General comments:

This study describes the vertical resolution dependency of the global storm-resolving simulation for the various vertical grid spacings between 800 m and 50 m. Because the time step is generally chosen for the vertical resolution, the authors further examined the time step dependency and the vertical resolution dependency with the fixed time step. The comparison for the vertical and horizontal resolution dependencies is added. Among the comprehensive results described, both the cloud liquid water and the cloud ice amount increase with the vertical resolution, and these do not converge until the 50 m vertical grid spacing. Cloud liquid water is equally dependent on the vertical resolution and the time step. The effect on the radiative fluxes is also documented, and the possible reason and the mechanism are argued.

This paper stands out for conducting numerous experiments across various vertical resolutions and effectively delving into the dependency on time steps and vertical resolutions as distinct discussion points. The analysis of mechanisms concerning the results is also reasonable and seems well-justified. With minor revisions, I believe this paper is ready for publication.

Specific comments:

1. p. 2, L.23: Clarify what kind of “the effect of a halving of the grid spacing”.
2. p. 9, L.243: Please insert references for “past studies”.
3. p. 9, L. 248: It is unclear where a strong stratocumulus signal contributes in “the South Pacific” in Fig. 4.
4. p. 10, L.260-267: Please quantify the correlation between cloud liquid water and reflected shortwave radiation and the other correlations. The correlation may be defined at each latitude, and the latitudinal profile of the correlation can be shown. Interestingly, reflected shortwave radiation correlates well with cloud ice in some latitudes, and OLR correlates more with the cloud fraction.
5. p. 11, L. 284: It should be clearer to rewrite as “the effect of a spatial refinement on cloud liquid water”.
6. p. 11, L. 315-381: In the Northern Pacific, the cloud liquid water effects vary regionally. The reviewer speculates that this regional variation is related to the synoptic condition simulated in this period. What do the authors mean by “the ventilation of the boundary layer”? It is not convincing that the numerical diffusion and “the ventilation of the boundary layer” explain the regional variation of the difference of cloud liquid water.
7. p. 12, L. 338-346: Figure 11f is not referred to in the text. Why the increase of upwelling (Fig. 10f) is related to the drying of the troposphere? It cannot be seen where atmospheric radiative cooling increases from Fig. 11f.

General comments

The study by Schmidt et al. performed a series of 45-day simulations using the global storm-resolving model ICON-Sapphire and analyzed the sensitivities of some basic “climate” features to the choices of vertical grid spacing and time step length. The manuscript documents sensitivities in the simulated atmospheric water amount, energy fluxes, air temperature, etc. The quantities presented include global integrals, zonal averages, geographical distributions, as well as mean vertical profiles in selected regions and different precipitation regimes.

I applaud the authors’ interest in carrying out such a study, and I appreciate the large amount of resources, including both computing time and human hours, invested in the study. The manuscript is relatively easy to follow, and the results can serve as a useful reference for other models of this kind. Therefore, my overall recommendation is to publish the manuscript in GMD after a round of minor revision.

The authors documented resolution sensitivities in this manuscript but did not provide in-depth explanations of the causes of such sensitivities. This is understandable given the complexity of the numerical model. Furthermore, while ICON-Sapphire is a global storm-resolving model, the manuscript focused heavily on spatially and temporally aggregated results without showing sensitivities in the simulated storm characteristics. It would be very useful if the authors could carry out some future studies in those directions.

Specific comments

Because of the many simulations discussed in the manuscript and the multiple ways of comparison performed (i.e., time and space combined and in isolation), it would be helpful to always include the time step length in the simulation names. I.e., in Table 1, in all figures except the first one, and throughout the text, it would be useful to say “L110-40s” and “L320-15s” instead of simply “L110” and “L320”.

By eyeballing Figure 1, we can easily discern the two or three sectors that constitute each vertical grid. As the number of grid layers increases, we see more and more examples of discontinuous slopes along the curves. I’m curious whether these discontinuities have been found to (or are expected to) affect the vertical propagation of waves. Could these discontinuities have contributed to the numerical instabilities in L320 and L540 simulations mentioned at line 133 of the preprint? If the authors had a chance to redesign and rerun the simulations, would they construct the grids differently?

