See attached comments

The authors derive ice particle maximum diameter as a new variable from the well-established 2D Video disdrometer. The new parameter is then used to compare with remote sensing retrievals of maximum ice diameter from two field campaigns.

Overall, I find the paper nicely written and the quality of the figures and tables is sufficient. One might wonder whether the derivation of an additional size variable is really enough content for a scientific publication. But considering the comparison with two interesting campaign datasets, I think the manuscript is worth to be published in AMT.

I have several comments and suggestions which I would like the authors to address. I think the paper needs some major modifications before it can be published.

Major Comments:

Maximum diameter is generally one of the most ambiguously defined parameters for highly complex particles, such as ice or snow. I am of course aware of the fact that Dmax is widely used in retrievals, model parametrizations, and in-situ observations. But I am missing in the paper a discussion  that Dmax has been defined in the past in very different ways. For example, most airborne in-situ probes derive Dmax as the size of the circumscribing sphere or spheroid. You decided in your manuscript to use the Feret diameter instead. I would like to see a comparison of your “Feret-Dmax” to more common methods, for example, the maximum size of a circumscribing ellipse. You also don’t comment on how Dmax is defined in the remote sensing retrievals that you are using in the intercomparsions later. In the worst case, discrepancies in your comparison of the retrieval results and the 2DVD might at least be partly due to different definitions of Dmax. I think this aspect should be discussed much more.

I have some problems with Fig. 5 and the discussion of it in the text: You have steel spheres with well-defined diameters as calibration objects. Wouldn’t it make sense to produce two sub-plots with D(sphere) vs Deq and a second one with D(sphere) vs. Dmax? You say in L. 307 that “Deq agrees better than Dmax with the true diameter”. But how can the reader judge that if you don’t show the true diameter? With two sub-plots you could also plot the data as box-and-whisker plots that can much better illustrate the distribution of points (in your current Fig. 5 it is hard to estimate how many particles are clustering together). You write in L. 307 that Dmax overestimates the true diameter by 0-10% for diameters larger than 2mm. If I look at Deq=6mm or 8mm I can see several points being 2-3mm in Dmax away from the 1:1-line (those points exceed your 10% a lot). Why does Fig. 5 show a general (systematic) overestimation for Dmax in comparison to Deq? For perfect spheres I would expect the points to be around the 1:1 line. You say that Dmax is defined in a way that it has to be larger than Deq, but this is not clear to me in the case of spheres. Finally, are you applying a correction to the 2DVD data based on your calibration results? Maybe you wrote this in the text and I missed it.

Minor comments and typos:

1. L 137: “a one-dimensional particle shape” Shouldn’t this be two-dimensional?
2. L 187: “For the display of single particles, an own programme was written”. By whom? The authors or the manufacturer? How is the programme called and where can the reader find the programme? It is not mentioned in the Code availability section.
3. L 202: Ice density should be 916 kg/m³. Shouldn’t the equivalent volume actually taking the much lower density of snow into account? I mean, if you assume the snowflake volume to be composed of pure ice, the precipitation rate will be huge, or?
4. L 235-236: How exactly are the parallel tangents drawn around the particle image? Are they supposed to be perpendicular to Dmax?
5. L 239: I would say you need a very trained eye to see a dendrite in Fig. 4
6. Fig. 4 (caption): There is no thick red line in the figures.
7. L 267-268: I am surprised about your statement that columnar particles would fall slower than plate-like particles. For example, if you look at Fig. 6 in Mitchell, JAS, 1996 you see that columns fall way faster than plates or dendrites. Please revise.
8. L 282: “This dominant crystal shape has to be presumed in advance” Which is a major weakness of the LIRAS-ice retrieval in my opinion.
9. L 295: “with further assumed shapes” Do you mean different shapes?
10. 296: “require further requirements” Consider different wording.
11. 324: You say that particles with O<0.6 which are shown in Fig. 8d can be assumed to be columns. This is completely unclear to me: If you observe a horizontally oriented plate or dendrite, its height will also always be smaller than its width (so small O). I would argue that you cannot reliably distinguish between columns and plates based on O. Even more confusing is the following: The #4 particle in Fig. 8d is identical to your example particle in Fig. 4 which you describe there as a dendrite (!). Also particles #5 or 6 in Fig. 8d look to me much more like aggregates than columns. Please clarify.
12. Figure 7: I stumbled over the 20dB offset between the MIRA and the W-band radar. This offset is huge, so I would like to know a bit more about it. Which of the two radars had the “malfunction” and was it related to the offset? Was one of the two radars properly calibrated and how? I assume your Dmax remote sensing retrieval depends on the absolute value of Ze, hence a discussion of this aspect is quite relevant.
13. L 329-330: What is the sampling area of HOLIMO. Up to what maximum size can particles be reliably detected by HOLIMO?
14. Figure 9a: “Z” is defined as the 6th moment of the raindrop size distribution. As you are observing ice particles, you actually show the effective radar reflectivity factor which is usually denoted as “Ze”.
15. L 372: You mention in the article several times (for example, in the abstract) that LIRAS-ice is evaluated “for the first time” against ground-based in-situ observations. I would assume that a proper evaluation of a remote sensing retrieval is presented in the same publication as the retrieval itself. I have the impression that you want to convince the reader how innovative your evaluation in this section is. I would suggest to leave that judgement to the readers or at least mention it only once.
16. Figure 10 (caption): Consider including a reference to Fig. 9.
17. Figure 11 and discussion: I see in panel b) a systematic underestimation of at least 15% of N by LIRAS-ice for all assumed shapes. How big is the uncertainty of LIRAS-ice in general? How much would a realistic uncertainty in radar calibration of +-1.5dB affect your N and Dmax estimate? This should be discussed.
18. L431 + L. 392: In L. 431 you mention a temporal shift of 2 minutes but in L. 392 you say the shift is only 1 minute (60s). Please clarify.
19. L430: How can you estimate the vertical wind shear from Fig. 11a? If you infer this from the shape of the fall streaks, you have to assume particles with identical vertical velocity over time.
20. L433: “the statistical results (…) are biased by the varying vertical wind shear” Why don’t you test the impact of different time shifts on your statistical results? It sounds plausible but it would be more scientific to show it.
21. L 435-436: “in order to make measurements better comparable to precipitation data by other instruments”. I agree, but then your Dmax definition should be comparable with the more common definitions of Dmax used by those instruments. Or at least you should show and describe possible differences (see also major comments).
22. L 461: I can imagine that wind deteriorates the 2DVD calibration. But why is no wind-shield used during calibration? The calibration “table” has no side walls that could hold off wind coming from the side.