See attached

This contribution by Pothapakula et al. is a comprehensive evaluation study of ICON in a new version. It focuses on important aspects of the global monsoon system and concludes, that grid-spacing still influences the climate modeling performance. It is an interesting and important contribution, which I find well worth publication after revisions.

Still, I have a few comments and questions as detailed below.

The abstract is quite long and it is difficult to identify the main findings. Is it a goal to evaluate the new ICON core (line 7, but not mentioned below) or just the grid-spacing dependence of global/regional monsoon representation? What is “organized moderate variability”? What is an “excellent representation”? Line 23: increased → decreased. Line 25: What “underscores” the region-dependent sensitivity?

Has the simple linear interpolation (line 120) of monthly SST data to daily data some impact on monsoon onset/offset? In the pre-monsoon phase this should slightly underestimate SST warming and this slightly increases the land-sea contrast. Perhaps this can explain a few days of monsoon onset/offset bias?

Significance: (line 330) “indicates that 40 km performs best”. This statement is based on a relativly small number of simulation years and the distributions shown in the violin plots (Fig. 4) are well overlapping. I assume here there is no conclusion about best grid-spacing possible.

Why do you exclude the southern hemispheric monsoons from the evaluation, but show them in the Figs.?

l632: I guess, here decreased not increased grid-spacing is meant? Does this worsening performance of precipitation simulation over ocean imply that the there is a tuning issues in the model? Later (l647) you state that the precipitation contribution by the convection parameterization independent of the chosen grid-spacing. The parameterization should contribute less at 10 km grid-spacing.

Minor issues:

l133: Tergen → Tegen

l165: “V” is horizontal (add) wind vector (and thus should be bold). Also here and in the following “kg/m-s” is uncommon. Correct would be kg/(ms) but we are used to kg/ms too. At line 441,2 you use inconsistently but correctly kg m-1 s-1.

l289: “skill scores” – usually a reference prediction/simulation (like a random, persistent ... simulation) is used in the definition of skill scores. In this sense you do not consider skill scores, just scores.

l325: “common bias” – references?

l389: dipole → dipole bias?

l656: “… variability …. show_s”

l837: incomplete reference