The study performs the analysis of the future simulations a hydrological model driven by the hydrometeorological input obtained from the ensemble of global and regional bias corrected climate models with the goal to investigate changes in flood quantiles and flood generating processes in France.

The manuscript is timely as it addresses a groining need for understanding future flood risk under changing climatic conditions, while accounting for the model uncertainties. However, the focus of the manuscript not quite clear to me. On one side, it seems that the focus is on the uncertainty that stems from simulations of different GCM-RCM pairs (most Figures in the main manuscript are very much focused on the model agreement and not on the changes themselves). But on the other hand, the description of the results is very much focused on description of changes and trends rather than quantification of uncertainty on the projected changes. Moreover, in my opinion the study lacks validation with the past flood observations and drivers. Please find my detailed comments below.

General comments

The theme of the study: As mentioned above the main theme of this study is not clear to me. The figures in the main manuscript almost exclusively focus on model agreement, while the Figures in the SI do show the changes in flood magnitudes (S5 and S6), but omit the information about the effect of model uncertainty. The description of the Results (Section 3) instead very much focuses on the changes themselves rather than the model uncertainty and its implications. If the focus is on the changes themselves a more comprehensive validation with the past of observations of floods is required in my opinion (see also my comment on the lack of validation), as well as a more comprehensive presentation of the trends in the main manuscript. Instead, if the focus is on the model uncertainty, I wonder how instructive is the bias correction in this case, as it essentially aims to equalize simulations across the models. Please clarify the main theme of this study and change/compliment missing aspects accordingly.

Lack of Discussion: Regardless of the choice of the topic, the Discussion is currently completely absent in this study. The section Results and Discussion provides only very few references to the previous studies. Putting own results in the context of previous research (either in regard of model uncertainty or in regard of explaining future changes in flood quantiles and drivers) would be extremely beneficial for this manuscript and can help to clarify the implications of the predicted flood changes.

Lack of Validation: Apart from a few notes in text that the simulation results are consistent with previous studies, I do not see any validation with the past observed floods. If the focus of the manuscript is indeed purely on the model uncertainty for future simulations, I find it acceptable, but since most of the results explained in the text rather refer to the future changes in flood quantiles compared to the past, the reliability of simulations for the past periods with existing observations have to be quantitatively proven.

Specific comments

Line 61: Floods with 20-year return period are hardly the “rarest”, please be more specific.

Line 101: I do not think that the studies of Froidevaux et al., 2015, Tarasova et al., 2018 and Kim et al., 2019 are relevant here, as they do not directly examine different flood types. I think the remaining 6 reference suffice.

Line 125-131: The purpose of this statement is not quite clear to me. Is it supposed to be the justification for the absence of validation? Regardless of the chosen fashion of model implementation (i.e., with or without calibration), validation is an essential part of any simulations study regardless of the choice of modeling chain configuration (even if the hydrological model allegedly plays minor role in the overall uncertainty). Please see also my major comments.

Line 230-232: Not quite clear how the baseflow contribution was estimated. Was it done for a single day that corresponds to the peak discharge of annual maximum flood? Please clarify.

Line 233-235: It might be warranted to search for the precipitation within 7 days before the start of the event to account for the time of concentration given the size of the study catchments, but it provides little justification for such search window relative to the flood peak, as in this case such search window represents not the time of concentration but the duration of corresponding rainfall or snowmelt event. This could be especially a limiting factor for snowmelt events, where build up time can be particularly long. Hence this potentially can lead to misclassification between “Rain and snowmelt events” and the rainfall events. A designated sensitivity analysis might shed light on the effect of the window search on the classification results.

Line 262-269: Some of these classification rules are rather ambiguous. Rule 1: it is not clear if volumes or rates are compared here. In case volumes are compared, it is not clear for which periods they are computed. Rule 4 and 5: it is not clear if these distinctions are only done when soil moisture threshold is not exceeded. Please clarify.

