The manuscript by Titus et al. investigates wind-cloud interactions using data from the spaceborne Doppler wind lidar onboard the Aeolus satellite. To this end, Aeolus Level-1A data are processed to create a cloud mask at 3 km horizontal resolution, which is then used to resample the Aeolus wind data to a finer grid than that available in the original Level-2B wind product. The resampling approach is validated using airborne Doppler wind lidar measurements as well as a regional weather model simulation.

Using the resulting Aeolus cloud mask and wind data at 3-km resolution, the authors analyze the relationship between cloud cover, wind speed, and vertical wind shear within the troposphere during boreal summer 2020, at various spatial scales and across different global regions. Among other findings, they show that wind shear under cloudy-sky conditions is significantly smaller than in the surrounding clear sky. Moreover, the wind speed difference between cloudy and clear-sky conditions increases with the clear-sky wind shear, particularly in regions dominated by cumulus and stratocumulus clouds.

General comment:

This study is of significant interest to researchers in atmospheric physics, especially those studying global circulation and its interaction with cloud formation and development. The manuscript is well-structured, and the figures effectively support the analysis. The approach, starting from the Aeolus L1A product to derive a cloud mask at full 3-km resolution, is particularly innovative and promising.

However, the methodology outlined in Section 2.1 raises several questions and appears prone to potential systematic errors (see specific comments below). Similarly, the subsequent resampling of the Aeolus L2B wind data using the derived cloud mask needs further clarification to prevent misunderstandings about the approach.

While the use of airborne wind measurements and model simulations lends some confidence to the method, it is based on a few examples from a restricted geographical area. This raises the question of how representative these examples are for the much larger dataset analyzed later in the manuscript.

The results regarding wind-cloud interactions at both regional and sub-100 km scales are insightful and illustrate the capability of Aeolus to detect such phenomena, despite its relatively coarse resolution. As such, the study opens avenues for further research into topics such as jet stream shifts and their coupling with cloud radiative effects.

I recommend the manuscript for publication, contingent on the authors addressing the specific issues outlined below.

Specific comments:

1. Lines 72 ff.: The authors correctly state that Aeolus measured the projection of the horizontal wind on the laser line-of-sight (LOS). However, the symbol u is used throughout the manuscript (e.g., in Eq. (3), Figs. 6–11) to denote the horizontal LOS (HLOS) wind speed, which may be misleading, as u traditionally refers to the zonal wind component. I suggest adopting the notation v_HLOS, which is commonly used in Aeolus-related literature.
2. Line 120: To my knowledge, the number of accumulated backscatter profiles averaged onboard is 18, not 16.
3. Lines 140 ff.: Please provide a reference for the hot pixel map corresponding to the end of the mission. Since only a subset of these pixels was active during the main analysis period (primarily boreal summer 2020), discarding all of them may be unnecessarily conservative. This approach could introduce systematic errors in the retrieval of particulate and molecular backscatter profiles, and in turn affect the cloud mask. This is particularly relevant given that only very few pixels are used in the retrieval.
4. 2.1 and Appendix A: The derivation of particulate and molecular backscatter from the L1A signal appears oversimplified. Notably: a) The correction for solar background is not mentioned, although it also contributes to the overall signal. b) The use of signals from 3 and 6 pixels as proxies for particulate and molecular backscatter is a rough approximation. In contrast, Donovan et al. (2024b) describe a more rigorous method that involves determining crosstalk coefficients and accounts for the non-uniform intensity distribution across the Mie spectrometer. While a full implementation may be beyond the scope of this paper, the authors should explicitly acknowledge these simplifications and discuss their potential impact on the uncertainty in the derived cloud mask.
5. Can the authors please comment on how the backscatter from aerosols, e.g., Saharan dust, is classified by their approach. I suspect that such regions are classified as “clear-sky”. What are the implications on their studies of wind-cloud interactions?
6. Line 180: How can the total backscatter take values of exactly zero? I assume that, in case of strong signal attenuation, the values would fluctuate around zero due to noise. If a threshold was applied to discriminate such values, please specify.
7. 282: Please clarify the term “closest L2B wind.” Is proximity evaluated in both latitude and altitude, relative to the bin center? This is important when duplicating L2B wind values, which are themselves derived from grouped L1B bins. I recommend to refer to the bin grouping algorithm described in the L2B Algorithm Theoretical Basis Document which should also be cited in the manuscript: https://earth.esa.int/eogateway/documents/20142/37627/Aeolus-L2B-Algorithm-ATBD.pdf
8. Line 296: Including statistics on how many of the resampled winds labeled as “clear” and “cloudy” correspond to L2B Rayleigh-clear and Mie-cloudy wind types would strengthen confidence in the cloud-masking approach.
9. Line 325 ff.: Are the reported variations in wind speed shown in HLOS units?
10. 5(c)-(e): The effective vertical resolution of the resampled winds appears similar to that of the L2B product. Is this due to wind speed duplication across adjacent 3-km bins? If so, this should be discussed. Additionally, a finer vertical resolution of 480 m was expected, but is not evident, particularly at higher altitudes.
11. Lines 552 ff.: The statement that “even at the coarse vertical and horizontal resolutions of Aeolus, it is possible to capture significantly different wind speeds within clouds and within their surroundings in shallow convection” may be misleading given that Fig. 9 presents airborne lidar data. This data has much lower random errors than Aeolus which enables to resolve the depicted features. Please clarify this distinction.

Technical corrections:

1. 270: Define the term “observation_type” and specify which values represent “cloudy” and “clear.”
2. Ensure a space between numerical values and their units throughout the manuscript (e.g., “480 m” instead of “480m”).
3. 7, 8: Extend the color scales to show the full range of wind shear (Fig. 7) and wind speed (Fig. 8), respectively.
4. Caption of Fig. 8: Insert a semicolon after “February 2020” to improve readability.
5. 528: Define the acronym “LES” at first use.
6. 550: Add missing punctuation at the end of the sentence.
7. Caption of Fig. 9: Specify that the data shown were acquired with the airborne Doppler wind lidar.
8. 4.3.2: At the beginning of the section, restate the time period covered by the dataset to aid reader comprehension.
9. 10: Use more distinct colors for “TrSc u_clear surrounding cloud up” and “Cu u_clear surrounding cloud up,” which are difficult to distinguish.
10. Caption of Fig. B1: The last sentence appears incomplete. Please revise.
11. Line 792: Correct to: “can be explained”
12. B6: Ensure consistency between the labels in the figure and the text: the terms “U_clear_neighbor_cloud_up” and “U_clear_neighbor_cloud_down” differ from the “surrounding” terminology used elsewhere.

Review of egusphere-2025-2065 ("Wind-cloud interactions observed with Aeolus spaceborne Doppler Wind Lidar") by Titus et al. 

The authors introduce an all-sky interpolated wind dataset from Aeolus, and showcase some very promising applications. My main criticism here is that, for all the technical prowess of putting the dataset together, the authors show a limited amount of results, and apply a relatively simple analysis on it. There is also a lack of depth in the discussions with earlier literature. At times I felt that, with the current amount of actual novel results on wind-cloud interactions (relative to the dataset's complexity), this manuscript was maybe fitting more in AMT. 

I still want to highlight the high ceiling of this study -- if the authors increase discussion depth and show more detailed analysis in some specific sections, this can become a very strong paper. Especially if specific comments 1+2 are addressed, this manuscript and follow-on studies will benefit a lot in my opinion. Below I list my suggestions to strengthen the current manuscript. 

I should point out I am not an expert in cloud mask derivation and backscatter signal handling, but ref#1 fortunately has good insight on that part. 

#####################################################################

#####################################################################

#####################################################################

# Specific comments:

#

#

# 1)

# section 2.1, l.160-165

# 

This is simply linear interpolation, correct? Call it as such.

The largest possible vertical mismatch between datasets (going from the original vertical grid to the interpolated grid) is half the vertical grid spacing that you define.

Since the 480m won't necessarily coincide with the RBS of Aeolus, how about interpolating both Aeolus and CALIPSO datasets to a better resolution e.g. 200m or better?

--> This way you minimize the portion of gradients that are lost in the interpolation, and avoid the largest mismatches in vertical location of the data.

While this aspect is not crucial for this particular manuscript (you show statistics of shear on 2km vertical scale in the end), you should make the reader aware that some of the gradients from Aeolus dataset are partially lost with these specifications. 

# 2)

# section 2.2.2, l.252-255: 

# 

Nice that you explain and detail this immediately, I was very intrigued by this feature in Fig. 2 from the beginning.

For a better comparison, would this motivate coarsening CALIPSO horizontal resolution to Aeolus'?

I would strongly encourage doing so, in combination with my comment about vertical resolution of the datasets (1).

--> In the end you want to avoid

a) the artifact of coarser horizontal resolution on cloud fraction values (which to some degree might affect clouds at other layers too, impossible to rule out or quantify in the current dataset)

b) losing any of the stronger wind shear values from Aeolus

If you performed your horizontal-vertical interpolation this way, results would become a lot more robust, and the comparison settings would be ideal in my opinion.

Follow-on work is planned by the authors with this dataset, and if the authors want to do more detailed research (which will eventually require shorter vertical scales than 2km), thoroughly addressing specific comments 1) and 2) would really help to get robust results out of their dataset. 

# 3)

# section 3.2, l.316-318:

# 

why not interpolate to 480m as in the previous comparison?

I would strongly suggest to make the same vertical interpolation as in section 2.

If you decide for 480m, you can average 5 consecutive bins with appropriate weighting to account for less representativity at the upper-lower edges.

# 4)

# Beginning of section 4.1:

# 

Not sure this part (Fig.6 and related discussion) provides really novel results, it could be moved to the appendix/supplement

See below some suggestions to strengthen this section (Fig.7).

# 5)

# Figure 7:

# 

This is not really representative of shear and shows basically what the circulation is at 10km.

Preferred format:

- take e.g. the 8-10km layer

- calculate ABSOLUTE shear between the levels within that layer from your resampled data --> you'll have a distribution for JJA

--> show e.g. the median of that

A figure 7 with panels like that for several height ranges e.g. a) 8-10km b) 12-14km and c)16-18km

--> would be a lot more informative of shear conditions and a more novel contribution to literature.

Note that can be easily obtained from your dataset as it is.

+ How come regions where the ground is above 2km are not masked out? What level is being used in those cases?

# 6)

# Method to average shear in section 4.3.2

# 

with this method, if you have large westerly and large easterly shear values within your sample, they will average out.

It'd be more interesting to show the profiles of S, with the averaging done over their absolute values. 

I.e. |S|_(z)

In Fig. 10, you only show shear within the cloud, and the equivalent clear surroundings. 

Could you show a third panel with the same but above the cloud top? I.e. difference between u_cloud_up and +2km. 

--> Would be very valuable to compare shear values within and right above the cloud

Again preferably as |S|_(z) 

# 7)

# l.612-620

# 

Very basic statistical concepts are attempted to be explained here, and they are coming out wrong. 

What you call your random error, decreasing by factor N^(1/2), is simply the standard deviation of your mean value.  

"The random error is thus smaller than the typical horizontal wind speed difference... " 

--> The only valuable use of your std of the mean of cloudy layer and (std of the mean of) surrounding clear sky; is to tell whether the difference between them (the two means) is significant. 

Which is the significance testing that you apply in Fig. 10? Two-sided t-test, Welch test? 

"documented to be several ms -1 generally." --> firstly, references are needed for this statement, and secondly, the documented random errors of Aeolus Rayleigh clear or Mie cloudy measurements are not calculated the way you do and thus are not comparable. 

I suggest to remove these lines. 

# 8)

# text of section 4.3.2 in general 

#

No discussion about tropical cloud or shear literature whatsoever. Needs to be expanded and highlight what this analysis adds upon previous works. 

# 9)

# 

#

Would it be possible to compare the wind on the cloud's uppermost layer from your all-sky dataset (that you show in this manuscript), compared to the collocated Rayleigh-clear value -- which you substituted by cloudy wind.

It would be super interesting to see whether there's agreement or some systematic differences between the two.

Especially as most studies rely on using Rayleigh-clear winds, this would really show the added value of using your dataset for a better estimation of wind shear around (and atop of) clouds.

#####################################################################

#####################################################################

#####################################################################

# Minor/technical comments:

#

# Title

This is more of a demonstration of the capabilities of your curated dataset, results on wind-cloud interaction itself don't really go in-depth.

I suggest adjusting the title accordingly, e.g.: "Demonstrating Aeolus capability to observe wind-cloud interactions with a merged all-sky dataset"

# l.20-21: please specify this is shear **inside** the cloud.

# l.63-68: please provide a summarized version for the intro, such detail belongs more in data/methods section.

Make the intro less technical, focus more on the key things you've done that make this novel approach valuable.

# same for l. 71-76

## Figure 1: Can be moved to an appendix or supplement, not really relevant/necessary here, geometry of HLOS is very well known.

# l.113: the airborne DWL, please state here from what campaign.

# l.124-131: you need to name here the data requirements for your study: why you process L1A yourself, adjust to CALIPSO vertical resolution for comparison, etc..

# l.134: it reads very awkward to me, that these 6 steps end up with letter 'g' as the final product... please number them 1-6 accordingly.

# l.190-194: please state here and/or at the beginning of this subsection, whether the method is similar to previous ones for cloud detection, and benefits of making your additions.

# l.212-214: this is the kind of detail that needs to be brought up much earlier, to reassure the reader that the method/comparison is not coming out of nowhere ;) and has been validated before

+ Also I think the AMT paper from 2022 by Feofilov should be cited instead of the 2024 book chapter.

## Figure 2: panel c) please mask out insignificant values.

+ adding a panel d) with same as panel (c) but standard deviation of the differences, would be very helpful in my opinion.

It would show which regions are the most uncertain. In (d) no significance testing would be needed.

# l.318-320: please start a new paragraph somewhere around this point, the current one is too long and information-dense.

# l.338: "not shown" --> please add figure to the supplement to support this statement.

## Figure 4: please increase the size of the axis labels and the legend

# beginning of section 3.3: Feels a bit like coming out of the blue, please motivate at the beginning of subsection 3.3 and at te beginning of section 3 that this case study serves as a great example of your resampling output.

Also, please name the tropical cyclone in the subsection title and text.

# l.382: 'cyclic season' sounds very awkward, do you mean hurricane season?

# l.383-386: state the name of the TC and the date in this paragraph

# Figure 5:

- caption: specify convention of positive-negative winds. (even if one can guess from the figure itself, and mentioned elsewhere in the text, should be specified in the caption)

- In panels d+e, I suggest adding a contour line (e.g. grey for contrast with dark blue) surrounding the cloud mask for a better reference

## Figure 6: you may cut the figure at 65° latitude then

+ un-saturate the color scale, you can go till 40m/s

+ I suggest to make color separations every 2.5m/s, as it is now, one cannot tell by eye from the figure whether it is 20 or 25m/s

# l.480: very awkward wording --> "As a result, IT is also .... WHERE Aeolus retrieves... "

# l.484: " that are typical of a strongly stratified free troposphere."

--> I'm not sure this is general knowledge in shear-related literature, do you have a reference backing up those numbers?

# l.505-506: is this a novel finding? I'd like to see some further discussion with more references in this section

# l.508: more up-to-date literature on Indian ocean cloud cover would be nice to discuss here as well.

# l.541: "We performed a sliding average vertically of 500 m" 

--> For a second time on the average profile, or is this just a repeated phrase? (you mention this on lines 538-539)

# l.549-550: stronger wind shear around the cloud-top inversion layer is a common occurrence, right?

# Figure 9: there is no (a) and (b) in the figure itself

# l.591: please show the map here, not in the supplement. 

# l.655: Is this really unexpected? You mention later this agrees with modeled results from the year 2000.

# l.658-659: Are you sure these values are not for R²? Fig. 11 shows anticorrelation...

# Figure 11: this is just a correlation plot, you're not really quantifying convective momentum transport here, just validating K-theory relations. Please rephrase the subsection title to make it more accurate. 

# Figure B1 caption: incomplete, ends abruptly

# Figure B2: this figure fits more in the main manuscript in the methods section - or in the corresponding results section

# l.970: reference doi link incorrect, should be amt-15...