General comments

This study presents changes in summertime precipitation over Europe, comparing historical simulations with reanalyses and observations, as well as future projections. It also relates them to changes in the atmospheric energy budget that sets physical constraints to the precipitation change. It is a well-motivated study, precipitation projection is highly relevant and models differ in their historical and future simulations. Applying a regional energy balance perspective is purposeful and a comprehensive comparison between CMIP6 models and two reanalysis data sets, as well as observations, is made.

Main conclusions are that reanalyses give a weak constraint on the energy fluxes, and that models differ even more for future projected energy fluxes and precipitation changes, but do agree on future precipitation reduction in summer over Europe. The greatest spread among the driving terms comes from the horizontal energy flux which is calculated as a residual.

I have some concerns regarding energy conservation in reanalysis, and the calculation of H as a residual. I would also appreciate some clarification of the formulation of the underlying energy balance equation. The authors could further strengthen their reasoning in some places, eg address more generally if/how from the model evaluation there is reason to trust the models in their future projections. They might explore more if the future projections could be constrained by skill in historical simulations of precipitation change or energy budget.

Regarding the presentation of results I see some shortcomings on statistical significance and use of multi-model means, and suggest some improvements to graphics presenting the results.

With these concerns addressed, I believe the paper should be well suited for publication in ACP.

Specific comments

1. The framing of the atmospheric energy budget (eq 1) could use some clarification. It would help to make clear where the energy budget is calculated and each of the fluxes referred to, as there is now a mixed discussion of in-atmosphere, TOA and surface fluxes. Also the sign convention should be made clear. Does plus/minus refer to up/down or gain/loss, etc?

For mean precipitation what matters is the energy budget within the atmosphere, where there is a net radiative cooling (from SW absorption, and larger LW emission) compensated for by latent and to some extent sensible heat fluxes. Therefore “shortwave cooling” is a bit counterintuitive, and it is not clear why longwave cooling (dLW) and shortwave cooling (dSW) have the same sign in eq 1. Not least as the example given is an increased absorption (presumably a negative cooling) leading to decreased precipitation, and the field apparently has both negative and positive values over the domain (Fig 2). It might be helpful to present the global mean balance, SW – LW  = LH + SH  (with SW, LW, LH and SH all positive) or the regional budget for that matter SW – LW  = LH + SH + H, leading to eq 1, where the d denotes change in response to climate perturbation.

Also, eq. 1 describes the changes in each of the fluxes (d) whereas the figures (fig 2, 3, 4,  5, …) are denoted without d.

The Discussion in Section 4 on comparison with observations is largely about TOA and surface fluxes without a clear link to the in-atmosphere energy budget perspective taken in the paper.

2. Energy conservation of the CMIP models is discussed, but do you have reason to believe that the reanalysis data sets are energy conserving? If not (and you already refer to Wild and Bosilovich 2024), can you check that and comment on if/how that affects the results? When H is calculated as a residual, all potential imbalance will be placed in this term, which might then not actually represent horizontal transport only. For the models too, this aspect of the method and potential influence on results could be more prominently discussed – currently there is a figure comparing residual and actual H for one model in the appendix. You might want to check/show other models and reanalysis, or quantify the bias between explicit and residual H, and thereby the error in estimated precipitation change.

3. In Section 3.1 and Figure 3.1, no measure of statistical significance is given in the discussion of regionally varying trends. For the CMIP6 mean there is indication of areas where a majority of the models studied agree in sign, but this is not a robust measure, especially given the model list including multiple models more and less closely related. Even the use of an unweighted multi model mean can be questionable (see eg Kuma et al 2023 on model code genealogy). Conclusions based on where models agree are not necessarily robust.

You might also comment on the similarity between the two reanalysis data sets, despite the difference in assimilation of observed precipitation data.

The same question regarding significance and agreement applies for section 3.2 and Figure 4.

4. Going from evaluation to future projection (Section 3.1 to 3.2) you don’t really address the question of what reason we have from the model evaluation to trust the model projections. From figure 5 it seems like models agree better on future precipitation changes (in sign) than on the energy budget terms controlling it. Why is that? Assuming the energy budget framework is physically sound, the question seems to be – why do the model precipitation not adhere to it, and what is it instead that controls the precipitation change in the models?

