Thank you for the invitation to review this interesting study that explore the feasibility of a space-borne Differential Absorption Radar that operates in the sub-mm region near the 380 GHz water vapor absorption line. The authors listed a number of advantages of the suggested approach, primarily the fact that the investigated frequency band is outside of the internationally regulated microwave regions of the electromagnetic spectrum. Given the high extinction of the signal for these frequencies its applicability is however limited to the upper portion of the troposphere. The feasibility study is performed mostly upon model and reanalysis data, focusing on the optimization of the frequency selection in order to minimize the estimated bias and uncertainty of the retrieval. Retrieved Cloud-Sat products are used to assess the expected observation yield in a realistic scenario.

Overall, I find this study interesting and worth to be published. The text is well written and easy to follow, the figures and the data are of good quality and I wasn't able to find any major fallacy in the logic or the approaches employed. I would like to suggest a few improvements that I think can elevate the usefulness of the study.

Line 6 - Any differential measurement approach implies pairs of measurements, here it is abruptly introduced the fact that triplets are needed. Perhaps it is worth saying already in the abstract that the use of 3 frequencies are necessary to account for the confounding effect of differential scattering and extinction by hydrometeors and absorption by dry gases.

Line 10 - here vertical resolution is 200m while in table 2 is going to be 50m. I guess the limitation comes from the vertical resolution of the model. But I do not think this discrepancy is well explained in the text. Perhaps it is worth mentioning it, and tell people that the result in the upper atmosphere are also sort of averaged over multiple range gates because of the limited vertical resolution of the model.

Line 71 - What exactly is used to represent ice and snow in the radar simulator? The text says hexagonal columns and dendrites, but the reference Leinonen and Szyrmer (2015) present a set of rimed and unrimed aggregates of dendrites and produce their scattering properties up to only 94 GHz. In Table 1 T-matrix and Mie scattering are mentioned which would imply spherical and spheroidal particles, is T-matrix used for ice and snow? Also, if Leinonen and Szyrmer (2015) shapes are used, are they consistent with the mass-size relation assumed in the WRF-1mom microphysics? Are the PSD references in Table 1 also the same in WRF or are they just assumed by the radar simulator?

Section 2.3 - I would appreciate a cleaner description of the retrieval, at the moment it seems to me unnecessarily convoluted. I understand the logic, but it took me a while to grasp the discourse between lines 109 and 119. Perhaps I was off-put because I disagree with the statement on line 109-110: eq.2 does not allow to fit the terms on the righthand side by measuring gamma at various frequencies unless one makes assumptions on the functional forms of the such terms. Anyway, why not just mention the immunity to calibration errors and move on directly to the concept of linearizing the 3 confounding terms with respect to frequency? Also, it is not immediately clear to me why the variability of the PSD is affecting the feasibility of a 2-frequency retrieval (line 117), the PSD is technically always the same (apart from beam-mismatches at various frequencies) and a third frequency would not help, the problem comes solely from the fact that all terms are frequency dependent so you need a way to transform the equation in to a system of equations with a limited number of unknowns.

Line 151 - Do I understand correctly? You simulated all 31 frequencies between 350 and 380 GHz (extremes included) at 1 GHz step. Then you made experiments considering all possible triplets among these 31 frequencies, correct? Which if I am not mistaken it accounts for 8990 combinations. Perhaps you can also mention this number.

Yield - I understand the definition, but I was wondering about a broader discussion on the benefit of this proposed observation platform in the context of existing satellite measurements. One of the great limitations of DAR techniques is the possibility of measuring only in clouds, this particular approach is even more limited to the upper portions of the atmosphere while DIAL can observe in clear-air. It might be interesting to make some quantitative assessment of the relative importance of the observational gap filled by this sub-mm DAR in the context of already established water-vapour retrieval techniques.

Trade offs - I found interesting observing the trade offs that one needs to make when selecting optimal frequency triplets. I was wondering if it would make sense to compensate for these limitations by introducing a fourth frequency? Not necessarily developing a full retrieval operating with frequency quads, but perhaps having a platform capable of operating multiple frequency triplets by selecting from a set of four available.

Precision - I understand this might be established jargon, but I find odd to use the term "precision" instead of "uncertainty". It just bothers me personally that this use of terms would imply that one would aim to reduce the value precision in order to get more precise results and the authors did a great job avoiding this problem by using terms such as "degrade", or "enhance", and so on. Maybe just mention this oddity at line 136 where precision is defined as the spread of random retrieval errors. Alternatively ignore this point all-together, I will be fine.

Review for revised manuscript of " Retrieval simulations of a spaceborne differential absorption radar near the 380 GHz water vapor line" by Millán et al.

This paper evaluates the potential of using spaceborne differential absorption radar (DAR) operating near the 380 GHz water vapor absorption line to detect water vapor in the mid- and upper atmosphere, especially inside deep convective systems. The authors identify optimal radar frequency triplets that minimize precision errors and biases using large-eddy simulations (LES) and end-to-end retrieval experiments. The manuscript is well-structured and addresses a significant challenge in atmospheric remote sensing. The methodology is robust, combining LES-driven simulations with retrieval algorithms to assess DAR performance. However, there are some questions and comments that may need to be further considered and major revision is necessary.

General Comments:

Abstract: The vertical resolution in the manuscript is inconsistent with the vertical resolution in the table 2. What is the reason for this difference? It should be supplemented in the manuscript.

Introduction: Please provide a clearer explanation in the introduction section of the advantages of using the 380 GHz water vapor absorption line over the existing 183 GHz absorption line for retrieval, emphasizing the importance of this absorption line.

Table 1: The radar forward model description references Table 1, but the function expressions of the assumed particle size distribution are not explicitly listed in the table. Please add it in the revised manuscript.

Specific Comments:

Line 189: Correct "radioocculation" to "radio occultation".

Lines 216–220: The precision improvement via along-track averaging (Figure 7) is compelling, but the discussion lacks a quantitative comparison to existing instruments. Adding a paragraph contrasting DAR’s precision after averaging with other methods would emphasize its novelty.

Line 258: Remove duplicate "the" in "the the types".

Line 285: The statement "biases are generally much smaller than precision errors" conflicts with Figure 6b, where biases for some triplets exceed 800%. Please check it.