In Figure 1, the curves are of similar colors and many segments are on top each other. It would be useful to use more colors - and reduce the marker sizes - to improve the visibility of all grid configurations.

The abstract only summarizes the numerical results. It would be useful to include some of the conclusions from the last section, particularly on the choices of grid configuration for future simulations.

This paper focuses on the effects of vertical resolution on climate simulations using the ICON-Sapphire global storm-resolving model (GSRM).  The authors find that while the model has sensitivity to vertical resolution (and time step), it is not as large as the sensitivity of changing the horizontal resolution.  I commend the authors for all the hard work put into performing and analyzing these simulations.  The topic of vertical resolution is often forgotten about when compared to horizontal resolution sensitivity and this is the first robust study (that I know of) for GSRMs.  Ultimately, I do feel this paper should be published, but I hope the authors consider my recommendations that I feel will increase the impact and clarity of this work.

Major comments:

• I strongly encourage the authors to provide analysis and discussion regarding the performance of their vertical grid configurations when compared to observations. I realize the authors admit in the conclusions they choose not to do this, but it makes the paper feel incomplete in my opinion.  Having no observational reference makes these experiments feel somewhat arbitrary and makes the paper feel a little boring in spots.  I do not feel this needs to dominate the paper, but having some analysis to answer the question of “does higher vertical resolution lead to better results (and by how much)?” would help to increase the impact of this paper and make it more interesting to other modeling centers.
• As the paper is written, it feels more like a sensitivity paper specific only to the Sapphire model. As a model developer myself, I wished the authors would have done a better of job of discussing and hypothesizing how generalizable their results may be for other GSRMs. I realize this is a rather difficult thing to do but some suggestions to bring this to fruition could be to 1) discuss if Sapphire’s parameterizations have been tested offline to satisfy vertical resolution convergence.   Whether they have or have not would be an interesting data point to other modeling centers about the type of sensitivity they may expect from their models if they followed similar testing approaches (or not). 2) Discuss if any parameterizations in Sapphire have explicit dependence on the vertical grid spacing (aside from discretization, of course).  Example: the SAM cloud resolving model (Khairoutdinov and Randall 2003) caps the SGS turbulence length scale to the local vertical grid spacing.  This inevitably leads to a vertical grid sensitivity.  If Sapphire has something like this (I see it uses a similar turbulence scheme to SAM but couldn’t find a paper with the specifics), this would be worthwhile to document and could explain some sensitivity seen in this paper.  3) The authors could speculate on the choice of SGS schemes to vertical resolution sensitivity (i.e. would a more complex turbulence scheme perhaps lead to a more robust solution?).
• While I found the paper to be well written enough that I could understand what the authors were trying to say, I did find many paragraphs and passages that were very awkwardly worded that required me to read them several times before I got the point. One (of many) examples of this is page 11, paragraph 305.  Therefore, I encourage the authors to read through the document and improve the wording and English.

Minor comments:

• This paper finds that the South America stratocumulus are the most responsive to vertical resolution change. This is interestingly very consistent with the findings of Bogenschutz et al. (2021) and Lee et al. (2021; 2022) in E3SM and I feel this should be explicitly noted.
• Further, the above mentioned work notes that significant improvement to marine Sc (in their models) is not achieved until the vertical grid spacing approaches ~15 m in the lower troposphere. Since observational references are not provided in this paper, it’s hard to tell if Sapphire’s increased vertical grid (which is not as aggressive in the lower troposphere compared to the aforementioned works) significantly improves marine Sc or not.  Once observational references are added, it may be interesting to tie the findings with that of the E3SM work.
• The authors may want to consider citing the work of Cheng et al. (2010) https://doi.org/10.3894/JAMES.2010.2.3 which is a nice paper that lays the groundwork for vertical and horizontal resolution sensitivity for CRMs in the doubly periodic framework.