see attachment!

Please see attachement

The paper is about determining temperature profiles with the help of elevation scanning MWRs during rain and how accurate these profiles are. Usually temperature profile retrievals during rain are not possible due to increased opaqueness of the troposphere within the V-band during rain and due to water accumulation on top of the radome. This paper introduces a method on how to retrieve accurate enough temperature profiles in the lower troposphere in spite of these conditions. Key aspects in doing so are only using off-zenith observations and only utilizing the four optical thickest V-band frequencies.

General comments

In my opinion, the necessary quantification of how well the proposed temperature retrieval performs during rainy conditions is missing. Most information is there within the figures but is not stated explicitly in the text.

Overall, the paper is written well and is easy to understand but sometimes details are missing. I will provide more detailed comments and suggestions on what to change below.

General question: What about snowfall? Temperature profiles are usually also not retrieved during snowfall, right? Maybe state in the introduction why you dismiss snowfall and only look into liquid precipitation.

Specific comments

Abstract: Quantification missing. What's the accuracy of the new retrieval in different rainfall scenarios?

32: retrieved Ts from NN approach and 1DVar technique? Here the flow of text seems to suggest that the 1DVar is a form of retrieval (NN and MLR), but it isn't, is it? I think it important to tell a little more what the 1DVar is/does or what it means.

41: reduced by how much? Your method reduces the error during rain EVEN FURTHER? Should make that clear.

Maybe cite Böck et al., 2024 in the introduction section. They look into external measurement uncertainties of scanning HATPROs and what these mean for retrieved temperature profiles.

45: "almost saturated" with what? is this quantifiable or is there a source? Or is this something you found out in this study?

61: better write: "in the order of seconds". I've seen for exact 5min measurements, that there are only ~250 data points and not 300, as expected. So not really a 1s resolution, rather ~1.2s.

62: K-band, not Ka-band! Please change this in the whole manuscript.

101-102: There are newer Rosenkranz models. Why did you use an older one? Explain.

Would using a newer gas absorption model make a difference? For showing what you want to show, the old model is sufficient I guess.

118: tbx/SPC Retrievals: Are there more details needed for how 13 frequencies predict the 14th?

150: Maybe explain shortly why only the upper 4 frequencies for elevation scans are used and the lower 3 frequencies for zenith (à optical thickness)

166: no range for moderate rain? Why exactly 2.7mm/h?

177-190: I'd wish for a little bit more quantification here; by how much do TBs differ? (It can be seen in the Figure, but it is not written anywhere).

And what is the threshold for significant difference (when does the pink shaded area start and why?)

196: "by the less and more transparent..."? get rid of the word less or rephrase.

Figure3: y-axis title: change it to Delta TB or something similar, to make clear you’re talking about a difference of brightness temperatures here. Maybe just call the shade of color pink instead of rose.

200: "degreeS of freedom". Please change in the whole manuscript. Also "gives the information content..." sounds strange. I would rephrase.

201: You always write "degrees of freedom of signal". Do you always need the word signal or can you omit it?

213-222: Can you quantify the differences a little more in the text? In general, you often describe Figures only qualitatively.

229-239 and Figure5: When talking about bias in this context, wouldn't it be better if you evaluate its variance/accuracy as RSME instead of standard deviation? (same for Fig.8). Or is this bias a mean we're talking about and the spread of this mean is then the SD?

Again: I think it would be better to also quantify your findings in the text.

240-250: Again: quantify also in the text.

251-260: Here you do quantify the differences in the text. You should do that everywhere.

Figure8: Again: In this context I'm not sure if you should rather talk about RSME instead of standard deviation. You should check that.

The values won't change much, as the only difference is that you divide by n and not n−1.

Why is the bias of the 4vz10phi that much worse above 1km in the no rain scenario?

276-279: Quantify: How much better does the new 4v9phi retrieval perform and/or with what accuracy during rainfall up to 2mm/h? You only quantify the case in the text with a rainrate of below 0.5mm/h.

280-end: Quantification is missing in the conclusion.

            How much Kelvin exactly is the new retrieval method better compared to the standard one?

            E.g. for slight rain below 2km: How much Kelvin is this different to non-rain conditions?

            And how much is it different below 1.5km for heavy rain?

            In general: What's the general temperature profile accuracy of elevation scanning state-of-the-art HATPROs for no rain scenarios and how does it compare to the new findings?

            I think this is important for the reader, so they can better classify/categorize the results of this paper.