Line 276-278: It is not clear how streamflow events were identified here, or are these streamflow rates? Are these relationships obtained from the observed or from the modelled data? Please clarify. Moreover, please clarify why particularly this threshold is expected to be appropriate for distinguishing “Soil water excess” floods.

Line 295-297: Although this study aims to analyze trends in the 2 and 20-year return period floods, the performance of the model chain is not shown neither for representing past floods with different return periods nor for representing their changes in the past. Please also see my major comment.

Line 302: Please clarify what is meant here by flood event characteristics

Line 306: I wonder if minimum 20 occurrences per process is sufficient for robustly detecting trends in floods with 20 year-return period.

Line 319: Please explain the acronym before using it.

Figure 1: I have difficulties interpreting the agreement between the models, because the meaning of the multi-model agreement index is not explained anywhere in the text and because it is not clear to me how this Figure and Figure S2 are connected and if they show the same information. Please clarify and make sure that all captions provide all the information needed for interpreting the corresponding Figures.

Line 350: Please clarify what is direct runoff in this case and how it is defined. Please also provide this information in the caption of Figure 3.

Line 357: Where these trends are displayed? It is difficult to judge the accuracy of such statements since the spatial trends are not displayed in any of the figures, only model agreement about the trends.

Line 396-408: I miss here the quantitative comparison with the flood classification using past observations in France. Please see also my major comment about the lack of validation.

Figure 5: It is rather difficult to judge the differences between there panels, correspondingly making it rather difficult to see the changes. Consider displaying these results in a different way. Please also clarify in the caption the meaning of the values on the x axes.

Line 452-455: This part sounds like a description of the Methods. Please move it to the corresponding section.

Figure 7: Please provide a detailed description of each cluster in the caption. Without it this figure is not very helpful for the readers.

Line 480-482: These changes are very difficult to see from that Figure. See also my comment to Figure 5.

Line 493-494: It is not quite clear which extremes are meant here. Please clarify.

Line 493-498 and elsewhere in this Section: I miss here completely a discussion about the anticipated changes, their possible causes and agreement with anticipated changes with other hydrometeorological variables. See also my main comment ton the lack of discussion.

Line 506: Please clarify if rainfall rate or rainfall volume was meant here.

Line 516: More frequent floods?

Line 507-522: It would be very helpful to put these results into perspective. See my comment above.

Figure 8: Please add explanation of clusters in the caption. Please also clarify what is meant here by maximum rainfall. Event? Rate? Volume?

Figure S2: Please clarify what MIA stands for and how its values can be interpreted.

Figure S3: It would be more instructive to indicate not the absolute SWI values that were reached during the Long events, but instead to indicate how often was the threshold that splits between Long events and Soil excess events exceeds.

Figure S4: It is difficult to judge visually agreement between the bars. Please provide a qualitative metrics that supports the statement that models consistently represent flood generation processes of the past. I repeat here my earlier comment again, it would be very useful to compare the consistency of these classifications with the classifications based on observations during the historical period.

Figure S5: Please name the clusters based on the corresponding geographical areas and indicate it in the caption, this would very much support interpretability of this figure, as well as the figures in the main manuscript.

The manuscript “Evolution of flood-generating processes under climate change in France” presents an analysis of trends in flood magnitudes and flood-generating processes for 3727 catchments in France. Using a hydrological model and 22 climate simulations, flood events were predicted for the time period 1975 to 2100. The authors find four distinct clusters of future flood changes in France. Flood magnitudes in the North are increasing. The driver in the North-East is a decrease in snowmelt floods and an increase in soil water excess floods. The driver in the North-West is less clear. The authors attribute the change to an increase in short rain floods, though the model agreement for this change is low (Figure 6). In the South, flood trends and drivers can be clustered into mountainous regions with a decrease in flood magnitude, explained by a decrease in snowmelt-influenced floods and an increase in short rain/long rain floods. The other cluster in the South collects the Mediterranean catchments, which show no trend or a decreasing trend in magnitude, explained by a decrease in saturation excess floods and an increase in short rain and long rain floods.

