Dear Dr Evangeliou and colleagues,
Thank you for your patience; we have struggled to find referees with time to provide a review, but I am glad that we have in the end got two informative reviews. I am pleased to accept your paper for ACP subject to the revisions outlined below, which derive from reviewers comments that I don't think you have quite fully satisfied yet,
Sincerely,
Rob MacKenzie

Abstract and elsewhere: to avoid the potential for misunderstandings, I suggest reporting the modulus of the FB accompanied by the word ‘underestimate’. You should report the RMSE results as well as FB in the abstract.
Lines 50-52: I don’t think you can use the terms “quite accurately” and “small discrepancies” when you have already reported |FB|>100% a few lines above. Please find a more specific summary of the model evaluation.
Lines 52-54. I don’t see how this final sentence follows from what went before. I suggest you state that your independent datasets were 50% lower than your own measurements at line 50.
Line 80 ppb here is presumably mass/mass (ng/g). Better to be explicit, I think, so as not to hold up the reader.
In the discussion on lines 120-124, and in Figure 1, please refer the reader to Table S1.
For Figure 1c, please refer to Silverstein et al. (2008) and try to find a better colour map.
Line 180. Do you really mean that trajectories were released at 20km altitude? This is well inside the polar stratosphere and so very unlikely to contribute to wet deposition at the trajectory endpoints.
Line 217 or thereabouts. Please insert a short subsection prefacing your approach to the statistics you will use to report measurements and the statistics you will use to compare model and measurements. You can move the material around line 260 to here to consolidate. This way you will set out your approach ahead of reporting the results. Readers may not agree completely with your approach, but at least they will know what it is and get used to it. In the current version, queries about the statistical approach get in the way of considering the scientific implications of the actual results. I would encourage you to draw the reader’s attention to figures such as S1 and S3 in this discussion, because these show clearly that summary statistics of central tendency and RMSE need to be interpreted with care. Figure S3 particularly shows the well-known property of the correlation coefficient that it can report a high value (goodness of fit) for an observation vs model scattergram with gradient very different from unity.
Line 223 and elsewhere. Please don’t mix resistant and robust statistical measures (percentiles) with non-resistant and non-robust measures (range, mean, and standard deviation). Report mean together with standard deviation, and median together with quartiles. It is perfectly possible to estimate percentiles for a sample of 10 or 11 – see a standard statistics textbook, or http://www.itl.nist.gov/div898/handbook/prc/section2/prc262.htm for a quick hint.
Lines 268-270. The discussion of FB gets very convoluted here, with positive and negative numbers in close proximity. I realise it’s awkward, but please look at simplifying, perhaps by talking about the modulus of FB.
Lines 296-300. Please report medians and percentiles from the literature, rather than just upper limits.
Line 392. Remove double full stop.
Line 402. Representing the Doherty et al. (2010) data as a mean with standard deviation leads you to reporting a standard deviation bigger than the mean. For data that is positive definite this simply illustrates the difficulties of working with assumed Gaussian statistics on highly skew data. I suggest you replace with median and interquartile range.
Lines 414-415. I would suggest replacing “no safe conclusions” with a statement that it is not clear whether the discrepancy arises as a measurement artefact (even though every effort has been taken in both papers to follow a robust protocol) or from real spatio-temporal variation.
Reference
Silverstein, J., Parsad, N., and Tsirline, V. (2008). Automatic perceptual color map generation for realistic volume visualization. Journal of Biomedical Informatics, 41(6):927.

Dear Nikolaos Evangeliou and colleagues,
Thank you for your prompt and thorough response to the minor revisions. I am now pleased to accept your manuscript for publication in ACP. I apologise once again for the extended review process, which is, I'm afraid, indicative of the pressures we are all under resulting in difficulties getting reviews,
Sincerely,
Rob
Ps. My comment to refer to Silverstein was poorly worded - sorry! - I meant have a look at that discussion of colour map perception and consider which colour map to use - the default colour maps in commonly-used plotting programmes have quite well known deficiencies!