The retrieval of cloud microphysical properties from passive visible-to-infrared satellite measurements is characterized by large uncertainties, and comparisons with in situ observations typically show limited agreement. In this study the authors make an attempt to attribute errors in satellite-retrieved cloud droplet number concentration (N_d) to errors in several underlying retrieval variables. The results show that there is a variety of reasons why the N_d retrievals can be off. The analysis is a useful contribution to the scientific literature.

General comments

Partly due to the limited number of cases the results are for the most part statistically inconclusive. I wonder if there are more aircraft measurements available that could be included, e.g. from the ORACLES campaign. In addition, it would be good to include a discussion on possible biases in the PDI measurements, since these are used as reference. Meyer et al. (2024) recently found systematic differences between PDI and  other in situ probes, which could be referred to.

Specific comments

P1, L15-16: This sentence reads strange: ‘cloud properties ..., such as cloud radiative effects, precipitation, and aerosol-cloud interactions.’. These are not really cloud properties but effects of clouds.

Section 1.1 is better placed in the Methods section rather than in the Introduction. The Introduction should also be extended by embedding the study in the existing literature, e.g., referring to earlier validation studies, and adding appropriate references.

Section 1.1: No details are given about the origin of the MODIS tau_c and r_e retrievals. Are these the ‘standard’ MODIS products (MOD06/MYD06 C6.1)? Which satellites (Terra or Aqua) were used for which cases, and did the authors analyse whether there was a (systematic) difference between them? Which shortwave-infrared channel was used? Was it 3.7 micron? These details are very important and should be included in the manuscript.

The writing in Section 1.1 is slightly sloppy and must be improved. Here are some examples:

-    The meaning of the symbols in Eq. (1) should be presented after the equation and not much later. Of course, further explanation can follow later.

-    It would be good to include the definition of effective radius, r_e, in this section.

-    ‘In order to estimate r_e, MODIS makes use of a weighting function’: this is not how it really works. The MODIS retrievals use shortwave infrared radiance measurements to infer r_e. And since the radiance in these channels originates mainly from near the cloud top (as quantified by the weighting function), the retrieval is representative of r_e in that region near the cloud top.

-    In Eq. (2), optical depth tau without a subscript appears. What is it?

-    ‘mu and mu0 depend on satellite position and correspond to the solar zenith angle and sensor zenith angle, respectively.’. Normally mu0 denotes the solar zenith angle. Also the solar zenith angle does not depend on the satellite position.

-    Eq. (4): r is not defined. Q_ext depends on r, but in Eq. (1) it does not.

-    Please make sure that regular words are not written in math mode. For example: ‘constant’ on line 45, top in z_top in Eq. (3).

-    L51-52: Physical cross section is usually called geometric cross section.

-    L71: Here r_e becomes a function of z, while before it was a retrieved quantity and not a function of z. Please correct the notation.

P101: ‘A discrepancy ... that agrees ...’: Is this sentence correct?

L114: This choice may make sense, but 60 to 90 m below cloud top does not correspond to the middle of the cloud (given that the cloud thickness is between 250 and 500 m, at least for the POST cases, Fig. 10). Please consider changing the term ‘mid-cloud’.  

L133-136: It would be useful to add some information on the typical time over which the aircraft measurements are aggregated. Given that MODIS sampling is instantaneous, the larger the time window, the larger errors due to temporal variability, including advection, become.

Eq. 11: This is a measure of uncertainty, but it is not the standard error. Furthermore, there is an implicit assumption that the errors of the individual 1x1 km2 observations are uncorrelated. Is that the case?

Fig. 1: in the axis labels, N should be N_d.

Fig. 1: Case 5 (08/14) is not labelled. (Same in Fig. 2. There also the leading 0s are missing from the labels.).

L171-174: Can you be more quantitative in what is considered good or bad agreement? A range of +/- 25% is depicted in Figs. 1 and 2, suggesting this is defined as the distinction between good and bad. However, Table 1 contains a case with an r_e difference of 17% which is considered bad agreement.

