I find that this is a carefully crafted paper, which I appreciate.  I can tell the authors put a lot of effort into error assessment and to describing their algorithm improvements.  They describe a v2 product that is a significant improvement over an already good performance by the v1 product.  However, I believe the authors vastly understate the importance of their soft calibration approach in the success of the OMPROFOZ product.  Nor do the authors adequately discuss the information content of their product given their soft calibration approach.  The derived ozone profiles are effectively normalized to equatorial MLS profiles once per year.  Therefore, the OMPROFOZ product is one that describes extra-tropical ozone profile variability and intra-annual ozone variability relative to MLS ozone profiles.  There is nothing wrong with this approach, nor is it even new to this version.  A product such as OMPROFOZ that better describes UTLS ozone variability is quite valuable.  But the authors should clearly state up front (i.e. in the abstract or in the conclusions) what this product is and what it is not.  In particular, it is not a product that independently measures long-term changes in ozone profiles.  Their soft calibration approach is central to the definition of this product, yet the authors treat it almost as an afterthought with little mention throughout this paper.  One gets the impression they themselves do not fully appreciate the role this normalization plays.

Section 1, Line 51

Remove "been"

Section 1, Line 56

Insert "to" before "evaluate"

Section 1, Line 66

Since this is the first reference to MLS in this paper, please indicate that you refer to the AURA instrument.

Section 2.1, Line 122

In the interests of full disclosure the authors should inform the readers that the row anomaly affects ALL of the UV1 channel.  There are no rows that are reliably free of the effects of the anomaly, though longer wavelengths tend to be less effected than shorter ones, and lower rows less than higher ones.

Section heading 2.1 is duplicated.  Section 2.2 is missing.

Section 2.2, Lines 171-175

The authors state that the measurement noise reported in the OMI L1B product underestimates the true noise.  This is a well-known problem across multiple instruments.  It occurs because pseudo-random systematic errors, either in the measurements or in the model, are much larger than detector noise.  The authors go on to state that their assumed minimum errors are uncorrelated between wavelength, but this seems to be a rather poor assumption.  Many modeling and measurement errors are spectrally correlated, e.g. cloud modeling errors.  Can the authors comment on the effect such an assumption has on their product?

Section 3.2, Line 298

Begin this sentence with, "In Version 1 these meteorological variables ..."

Section 3.5, Line 358

Unlike other BUV instruments, normalizing by the OMI measured solar irradiance does not reduce optical degradation errors.  With OMI there are several optical elements not shared in common between Earth radiance measurements and solar irradiance measurements.  In the case of solar irradiance these elements (diffuser and folding mirror) represent the primary sources of degradation.  As a consequence, the Coll. 4 degradation corrections for radiance and for irradiance are completely separate.  Furthermore, the degradation correction for irradiance measurements was derived by assuming constant solar irradiance over the mission (Figure 5c demonstrates this point).  This is a reasonably good calibration approach for wavelengths longer than 300 nm, but not for the UV1 channel.  There are clearly benefits to normalizing OMI radiances with a time-dependent solar irradiance, but cancellation of long-term optical degradation is not one of them.

Having said all this, it's not clear why the authors are concerned about optical degradation given their soft calibration approach outlined in Section 3.8.  Perhaps the authors should make it clear in Section 3.5 that their interest in improved solar irradiance measurements is solely to address seasonal variations in instrument calibration.

Section 3.1, Lines 369 - 371.

What does "implementations are identically applied" mean?  If you mean to say that Coll. 3 and Coll. 4 data are treated the same for this experiment, please say so.

Section 3.6

The authors could instill confidence in their pseudo-absorber approach to scene inhomogeneity if they were to demonstrate that their empirical parameters correlate with scene reflectivity changes or the small pixel column results contained in the OMI Level 1B product.  The largest slit function errors will occur in the along-track direction at cloud edges, which can be identified via scene reflectivity or the small-pixel data.  However, if such a comparison has already been shown in the referenced paper the authors may ignore this comment.

