Please see the attached document for my recommendation and comments.

The paper documents the long-term performance of the Aeolus wind measurements tracking the evolution of the performance across the mission. In particular a detailed assessment of the atmospheric area covered by the Aeolus Rayleigh and Mie measurements is discussed. Errors estimated from the differences between observations and short-range NWP forecast are compared with error estimates based on SNR and number of averaged samples. The topic of the paper is certainly appropriate for AMT. The paper is of great interest for Aeolus users and for preparation studies of the Aeolus-2 mission. 

The purpose of the paper is clear and well stated. However, I - not an Aeolus user- have found the paper difficult to follow in some parts.

Sometimes the used terminology and notation does not help the reader;

there are some inconsistencies across the paper (e.g. different model background errors are assumed in different figures);

it is not clear what are some mean or median values computed from (e.g. Fig.2 depicts daily average random errors computed across what measurements? All of them? not exluding anything?).

Also the SNR terminology is in my opinion kind of misleading. The SNR is the SNR, if we start scaling it with the square root of the number of measurements is not an SNR any more (of course the latter is the quantity that is relevant to the measurement accuracy, by averaging we are not improving the SNR we are improving the sensitivity). In fact it is named "scaled SNR" (but again to me the naming is not really helping the reader).

Here a list of suggestions for improving the paper:

Line 236: To avoid confusion I would name it normalised atmospheric signal (NAS) because it is normalised to a specific date.

Line 247-248: not sure what you are alluding to. I do not see any peculiarity at that time

Fig:3 really difficult to grasp the significance of this figure. Also not clear to me what is the meaning of the different labels with the different h. Simplify the figure or even delete?

Fig.3: h is not a very good name for an horizontal bin length (why not using Dx?). Later on I see you use d_bin for the vertical bin thickness. In my opinion this is not a good selection of variable names. 

Eq.(2) not fully clear to me over which dataset the median is computed from (all altitudes?)

Fig.5: I am not sure I see three tones of red. The faintiest red looks more like a white.

Fig.7 why there are no LIPAs simulations in the polar regions? 

Line 528 and Fig. 12 "the altitude dependence of the Mie and Rayleigh SNR per wind result". What dose it mean? Is this a mean or median value computed across valid data?

Eq.3: I cannot actually find in App.B a formula for the EE (apart from saying it is inversely proportional to the SNR).

Fig 15: the figure is a little bit messy. I would probably reduce the number of lines to the most significative ones (2 or 3 out of 6). Also I would expect that the errorbars became longer and longer with smaller errors (which seems to occur for panel b but not for panel a)

Line 604: "the actual error" : I am not sure we can refer to that as the actual error. It implies that ECMWF background wind is the "truth" (particularly not applicable for the Mie winds in presence of clouds and convection).

Line 609-610: not sure I understand this extrapolation. I thought the background error (assumed 2 m/s) should be already subtracted from the y-axis values. Maybe it is worth properly defining epsilon somewhere. In fact how does the extrapolation work when comparing Rayleigh and Mie winds? Actually I see some explanation later but does this actually mean that the background error for the Mie winds should be taken much larger than 2 m/s or otherwise that the EE is underestimated at high SNR?

I would actually rephrase line 610 accordingly.

Appendix

Figure A1: I have to say that I struggle understanding the meaning of the last box "Dimensions". It is not very intuitive for somebody who is not a Aeolus user to figure out what is meant with observation, measurement, pixel, wind result (#profiles #range bins, # averaged profiles, ... are simpler concepts to grasp)

Figure A2 is also difficult. A simple schematic with the terminology used could help.

Line 613: "yielding unrealistically high values", I am not sure this are unrealistic values for the errors, to me they may be  very realistic (rephrase?)

model background error: sometimes it is assumed 2.0 m/s sometimes 2.5 m/s . I would try to keep consistency across the paper.

Line 677: I would recommend to introduce also the reference Illingworth, A. J., and Coauthors, 2018: WIVERN: A New Satellite Concept to Provide Global In-Cloud Winds, Precipitation, and Cloud Properties. Bull. Amer. Meteor. Soc., 99, 1669–1687, https://doi.org/10.1175/BAMS-D-16-0047.1.

Lin373: ; ==> ,