The publication describes an improved (the final) data product of MIPAS, ORM V8, which is available to a large user community. It should therefore be of great interest to the AMT readership. The improvements concern the forward model (horizontal gradients, new spectroscopic data), the inverse model as well as an improved inclusion of calibration data and improved cloud filtering and filtering for bad quality data. In addition, there is an improvement in the temporal stability by a reduction of instrumental drift. Discontinuities in the measurement series are reduced by a better exploitation of calibration data (daily gain factors). Overall, the paper is very convincing and should be published after some minor changes.

My main comment concerns the treatment of horizontal inhomogeneities. A first order model is used, which only accounts for linear variations using a horizontal gradient. Statistically, improvements are achieved, e.g. with regard to day/night differences (ascending/descending orbits). However, in situations of high (small scale) variability, e.g. due to wave breaking, using a linear gradient may even be counterproductive. For example, the authors state that closest coincidences of ERA-Interim ECMWF data for each MIPAS profile were used for computing the horizontal gradients of temperature, water vapour and ozone. How is the gradient then calculated? By simple linear interpolation? Since the horizontal resolution of the ERA-Interim data is higher than that of MIPAS (distance of the tangent points), the calculation of an effective gradient would make sense, which also accounts for smaller-scale variability along the line of sight. If a simpler procedure was chosen, the authors could briefly discuss this and state that essentially a statistical (climatological) improvement was targeted.

Other comments:

l.11 Introduce IG2 (Level 2 Initial Guess). The abbreviation is used a few times, but not explained until l. 318.

l.23 The effects are not limited to the region from the surface to the mesosphere. For example, solar activity also affects the thermosphere.

l.26. I find the Kidston et al. citation a bit too specific here. I would therefore also use another citation, e.g., from the IPCC.

l.27 A citation for the air quality aspect is missing.

l.30 ... period (Fischer et al., 2008).

l.33 Indicate approximately how large the resolution is.

l.42 The statement does not apply to every tangent point since a global fit procedure is used?

l.78 I would put the Fischer quote at the end of the sentence.

l.85 A sketch of the measurement geometry would be helpful for some readers.

l.88 What is meant by revisit time of three days? It should be about a month.

l.112 Both ... vary

l.130 I would say what is meant. Horizontal gradients of temperatures and trace gases.

l.140 ... and lower thermosphere?

l.143 I assume that this sequential procedure is also iterated?

l.165 Briefly state the purpose.

l.171 Statistically, it shows an improvement.

l.193 mean retrieved profiles

l.221 in about >> about

l.257 better "between observed and simulated spectral features".

l.258 consistency check is o.k., but the 7.6 mm region should not be further included in the HNO3 retrieval itself by just to account for HNO3 interferences in the 7.6 mm region.

l.268 ... improves the spectral simulations...

Legend Figure 10: Perhaps spell out FR and OR again.

l.331/332 The sentence is difficult to understand.

l.333 If necessary, give some details for priority system.

l.337 mb >> hPa

l.357 Delete space after "range".

l.364 What do you mean by 0 hPa?

l.406 I don't understand this big difference between models levels and pressure levels right away.

l.486 I would somewhat rewrite how you reduce outliers after the retrieval, e.g. ... after the retrieval based on a more sophisticated quality flag.

l.620 long-term evolution

Figure 23: Improve quality of symbols and lines in the plot.

l.655 ... observed scene multiplied...

l.840 Remove the first parenthesis.

Review of amt-2021-235 “Level 2 processor and auxiliary data for ESA Version 8 final full mission analysis of MIPAS measurements on ENVISAT by Raspollini et al”

This paper describes the final ENVISAT/MIPAS reprocessing using ORM v8. Improvements from level-2 noted include the areas of line-of-sight-gradients, cloud contamination, data quality metrics, spectroscopic data, and a priori atmospheric data. Improvements from level-1b include reduced instrument drift and radiometric calibration. MIPAS v8 products include more atmospheric species with better accuracies and temporal stability than in previous versions.

I recommend publication with attention given to the following minor corrections / discussion points.

L11

IG2 not defined

L23

… could also mention pyroCb events arising from wildfires e.g. Australian New Year fires in 2020.

L23

[Taken together] these changes

L29

“of the atmospheric emission” … perhaps you meant to say “of the infrared emission” otherwise better to just say “of the atmosphere”

L30

Which is correct … ENVISAT (as in the title and elsewhere) or Envisat?

L33 probably should jam “infrared” in this line before “emitting”

L34

10 years is a long time… CO2 increased quite a bit - how was that handled for the retrieval of temperature?

L67

I’m sure you have stories of “what not to do” that would be similarly helpful!

L80

10 hour => 10:00

L97

maximum [interferometetric] path ?

L100

reducing the spectral resolution to

L107

MIPAS was not operating routinely for a long period. Since a 20cm path was causing mechanical problems, what was the largest path thought to still be safe? Was a new max path limit determined from in-orbit testing?

L108

In the OR scan the lowest tangent height was increased from 6 to 7 km. Was that for an engineering mechanism reason or was it decided to trade-off the the 6-7 km region in order to add 4km in the mesosphere?

L139

non-LTE

L145

What about the (approximate) reference height so you can vertically locate the limb scan. Where does that come from? Do you have an internal a geopotential height product?

L210

Could you give some indication of the error incurred in assuming no horizontal gradient across the line-of-sight in the refractive index? Also this is the only mention of the Curtis-Godson approximation (CGA) and should have a reference to the original work at least and possibly to other documented retrieval codes for ir limb sounders. What led to the choice of CGA and were any other alternatives considered e.g. Emissivity Growth Approximation (EGA) or Correlated-k (presumably not needed as the micro windows are quite narrow)?

L211

Regarding the ray tracing… What is the spacing of the angular grid used to represent the climatology along the line of sight?

“With an adaptive step that depends on the curvature of the ray paths”… probably worth to point out that this describes the treatment near the tangent point where finer vertical levels are required for accuracy.

L228

Do you really have the horizontal resolution to achieve a 2nd order correction?

L342

What is the effect on temperature on the discontinuous CO2 concentration change on crossing a year boundary?

L455

How about a joint retrieval of H2O and the HDO/H2O ratio?

L454

Where does the magic number 31.6 come from? Need some context here e.g. over what values does the continuum range?

L470

What is the size of the matrices that need to be furnished to accomplish this feat. Would representative matrices suffice i.e. if they be aggregated  over seasons and latitude bands?

L593 and L683

not-good -> bad I suppose

L594

So are the VMR retrievals also marked as “bad” or some other specific flag that indicates missing data because the temperature retrieval failed?

L603 and Page 33 Fig 22

Representative retrieval [product]? What product and OR or FR? If this distribution appears (or not) in Fig 1 or 2 then please indicate which one.

L608

the values of the thresholds [for each product]

L601

derive [atmospheric] trends

L636 and Fig 23

Fig 23 is not very good for showing the temporal variation. The sequence of colors/symbols bears no relation to the time domain. I suggest taking only three pressure levels for UTLS, LS and MS and plotting the Tdiffs vs time.

Also then you could indicate where the decontaminations (L653) occur on this figure.

L686

Some indication of file sizes would be useful

L698

If 2nd order effects are a big deal (I doubt that the auxiliary model data can be used to correct to this order anyway) then you should indicate that these gradients are limited to a linear assumption along the line of sight.