The manuscript by Lenderink et al presents a comprehensive evaluation of convection-permitting model (CPMs) performance for simulating rainfall (and extreme rainfall) relative to regional climate models (RCMs). The distributions of the rainfall, the rainfall-temperature relationship, and the rainfall-relative humidity relationship, and the differences between climatic regions are all better respected in CPMs as compared to RCMs. Moreover, the authors discuss in a very transparent manner the possible short comings of CPMs which will surely lead to improved model performance in the future as these short comings are addressed in future research.

As we move to a world where planning decisions are made on the basis of including more and more climate model results (rather than historical observations) building trust and confidence in the model results, as well as improving model results, is paramount. This manuscript addresses both these aims making it an important and impactful manuscript. This is a very thorough evaluation and hence I unreservedly look forward to this manuscript’s publication.

I have comments below which I hope the authors will consider, but I should note, apart from errors I spotted in the supplementary figure numbers, the rest of the comments should be treated as suggestions. Although I feel it would be remiss of me to not suggest references that I am aware of, the authors should at no point feel obliged to cite them unless they feel they would enhance their manuscript. So again, I reiterate my comments should be treated as suggestions, not prescriptions.

General comments:

The robustness of the super Clausius-Clapeyron scaling wasn’t explicitly stated in the abstract or conclusions and this could be added if the authors wish.

The interesting result of rainfall being more intense at lower RH appears is well discussed. I had a thought that this could also be affected by the drop in temperature across events which becomes greater with rarer events (Figure 4 in Barbero et al) but I assume that picking the temperature before the rainfall event negates this effect? The authors may wish to comment on this in the manuscript but shouldn’t feel obliged to. Apologies if this was already mentioned and I missed it.

Barbero, R., Westra, S., Lenderink, G., Fowler, H.J., 2017. Temperature-extreme precipitation scaling: a two-way causality? Int. J. Climatol. 38, e1274–e1279. https://doi.org/10.1002/joc.5370

Line by line comments:

Line 45: The reference below showed RCM parameterisations fail to capture spatial dependencies of rainfall.

Li, J., Wasko, C., Johnson, F., Evans, J.P., Sharma, A., 2018. Can Regional Climate Modeling Capture the Observed Changes in Spatial Organization of Extreme Storms at Higher Temperatures? Geophys. Res. Lett. 45, 4475–4484. https://doi.org/10.1029/2018GL077716

Line 52: Where you say “when future changes are dominated by simple thermodynamics” when is this exactly? I think it is for long duration (daily) rare rainfalls, but it would be nice to state this explicitly. Having said that, if a short duration event is embedded in a long duration event, then convection permitting modelling remains crucial.

Line 72: “depend on the temperature measure used” – I wonder if something could also be said for the need to make sure the temperature is the one driving the event by making sure rainfall events are independent when captured. When this is done this gives reduced variability in the rainfall-temperature scaling. This conclusion in Visser et al is ultimately the outcome of these three manuscripts.

Barbero, R., Westra, S., Lenderink, G., Fowler, H.J., 2017. Temperature-extreme precipitation scaling: a two-way causality? Int. J. Climatol. 38, e1274–e1279. https://doi.org/10.1002/joc.5370

Schleiss, M., 2018. How intermittency affects the rate at which rainfall extremes respond to changes in temperature. Earth Syst. Dyn. 9, 955–968. https://doi.org/10.5194/esd-9-955-2018

Visser, J.B., Wasko, C., Sharma, A., Nathan, R., 2021. Eliminating the “hook” in Precipitation-Temperature Scaling. J. Clim. 34, 9535–9549. https://doi.org/10.1175/JCLI-D-21-0292.1

Line 95: it might be worth mentioning that such reductions in relative humidity have also been observed and not just modelled?

Denson, E., Wasko, C., Peel, M.C., 2021. Decreases in relative humidity across Australia. Environ. Res. Lett. 16, 074023. https://doi.org/10.1088/1748-9326/ac0aca

Line 170: I could be wrong here, but was Pmean used at all in the manuscript? I got the impression that only Psample was used so I was a bit confused to why these were both defined.

Line 226: Is the “all time-steps” including zero rainfall time-steps or just the wet time steps?

Section 3.1: There was a really comprehensive discussion of why some models were better/worse for the RCMs but for the outlier for the CPMs (UKMO-UM), it wasn’t discussed as to the reason why this might be the case – this could be added around lines 235-240.

Figure 3 caption: The parentheses for the last line are probably not required.

Line 251: The higher intermittency I assume may lead to drier soils and less humidity? The link to the above sentence that discuss model performance was missing because the above sentences talk about the model performance in terms of simulating temperature.

Lines 252-255: I wasn’t really sure what figure I was supposed to be looking at, so I didn’t really understand this part I am sorry. I think Figure 3 was introduced but then the discussion shifted to discussing Figure S2 and this wasn’t clear.

