Suggestions for revision or reasons for rejection  I reviewed a previous version of this paper. In general, the authors have addressed my concerns, and I commend them for their efforts in the revision. I have a small number of remaining comments, but after that feel the paper can be accepted for publication in ACP. I am happy to read a future revision if the Editor would like, but I don’t feel that it is a necessity. The comments mostly relate to the presentation of some of the Figures, where what is shown is not clear from what is written.
General: I had suggested adding an evaluation of model Ångström exponent to the paper. The authors respond that the model does not provide this, only the AOD at 550 nm. That’s fine, but I suggest noting this clearly somewhere up front (like at the end of the introduction and/or the start of section 2.3). Otherwise a reader may wonder if you chose to ‘hide’ model Ångström exponent or something.
Page 3 line 20 and page 5, line 25: the authors cite both Holben et al (1998) and Dubovik et al (2002) for the AERONET network here. The Holben paper is the correct reference for the network as a whole. Dubovik presents some climatological results for AERONET’s aerosol inversions, which are not the data set the authors use (they use the direct Sun data). So this is not really relevant. I suggest the authors cite only Holben et al (1998) in these two places.
Page 7, lines 1819: “The Ångström exponent values presented here are based on aerosol extinction.” I suggest deleting this sentence. It seems out of place, and anyway this is the standard definition which was already given in Equation 2.
Figures 3 and 4: I understand now that this is AERONET data, not model simulations. If this is the seasonal cycle and multiannual means and standard deviations, then why are there standard deviations for sites and months where only 1 year contributed? For example in KAUST there are data for MarchJune, the numbers above the points indicate 1 year contributed to each month in the seasonal cycle, but there are large standard deviations. If there is only 1 year contributing then the standard deviation can’t be calculated, by its definition, because you only have 1 data point going into the mean. Is this an error or is this not in fact the mean seasonal cycle? Perhaps it is calculated as a mean of instantaneous data or of daily means instead? In that case the standard deviations would include both day to day and year to year variability (which is not what is implied). Either the plot needs correcting or the caption updating to reflect what is shown. My preference would be to take the AERONET daily product and average it to monthly means. Then, calculate the annual cycle (multiyear mean and yeartoyear standard deviation) based on the monthly mean values. In this case sites like KAUST will end up without standard deviations, for that reason. But the current presentation of the figure is not consistent with what it says. Or have I totally misunderstood?
Figure 5: this appears before Figures 3 and 4 in the manuscript file uploaded (acp2017250manuscriptversion5.pdf).
Figure 6: the same comment about averaging for Figures 3 and 4 may apply here, it is not clear from the description.
Figure 9: it would be good to have the top and bottom rows on the same scale (top is 01.5 now, bottom is 01) to more easily compare the two time periods. Maybe putting both on the range 01.2 is a good compromise?  

Dear Authors,
I have one more referee report with minor comments. Please try to address them prior publication. Please note the comments about the figures, where either the description is incorrect, or the plot itself.
Thank you in advance,
Kostas Tsigaridis 