L184-185: In Fig. 1 there are two POST cases for which satellite and in situ observations are well within the +/-25% lines. Why is one of these not considered as good agreement?

Section 3.3: For reference it would be useful to include the cloud optical thickness of the POST cases.

Fig. 4: Liquid water content is referred to as LWC. This should be L, as elsewhere in the paper.

L207-212: It might be added that, consistent with the underestimate in MODIS cloud optical depth, also the inferred cloud geometrical thickness is too small, i.e. cloud base is too high.

L214: Remove ‘a’ before ‘twice’.

L216: PDI does not observe tau_c as such (but it can be inferred from the beta_ext profile), so this sentence is not correct.

L247: ‘MODIS assumes ..’: rephrase (an instrument does not assume anything).

Fig. 9: Here we seem to be looking at two POST cases that were not discussed before. Where are these located in Figs. 1 and 2? To improve clarity, you could consider to label every POST day (and not just the five of Section 3.3) with a case number (1 to 8), and include a table with the relevant MODIS and PDI cloud properties.

Section 3.4: Would it make sense to be a bit more quantitative about the impact of k on N_d? If k is slightly larger than 0.9, the deviation in N_d is between 10 and 15%. I would agree that this is indeed ‘not large’ (L251).

L274-277: ‘MODIS predicts ...’: rephrase

L280: Add ‘even near the middle of the cloud’ (deviations near top/bottom are common as mentioned before)

Reference

Meyer, K., Platnick, S., Arnold, G. T., Amarasinghe, N., Miller, D., Small-Griswold, J., Witte, M., Cairns, B., Gupta, S., McFarquhar, G., and O'Brien, J.: Evaluating spectral cloud effective radius retrievals from the Enhanced MODIS Airborne Simulator (eMAS) during ORACLES, EGUsphere [preprint], https://doi.org/10.5194/egusphere-2024-2021, 2024.

The authors assess the accuracy of cloud droplet number concentration (Nᴅ) derived from MODIS satellite observations by comparing it to in situ measurements from three field campaigns. Their main finding is that MODIS tends to overestimate Nᴅ, with discrepancies of ±50% or more. The authors suggest that these errors stem from variations in microphysical and radiative variables, though no single source of error is identified. However, the results are inconclusive due to the limited dataset, and further high-resolution vertical sampling is needed to establish statistical significance.

Review:

As a scientist less familiar with this specific topic, I suggest a minor revision. While the study presents important findings, its clarity and structure can be improved, particularly for broader audiences.

General comments:

The abstract does not clearly state the main aim or conclusion. Consider adding that this study focuses on evaluating the accuracy of MODIS retrievals using aircraft observations. At the end, include a conclusion stating that more data are needed to better assess the accuracy of MODIS retrievals.

Nᴅ is a key focus of the paper, but it is no really clearly defined in the introduction. Briefly define and explain why Nᴅ matters. Similarly, explain more terms like cloud optical depth and cloud effective radius for readers unfamiliar with the field.

The introduction is brief and would benefit from:

• A discussion of other methods or campaigns measuring Nᴅ and whether this study is the first of its kind.
• Mentioning other satellite measurements and the importance of accurate Nᴅ values.
• Including citations to situate the study within the broader research landscape.

In section 1.1, several parameters in the equations are not defined immediately and are only explained later in the introduction. This can make it difficult for readers to fully understand the equations when they first encounter them. It would be helpful to define parameters consistently and right when the equations are introduced. For example, Qext is first encountered in line 29, but its definition is provided later, in line 51.

