Minor comments

• l. 90 - "modelling of the optical properties of cirrus clouds is a formidable problem..." I share the authors' pain. The rest of the paragraph provides a nice review of previous efforts tackling this problem, but I'm not sure it brings a lot to the current discussion.
• section 2 -- the number of instruments is a bit large, and a table or a figure summarizing them would be nice.
• l. 132 - "the resolution is 10 s" -- I was surprised by this (relatively) very long sampling interval. It is worse than the 5 s mentioned in Cairo and al 2011. This means the MAS samples one point every 1 or 2km, which is quite a large distance. Are there ways to improve this?
• Figure 1 (1) : it is unclear to me why in Figure 1 the highest concentrations of points are not located at BR=1 ? Since clouds show up on 7 hours from the entire 35 hours of measurement, I would expect most of the points sampled by the MAS to be cloud-free, and thus to produce a BR=1 measurement. Could you please clarify my misunderstanding?
• Figure 1 (2) : could you specify in the legend the total number of data points?
• l. 158: "a negative trend with respect to temperature can be discerned" this is also reportedly the case in CALIPSO data, Sassen et al 2009 -- but I agree this trend is frustratingly hard to document quantitatively.
• In Methods : couldn't the CCP probes provide info about the particle aspect ratio? Such info could then be compared to the AR retrieved from the beta fit. Or do you think the aspect ratios from probes are too uncertain?
• l. 250: there is something wrong with the URL, I'm guessing the underscores were swallowed by the formatting system.
• l. 255: clicking the URL here brought me to the American Express website.
• l. 263-266: there is a problem with formatting here
• l. 272: missing period
• l. 371: this paragraph suddently mentions beta NC. I think this is the first time in the paper that beta NC is mentioned, and I don't think it is defined. Is beta NC the best match for the measured beta among the simulated beta AR NC ?
• Figure 3 (1): have you tried to create a scatterplot of depolarization vs AF? vs AR? of AF vs AR? Could you perhaps write the various labels of the vertical axis in the same color as their associated plot? It would make it easier to parse the figure.
• Figure 3 (2): is the AR retrieval at each time stamp completely independent from the previous retrievals? In some parts of the plot it looks like the retrievals are "stuck" at AR=3.
• Figure 3 (3): Figure 3, like most of the figures, is a bit on the small / crude side. This makes reading the figure harder than it should. Could you please produce higher-res versions of  figures next time?
• Figure 3: I understand this is a bit outside the scope of the paper, but it would be very interesting to document the spatial scales over which depolarization is homogeneous.
• l. 400: "(19000s and 21000s)" what are those numbers? Can they be found somewhere on a figure? On Figure 3 times only go from 33000 to 38000.
• l. 400-405: Are the MAS/MAL depolarization ratios similarly correlated? Does the results from figure 4 suggest it is possible to use MAL measurements instead of MAS when more practical?
• Figure 4: Using a square aspect ratio in Figure 4 would make sense
• Figure 5 and 6: Having both as subplots of the same figure would help the reader. In Figure 5 I find it confusing that the y-axis says "Backscatter coefficient NIXE-CAPS" -- the NIXE-CAPS does not measure the backscatter coefficient. Same for figure 6.
• l. 419: "From the inspection of Figure 6 we can see an attempt to achieve a 1-1 correlation between the two datasets". I don't understand this sentence.
• l. 421: "the optical modelling completely fails" -- completely is perhaps a strong word, sometimes the retrieval is correct. But I agree it's perhaps worse than chance, the performance is very poor.
• l. 444 : "most of them are so called irregulars" how do you know this? Is this based on CCP imagery?
• l. 476: "We remind again that" perhaps with some editing it would be possible to avoid two reminders
• l. 477-485: This discussion is great. Based on Figure 12, wouldn't it be possible to combine the measured volume depolarization ratio and BR to retrieve the particulate/aerosol depolarization ratio (by extending the "line" these two measurements fall on all the way to the right), and use it instead of the volume depolarization ratio throughout your study? You might find better correlation with the AR time series.
• l. 490-493: It is not clear to me what is learned from Figure 13. The strong correlation between BR and Nice has already been shown before.
• Figure 15 is fascinating. Could you comment on whether particle orientation could have an effect on the volume depolarization ratio measured by the MAS, considering the specific MAS setup? Electric charging linked to lightning activity could lead to vertically-oriented particles, that would show planar faces to the MAS. Do you have any idea if the variable dynamic conditions (lines 500-504) could be disambiguated somehow by observation?
• l. 508: "a measured aspect ratio would be an obvious candidate" -- I was under the impression imagery from the CCP could provide this kind of retrieval.
• l. 579: From the start, it is not clear to me what is to be gained from following this path. Could you expand a bit on what you hoped to achieve with this?
• l. 590: "this may induce an unquantified bias in our presented statistics" -- I'm not sure I understand the point this paragraph is trying to make. Do you wish to state that the results you present here are not necessarily representative of all cirrus that can be observed from space? This could be said in a more straightforward way. Also, this is not a knock against your study -- 7 hours of in-situ sampling cannot be expected to cover the entire range of cloud variabilities.

Technical comments

• l. 408: "rangeis"
• l .441: "are hardly change"
• l. 461: "as instance" -- for instance?
• l. 475: "and and"
• l. 537: "MAS,."
• many references contain two URLs, one for the DOI (see next comment) and one that leads directly at the publisher's website. The DOI link should (in theory) always resolve to the publisher's website, so both are duplicates, and one could be omitted without loss.
• many DOI links appear as https://doi.org//https://etc (i.e. https:// appears twice). Clicking them goes to the right place, so I'm guessing this is a display issue.