This paper presents a modified version of the WFDOAS algorithm to retrieve total ozone columns from the OMPS-NM instrument. The ultimate goal is to combine the nadir and the limb OMPS instruments to derive tropospheric ozone columns. In addition to the realized algorithmic developments, the study also investigates the quality of the retrievals by comparing them with ground-based data as well as with independent OMPS and TROPOMI retrievals.

The topic of this work fits well for AMTD and, from the presented results, the product quality appears to be good. However, I find that the section on the retrieval algorithm description currently lacks of details, justification and illustrations for the different choices that have been made. Details are provided below as major comments. I encourage consolidating the manuscript and to submit a revised version to AMTD.

Major comments:

Retrieval algorithm:

• The modifications made to the retrieval algorithm should be better justified and illustrated. I understand that the lower spectral resolution of OMPS may lead to more important cross-correlation and that there is a need to extent the fitting window. However, why considering one every two wavelengths only, which is expected to reduce the benefit of extending the fit window? Could you illustrate the cross-correlation between fitted parameters for the different options (small and large windows, wavelength selection)? In addition, it seems very strange that taking all wavelengths or the other set of wavelengths has such an impact. How do you justify this? Is there any physical or instrumental reason to do so?
• It would be beneficial to present the product issues with the original WFDOAS algorithm and the impact of all individual changes and/or choices made.
• In addition, it is mentioned before that only the central FOV bins, 17 to 20, are used. Can you really assess the across-track variability using only 4 measurements?
• It is also unclear what is the additional fitted parameter to account for the slope of the ozone absorption signature? Is the polynomial of first order and not a constant? This parameter is not mentioned any longer in the list you give at line 136. Please clarify this.

O3 profile climatology:

• While the total column dimension may reliably account for the O3 variability in the stratosphere, this is not the case in the troposphere. There may be significant longitudinal and time variability in the tropospheric ozone content, which is not covered by the climatology as currently built. For example, in Tropics, there is a significant wave-one pattern in the tropospheric columns with an amplitude of up to 20DU. This should be acknowledged and further discussed in the paper. What are the consequences of this limitation on the retrieved total ozone columns?

Validation:

• Why do you limit the validation to the period 2016-2018? In section 2, you mention that data from 2012 to 2018 have been processed. It would be beneficial to extend the ground-based validation to the full available period to better evaluate the product stability.
• The added-value of Fig. 8 is limited, especially for differences seen on the orbit edges where sampling differences likely dominate. I encourage making similar plots based on a larger amount of data. The latitudinal dependence aspect is already covered by Figure 9. In Figure 10, the discussion would be easier to follow if you would show similar time series of the relative differences in addition to the current panels. As for the ground-based validation, extending the comparison with the operational OMPS product for the full period 2012-2018 would be beneficial.

Minor comments

• p. 2 Line 28: Chiou et al. does not include any DOAS algorithm, but relies on the GTO/Direct-fitting CCI (not DOAS-based) algorithm. The latter is not cited at all while applied to most of the sensors mentioned here. Please correct this and add proper references for DOAS algorithms for each of the sensors.
• p. 2 Line 41: please add a reference for GOME-2.
• p. 2 Line 56: replace as “a three-part instrument, namely a nadir mapper (OMPS-NM), a nadir profiler (OMPS-NP) and a limb profiler (OMPSLP), collecting data since January 2012.”
• p. 3 Line 66: As the limb profiler is not described, it is not clear to the reader what are the implications to have only data “from the central of the three vertical slits”. What does it mean in terms of coverage/spatial resolution?
• p. 5 Lines 110-115: Please better specify what temperature parameter you fit exactly since I do not think you fit a T° profile but most likely a single parameter.
• p. 8 Line 192 : 3.5 km x 5.5 km since August 2019.
• p. 8 Line 192: Unclear sentence. OFFL and RPRO are produced similarly and both include a cloud correction. Heue et al., 2016 is not an appropriate reference for LIDORT I believe.
• p. 10 Line 234: “De Bilt” instead of “Debilt »

Review comment on

Andrea Orfanoz-Cheuquelaf, Alexei Rozanov, Mark Weber, Carlo Arosio, Annette Ladstätter-Weißenmayer, and John P. Burrows

Total ozone column retrieval from OMPS-NM measurements

The paper describes a total ozone Ozone retrieval form OMPS-NM data using a modified DOAS approach.