Overall, I find distinguishing future flood development by generating process highly relevant, and the method of analysis very good. However, the reason I summarised the four clusters, is that the manuscript could benefit from more clarity and consistency in describing the trends, drivers, and clusters. At the moment, there are some statements that contradict each other, or contradict the results shown in the figure (see detailed explanation below). Since I find the underlying method, analysis, and figure to be sound, I am sure the authors will be able to easily address these points.

Specific comments:

The statements regarding trends in the contribution of short rain floods are not consistent between the abstract and the results/figures. In the abstract, it says:

“The proportion of floods linked to soil saturation excess is decreasing while the proportion of floods linked to infiltration excess related to extreme rainfall is increasing, particularly in the southern Mediterranean regions.”

Whereas in Section 3.3 and Figure 6, it is clearly discussed that the strongest increase in most catchments is the increase in saturation excess floods, with the exception of the Mediterranean region, which only makes up a small number of catchments compared to the catchments in the North.

Since there is quite a lot of information being presented (trends in magnitude, trends in drivers, trends in process frequency, correlations between drivers and magnitude) that all contribute to explaining changes in flooding, I would recommend increasing clarity by:

1. Being very clear about if a mentioned increase/decrease/trend is referring to changes in magnitude, or frequency.
2. Use consistent terms for the flood drivers (e.g. infiltration excess vs rainfall on dry soil, soil water excess vs saturation excess)
3. Giving the clusters more descriptive names as well as showing how many catchments are included in each cluster. In Section 3 the clusters are referred to only by number, but then in the conclusion, they are suddenly only mentioned by regions (Mountainous, Mediterranean, North, temperate oceanic zone…). I find these descriptions actually more helpful and would recommend adopting them earlier, together with a visual summary of what the distinguishing aspects of each cluster are.

Further comments:

Basins located in Cluster 1 show a clear increase in short rain and long rain flood contributions (Figure 6). Why do the flood magnitudes decrease (Figure 1), if particularly short rain floods are associated with higher magnitude events (Figure 5, line 568/569)?

It is unclear, both from the abstract and the methods section, if the model simulations are part of the work done for this analysis, or if they have been published previously. From the cited reference Tramblay et al, 2024, I would assume the latter. Please make it clear, that existing simulations are being used. Otherwise, I would expect more model evaluation to be part of this manuscript.

In regard to the flood process classification, how influential is it, that the antecedent soil moisture is extracted the day before the flood event? Kemter et al (2020, 2023) extract soil moisture the day before time of concentration. Similar for Tarasova et al (2020) where antecedent moisture is extracted before the separated runoff event, and Stein et al (2020) who extracted antecedent moisture 7 days before the flood event. Since the time of concentration based on rainfall distribution is extracted anyway, why not extract antecedent moisture the day before the rainfall event? Otherwise, antecedent moisture will have already been influenced by the rain.

In line 514 you discuss that “correlations with SWI are almost systematically stronger for frequent floods (2-year), than for rarer floods (20-year).” Please discuss how that matches previous findings that frequent floods are more likely to be soil moisture influenced than rare floods (e.g. Wasko et al, 2021).

Technical comments:

L60: What does “The trends in flood magnitude are contrasted, with increasing trends only in the northern regions of France” mean? Currently the sentence does not make sense.

L118: space missing

Figure 5: Please keep the x axis consistent between the three plots.

Figures: Please ensure for all figures that colors are accessible for people with color vision deficiency (Stoelzle & Stein et al, 2021). This is particularly and issue in Figure 5 and 7

L568: The correct term should be “In contrast, …” not “On the contrary, …”.

L568: Missing letter in “soil”

Stoelzle, M., & Stein, L. (2021). Rainbow color map distorts and misleads research in hydrology–guidance for better visualizations and science communication. Hydrology and Earth System Sciences, 25(8), 4549-4565.

Wasko, C., Nathan, R., Stein, L., & O'Shea, D. (2021). Evidence of shorter more extreme rainfalls and increased flood variability under climate change. Journal of Hydrology, 603, 126994.