Liquid water cloud plays an important role in the Earth's atmosphere, while a great deal of uncertainty still exists in observational cloud properties. Cloud droplet number concentration (N_d) is one of the most important cloud properties, which associate clouds with aerosol. This study compared four ground-based N_d retrievals from both lidar and radar retrievals with in situ measurements and investigate seasonal variations of N_d and r_e. Their results showed good agreement between ground-based retrievals and in situ measurement for overcast conditions. Also, the consistency between N_d retrievals and in situ measurement struggles with broken or low LWP clouds. By extending these retrievals to longer time period, obvious seasonal variations of N_d (r_e) values exhibits and are consistent with previous researches. I believe their evaluation promote our understanding of uncertainties of remote sensing data. However, the paper needs to be improved to be qualified for publication by addressing the following comments.

General comments:

1. Line 93-94: I think you need add more details about why you choose these four ground-based N_d
2. Line 121: literatures or documents of the instruments’ information showed in Table1should be cited here.
3. Line 207-210: This sentence is not easy to read. You may consider reorganizing the sentence structure to simplify and make it clearer.
4. Line 214: you assume a linear increase of LWC in radar retrievals. Are there any impacts of this assumption to the resultswithout regard to f_ad in this situation?
5. Line 218: you missed ρ_winequation 7 accordingto Mace (2000).
6. Line 232-233: I think you should explain more about the meaning of k* and point out why use k* to replace k.
7. What do the black circles mean in figure 4b?
8. Line 365: the word “greatest” may cause misunderstanding. You should replace it with another word.
9. Line 380: I notice that the higher N_dfrom in situ measurements actually appear on 02/07/2018, 06/30/2017 and 02/12/2018. If you have a specific criterion, you should point out here.
10. For more intuitive and easy reading, I think you should label the broken conditions in Table 3 and other figures that appears the date of 12 flight days.
11. Line 412-414: what are the possible causes ofthe inconsistency of r_em and N_d retrievals of NDROP VAP compared to FCDP?

Detail comments:

1. Line 28: delete the repeated “using the”.
2. Line28-30: this sentence has a linguistic flaw. I suppose you may want to begin a new sentence from “given”.
3. Line 59: cloud optical -> cloud optical depth
4. Line 95: 2018 -> 2017
5. Line 219: Miles -> Mace
6. Line 289: missing ‘cloud depth’ in your statement of figure 1.
7. Line 293: figure 1c -> figure 1d

Line 421: full name of TSI should be presented in your main body

This study aims to compare and evaluate four ground-based remote-sensing methods for retrieving cloud properties, with a focus on CDNC retrievals. CDNC is crucial for studying aerosol-cloud interactions and for understanding cloud processes but its retrieval from remote sensing still suffers from significant uncertainties. Numerous methods exist for CDNC retrieval that rely on a number of assumptions often unclear to the community. Therefore, this effort to summarise and evaluate here these methods against in-situ observations from 12 flights is timely and valuable.

The manuscript is well-written and appropriately cites current literature. The authors carefully describe the four established retrieval techniques and their respective assumptions. Comparisons to in-situ observations are done meticulously. I find the study convincing and well within the scope of AMT. I advise for publication after addressing the following comments and suggestions:

General comments:

1.  l. 146-147: This sentence is somewhat misleading as it could imply that this formula is independent of the DSD shape. While the equation may not require assumptions about the DSD shape, these will be needed when retrieving or computing the extinction coefficient and liquid water content. If that is what the author meant I would suggest rephrasing for clarity. If not, please provide more explanations.
2. l. 216: Does the logarithmic width of the lognormal distribution (0.38) relate to the k parameter? It would be interesting to know its equivalent value if so.
3. l. 233: Briefly explain the physical meaning of the cloud system k parameter, as defined by Brenguier et al., and why it's used for the VAP retrievals but not the other approaches.
4. l. 240-245: Specify the field of view for the instruments or the spatial resolution of the COD and LWP retrievals. It would also be useful to briefly state what the basic assumptions are for these retrievals, especially regarding the vertical distribution of cloud properties: are clouds assumed to be vertical homogeneous? In that case there would be an inconsistency with the assumptions from Eq 8, where the COD and LWP are used. This should at least be mentioned, as it can also partly explain why the VAP Nd retrievals are often different from (and more uncertain than) others.
5. l. 256-263: Opting for 9/5 over 3/2 would align better with the adiabatic assumption in Eq. 8. You mention Chiu et al. (2012) found better results using 9/5 but still chose to use 3/2. Please justify this choice.
6. l. 280: Indicate if clouds were fully profiled vertically, to clarify the later definition of <Nd>.
7. l. 317-324: It's unclear whether values of fad > 1 were set to 1 before computing the mean of 0.76. This can be problematic because the mean value may then not be very meaningful (the distribution would be far from normal). Why not use the slope from Fig 2d instead?
8. l. 384: Are these truly seasonal variations, or are rather day-to-day as mentioned later?
9. l. 501-503: Do you think the issues faced by VAP method are comparable to those faced by typical radiometer-based satellite retrievals of Nd? This conclusion may not be straightforward to draw but could be an very interesting message for the community. As you note in the next paragraph, ground-based retrievals are very valuable to evaluate the global satellite dataset.