. Why not use MODIS LU?

I thought that you addressed these comments very well but I would like to see you use more of your justification for scaling the friction velocity included in Section 2.1.

3. Reviewer 1 - Page 6: I don’t understand what is the point of using RCP projections when modeling retrospectively. It seems that 2008 emission inventories are used for more detailed spatial-temporal allocation. Then why not just use these inventories? What is an RCP projection for past years? Please explain the logic here

Can you clearly add somewhere that the CMIP5 historical emissions (Lamarque et al., ACP, 2010) only covered up to 2005 and that the RCP emissions covered the period 2005-2100. Then it will be clearer why you are using RCP4.5 for those sectors not included in MIX2008.

4. Reviewer 1 - P8ln28: Are the results shown in Fig5 averages for all 5 years?

Can you be explicit about this in the figure caption i.e. multi-annual annual means of CF,AOD etc.. for the time period 2006-2010

5. Reviewer 2 - (3) In general, the discussion of model output against meteorological and compositional variations is (i) vague, i.e. non-quantitative (using phrasing like agreed well, satisfactory, etc.), and (ii) lacking explanatory insight, i.e. lists of potential reasons for discrepancy are given which could be written down as potential explanations without needing to do these comparisons. The authors should endeavour to provide more quantitative assessments of model performance, including how their mod-obs statistics compare with expectation and with other studies, and also to provide some more informed analysis of what is the driving explanation for mod-obs discrepancies for particular variables or circumstances.

I think you’ve addressed this quite well, by adding in the commonly used root mean square error as part of your statistical assessment as well as placing the WRF-CMAQ results in the context of other model results. Can you correct line 10 on page 12 i.e. GEOS-Chem and CESM tend to under-predict PM2.5 concentrations (by about -50%...). If the term “under-predict” is used, then there is no need to use -50% … it should be 50%