This paper studies changes in spatial flood extents across Europe and its drivers using a state-of-the-art hydrological model. The paper reports that, on average, there has been an increase of 11.3% in flood extents across Europe over the past 70 years, with regionally varying trends and regionally varying drivers. The work is well-presented, comes across as robust (though with some of the extensive modeling dependencies and relatively low NSE scores for many regions that’s hard to truly assess), addresses a relevant topic, and goes clearly beyond the earlier works on this topic (that are also well cited by the authors). I commend the authors for submitting such an organized manuscript and I do not think that my comments would further improve this paper. Well done!!

The manuscript “An increase in the spatial extent of European floods over the last 70 years” by Fang et al., presents a large-scale analysis of the spatial extent of floods and its spatial and temporal variations in the last 7 decades. The analysis is based on model simulations from the mHM model driven by observational data. The study finds that increased on average over most parts Europe and attributes its changes to changes in the magnitude or the spatial dependence of its drivers. The manuscript is well written, and the analyses and results are presented in a convincing way. Please find my comments below:

Major comments:

Comparison with HANZE: at best the detection rate of the. 100 largest flood events flood events compared to the HANZE database is 50%. This low detection rate raises questions on the threshold used for the definition of flood events. The authors define flood day using the 99^th percentile. This translates in more than 3 events in a year on average in each pixel. Many of the selected ‘flood days’ are therefore ‘just high-flow days’ (not even all annual maxima cause inundations). This could be a cause of this big discrepancy with the HANZE database and could suggest the use of a higher threshold for the definition of flood days (e.g., the 5- or 10-year flood).

Figure 9c and L241 “changes in soil moisture seem to play a minor role in the detected changes in flood characteristics”. This is in contradiction with the findings of Blöschl et al. (2017, 2019) and Bertola et al. (2021), Tarasova et al. (2023) who find that antecedent soil moisture is relevant to explain negative trends in flood magnitudes (and shift in timing) and increase in flood-poor periods in the Mediterranean catchments. How do the results of this analysis compare to this literature? What are the reasons of this discrepancy? Are the changes in flood extents caused by different drivers than trends in flood magnitudes and temporal clustering of floods? How is soil moisture estimated in this analysis?

L346-348: only drivers occurring in the spatial area of flood events are considered. This has big implications in the attribution analysis especially for very large rivers, as contributing drivers occurring within the actual catchment area are not considered (often drivers occurring in one part of the catchment, e.g. snowmelt or rainfall, cause flooding downstream where these drivers do not necessarily occur). What are the implications on the attribution results?

Specific comments:

L39-46: it is true that “other studies rely on observations and may miss important information due to uneven spatial distribution of stations”. On the other hand, models (like the one used here – L65-72) are calibrated and validated using observations so have this intrinsic limitation too. Furthermore, models have other limitations, e.g. modelling uncertainty and resolution of simulations (in this case gridded runoff simulations that are quite coarse – 0.125° =~ 11km …). Furthermore, for the attribution analysis the study relies on EOBS (L79-80).

L21 “spatially compounding river floods”: the use of “compounding” seems inappropriate in this context as we are talking about spatially widespread events and there is no mention of simultaneous occurrence of flood drivers in this section (or elsewhere in the paper). I suggest substituting “compounding” with “widespread” or similar.

L52: I suggest citing Lun et al. (2020) – the first study detecting flood-rich and flood-poor periods in Europe.

L100-106: are flood days defined at each grid cell separately? Spatially connected flood days (i.e. pixels) and overlapping flood patches are ‘further combined’. Does this mean that they are considered as one single event? Please clarify.

L125: E_past and E_pers denote the AVERAGE flood extent ?

L158: 244 flash floods from the HANZE dataset are captured in the dataset. However, flash floods typically occur over small catchments (a few km2) while the catchment area that is captured is at least of the order of magnitude of 1000 km2 (due to the resolution of the gridded simulations). Similarly for the time dimension, i.e. flash floods typically last less than 24h, while the resolution of the simulations is daily. How can mHM model simulations capture such flash floods? What are the implications in terms of such identified events?

Figure 2: is this figure only representing the events in the HANZE dataset or does it contain results of this analysis? If it does not contain results of this analysis, it should be moved to another section (e.g. appendix).

Figure 9: it is not clear if the maps show changes in snowmelt, rainfall and soil moisture over all days in the two periods or if they refer to changes for flood events only (i.e. only the rainfall causing flood events or changes in rainfall in general?)

L329: “aligns closely with an independent impact-based

Figure A1: labels and titles of the plots are not fully clear and not explained in the caption.

The manuscript titled “An increase in the spatial extent of European floods over the last 70 years” by Fang et al., aims to identify large spatio-temporally connected flood events over the last 70 years in Europe using the mHM hydrological model.

Overall, the manuscript is well organized and written. For some of the methods used in the study a better justification should be presented so that the reader is better aware of the reasons for certain choices.

For further details find my comments below.

General Comments:

Naturally, as with such a complex study, a lot of data sets and models have been used so that at some stage the reader might get lost what has actually been done. I think it would be helpful to add a schematic/graphical representation showing the inputs, models used and outputs and their ways of comparison in one to two panels to make things clearer.

Additionally, it is unclear why the period beginning with 1951 was used. The mHM model has a spin-up period of 1940-1959. Hence, I think the analysis should be conducted only after… This would also resolve the issue that as described in L 126 10 years of data are discarded (1981-1990). So the “attribution” could then be done using the two periods 1961-1990 and 1991-2020.

Also, the section 2.5 on “attribution” of changes in flood events needs to be re-written. A lot of assumptions are being made in this section, but it remains illusive why certain choices are being made (suggest adding some explanations instead of just citing references that have done the same previously). Additionally, the authors should aim to add some “physical reasoning” to this mainly statistically based attribution exercise. Additionally, the attribution reminds of the use of conditional probabilities. Please elaborate/discuss why this approach has not been considered instead, i.e. what are the advantages of the current approach…

Finally, I am missing a discussion of the presence of human effects of the hydrology/floods through dams etc. particularly with regard to the conclusions drawn with regard to the attribution of changes in flood extent. With more and large dams that have been built in the recent decades in south of Europe a lot of water will be held back and is no longer available for flooding.

How do the authors reconcile the reality with the modelled results… Please elaborate and discuss.

I also in general agree with most of the comments of Reviewer 2, therefore I will not list similar comments/concerns in this review.

Specific Comments:

L 28: Suggest replacing “a future climate” with “future” as a future climate is not the only important factor.

L 66: Please add which version of E-OBS was used.

L 70-71: Please specify what percentage “most”  refers to and what quantifies as “low” station density. Suggest also showing in the appendix area and stations that have been excluded from the analysis.

L 82: Please specify what method was used to downscale the resolutions.

L 105: Please elaborate with a short sentence, why 0.4.

L 112: Please elaborate why 1000 km2.

L 122: Please elaborate why 0.7.

L142: suggest replacing “well” with “satisfactory”, as this is the terminology mainly used associated to the use of NSE values.

L360: I´m not sure if one can conclude that “floods are more widespread in low-lying regions, such as parts than in high mountainous regions like the Alps.” As the authors have a priori excluded smaller catchments which would naturally be found in the mountainous areas…

Figures: In some Figures red and green colour coding is used. Please avoid using this as it is difficult for colour blind readers to discern the Figures. Suggest using “colour-blind safe” colours instead.