This study investigates how climate conditions may impact the rain-on-snow (ROS) floods in Germany using an LSTM model trained on downscaled climate projections and explainable artificial intelligence techniques. The results suggest that while ROS flood frequency will decrease in most regions, the severity of extreme ROS floods is expected to increase across all major river basins, and the role of snowmelt in these floods will become less significant. Overall, this study is comprehensive and well-designed. However, I still have several questions and suggestions for improving the current work.

1) The "magnitude" of floods is mentioned in the title. However, the absolute magnitudes of floods are represented by the relative ranks.

2) It is suggested to clearly define the “trans-basin” floods. How to distinguish whether the flood is "trans-basin" or "within-basin"?

3) Lines 5 and 82: The LSTM model may not be the state-of-art deep learning approach nowadays. It would be better to compare its performance to that of the emerging alternatives, e.g., physics-informed deep learning, and the ensemble-based approach is also beneficial given the limitations of any single deep learning model.

4) Lines 18, 471-472, and 523-524: The conclusion that the discrepancy is mainly due to different snow dynamics in climate models should be supported by more detailed arguments and evidence.

5) It would be better to make the Introduction Section more concise and to clearly state the research gaps and study objectives.

6) Figure 1 and the other maps: It is suggested to add a north arrow and a scale bar.

7) Line 98: How to define the "subsurface runoff" and the "day length"?

8) Line 108: The evaluation metric, NSE, has some inherent limitations. Also, given the sampling uncertainty and measurement errors in both temporal and spatial data, it is suggested to present the values of metrics through a statistical distribution instead of a fixed number for a single evaluation period and to apply the metric to the variable of interest like specific flood peaks. The authors can refer to the article below for more information about the limitations of some commonly used evaluation metrics in hydrology.

Reference: “Beyond a fixed number: Investigating uncertainty in popular evaluation metrics of ensemble flood modeling using bootstrapping analysis” (https://doi.org/10.1111/jfr3.12982)

9) Table 1: Could you add a reference for the criterion, "SWE>15mm"?

10) Line 230: How about the potential impacts of human activities on the other gauges? Any thoughts about how to incorporate those infrastructures into the modeling process, especially for the urban areas? 

11) Figures 3, 7, 8, and 9: The horizontal axis, "Stations", was sorted by different criteria, such as the mean NSE and the median number of ROS floods. Is it possible to keep it consistent? So it is relatively easy to identify each gauge station.

12) Figure 8: Please add the denotation for the shaded areas to the figure caption.

13) Line 370: It would be better to specify the exact p-value rather than "P<0.05".

14) Line 479: Is it "GMCs" or "GCMs"?

The study by Czakay et al. applies an LSTM model, trained by ERA5 data, in combination with an ensemble of 31 downscaled CORDEX CMIP5 to investigate how climate change may impact rain-on-snow floods in Germany. The study is well-designed and comprehensive, and it provides some interesting new findings. Still, I have some questions and comments that should be considered in a revised manuscript.

General comments

• My major concern is about the detection of peak-over-threshold events and the definition of floods. The authors use a threshold value to derive on average five peak flow events per year (POT5). A flood is then defined as POT5 events exceeding a return period of two years. How is the independency of flood events guaranteed? Two subsequent peaks may actually belong to the same flood event. This could have major impact on the results and their interpretation.
• Flood magnitudes are displayed by ranks. This requires better introduction, explanation and guidance for the reader when explaining the results.
• The authors could elaborate more about the uncertainties of their simulations and projections. In particular, the role of the deep learning approach on the overall uncertainty could be discussed in some more detail.

Specific comments

• The results presented in the abstract could contain more specific, quantitative information. The author may also consider adding details on the data and methods used (which deep learning model used etc.)
• The introduction lacks a clears research question or research objective. Furthermore, the introduction should also introduce and explain trans-basin floods in some more detail. I also suggest to highlighting the relevance of rain-on-snow events compared to other flood types in Germany in some more detail.
• L39 add direction of change
• L54 why?
• L55-56 this sentence is not complete
• L70-78 on which spatial scales? What about uncertainties in the (hydrological) impact models?
• L93 already mentioned
• L96 add spatial scale of ERA data
• L107 would it make sense to apply an ensemble of LSTMs to learn more about the uncertainty of the deep learning approach?
• L135 check for consistent use of IGsm and IGsnw
• L186 the authors should provide more information on how they performed the bias correction.
• L187 This could need a clearer explanation. Maybe highlighting the difference between ERA5 and ERA-Interim could help.
• L230-235 What about anthropogenic activities in the study catchments in general? I would guess that this is an issue. How does this impact the results?
• Figure 3: I don not understand the orange line in combination with lines 240-244.
• Figure 4: The LSTM predicts floods not contained in the observation data. What does this imply for future projections?
• L290-295 I think these data sets need a better introduction in chapter 2.
• Figure 7c, f (plus corresponding text, L293): I do not see this improvement.
• With reference to the skill in predicting high-flows, this should read a bit more critical. How well does the deep learning approach compares against a regular hydrological model? Is it possible to assess/discuss this?
• L344 add a reference to Figure 10 underlining the spatial information
• Figure 10: is it possible to indicate the significance of changes here?
• Figure 11: notches in the boxplot may indicate the significance of changes; Fig. 11 c) why “all” smaller than individual river basins
• Figure 12 (c): Do I understand correctly that ROS events are currently by far the most dominant flood types in Germany? I can hardly believe that.
• Figure 13: maybe the small statement in the text on changes in the seasonality is enough and does not require this figure which does not provide so many extra information.
• L449 I wonder if percentage change is an appropriate measure here. If we assume the same absolute decrease in the number of days with snow cover in lowlands and the Alps, then the percentage decrease is of course smaller in the Alps since the total number of snow days is generally larger high-mountain regions.
• L486-490: repetition, this has already been mentioned in the method section