Review

Included here are the more substantial and/or longer comments (i.e., notes, rather than comments, from the attached manuscript file). Grammatical corrections and suggested wording for clarity are included as annotations on the attached manuscript file, but not here. Only the yellow highlighted text and red comments need be addressed directly from the manuscript file. The associated line numbers of those comments are listed at the end of this document. 

General comments:

The subject of this study is microphysics in tropical cumulus congestus, with a focus on the number concentration and effective radius of cloud droplets near cloud or thermal tops. Retrievals of these quantities at/near cloud top obtained from the airborne Research Scanning Polarimeter (RSP). These were compared to analogous quantities from a set of simulations with bulk 2-moment microphysics and with bin microphysics. In-situ measurements of the DSDs were also compared with simulated DSDs. In addition, a thermal-tracking analysis method was also used to gain insight into the evoluation of microphysical quantities in thermals, and how it varies with microphysics. 

The manuscript is generally well written, well organized, and contains excellent graphics. Only minor revisions are recommended. 

The more substantial and/or longer (but still minor) comments:

1. line 149: why is k needed if Reff and veff are both retrieved? Do you want k to relate Rv to Reff?

2. lines 420-21: It is not clear how this works; explain. Do you mean to say: weak pcp production allows cloud droplets to accumulate in an outflow layer?

3. lines 455-58.  An alternative conclusion: dynamics matters,  but its effects are accumulated or averaged. Drops are a product of their histories not their instantaneous environment.

4. lines 516-17: Fallout of rain would not affect supersaturation, right? But rain is a sink of Nc so increase of Nc due to activation is not clearly evident in presence of rain.

5. section 4.5: Where in a cloud (relative to the current cloud top) is a thermal typically located? At cloud top?

6. Fig. 13 might be less cluttered if streamlines were shown in a separate row, and omitted otherwise

7. Figure 15: Because thermals presumably entrain at different fractional rates, I would expect the variability about the means plotted in Fig. 15 to be large. I recommend giving the reader some idea of the variability among thermals by adding shaded error bars to these plots, for the CNTL and FIXED_AERO_NO_AC profiles.

Ideally, one would stratify the thermal-tracking analysis by fractional entrainment rate. What is plotted in Figure 15 is what a bulk cumulus parameterization would be asked to predict, as opposed to one based on thermals with different fractional rates, such as Arakawa and Schubert (1974), and examined in terms of parcels with different entrainment rates by Lin and Arakawa (1997, Part 2) (https://doi.org/10.1175/1520-0469(1997)054%3C1044:TMEPOS%3E2.0.CO;2).

I recommend making the suggested change to Figure 15 and to add a discussion of what is plotted in terms of a bulk model vs a 'spectral' model.

8. lines 535-36: It must be activation of newly entrained CCN. The concentration of CCN from cloud base will be diluted by entrainment, and these CCN alone cannot maintain a constant Nc (per unit mass) with z. 

9. lines 537-38: Define secondary activation. If this means activation of

newly entrained CCN, then this hypothesis is the same as the first hypothesis. If it means reactivation of CCN from lower levels (i.e., not entrained at the current level), then it  cannot offset entrainment dilution.

10. Section 5.1: This short review of convective microphysics in large scale models may be useful to some readers. However, there are no specific recommendations. It could, and perhaps should, be omitted in the interests of reducing the length of the manuscript.

If you do retain this subsection you may want to point out that adding complexity to convective microphysics schemes usually entails tuning of the microphysical parameters, in particular, rates of conversion to precipitation, rather than using results from cloud resolving models.

Are you aware of any convection microphysics schemes that are aerosol or CDNC aware?

11. lines 565-566: Explain/justify/revise this statement. 

Condensed water amount is almost entirely a function of altitude not updraft speed.  That is why updraft mass flux is generally sufficient to predict the large-scale convective condensation rate rather than requiring vertical velocity and updraft fractional area separately.

12. lines 569-581:  Before such details as you discuss in this paragraph are included convective microphysics schemes, we may be using convection-permitting models. 

13. Lines 582-591: Perhaps this paragraph should be moved to some other section because the topic seems to differ from that in the rest of the paragraph.

14. Section 5.2 heading: "Training" sounds like ML. Is that intended? If not, then perhaps you could use ‘Evaluation and improvement' which is of course what training is, but doesn't sound like ML.

Other comments: (see annotatations in manuscript file)

Lines 69, 80, 116, 123, 127, Eq. 2, 145, 157, 211, 221, 254, 282-284, 291,

343, 361-362, 379, 396, 399 (3 comments), 412, 419, 420, 425, 429, 460, 497, 498, 500, 504, Figure 13 caption, 521, 560-561, 565, 568, 636 (2), 642, 643, 645 (2), 650, 653-654, 664, 667, 674