Summary: Zhang et al. describe a novel conditional Monte Carlo subsampling approach (cMC) to investigate the role of solar heating on marine stratocumulus cloud evolution. This approach allows them to artificially inflate the ensemble of large eddy simulations. In particular they investigate how solar heating changes the relationship between cloud droplet number concentration (Nd) and cloud water path (or liquid water path, LWP, in this case). They find that the Nd-LWP relation has a negative trend during the night time (in the absence of solar heating) and becomes less negative during the day (in the presence of solar heating), converging to a value around -0.2 in the late afternoon. This so-called "buffering effect" whereby the Nd-LWP relation is buffered back towards zero by the solar heating is attributed to the strong dependence of cloud absorbed SW radiation and cloud LWP; or simply, that thicker clouds absorb more strongly and thus thin at a faster rate than thinner clouds. They discuss the implications of these results for the time-dependent efficacy of aerosol injection, for example in the case of climate intervention via marine cloud brightening (MCB).

General comments:

Some limitations of this study, especially in it's relevance for MCB, include:

1. The simplistic and unrealistic assumption of the size and composition of the aerosol particles (ammonium sulfate, lognormally distributed with mean radius of 100nm). A more realistic MCB experiment would include seeding from larger, more hygroscopic sea salt spray.

2. The fixed (across the ensemble) prescribed SST and large-scale divergence which under-samples the relevant dynamical space these clouds occur in.

3. Of course, the mentioned restriction to the non-precipitating regime.

Overall, I think this paper is very through, well-written, and will be of interest to the aerosol-cloud interactions community. The cMC approach is also quite interesting and may be of broader interest outside the aerosol-cloud interactions community.

Specific Comments:

- cMC sampling:

    - L144: Do you impose any threshold on the correlation coefficient for the regression when you build these subgroups, or just the slope? 

    - Fig 1. I'm curious if it's possible to give an indication of the strength of the correlation in this figure. How robust is the slope over time? Is the r-value similar across these sub-ensembles? across time? Is the r-value always fairly large? If not, what does that indicate? And can that be shown in the plot? Maybe when r < 0.5 (or some other value state) you could make the lines more transparent?

- L173: Can you quantify the entrainment velocity from your output? How does entrainment velocity quantitatively depend on droplet radius in these simulations?

- L204: If phi_ENT tendency is calculated as the residual between the total phi tendency and the phi_RAD tendency, then what is the "residual" referred to here?

- Fig 2. I would just recommend using a different color palette in panel a) to distinguish these RAD and ENT components from the coloring of the sub-ensembles used in the other panels.

- Fig 3. Can you use the same colors (shade of blue) as the sub-ensembles that you are referring to? It says in the caption that blue refers to blue, but since it's a different color blue this is a little confusing.

- L260: This should be caveated with the assumption that all other conditions are unchanging over time, besides Nd.

- L268: Can you clarify the details of your regression two-sided t-test to determine the near-zero slopes? What is the p-value?

- L272: Re: my comment of limitation #1, I suggest adding a comment here in the text to clarify that this implication for MCB is limited by the opportunistic sampling strategy. Because you do not simulate actual injection, the "injected particles" necessarily come from the same underlying distribution as the background particles. However, in a realistic MCB simulation you would probably seed with larger, more hygroscopic particles to resemble sea salt. The distinction here may be subtle, but so is the prospect of MCB efficacy. The "aerosol perturbation" then referenced is really more similar to a perturbation in the background aerosol, some co-variability between meteorology and aerosol, than a deliberate MCB seeding.

- Fig 6. Is the "aerosol perturbation" time the same as local time? Can you add back the grey shading that you have on all the other figures to indicate night from daytime?

- Fig 8. How is the cloud aspect ratio defined? Where does the scaling come from? What are the assumptions that go into this scaling? Please give more explanation and a citation, if one exists. 

- L364: You discuss how advection will change the large-scale forcing (SST, subsidence). But again, re: my earlier point about the limitations, you also should add to the discussion how the variability in initial large-scale conditions may alter these results. Some of these limitations may be introduced earlier in the paper.

This paper uses a conditional Monte Carlo subsampling approach to analyze the diurnal response in the liquid water path (LWP) adjustment to solar heating in non-precipitating stratocumulus using large-eddy simulations, finding that LWP has a strong dependence on shortwave heating which act to modulate the overall adjustment. Overall, I think this is a strong paper that is well written and only have a few critiques that I want the authors to address prior to publication.

General Comments:

1. L80-81:
   1. Assuming you chose 24-hour simulations because you want to investigate the diurnal variability in the LWP adjustment, would there be any benefit in running longer simulations (e.g. 36, 48, or 72 hrs.)?
   2. All your simulations start at 18:40 local, do your results depend on when the simulations start (i.e. the overall trends in figure 1)?
2. L141: Why did you choose the thresholds on cloud-top height, surface sensible heat flux, and 800 hPa relative humidity listed here?
3. L352-360: How frequent are non-precipitating stratocumulus and, given recent observational studies demonstrating the diurnal impact of cloud-top entrainment on the LWP adjustment may also be modulated by precipitation (e.g. Smalley et al. 2024), how representative are your simulations of the real world? On a side note, are there any plans in the future to do similar analyses of precipitating stratocumulus cases?

Minor Comments:

1. L20: “lead to more, smaller” sounds awkward. Maybe change it to “leads to an increase in smaller”
2. L35: “Making the quantification of LWP adjustment” should be “making the quantification of the LWP adjustment”
3. Figure 1: For ease of interpretation, could you move the threshold values of dln(LWP)/dln(N_d) listed in the caption to a plot legend instead?