The discussion of the results has been improved in this revised version and most of the reviewer’s comments have been satisfactorily addressed. However, there are more points that could further improved and are detailed below.

With regard to the reviewer #1 comment on the representativity of the year 2006 for the period 2000-2015. As a reply, I would expect, in addition to the authors reply, discussion on how ‘typical’ were the weather conditions including winds, temperatures, precipitation, as well as the fire emissions.
AOD calculations: It would be good to provide a supplementary table with the information on the optical properties used for different aerosol components.
Page 6, line 3, please provide order of magnitude of the spread among models fund by Koffi et al. (2016).
With regard to the overestimate of sea-salt and dust by GEOS-Chem and GOCART, there are published studies specific to these models and the representation of seasalt and dust with model evaluation, please use them to complement the discussion in page 8 (section 3.1)
Please add the reference of Stohl 2006 in the reference list.
Furthermore for the overall discussion, discussion of budget estimates in the polar regions would be very enlightening.
Page 8,line 22: please change ‘AND’ to small letters
Page 9, line 19 & 20: can you comment on the simulated decreasing trend of AOD? For instance, is this numerical or is it related to the emissions in the models or something else?
Page 10, lines 28-29: can you comment on the nitrate simulations? And put these results in the context of the recently published in ACPD AEROCOM paper on nitrate intercomparison (Bian et al., acp-2017-359, 2017)
Page 12, lines 24 & 25, decreases.
Page 13, line 22: please explain why.

Please re-revise carefully your manuscript based on the above recommendations.

Thank you very much for the additional revisions to your manuscript that are improving its quality. However, I would still miss 'numbers' to the remplies to :
---Page 6, line 3 comment on the findings of Koffi et al - please provide numbers/order of magnitude/ percentage or absolute bias whatever is more convenient for you
---on your reply to the next comment, you mention 'The model bias was reduced by adding a SST-dependent source function' please provide number quantifying the improvement.
---Page 10, lines 28-29 comment - similar to above, please be more quantitative and provide numbers.
with kind regards, Maria Kanakidou