As preparational work a new ozone climatology has been generated. Which might be interesting in itself, is this available for other user? Has it been compared to existing data sets?

The algorithm is based on the Weighting Function Differential Optical Absorption Spectroscopy algorithm (WFDOAS), that is adapted to the OMPS-NM. However OMPS-nm has a spectral resolution of 1 nm and a spectral sampling of 0.42 nm, hence the slit function is represented by ~2.2 measurements points, for a classical DOAS analysis this might cause an undersampling issue. The authors solve this issue by skipping half of the spectral points (using only the odd spectral channel) - which gives reasonable results compared to other observations, however no real explanation is given why the approach is working. Moreover, when the other half of the data is used the comparison shows stronger deviation.

The comparison showed a good agreement with operational OMPS and TROMPOMI data sets, as well as with ground based Brewer and Dobson measurements.

general comment

The analysis is applied to roughly 44 data points (316-336nm) where only half of the data is used. The algorithm is described to become unstable if the complete data range is used. Is its possible that the major deviation are caused by just a few data points? To check this possibility I suggest to run the analysis for one orbit skipping one even data point after the other. A combination might also be possible but this might easily end up in larger study.

minor comments:

5.3 S5P/TROPOMI

L 197 This reference is about a tropospheric ozone retrieval but in this context it seems to be a reference on the RTM LIDORT.

L 207 why gridding data from two algorithms applied to TROPOPMI spectra for the comparison? Both resulting VCDs have identical coverage. So a direct mapping seems easier.

6 Validation

L 218 For the comparison of OMPS with TROPOMI the data are again gridded, this probably can not be avoided. But I suggest to use only one gridded TROPOMI data set here.

Figure 5: The seasonal map shows a strong orbital pattern, which seems surprising when 4 years of data were averaged.

6.2 Comparison with OMPS-NM operational product and S5P/TROPOMI

For the OMPS-NM data set only the central field of view was used in the comparison 150km, while for TROPOMI the complete swath was taken into account ~2600 km. I suppose the comparison will improve if also for TROPOMI only the central pixels (~210 to 240) are used.

technical comments

Figure 2: [VMR] stands for volume mixing ratio and is hence not a correct unit, please change to [ppm]

Figure 4: in Figure 3 a positive bias between S5P/TROPOMI WFDOAS relative to the ground based observations is shown, here (in figure4) it seems there is a negative bias of the Operational OFFL data relative to the WFDOAS. I suggest showing WFDOAS - OFFL instead, to have more consistent figures.

Figure4: The minus sign at (-10) has disappeared from scale.

General Comments

The authors have implemented a total ozone algorithm for the Ozone Mapping and Profiler Suite (OMPS) Nadir Mapper (NM) based on a technique they call the Weighting Function Fitting Approach (WFFA). The purpose of the algorithm is to estimate total ozone from OMPS NM coincident with OMPS LP stratospheric column ozone, and compute the tropospheric column by taking their difference. As such, the algorithm has been applied only near nadir and in cloud free conditions. The authors explain the WFDOAS approach is not the optimal to retrieve total ozone (TO) from OMPS NM due to the relatively low spectral resolution of the instrument which negatively impacts analysis of the differential spectrum. In the WFFA method the primary algorithmic changes increase the width of the fitting window and reduce the low order polynomial to a constant term. With these modifications, the WFFA method fits the spectral slope of the ozone absorption in addition to higher order structure. The first half of the paper describes the algorithm. The second half details validation of the total ozone retrievals using other datasets that are well known in the field. The total ozone results presented in the second half of the paper look quite good and therefore I feel this algorithm is very promising. However I have some questions about the algorithm and how it has been presented. I recommend the authors address the following to significantly strengthen the paper.