Section 3.8

Per the authors' description in lines 454-463, OMI calibration is adjusted so that the retrieved ozone profiles match MLS + LLM profiles in the tropics.  Per the description provided, this normalization occurs during the northern summer every year.  While such an approach should help deal with systematic biases caused by the row anomaly, a once-per-year correction is inadequate to deal with the variable nature of the row anomaly.  The authors should address the question of what intra-annual TCO errors may remain after the once-per-year soft calibration corrections.

Section 3.9

The authors should attempt to provide a physical explanation for the observed intra-orbital variations in residuals.  Without a reasonable explanation how can the authors or the readers be confident a static CMC correction is approriate and adequate? The most likely explanation for the observed residual variation is additive errors (e.g. stray light) and the row anomaly.  Will the CMC as the authors have implemented it address errors introduced by stray light and the row anomaly?

Figure 13

The MB and Std. Dev. labels appear to be reversed.

Figures 13 & 14

Given that the OMPROFOZ product is tied via soft calibration to MLS+LLM, it will be helpful to show readers similar comparisons of MLS+LLM to the same ozonesonde measurements (or at least measurements from the same stations).  This may provide insight into how much of the observed TCO-sonde difference arises from the choice of soft calibration.

Section 4.0, Line 583

The authors should avoid referring to the flags as "TOMS-based" and instead continue to reference the OMUANC product as the source of these flags.

Section 5.0, Lines 628-630

This brief mention of the soft calibration understates the role it plays in the performance of this product.  The text suggests that its role is to merely keep the simulations close to the measurements, perhaps to keep them in a more linear regime.  Surely, in an iterative retrieval algorithm that typically requires no more than 2-3 iterations (Section 2.2) dependence on the initial guess is not strong.  The authors should acknowledge that the primary role played by their soft calibration is to eliminate the long-term drift observed in v1, and to remove some of the static and slowly varying row anomaly errors that have hitherto stymied all other attempts at retrieving ozone profiles from OMI data.

Section 5.0, Lines 643-648

The authors imply that the improved long-term drift is somehow related to switching from Coll. 3 to Coll. 4 and to some unidentified implementation details.  Given the soft calibration approach in v2 it is unlikely that the improved calibration in Coll. 4 plays any role in this.

The second version of the OMPROFOZ research product (OMPROFOZ v2) has been introduced in this paper, which incorporates several improvements to enhance the accuracy and long-term consistency of ozone profile retrievals from the OMI instrument. The retrieval quality of tropospheric column ozone has been improved.  The presented methods are presented clearly and the paper is generally well written.

General comments:

1. Could the authors also provide a comprehensive discussion of the limitations of the OMPROFOZ v2 algorithm and the potential sources of error in the ozone profile retrievals, and plans for the next version if there will be.

2. Can you address the potential impact of cloud and aerosols on the accuracy of the ozone profile retrievals and is it possible to derive reliable near-surface ozone from UV measurements after significant improvements in the calibration and retrieval algorithm?

3. Better to summarize these analysis of the uncertainties associated with the improved algorithm updates and their impact on the accuracy of the ozone profile retrievals, maybe in a table.

Specific comments:

 Line 182-183: “three kinds of parameters are newly added to implement the slit function linearization and common mode correction as a pseudo absorber.” Please clarify in this sentence if the slit function linearization parameters are also implemented as pseudo absorbers?

Line 186: It’s better to describe how the “covariance matrix” be constructed or refer to some references that has described it (May be Liu et at., 2010).

Line 237: I guess 𝐼_𝑒 represents the measured random noise errors.

Line 283: May be add a plot of information contents or averaging kernel in Figure 2 for the layer of tropopause helps understanding the sentence “In the subtropical region, LLM may also provide incorrect information in the presence of high tropopause height…”

Line 357-359: How does the monthly averaged irradiance spectrum cancel out the common degradation existing in radiance and irradiance?

In section 3.6, how about the correlation between slit functions parameters and ozone parameters in the retrieval?

In section 3.9, does the common mode correction improve the ozone profile retrieval other than just improving the fitting residual?

