We thank referee #1 for the comments, corrections and suggestions, which we appreciate very much, and we hope that our respective implementations will increase the readability and quality of the manuscript.

Original questions/comments of the referee are marked by RC: and set in italic font.

RC: This paper presents a new version of the MIPAS stratospheric ozone profile data set. The work is well motivated by changes in the retrieval strategy, which will be important for some community users of the data set, and by the new calibration of the level-1b spectra. The paper is quite detailed although some elements listed below are difficult to follow for those outside the project. The work fits well within the scope of AMT and should be published after consideration of the minor points below.  It is recommended that the abstract be rewritten as it is currently quite lengthy and should be focused on the most important points of the new dataset.

Reply: As both referees unanimously raise this issue, we will rewrite the abstract in a more concise manner, leaving out statements which might be too detailed and concentrating on the major methodical and input data changes and on the major findings.

RC: Section 3.1 on the preceding temperature retrieval is not clear. What is actually done in this step? Who are "these authors"?  A better (not longer) summary of this step from Kiefer et al 2021 is needed as it comes up several times in following sections.

Reply: We agree to restructure and rewrite this section for clarity.  The new text of Sec. 3.1 will, among other changes, use the reference Kiefer et al. (2022) instead of "these authors".

RC: Can the authors provide any insight about why the inclusion of horizontal gradient did not work?

Reply: Unfortunately, we are not very clear on this issue.  We know from the analysis of the horizontal averaging kernels that the horizontal range to which MIPAS is sensitive is only a few hundred km (typically below 400--500 km, according to the measurement mode).  We suspect that MIPAS is not very sensitive to ozone variations along such a short distance. In exceptional cases (e.g. vortex edge) there might be larger horizontal ozone variations to which MIPAS might indeed be sensitive, but in such situations the parametrization of the horizontal structure by a simple gradient would be inadequate to represent the true horizontal structure, and the retrieval of additional parameters to describe the horizontal variability would destabilize the retrieval.

RC: Why is it better to jointly fit water vapor instead of using the regular product?  Explain what is meant by "In order to avoid propagation of assumptions on the water vapour profiles on the ozone data product".

Reply: Our retrievals are done sequentially. After shift correction there is the T+LOS retrieval, which is followed directly by the ozone retrieval.  Hence, there is no regular product of V8 water vapour available at the time of the ozone retrieval.  Still, V5 water vapour data could be used.  However, ozone spectral lines used for the retrieval are appreciably interfered with water vapour spectral lines. The spectroscopic data of these lines may be inconsistent with the spectroscopic data used for the water vapour retrieval.  Thus, the spectral signal caused by the water vapour lines in the ozone microwindows will not be accurately be accounted for if the water vapour concentrations from the water vapour retrieval are used. To avoid propagation of related errors on the ozone results, water vapour is fitted jointly with ozone. We will add this information to the manuscript.

RC: It would be useful to connect the weakening of the smoothing regularization to impact on vertical resolution.

Reply: Yes, that seems very reasonable.  We shall add appropriate text and quantification in Sections 3.3 and 5.1 in the revised version of the manuscript.

RC: Numerical issues: what is meant by the "result flipping back and forth".  This statement needs more rigor and explanation.

Reply: Text similar to that in Sect. 3.12 of the MIPAS V8 temperature retrieval paper by Kiefer et al. [1], which gives more details on this issue, shall be implanted in the manuscript.  The new text, replacing parts of lines 182-183 will be: In the retrieval code an `oscillation detection' has been activated which identifies failure of convergence in the sense that the iteration flips back and forth between two minima of the cost function according to x_i+1 ~ x_i-1 and x_i ~ x_i-2. In this case x_i+1 is set to (x_i+1 + x_i) / 2, and further iteration steps are performed.

RC: What causes the strange arching shapes in the noise-component in Fig 1?  Is this due to the mismatch between retrieval and measurement grids?

