Dear authors,

I would like to elaborate on a problem with the paper in the hope that you can modify the paper using model data from the existing simulations LR450/LR150 without the need for any additional runs. My suggestion is based on what I perceive to be the main purpose of the paper, to compare how the representation of convective processes is improved going from LR450 to LR150. I am not concerned with domain or boundary-condition sensitivity.

There are two processes highlighted by the comparison of LR450 and LR150: (i) cool-pool generation, and (ii) cloud-edge entrainment. For both issues, I think we agree that LR150 is the more realistic. However, both issues require some clarification and further illustration. Verbal comments in the text do not illustrate the difference in causes of cold-pool generation and effects thereof. Comments regarding entrainment are confusing and possibly wrong (page 16, line 27).

Unless the above sentence "...indication of stronger entrainment..." is a typo, you appear to suggest that LR150 may have stronger mixing along the vertical cloud edge. This reasoning seems incorrect to me, but of course I may be wrong about it, or may have misunderstood your intent. Higher resolution invites sharper cloud edge with less, not more, transport to/from cloud and its clear-air environment. This is appreciated by means of a simple scaling argument: eddy diffusivity being proportional to grid-scale squared.

Given that the environment is dry and therefore entrainment dilutes buoyancy, less numerical mixing (i.e., at finer resolution) is better for buoyancy-induced updraft and further vertical development of cloud. This sentiment is echoed precisely on page 19, line 4. The notion of undilute "hot towers" figures prominently in our understanding of tropical deep convection in MCSs and is instrumental in midlatitude warm season also.

Comparing the two simulations, we should not expect clouds to look the same or to appear in the same time and place exactly. That is ok, however, as it is straightforward to infer general conclusions about how cloud properties differ between the simulations. Here are some suggestions:

1) create a suitable metric of microphysical gradient(s) along the vertical cloud edge
2) take an ensemble of clouds from each simulation, construct a CFAD = contour frequency by altitude diagram for one or more of these metrics

https://www.google.com/search?rlz=1C1GCEA_enUS783US783&sxsrf=ALeKk00JT3HlVelP3K5mlfil_Gq8R-Grvg:1587532072999&q=contour+frequency+by+altitude+diagram&tbm=isch&source=univ&sa=X&ved=2ahUKEwj38Z7qofvoAhWTtZ4KHan1DyEQsAR6BAgJEAE&biw=1366&bih=657

3) similar diagrams can be made for in-cloud motions, e.g., up- & downdrafts
4) ideally the same diagram can be used to infer differences in sub-cloud behavior (next point)

For the cold-pool aspect, some illustration of how the two simulations differ is needed. There is too much speculation of how, e.g., horizontal motions within or around the cloud complexes differ between the two simulations. Please provide some illustration.

If it were not too much trouble, I would welcome a Helmholtz decomposition of horizontal velocity into rotational, divergent and Helmholtz residual. But if it is not possible that is ok.

As for where to place additional figures, it is at your discretion whether to incorporate a Supplement, or to place in the main text.

Thanks for your submission. The quality is already high, but the science needs an improved demonstration to isolate the most important points.

--Tim Dunkerton
(tim@nwra.com)

Good improvements, thanks for the extra effort. --TD