In this work, Bak et al. provide a thorough description of “An improved OMI ozone profile research product version 2.0 with collection 4 L1b data and algorithm updates”. Although it is of significant scientific value to the space-borne atmospheric monitoring community, its presentation can be substantially improved, both in terms of research results and general phrasing.

Comments:

Line 59: A degradation is provided in %, but the time range is not specified. Is this per decade, or 'now' with respect to beginning of mission? Idem for the wavelength stability in nm (also in line 155).

Line 62: “satellite ozone profile products have not been reliable for long-term analysis” sounds too strong. This should be rephrased.

Line 68: “latitude/season/viewing geometries” can be combinedly referred to with an optical path dependence? Please discuss appropriately.

Lines 84-85: “Note that OMI measurements have been reprocessed to deliver the new collection 4 dataset” Does this apply to the entire time series, or a limited period only?

Lines 109-110: It is not clear where the codes between brackets refer to.

Lines 126-127: Does the “KNMI flag” name apply in general, or in this work only?

Lines 135-136: “In addition, OMI total column ozone product (OMTO3G) is used in deriving empirical correction spectra.” This sounds too vague. A reference or more detailed explanation is needed.

Line 149: “48 x 26” instead of “38 x 26”?

Figure 1 may be too detailed or might need an explanation of the variables used.

Line 175: “assuming that measurement errors are uncorrelated among wavelengths” Is this a valid / common approach then? Please specify or provide referencing.

Lines 186-187: Explain with reference(s) how the correlation length is applied in the covariance matrix.

Lines 207-209: What happens to levels below the effective surface pressure?

Line 276: Where is this tropopause pressure obtained from (also in Figure 2)?

Lines 303-305: “However, the data transmission has been accidently halted since June 2011 and hence climatological monthly mean data have been used as a back-up in the data processing.” Not clear whether this applies to v1 or v2.

Lines 327-329: “BW measurements were better parameterized as quadratic temperature-dependent coefficients with uncertainties of 0.25-2 % whereas for BDM measurements fitting residuals of 2-20 % remains.” Also lines 346-347 “radiometric uncertainties of the new reference spectrum are below ~ 1 % whereas for SAO2010 those range from 5% in the longer UV part to 15 %” That’s two times a factor of ten improvement, and might be the most impactful changes. This might be stressed in the conclusions and/or better explained and referenced?

Why are different temperatures chosen for plots (a) and (b) in Figure 4?

Line 401: Explain cross-product and I_h in the equation.

Lines 411-412: A factor f_c seems to be missing the second term of the formula (while on the other hand f_c is not explained).

Lines 412-413: “finer grids than at least 4 pixels per FWHM so that the spectral convolution is applied to account for OMI spectral resolution” This is too brief to be clear.

Table 4 and Table 5 seem to be missing?!

Lines 428-431: this is unclear if half-streams are not explained.

Line 433: “(2 vs. 6) and number of layers (24 vs. 72)” which versions are compared here?

Lines 435-438: “Note that the Ring simulation remains unchanged from v1 algorithm; the spectral structure of the Ring signal is externally simulated with the iterative fitting of amplitude of the Ring spectrum and then subtracted from the measured spectral reflectance.” Please provide reference.

Lines 446-447: “along-track stripes that are commonly found in OMI trace gas products” Should this be clear from the plots? Please provide references.

Section 3.9: Please explain important “pseudo-absorber” concept for clarity of common mode correction.

Section 4: Why the severe limitation of having 3 sites above Central Europe only? To what extent are this results globally valid?

Line 545: A 3 sigma for ‘extreme values’ seems quite low. Real outliers are typically beyond that.

Line 611: What does “continued externally” mean?

Technical corrections:

Line 51: “has been contributed”

Line 113: “shorter” refers to “shorter-wavelength”?

Line 132: “stricter and more reliable”

Line 133: “flags are raised” instead of “flags are flagged”

Please explain norm symbols (with lower and upper 2) in equation (2).

Line 471: “stripes”

Line 549: “relatively not serious” is unscientific phrasing.

Figure 13: The legends for mean bias and standard deviation seem to be reversed.

Lines 650: please put dates on “in progress” for future reference.