Specific Line Edits:

• Line 7: “We find that MODIS Nᴅ is best interpreted as representative of the mid-cloud region, as there is almost always a greater discrepancy from in situ values near cloud top and cloud base ” This should be rephrased. Larger differences near the cloud top and base do not necessarily make the mid-cloud region more representative. Please clarify why the mid-cloud region is considered more representative.
• Line 15: Correct “cloud drop” to “cloud droplet.”
• Line 22: Rephrase “When we find…” for clarity.
• Line 33: Instead of indirectly describing the weighting function, state directly: “A weighting function is used to weight the impact of measurements on the satellite-derived variables.”
• Line 38: Add “the” before “cloud top” and provide a citation for the statement about peaking regions. Clarify how cloud optical depth is addressed in this context.
• Line 42: Explain that “r” represents droplet radius.
• Line 93: The term “level legs” is unclear. Define it for readers unfamiliar with flight terminology.
• Line 106: Rephrase “for each flight analyzed” to “for each analyzed flight.”
• Line 111: Clarify criteria used by flight scientists to select top legs.
• Line 112: Reconsider “Therefore” and provide justification for the liquid water content threshold (L = 0.05 g/m³).
• Line 117: Define “leg” where it first appears (Line 93).
• Line 130: Clarify what “Lad(z)” refers to, as only L(z) is defined earlier.
• Line 192: Explain what is meant by “cloud drop evaporation due to entrainment.”
• Line 236: Change “PDI observation” to “PDI observations.”
• Line 247: Rephrase “ranges 0.7 ≤ k ≤ 0.9” to “ranges from 0.7 to 0.9.”

This study evaluates MODIS cloud droplet number concentration (Nd) retrievals by comparing them to in situ measurements from three aircraft campaigns. The results indicate that MODIS tends to overestimate Nd, with discrepancies of 50% or more. The authors explore potential sources of retrieval errors and conclude that no single factor dominates the discrepancies. This study is a valuable contribution to the validation of satellite-derived cloud microphysical properties, although the limited number of cases constrains the statistical conclusions.

I recommend publishing this manuscript, as it presents valuable findings. The study is well-structured, and the results contribute to our understanding of limits and merits of satellite-based cloud’s microphysical retrievals. However, the authors may consider the following comments and suggestions to further strengthen their analysis and discussion:

The PDI is treated as the reference dataset in this study, but systematic biases in PDI measurements should be acknowledged and discussed. The authors may consider exploring how potential biases in the PDI might influence the comparison with MODIS retrievals, particularly in the context of cloud droplet number concentration estimates. A discussion of known uncertainties in PDI data and their implications for the conclusions drawn in this study would enhance the robustness of the analysis.

The introduction would benefit from a more detailed discussion of past validation studies on cloud droplet number concentration retrievals. A clearer connection to previous validation efforts would help contextualize this work within the broader field of cloud microphysics and remote sensing. The authors may also discuss how their findings compare to and extend prior research, highlighting the specific advancements made in this study.

The authors may investigate whether there is a correlation between the bias in cloud base and other key parameters, such as the optical depth or the profile of liquid water content. If such correlations exist, they could provide insights into the nature of systematic retrieval biases and their physical causes. Addressing this point could strengthen the interpretation of the results and provide a more comprehensive view of potential error sources.

Moreover, entrainment can affect the assumptions made about both k and f. The authors may consider expanding their discussion on how entrainment variability influences these quantities and whether it introduces systematic biases in retrievals. A quantitative or qualitative assessment of entrainment’s impact on the assumed values would provide additional clarity. Given that entrainment may contribute to the observed mismatch in retrieved quantities, it might be beneficial to attempt an estimate of entrainment at the cloud top using MODIS data. Specifically, high spatial variability in brightness temperature or effective radius over some area of the cloud deck could serve as a proxy for entrainment activity.

The authors may consider evaluating whether this estimated entrainment degree correlates with biases in the retrieved quantities, which could provide further evidence of the role of mixing in retrieval uncertainties.

These comments are intended to refine the discussion and improve the clarity of the manuscript. Overall, I find this study to be a valuable contribution to the field and recommend its publication. I believe the manuscript remains suitable for publication even if not all the above suggestions are fully met.