Review of Pukite et al. submitted to AMTD:

Retrieval algorithm for OClO from TROPOMI by Differential Optical Absorption Spectroscopy

The authors present a new algorithm to retrieve OClO from measurements of the TROPOMI satellite instrument. Overall, the paper presents valuable information and recommended DOAS settings for OClO retrievals from space that deserves to be archived.  I believe the paper should be published in AMT. The authors present convincing comparison between TROPOMI and ground-based OClO SCDs, which clearly demonstrate the value of the satellite dataset. The authors present a number of tests and new settings. Some of these are new and make a lot of sense. However, the impression is also that many complication is introduced that is not really necessary. What is missing is a summary of the retrieval settings and choices that has the largest impact on the accuracy. The English is not very good and could be improved.

-Retrieval settings: Section 2.1

-Calibration: It is unclear whether it is performed on solar measurements or on mean radiance spectra. In the latter case, atmospheric SCDs should be fitted as part of the calibration step? The description in Appending A1.1-2 is unclear. I don’t understand what is tau_i and I_i and what the proposed weighting (section A1.2) is supposed to solve.

-The sentence on l106 “The effect for this application is however negligible” is strange. Why introducing something in the text which has no effect?

-the description of the Ring effect is unclear. It is explained that Ring spectra are calculated at 2 Temperatures from the reference spectrum. Do you mean the reference for SZA 60-65°? If yes, I don’t understand how it is calculated. Are the 4 Ring cross-sections fitted?

-Section 2.2

To help the reader, I suggest to add directly in Fig1 the indication of which days are expected to have enhanced OClO or not. In Fig1 left, the marker “x10” should be “divided by 10”? 

-Figure 3: it would make sense to show the standard error also or instead of the std (which is already shown in Fig1).

-Section 3:

-the ground-based data are not analyzed using the same settings as used for TROPOMI. It is not fully clear to what extend and how this can explain the observed differences. E.g. in Fig6 left, there is a clear offset of ~1e13 cm-2 between TROPOMI and ground-based data for low SZA. Is this related to different DOAS settings, sampling bias, other? Please discuss this in the text.

-Interestingly at Neumayer, the scatter of the SCD differences is much higher than for Kiruna. Is it because the SCDs range is larger? Or is there an instrumental related difference? Or something else? Please elaborate.

-Section 4: it would be good to understand what is the dominating factor explaining the offset between the 2 OClO data sets. I imagine it is probably related to the use of irradiance as reference spectrum and it is likely the largest source of error of the retrievals.

-The Appendix B is hard to digest. I suggest to add a summary table (extending Table B1) in the main text with typical errors on the SCDs coming from the main sensitivity tests so that the reader can have a rapid idea of what matters and what not. How the errors from the sensitivity studies are relevant compared to the typical OClO values and the differences from the validation exercise?

-Sensitivity studies 5, 6, 13,14,15 have very little impact on the results. Consequently, one could argue that the related settings introduced are not really necessary. E.g. the mean of normalized earthshine spectra, the offset correction quadratic term could be optional.

Minor comments

-Abstract: the first 10 lines are too generic for an abstract and should belong to the introduction section. “OClO” is defined twice in the abstract.

- “so called” -> “so-called”

-wording such as “Last but not least” should be avoided.

-lines 315-316: a reference to a next section (Sect. ??) is erroneously made. Please remove.

General Comments:

In this work, a new retrieval algorithm of the slant column densities (SCDs) of OClO is proposed. This algorithm, aimed to be applied to TROPOMI DOAS measurements, takes into account different spectral effects not considered in previous retrievals. A corresponding error analysis of some retrieval settings has been performed. The authors also present a comparison between OClO SCDs, obtained by TROPOMI trough the new algorithm, and from ground-based zenith DOAS measurements at Kiruna and Neumayer stations. The results show a very good agreement with these instruments (especially at Kiruna). The SCDs of OClO obtained in this work have been also compared to preliminary S5p+I OClO products during different periods of the year, showing similar SCD evolution but presenting an offset between both datasets.

All the manuscript (text and figures) is clearly presented. The new concepts and settings introduced in this new algorithm are exhaustively explained (appendixes), as well as the corresponding error analysis and sensitivity studies. I think that the results exposed in this work will be useful for the treatment and analysis of the OClO SCDs obtained by TROPOMI. Thus, I think the the paper should be publish in AMT. However, I think that some questions should be better clarified.

