This paper discusses the connections between antecedent conditions, event rainfall and flooding in 20 large catchments in Switzerland, making use of stochastic weather generation and the HBV model.

It makes use of modelled data to overcome lack of historical flow data, which also allows the use of hourly time-series of precipitation and flow, but acknowledges the shortcomings of investigating short-duration high-intensity rainfall. Return periods of floods and rainfall were explored and quantified using random-forest classification methods, to mixed success. Interesting patterns of the lack of coherence between annual maximum flow and annual maximum event rainfall. 

This paper uses one set of parameters which model the "median flood frequency curve". It would be interesting to see the more extreme flood frequency curves made use of, to see the variability of the outcomes.

Overall this is an excellent paper, which only requires a small number of changes to finish.

Major issues:

 - A discussion of the variability of the ACRT would be useful to put some of the other values into context, as longer ACRT may have larger volumes.

 - l125. Were the rainfall and flow records tested for stationarity? Although climate change is not the focus of this paper, understanding whether the stationary stochastic generator is based on stationary data is important. Similarly, if the AMAX flow observations are not stationary, then the relationships between AMP and AMF in reality may also be changing over time.

 - l125. How long were the rainfall, temperature and flow records used to extrapolate out to 10,000 year simulations? If short, then the overall distributions may under-estimate the variablity of these observations.

 - l265. Forcing to link the AMP and AMF might lead to problems when close to the start or end of the hydrological year, leading to the pattern of "AMP after AMF". Can this be checked to see whether this is a problem?

Minor Issues

 - l97. Put numbered lists on new lines for ease of readability

 - Fig1. Figure could be full page width, or blue point made bigger.

 - l126. What do you mean by "pseudo-observed"?

 - Table 1. Are those standard terms for the regimes? If so, provide a reference. Otherwise define the regime terms.

 - Table 4. Please add more labels to the first row and column.

The article analyses the role of antecedent moisture conditions on the estimation of extreme floods on a basin in Switzerland. The authors use a rainfall generator and a hydrological model to generate long time series on which their study is based.

I found this is an interesting paper. I have only some minor comments detailed below.

Detailed comments

1. General: Some parts of the article should be corrected by native English.
2. Lines 33-35: Mathevet and Garçon (2010, https://doi.org/10.1080/02626667.2010.503934) also discussed this issue. Their analyses could be shortly commented.
3. Introduction: The following works may be interesting to cite in the context of this study :
   • The work by Cameron et al. (1999, https://doi.org/10.1016/S0022-1694(99)00057-8) may be worth citing here because it considered uncertainty in flood estimation using continuous hydrological modelling.
   • The work by Merz and Blöschl (2009, https://onlinelibrary.wiley.com/doi/10.1002/hyp.7168) on the controls of flood events
4. Lines 67-69: The authors could shortly discuss multi-model approaches in the context of flood estimation.
5. Lines 94-99: I found the authors do not clearly show the originality of their work compared to the previous studies they cite in the introduction or others they discuss later in the article. The authors should more clearly state the novelty of their study. What are the gaps it intends to fill compared to previous works?
6. Section 2.1: For the readers who do not know the Aare basin, I found a short physical description is missing.
7. 1: A horizontal scale and north are missing.
8. