General Comments:

This manuscript presents a new dataset containing triple-frequency, vertical-pointing radar (CloudCube) data during the EPCAPE field campaign along the California coastline. This dataset is both unique and useful in that it contains co-located G-band observations (238.8 GHz) with lower frequency channels (Ka- and W-band). Importantly, these are the first atmospheric observations at this frequency (238.8 GHz), which can detect much smaller hydrometeors (cloud and precipitation) than the Ka- and W-band radars. Though I’m unfamiliar with calibration techniques for radars, the explanation is clear and sufficient. The dataset itself is easily accessible on Zenodo and complete, with any missing data explained (Fig 3). This dataset can help inform on choosing the appropriate radar frequencies for future field campaigns or satellite missions, such as the Atmosphere Observing System (AOS) mission. I recommend the manuscript be accepted with minor revisions for clarity.

Specific Comments:

L9-10: Though it’s implicit, it would be helpful to specifically call the radar “ground-based” at some point early on. The name CloudCube initially gave the impression that this could be a spaceborne instrument.

Fig. 5 caption and L211-213: “short-range horizontal streaks” and “sporadic vertical streaks”: It’s difficult to see what features are being pointed out. Perhaps describe where in these plots (time, height)

L71,140 and elsewhere: the use of “range” to describe the height above the radar is somewhat confusing. I think explicitly stating somewhere that range = distance from radar would be helpful to better understand “blind range,” “close-range,” “unambiguous range,” etc.

Technical Corrections:

L38: change “this kind” to “these kinds”: “...suitable fit for these kinds of measurements…”

L287: “...blue areas respond to particle diameters below the Rayleigh limit…” should this be “blue areas correspond”?

Multifrequency radar observations of marine clouds during the EPCAPE campaign

ESSD-2023-454 (Earth System Science Data)

Reviewer Comments Document

General Comments:

The paper by Socuellamos et al., describes a new multifrequency radar dataset that has important implications for future remote sensing precipitation missions. The dataset appears to be of high quality, and is unique in that it offers the first look into clouds using a high frequency G-band radar. These types of instruments are incredibly powerful for viewing very light intensity precipitation that is otherwise unobservable using traditional radar systems. The fact that these observations are combined with other, more common frequency radars, opens the door to exciting new possibilities in precipitation research. Overall, the paper is well written, and doesn’t contain any major technical flaws. While this is an important dataset to release to the public, and the topic is certainly relevant to the readers of ESSD, there are some improvements to the dataset itself (primarily related to documentation, missing data choices and data artifacts) that I’d like to see improved upon before I can fully recommend this for publication. Further, due to the different radars used here, and the multiple quality assurance (QA) and calibration techniques that were applied, this manuscript would greatly benefit from a data processing diagram to summarize the entire process in a digestible manner. With these changes, I believe the paper would be an excellent addition to ESSD.

Decision:

Major Revisions

Major Comments:

1. The most substantial comment I have for this paper is related to the general structure, formatting and description of the dataset itself (which is the focus/primary product from this work). While the data contents are certainly useful, its current state could be improved and made more accessible by following CF Metadata Conventions, and by addressing a few of the remaining data artifacts I encountered. For a detailed list of these comments, please see my specific notes in “Dataset Comments” below.
2. Further, there was a fair bit of QA done to improve the quality of the final datasets. This is great, but when combined with the fact that there are multiple different radars here, it can be challenging as a reader to follow what exactly was done to which product. It would be beneficial to provide another figure early on that summarizes the data processing and QA steps. This is something that you could refer back to throughout the document to make the process clearer. I add this as a major comment, as it might be a bit of work to concisely condense all of this information within a single figure, but I believe it would be a valuable contribution.

Dataset Comments:

Each of these comments are in reference to the provided NetCDF datasets available for download on Zenodo.

1. Standardization: I would recommend rewriting the data variables in the NetCDF files using CF Metadata Conventions where possible (https://cfconventions.org/). Using standard properties like standard_name, long_name, missing_val within the file metadata makes accessing and dealing with the data much easier for consumers.
2. Data Artifacts: I noticed an odd, unphysical looking band of reflectivity in the G-band data on the following days: 040123_004927_Multifrequency.nc, 033123_194324_Multifrequency.nc, 033123_224802_Multifrequency.nc, 033123_234344_Multifrequency.nc. Do you have some insight into what is going on in these cases? I wonder if there is a way to mask these regions, as they might lead to further derived issues later on (e.g., you can see problems in the reflectivity ratios too). Perhaps these issues are discussed in the paper and I just missed it?
3. Missing Values: I also noticed an issue with the missing values prescribed on the following days: 041323_232910_Multifrequency.nc, 041323_235527_Multifrequency.nc, 041423_000835_Multifrequency.nc, 041423_002144_Multifrequency.nc, 041423_003452_Multifrequency.nc, 041423_004800_Multifrequency.nc. It appears that when the blind zone was being masked, instead of NaNs, -inf values were assigned. I would recommend sticking to a single missing_val type (NaN), as -inf values are quick to saturate data scales and make plots generally unreadable.
4. Timesteps: I would also recommend avoiding non-standard time steps. It can be a bit of a hassle to implement this, but it saves users a lot of time trying to figure out when time step 0 starts in a file (reading that info from the filename is lost if ever stripped/renamed). Ideally, the time axis would have a UTC variable that begins at the same point for each file (e.g. midnight), and would have exactly 86400 time steps (seconds in a day) with missing NaN columns. A similar comment for vertical extent, why do some files (e.g., 041123_220747_Multifrequency.nc) have a larger vertical extent than others? Shouldn’t this be set to a fixed value for consistency? If it is to save on file size, you could also deflate the NetCDFs considerably (e.g., deflate level 2) as the float64 double you are using here is likely excessive in terms of required precision for these observations.
5. Dimension Variables: I don’t think the height and time variables are correctly set as data dimensions (files are currently using Nr and Nt). To be clear, it is fine to have those variables as dimensions, but they are unitless in this case, and data users shouldn’t have to go look at another variable to check what the heights are (i.e., you can package that together into a single dimensional variable in the NetCDF). 
6. Metadata: This is sort of related to point 1 on CF-conventions, but the variable metadata is generally lacking in detail, and there is no contact information in the global attributes. Ideally you want your dataset to be able to somewhat stand on its own without the associated manuscript and I would add these descriptors for that reason.

Minor Comments:

1. Can the authors comment on the applicability of the G-band radar for very light intensity snowfall? This technology seems incredibly powerful for observations of fine ice crystals, for instance.
2. While I realize that the experiment lasted 6 weeks, there are really only 6 days of comparable observations across all radars. I think this fact should be brought up earlier in the paper, as I was expecting there to be much more data than what really exists for all three radars.
3. What are the vertical extents for each of the radar instruments? Is this provided somewhere because it wasn’t clear to me from the information in Table 1?
4. The Figure 3 colors make this challenging to read if you are colorblind. I would recommend changing the palette here for accessibility. I would also add a bit more space between hatches on the ‘Data available but not provided’ for further clarity.
5. Figure 5 is really neat, what is this horizontal feature at about 1.75 km? Also, what date/time is this occurring at? I would include the date/time in the figure somewhere.
6. The process for Ka and W-band calibration in Section 3.3 is quite interesting. Is this a fairly standard procedure? I noticed you mention different assumptions (i.e., “we can assume that the hydrometeors are in vertical dynamic equilibrium, and that the population of particles contained within the G-band radar volume resolution are all the same size”), and this process must therefore have some associated uncertainty wrt. the calibration that I feel should be discussed.
7. Figure 10, why do panels (a) and (d) have different amplitude scales? Should these not be the same, since we are comparing between the two?
8. Figure 11 is great, and I feel that you could have a version of this figure earlier on (without the reflectivity ratios) or split this figure up, to illustrate to the reader the primary differences in the retrieved signals from each radar for the same cloud system. I was disappointed that this was left until the last figure, as it is excellent motivational material.
9. Code availability. Is the processing code/QA code publicly available in some repository alongside the dataset? This is always nice to provide when possible.

Specific Comments:

These are mostly small grammatical changes I would recommend for enhancing the overall readability of the manuscript.

Line 38: “this kind” -> “these kinds”

Line 118: during -> for

Line 164: What is ITU? I assumed it was a reference, but I don’t see it in the reference list? Or was this defined somewhere else that I missed?

Lines 234-236: I would reorganize/rewrite this sentence, as it isn’t clear to me what you mean here (I’d also delete “deep detail”).

Line 237: I would also rewrite this sentence as “one can identify the times of most interest” is a bit verbose for what I think is being said.

Line 242: “using for that purpose” -> using

Line 287: respond -> correspond

Lines 335-336: “can reveal valuable insight about the particle size and size distribution” -> “can reveal valuable insights into particle size distributions”

Line 336: performing -> deriving