5. The potential for model constraint is not fully explored. Do the models that fall within the reanalysis precipitation range have smaller error included in their residual term (better agreement between balance-calculated and model-derived H)? Do the models that fall within the reanalysis range for the fluxes produce a more constrained range for precipitation than the full ensemble? Can we in fact learn something about precipitation representation in these models, from the evaluation done here? Could the historical precipitation change be applied to create a constrained future projection span, using the models that perform better (ie using Figure 3 to constrain Fig 5)?

6. The attempt to relate results to model ECS (Figure 6 and discussion) is not fully motivated. Is there a reason for the flux divergence to be related to sensitivity, or is the choice of relating H to ECS based on the large spread in H among the models? Given that H is calculated as a residual, it is even less clear why this should have a physical relation to sensitivity, and the result of low correlations is thereby not surprising. The weak relation between change in precipitation and model sensitivity would be worth commenting on in relation to previous literature, how come the general features of regional drying and wettening don’t scale with model sensitivity?

I think this section, if you want to include it, needs some more explanation and elaboration. Currently the statement on L257-258 that the signal strengthens with higher ECS seems a bit strong. You might also want to relate to studies like Barnes et al (2024) who look closer at regional climate projection and to what extent it can be attributed to model climate sensitivity,

7. For figure presentation, please add measures of statistical significance where possible, and please make figures and fonts larger (especially figures 2 and 6). In Figure 3, the symbols for the two reanalyses are inseparable, as are the differently shaded grey bars. It would help to choose these symbols and colours differently. The = sign in the chart is a bit confusingly placed, and would perhaps fit better in the x-label if you want to include it. See comment above regarding d, and sign convention for the terms.

Technical corrections

L13 compares -> compare

L14 in the atmospheric energy budget

L15 “diversity” has positive connotation, when what is actually described is models deviating from each other and thereby from observations. Maybe spread is a more neutral word

L33 each degree of warming

L87 (DeAngelis, … Wild) references repeated

L91 this is pretty much a textbook statement regarding radiative effects of clouds, so the choice of references seems a bit arbitrary

L107 monthly files, please phrase more specifically

L242 is notable other studies -> is notable. Other studies

L242 “they” -> models?

L277 over central Europe

Review of “European summer precipitation” by Birthe Marie Steensen, Gunnar Myhre, Racheal Byrom, Ada Gjermundsen, Caroline Jouan, and Camilla Weum Stjern

General comments

The manuscript considers changes in summer precipitation over Europe and analyzes these changes using a regional atmospheric energy budget framework. It compares historical simulations in 35 models from CMIP6 to both observations (E-OBS daily gridded observational dataset) and reanalysis datasets (MERRA2 and ERA5), finding that the reanalysis datasets generally match observed rainfall but also exhibit large discrepancies with CMIP6 models when considering other terms in the atmospheric energy budget. Models generally agree that European summer rainfall will decrease with anthropogenic forcing.

Overall, the manuscript is strong, and the regional atmospheric energy budget framework which the authors apply is both sound and insightful. There are a few changes which might strengthen the manuscript which I hope the authors will consider and which, once addressed, would make the manuscript suitable for publication in ACP.

Specific comments

• In the atmospheric energy budget framework, the horizontal energy flux H is calculated as a residual. While the manuscript validates this approach for the NorESM2-MM model in Appendix A1, this approach is not validated in other models (nor the reanalysis datasets) in the same way. Of any term in the atmospheric energy budget, delta H appears to be the main driver of changes in precipitation, so, if not overly cumbersome, it would be helpful to further validate the manuscript’s treatment of H as a residual.
• It is unclear why delta H is compared to ECS in Figure 6. While these quantities are statistically related, one would assume there is some physical reasoning which connects these quantities in a more mechanistic manner as well. If such a connection exists, the manuscript would benefit from making this connection more explicit; if not, it would benefit from clarifying that these are only correlations (and thus may be spurious).
• The labels in several plots are difficult to read and the color choices for several quantities are difficult to distinguish (e.g., gray bars/regions in Figure 3). Consider increasing the font size of particularly small labels and changing especially similar colors/markings to be more easily distinguishable.

Technical corrections

• Figures 2-6 should label each term in the atmospheric energy budget as the change in that term (e.g., delta SH) not the term itself (e.g., SH).
• Change to “compares” on line 13
• Add “budget” at the end of sentence on line 14
• Missing period on line 242
• Replace “being” with “to be” in line 245
• Replace “strengthening with” with “stronger in” in line 257
• Replace “while” with “with” in line 263