|Review of ‘Global free tropospheric NO2 abundances derived using a cloud slicing technique applied to satellite observations from the Aura Ozone Monitoring Instrument (OMI)’ by S. Choi et al.|
The manuscript by Choi et al. has somewhat improved, and remains in principle an innovative piece of work. Some of my earlier concerns have been addressed and the discussion is more balanced now. However, I cannot understand why the authors persist in their use of the geometrical AMF. Their Figure 4 is the best encouragement to substitute the geometrical AMFs by the near-Lambertian AMFs in the retrieval method. I agree with the authors that the use of the latter will probably result in similar spatial and temporal patterns, but the absolute magnitude of the estimated OMI VMRs would no longer suffer from an avoidable systematic error as they still do now. The end-to-end test in Fig. 4 shows as much. Moreover, the bias with the INTEX-data would be improved when using the near-Lambertian AMFs. The problem I keep on having with this paper is this: while the authors show that the near-Lambertian AMF is a clear improvement over the geometric AMF, they do not make this clear in the text, nor do they follow up on this result by actually showing the near-Lambertian AMFs results in the manuscript (they bury it in the Appendix). Therefore I think they should revise their method and text and present in a revised manuscript a method/dataset that does not suffer from an obviously avoidable systematic measurement error. If for some reason they really need to stick to using geometrical AMFs, they should make unambiguous statements or disclaimers throughout the paper that the absolute values are biased high because of the geometric AMFs, and that using proper AMFs would make a clear improvement to the method and the dataset. For instance, in section 4.1, the authors should also include a statement on the mean OMI VMR when the near-Lambertian AMF is used, and how this reduces the discrepancy with INTEX-B (around lines 391-394). Also, in Figure 7, GMI and OMI above-cloud VMRs would agree better if the near-Lambertian AMF is used.
My other concern is the use of a uniform C-shaped profile, which will not do a good job in the AMF calculation, even when taking the near-Lambertian approach. Especially for situations with outflow or lightning or convective transport of BL air to high altitudes, the profiles will be very different from the single C-shaped profile, and this deserves much more discussion. In the current manuscript this issue is barely touched upon at all, and the answer to my previous comment was not particularly helpful.
P7, L205: To obtain the contribution from lightning to the NO2 column, the authors have done a difference run with lightning switched on and off. When we try something similar with two CTMs we are running, we get results that are very difficult to interpret for the lightning ‘off’ run, possibly because of a lack of ozone production in the high troposphere, and chemical feedbacks on NOx. It would be reassuring if the authors could check whether their model run with lightning ‘off’ is indeed straightforward to interpret.
P8, L262: please clarify what the C1-model is. I don’t think many readers will be familiar with the Diemendjian-paper.
P9, L276-278: please indicate if the cloud fraction was also larger than 0.9 here.
P9, L297: typo ‘near-Lambertia’
P10, L330: typo ‘measureme/nts’
P18, L622-625: what fraction of useful samples is retained when this outlier check is being performed?