This paper by Struglia et al. presents a comprehensive investigation into the effects of increased spatial resolution on climate simulations for the Mediterranean region, with particular focus on precipitation patterns over Italy. The authors perform dynamic downscaling of CMIP6 global climate projections producing a coherent set of high-resolution multi-scenario climate simulations based on the WRF-ARW version 4.2.2 model with a double-nesting technique. Initially, the model's resolution is set to 15 km across the Europe CORDEX region before being enhanced to 5 km specifically for Italy and the northwest Mediterranean. The multi-scale strategy represents an advanced approach to close the gap between global climate and regional climate projections. This paper represents a high-quality scientific contribution to regional climate modeling and it cam be published after minor modifications.

Minor 1: If the authors could remark on how the donwscaling impacts the climate change signal of the global climate model over Italy, it would greatly enhance the manuscript's overall added value.

Minor 2: Please add the colorbar units in figures 12-15

The authors run two WRF models, a 5-km centered on Italy and a 15-km centered on Europe.  The nested models are forced with ERA5 boundary conditions for an historical run and with 3 SSP scenarios using the MIP-ESM model.  They describe this a CMIP6-driven, but really it is only a single model.  For that title I would expect multiple CMIP6 models to be used.  Please remove CMIP6 from the title. 

Basically, this is a paper not about model development, but about the simple application of an off-the-shelf WRF model run at two resolutions.

The authors spend much time in the abstract (where it is inappropriate) and in the introduction describing convection permitting (CP) models that need scale of <4 km.  But then, they use models that are not at CP scales, so why waste this space which should be better used describing what they did here.  There is also a lot of talk about ECVs, but the comparisons are only to T2m and Precip., so why open discussion on ECVs here?

The two model simulations WRF-15km and WRF-5km are compared against observations (with the ERA5 forcing) but there is less comparison and emphasis on the overlap region, the only region of interest here since the case being made is what is the difference between 5km and 15km.  In fact the experimental design is poorly thought out, Both 5km and 15km should have been run over the same domain, because making the 15 km domain so much larger changes the way the ALPS domain sees the boundaries.  Better to run both for the same Italian region.

L34 – "demands for high-res" is confusing use of English

L35 – Really, so many regions over the globe are "critically prone to the impacts of local-scale and severe weather"   Certainly the mountains are difficult to model, but not unique in tough weather.

L39 – No one is seriously running 200 km global climate models these days!

L43 – what does "(120 divided by 20 km) mean?  6 km?

L66 – CP is not defined except in the abstract (does not count as text)

L73 – "model uncertainties" or model errors?

These are some examples of why this paper is difficult to read and interpret.

Figure 5 – the description of this figure in the paper is contrary to what I see in it.  The summertime dip in the 5km model is described as producing a much better seasonal fit than the 15km one.  I see that the summer (months 6-7-8) in 5km is better, but for the rest of the year, it is still just like the 15km and overestimates the rain.  "more realistic" is a marginal call.  " better in summer" is OK

Figure 6 is even more of a shock.  The 5km run has NO cumulus rain, only annually uniform stratiform rain.   The 15km run develops cumulus in summer, but the total (Fig 5) is far too large vs. observed.  This indicates the your scale-aware convection is totally failing:  as we go form 15km to 5km your cumulus convection parameterization shuts down.

Figure 8 – The identical power spectrum for 5km and 15km tells me that both are doing the same thing because both have the same parameterization??

Figures 12&13:  If you want us to learn anything from this experiment, both should be plotted on the same grid so we can compare.  As is, there is no useful information here.

L381 Conclusions:

               I would say this is a downscaling example, but certainly not a strategy.

              We have seen no evidence the your G-F cumulus scheme mimics a CP scheme.

                           In fact it seems to shut down.

              The 5km runs improved a wet bias ONLY in summer, and they did that oddly (above)

              How does this work show that your 5km or 15km runs are "adequate to provide the boundary conditions for" CP scale and further downscaling.  This is simply not shown.

I am confused as to why this paper is a GMD paper, there is really no model development.

An odd question:  since you are forcing at the somewhat distant boundaries, do you not have synoptic climate variability in the interior, especially with regard to convection?  Do you force internally?  Therefore, would different ensembles produce different statistics?

The code and data availability is very weak.  Is the exact WRF version used here the "current version" with the github reference?  Were any mods made?  I agree it Is not the responsibility of these GMD authors to archive the ERA5 data.  But where is the reference to the MPI-ESM data? (also on another archive).  What is missing here is the actual values from the figures posted here.  That is a useful minimum requirement.  What code was used to process the data sets?