The authors investigated how the “hot model problem” is critical for the future changes in mean, high and low flows in the North America. I have some minor comments.

L20-23: At the same time, inclusions of hot models lead to critical risks of significant overestimations of climate change impact in some areas.

Please add references of ESMs in the table 1. Did you compute ECSs by yourself?

Line 56: Shiogama et al. (2022, Nature) constrained future annual precipitation changes. Please see their Fig. 3b. Hot models overestimate annual mean precipitation increases in Alaska, Canada and the West US where your spread of flow changes significantly decline by removing the hot models.

Shiogama, H., Watanabe, M., Kim, H. Emergent constraints on future precipitation changes. Nature 602, 612–616 (2022). https://doi.org/10.1038/s41586-021-04310-8

Line 61: Shiogama et al. (2022, ERL) showed that hot models could cause overestimations of aggregated economic impact of future climate changes.

Shiogama, H., et al. (2022) Uncertainty constraints on economic impact assessments of climate change simulated by an impact emulator. Environ. Res. Lett. 17 124028

https://iopscience.iop.org/article/10.1088/1748-9326/aca68d/meta

L183: How did you define outliers?

L322: High ECS of NESM3 does not necessarily mean greater regional warming (Fig. 3).

Because the total spread ratio analysis would be not sensitive to the removal of the second largest model, it may be better to show standard deviation ratios in Figs. 12 and 13.

Thank you for inviting me to review paper: “The Dilemma of Including Hot Models in Climate Impact Studies: A Hydrological Study” by Asenjan et al. At the outset, I should mention that I am a climate researcher (rather than a hydrologist).

The authors address an issue of maximum concern to climate change research, that at present, the CMIP6 ensembles contain a few models with an especially high climate sensitivity (“hot models”). The concern is that such simulations may project impacts that are especially difficult to adapt to – causing alarm - yet, we may find in the decades ahead that these hot-running ESMs are unrealistic.

The authors focus, in particular, on testing the effect of especially warm ESMs on streamflow. I like this approach because the emphasis is on a major impact and of much concern i.e. flood risk.

Based on the Abstract alone as a start, in some ways, I like the catchy title, but using the word Dilemma implies something unresolved. In fact, this paper provides definite findings. Maybe something a bit more factual such as e.g. : “Including Hot Climate Models in estimating Streamflow does not alter the assessment their future variability”.

Second, in general and as the world warms, there will be a hydrological intensification for the planet. Hence, it might be expected that hot models generate higher river flows. The authors need to give a good reason for concentrating on variability, rather than mean trends. However, reading further into the manuscript, then mean changes e.g. “Qmean” are considered – maybe reword the Abstract to “mean changes and changes to variability”?

It is often asked: “Given the problem of climate change is now emerging strongly in the measurement record, why is it so difficult to constrain future warming estimates?”. The main reason for this is the historical record also includes a strong cooling aerosol effect. There are a few papers out there that make this point, and could be worth citing around line 45? In other words, we do not know from present-day measurements if we are in a high sensitivity fast warming world with strong contemporary aerosol cooling, or the opposite. That is the usual reason why we cannot reject outlier ESMs.

The description of the methods feels a bit too short in places, which is a shame because the authors have worked especially hard to entrain data, bias-correct a hydrological model against ERA5, bias-correct the ESMS. The test data is especially comprehensive (~14K sites). Would a schematic work that can capture some of this activity?

I had not heard of the KGE metric / statistic before, and where it is introduced, there is no citation. Can a reference be provided, and maybe a short sentence that explains its basic features? Presumably, the statistic allows an assessment of which hydrological models are performing best (based on aspects of variability in timeseries?). Please also check the order of presentation. Figure 1 is cited in the text (line 84) before the modelling is described in full (around line 124).

There could be improvement of the manuscript around page 7 (lines 135-158, and Figure 2). Can it be confirmed that Figure 2 is a generic plot, referring to any metric of interest? Line 143 “Figure 2 present the three dispersion metrics….streamflow”. This is too vague – does the “y” axis of Figure 2 related to Qmean, Qmax or Qmin? Apologies if I am missing something obvious.

In results, line 186, there is reference to ECS. However, neither here, or in the caption to Figure 3 (or caption to the Table) is a reference to the source of the ECS values. For Figure 3, you could potentially use some sort of marker to differentiate the five warmest models on the left. For instance, a horizontal arrow under the first five models, with the words “warmest models”.

In places I thought the captions to diagrams and tables were very short. Obviously, captions cannot repeat everything that is in the paper text, but a little more information in places might help the reader (this is important if, upon publication, people extract diagrams and captions to place in powerpoints for talks).

This might be something for the HESS editor to advise on, but in my view, the main plot of the paper is Figure 5. This presents very clearly the geographical effect of removal of models with a high ECS. After that, there are many helpful diagrams, but I am not convinced that all are needed. Or some sort of multi-panel plot might help, e.g. merging the box plots. The authors could consider placing some of the additional information and plots after that in an SI – e.g. Figure 12.

One small frustration with this manuscript is that once variables are declared, and given abbreviated names, often such names are not then used systematically through the paper. This is especially noticeable when presenting keynote diagrams, such as Figure 5. Hence, the caption for Figure 5 should read something like: “Total spread ratio, TS_nd, for Q_mean…..”

Would the authors like to comment on the spatial cohesion in plots such as Figure 5. This diagram in particular seems to show that it is the higher latitudes that will see the bigger effects if the high ECS models are correct? Certainly the largest levels of warming will be seen towards to poles, which is a generic result for ESMs. That is the ESMs with the largest ECS which show the most pronounced warming poleward. This might also be true for precipitation?

It is noted that the reference list is balanced and comprehensive. I also like the Introduction, and the references cited where attempts have been made (either via emergent constraints or historical data) to constrain the bounds of Equilibrium Climate Sensitivity (ECS) bounds.

This paper contains a wealth of important information, addresses an important problem of how to deal with ESM outliers and in the important context of such model impacts on streamflow. I am very happy to see any other revised paper version.