My main concern is that the authors of this work claim that the new algorithm improves the retrieval results, but looking at the comparison between SP5+I and the results of this work, can that be really stated? The new algorithm takes into account several fine effects that, in principle, should improve the OClO SCDs retrieval and decrease the corresponding errors. But, in practise, how can we say that the results of these work are better than those of SP5+I? For high SCDs, results are very similar, and for low SCDs the offset between both datasets cannot be explained. It is true that, as the authors explained, OClO observation is not expected when the temperatures are still warm, as it is observed in the results of this study (Figure 10). Contrarily, SP5+I results show a background level of OClO. But, it can be affirmed that the results of this work are better than those provide by the SP5+I? Did the authors compared the results of both algorithms with independent measurements (as those of Kiruna or Neumayer)? Please, explain better.

Specific Comments:

I would like also to clarify some questions:

• Figure 1: Some days of different periods of the year for both, NH and SH, have been presented. Are those days representative of the corresponding periods? If it is the case, do you think that the bias introduced by SZA at each season could be, at least partially, corrected somehow?

• Page 11, line 225: why for clear sky cases the signal to noise would be lower?

Technical Corrections:

• Page 4, line 102: “Earthshine” instead “Earth-shine”, for coherence.
• Page 6, lines 123-124: A wavelength l=379 nm is selected for evaluation. Please, explain briefly why.
• Legend of Figure 1: “.. indicated in the legend on the right.”, instead “left plot”.
• Page 7, line 161: “In an ideal case,..”
• Page 17, line 316: “(Sect. ??)”
• Page 25, line 476: “depends on both”, instead “depends both on”.
• Page 25, line 489: “Within the chosen OClO fit window,..”.
• Page 25, line 490: “cross section” instead “cross-section”, for coherence.

Review of 'Retrieval algorithm for OClO from TROPOMI by Differential Optical Absorption Spectroscopy' by Pukite et al. submitted to AMT

This manuscript presents a retrieval algorithm for OClO slant columns from TROPOMI measurements using the DOAS technique. To improve the accuracy of the retrieved data, the authors introduce additional fit parameters accounting for spectral effects which have previously not been accounted for and they provide a discussion of the uncertainty estimates including a novel application of an autocorrection analysis.

The authors show that their retrieval of TROPOMI OClO slant columns is in good agreement with ground-based zenith sky measurements made at two polar stations. They also compare their TROPOMI product with preliminary data retrieved with the operational TROPOMI OClO retrieval algorithm and discuss the observed differences.

The study is clearly presented in the manuscript, and in addition, the authors also provide substantial material describing relevant retrieval concepts and settings (Appendix A) and an extensive sensitivity study investigating the effect of the different retrieval settings on the OClO slant column data in comparison to a standard scenario (Appendix B). The paper is recommended for publication in AMT.

General comments:

While some of the aspects included in the uncertainty analysis of the retrieval include a novel approach (the application of autocorrelation for the systematic error analysis), stating that this is overall a new retrieval algorithm seems to me somewhat exaggerated since my understanding based on the manuscript is that the difference to existing algorithms is mainly that additional fit parameters have been used. If that is not correct, and the algorithm is indeed novel then please describe this clearer in the text.    

Also, in the conclusions, the authors state that ‘the detection limit is similar to the detection limits of earlier instruments’ – i.e. that this has not really improved – but then, also in the conclusions, they state that ‘Including these terms improves the retrieval results especially for low OClO SCDs’. Aren’t these 2 statements contradicting each other?

Specific (minor) comments:

Page 1, line 5: Should read ‘From the measured spectra, highly resolved …’

Page 1, line 11: Just use OClO since this has been already introduced in the paragraph above.

Page 1, line 15: ‘… effects, a higher order …’

Page 1, line 21: typo: ‘operational’

Page 10, line 125: left bracket is missing

Page 12, line 250: ‘… zenith sky …’

Page 12, line 252: ‘… in a fit window of …’

Page 12, line 252: replace ‘considered’ with either ‘included’ or ‘used’

Page 17, line 306: ‘.. are listed: The retrieval …’

Page 17, line 309: ‘.. terms are applied (or used).’

Page 17, line 311: ‘… within the 89 ….’

Page 17, line 316: ‘(Sect. ??)’ needs to be fixed

Page 17, lines 318-319: ‘… with the correlation … has an offset …’

Most pages have sentences where commas are missing but I assume that this will be addressed anyway during the proof-reading phase.