Lines 266-274: I am not sure the results of the DPD for median and 80^th percentile (middle and right columns) were contrasted in the text?

Line 319: “similar to a quantile difference plot” - I think an “a” is missing in this sentence.

Line 302: Again, the link of soil memory and runoff to the temperature simulated is missing.

Line 348: Can the figure numbers in Supplementary be listed? I think this sentence needs rewording because when you say “plots for individual models are presented in supplemtnary” this implies that Figure 5 is pooled model results, but they are just observations correct? Maybe saying “plots comparing observations to individual models are presented in supplementary and are summarised in Figure 6?” would be clearer.

Line 349: “For further investigation…” should this say, “For further investigation and comparison to the model simulations”? I think it could be made explicit that you are now shifting from observations to also looking at model output.

Figure 6: I think the caption is wrong – it says NL is presented, but the text at Line 353 says NL is in the supplement. The caption should also mention the percentile analysed. Finally, I think you mean Figure S10 and not Figure S6?

Line 350: Although the split of low and high RH was explicitly stated in the methods it could be mentioned here again to remind the reader?

Line 417 – I think you should mention the figure number in supplementary.

Line 446 – “Area” could maybe be “regional” differences?

This manuscript provides a comprehensive analysis and evaluation of ensembles of RCMs with parameterized convection and CPMs with explicitly resolved convection for two regions that are likely to see changes in rainfall convective rainfall extremes.  The topic is well researched with a comprehensive background section, the analysis is reasonable, and the results are convincing.  Further, the authors provide an insightful discussion on their results and the paper is almost entirely devoid of grammar issues.  I think the paper is valuable and an interesting contribution to the literature and, accordingly, I recommend the paper be published pending the resolution of a handful of minor comments.

I have two key questions that the authors should address so as to better contextualize their results and conclusions.

1. What is the specific resolution (or range of resolutions) of the CPMs used in the study? The RCM resolution is listed at 12km, but I don’t recall seeing the CPM resolution described in either the manuscript or the supplementary material.  This is important as numerous studies show that for convection, diurnal convection in particular, 2km resolution is generally superior at reproducing the timing and intensity of diurnal convection (e.g.,  https://agupubs.onlinelibrary.wiley.com/doi/full/10.1002/2014RG000475).  If the models used in this study are coarser resolution than 2km, I think that's fine since the main consensus within the community is at 4km cutoff.  That said, some context here would really improve the manuscript.  On a similar note, any comments the authors can provide on the LSMs used in the models described in study would be helpful since at sub 4km resolutions, simulated convection becomes increasingly tied to the LSM. (e.g., are there any known biases in the LSMs that might impact the results or explain some of the shortcomings in the CPMs?).
2. What is the generalized modality of the convection that makes up the extreme rainfall events (95 percentile, e.g.) in the regions described in the study? Diurnal air-mass “pop-corn” ordinary cells? Organized MCSs?  Terrain-initiated convection? Convection forced by or embedded within synoptic scale systems (e.g., fronts)?  Mix of all of the above?  The authors provide an excellent discussion on the complexities of convection and extreme rainfall and how changes in absolute and relative humidity in the surface can influence or be related to cloud-scale dynamics, however without information on what convective modes are being simulated, this discussion is without context.  For example, the authors discuss convective plume size vs. dry air entrainment as a physical reason for the observed behavior of extreme rain vs. dew point depression.  However, my understanding is that this process is mostly only relevant for continental convection associated with very dry mid-level air where wider convective plumes associated with larger surface DPDs can protect the convective core from dry-air entrainment in the mid-levels.  If, for instance, the convection most associated with extreme rain in the NL or SFR was more characteristic of air-mass convection with very moist conditions throughout the atmospheric column, then this process likely isn’t a great explanation for the DPD/extreme rain relationships observed.   To be clear, I am not arguing that this process isn’t relevant in this instance, I’m just arguing that without added context on storm modality, the authors’ speculation on physical processes carries less weight.  A brief discussion of predominant storm modality, and perhaps an example figure showing a snapshot of simulated rain-rate compared to radar image for an extreme rain event would be extremely helpful to the reader and would improve the manuscript.  Such a discussion would also relate back to CPM resolution, since the differences between 4km and 2km resolutions would be quite different for a synoptically forced MCS vs air-mass thunderstorms.

Minor comments:

1. Why not look directly at CAPE or CIN? When discussing /speculating on some of the convective processes and how they relate to changes in absolute and relative humidity? Were these data simply not available?
2. Why use DPD instead of RH directly?
3. I’m unclear as to the use of the 5x5 pixel sampling, and how it was used in the data comparisons, the mean was computed, but I don’t know if it was actually used?
4. Line 246/247: dew-point temperature, TD, and TDD should be modified: TD and TDD. (no need to repeat dew-point temperature).
5. It is my understanding that the analysis performed comes from simulations that the authors themselves did not perform, but this point is not entirely clear as it relates to the CPMs.  I suggest the authors clarify this point.