Specific Comments



1. The authors have made an unusual accommodation to get good results from their OMPS NM retrievals - they have achieved their results using only alternating pixels in the OMPS NM spectra. Oddly, this works when odd-numbered pixels are used in the retrieval. When even-numbered or all pixels are used, the results are unstable from one retrieval to the next and a bias is  observed. I think it is important in this paper to provide further information on this instability. The issue raises questions about the performance of the algorithm, the instrument, or both. Do the authors have an idea why the WFFA method gives reasonable results in the one specific case? Perhaps the fitting of the spectral slope to determine TO is affected by end-point sensitivity of the fit? The authors use a fitting window of 316-336 nm in their algorithm. Does an adjustment of this window to include/exclude 1-2 spectral points at the window edges produce a more stable retrieval when all spectral pixels are used? Are there any particular spectral features at the edges of the fitting window that complicate a reliable spectral slope determination that might show as a noticeable pattern in fitting residuals? If the authors think the issue is related to quality of OMPS NM spectra, this should be stated. It is worth nothing that colleagues at BIRA have successfully retrieved total ozone from OMPS NM with the GODIFT v4 algorithm to produce data consistent with the GTO-ECV record. I am aware of no similar issues with processing OMPS NM data.

2. The explanation of the insensitivity of the WFFA algorithm to absorbing aerosols and other broadband contributions should be explained better. The WFFA approach fits the spectral slope to estimate the ozone absorption signal, but several other geophysical effects may also affect spectral slope. The authors assume an aerosol-free atmosphere in their forward model and retrieve an effective scene albedo at 377 nm using the LER approach, so albedo wavelength is 40 - 60 nm from the edges of the fitting window region. Absorbing absorbing aerosols can produce several percent in spectral dependence in the radiance signal in this spectral region. The authors state the aerosol effect is largely accounted by the effective scene albedo, but I feel given the nature of the algorithm this may be an oversimplication. How can we be better assured of this? It is true that results shown later there are no significant  ozone anomalies in regions of high aerosol load. But I cannot explain why.  The WFFA algorithm may well be as insensitive as authors claim, but it would be useful for the reader to know the reason(s), and clarify circumstances where residual error may grow to be significant. Absorbing aerosols are common in the tropical regions and these are regions where tropospheric ozone is of particular interest.  Since tropospheric column is a relatively small fraction of the total column, small errors for TO can be non-negligible for tropospheric ozone determination. 

3. The sensitivity of the algorithm to tropospheric ozone is not discussed in the paper. This should be addressed in some fashion given the goal of the algorithm.

4. Some discussion of algorithm uncertainty and sources of error would strengthen the paper considerably. 

5. It is unclear why S5P/TROPOMI results from different satellite algorithms are compared. How do these comparisons relate to the OMPS-NM WFFA TO algorithm in the present manuscript?

6. The title is very general. A more specific title will help readers distinguish this work from that of others.

Techincal comments:



Line 2: its -> the

14: delete "characterizes the stratosphere. In turn,"

15: remove "On the other hand,"

19: remove "Among others," 

20: "1970's, have provided"

23: specify Suomi NPP OMPS

22: change 1994 to 2005

24: named -> known

24: (all) -> (all instruments)

29: giving -> which is useful to establish 

32: this -> that

57: sensor (no s)

58: radiation instead of radiances?

64: 150 km wide swath

95: "linearly"

101: remove comma

Eqn. 1: is this C  or C_i? 

113, 137: same comment as for Eqn. 1

118: please revise statement in light specific comments above

127,128: readouts -> pixels

140: define RTM

149: this first sentence seems out of place; can safely remove.

174: may be V8.6

203: "from the" -> "reported with"

212: Is this the IGACO3 recommendation?

221: Fig. 5 shows ozone lower over Antarctica than tropics during SON.

250: "OMPS-L2" does not indicate a specific product. Please clarify which product.

284-286: cloud contaminated scenes would generally have low bias, not high

291: define TOCS

305: th -> the

327-328: more should be said to justify this statement. What are requirements for retrieving tropospheric ozone from the limb-nadir matching technique?

Fig. 5: striping in these TO maps seems large for a 150 km wide swath. 

Fig. 10: cannot find a reference to this figure in the text. Please define what the shaded areas represent. What is the difference between the grey and the very light green shaded areas?

* Minor editing note: the use of plural nouns is not needed in a number of places.