Reply: We assume that the referee refers to structures mainly visible below 20 km in Fig. 1b.  In general, the mismatch between retrieval an measurement grid are causing the stripes.  This is well visible in Fig. 1a (FR data) for virtually all altitudes: The smaller the distance between retrieval and measurement altitude is, the smaller the noise error will be.  We think that there are several components which contribute to the "arcs".  First, the mismatch as just mentioned.  Second, the change in the vertical scan grid with latitude, which is, on purpose, much stronger for RR data than for FR (Actually, above 30 km the structures visible in Fig. 4 give an impression how the measurement altitudes change with latitude; below, the "arc-effect" seems to dominate the picture, too.).  Third, the strong difference in the spacing of the vertical scan grids of FR and RR data at lower altitudes.  There RR measurements are 1.5 km apart (FR approx. 3 km), while the retrieval grid is 1 km.  We

think that some sort of Moiré effect might emerge from these three ingredients.

RC: The discussion of spectroscopic errors and the overall link to the TUNER recommendations is appreciated.

Reply: Thank you!

RC: Including a plot of the vertical resolution with the cases shown in Fig 3 would be useful.

Reply: A good suggestion; This will be done.

RC: The discussion on the "improvement" shown in Fig 5 due to reduction of temperature propagation errors from high altitudes sounds good.  Is it verifiable that v8 is in fact better?

Reply: We know that nominal retrieval results do better coincide with middle atmosphere results (which use measurements up to 100 km) for V8 data than for V5 data.  However this is not mentioned in the paper, it leads us to believe that V8 ozone will be superior to V5 data at high altitudes.  Currently dedicated ozone validation studies are under way. Looking at northern polar winter conditions will be a part of these.

RC: Should the new maximum likelihood representation for the analysis of the drift? Why not?

Reply: We assume the question is whether the new maximum likelihood representation should be used for the assessment of the instrument drift.  As pointed out in the first paragraph of section 6, these new data are expected to be void of time-dependent effects in the vertical resolution and hence better suited for, e.g., trend studies.  So, when comparing to other instruments in this respect, which could be used as a way to assess the residual drift, the new representation clearly should be preferred.  Inter-version comparisons, e.g. V8 versus V5 will not be possible, since yet there is no maximum likelihood representation for V5 data.

[1] Kiefer, M., von Clarmann, T., Funke, B., García-Comas, M., Glatthor, N., Grabowski, U., Kellmann, S., Kleinert, A., Laeng, A., Linden, A., López-Puertas, M., Marsh, D., and Stiller, G.P.: IMK/IAA MIPAS temperature retrieval version 8: nominal measurements, Atmos. Meas. Tech., 14, 4111-4138, https://doi.org/10.5194/amt-14-4111-2021, 2021.

We thank referee #2 for the comments, corrections and suggestions, which we appreciate very much, and we hope that our respective implementations will increase the readability and quality of the manuscript.

Original questions/comments of the referee are marked by RC: and set in italic font.

RC: The paper by Kiefer et al. reports on a new version (V8) of ozone retrievals from the MIPAS satellite dataset. Most importantly, the new retrievals are driven by a new calibration version of the underlying spectra. Essentially, the paper provides a reference document to the community what the changes are wrt. to previous MIPAS ozone retrieval versions and how performance has improved. These improvements are particularly important for the usage of MIPAS data in trend analyses and studies of the upper atmosphere. The paper is very well written and all the analyses are rigorous. Therefore, I do not have any but a few technical comments.

RC: Abstract: I would support reviewer 1 in requesting the abstract to be rewritten in a more concise way highlighting the most important findings.

Reply: As both referees unanimously raise this issue, we will rewrite the abstract in a more concise manner, leaving out statements which might be too detailed and concentrating on the major methodical and input data changes and on the major findings.

RC: L5: temperatures -> temperature data L8: Ozone lines -> ozone absorption lines

L39: mod ededicated -> mode dedicated

L79: As the cause... rephrase -> An inadequate... was identified to be the cause ...

L181: was set to -> was set to a spacing of

L236: to allow the correct treatment -> to allow for the correct treatment

L313: not changed in the retrieval -> not varied in the retrieval

L436: was done -> was carried out

Reply: These corrections and wording suggestions are well justified and shall be implemented in the manuscript.

RC: L53: There is no section 1.2. Is a section header "1.1" really required?

Reply: We see the point. Subsection 1.1 will be promoted to become a full section (Sec. 2).

RC: L105: The preceding temperature retrieval -> The general reader might be confused by what "preceding" refers to. Why not just "Temperature retrieval"?

Reply: Preceding is meant to allude to the fact that we do the retrievals of temperature and the diverse gases in a sequence, one after another. Hence, we prefer to leave it as it is.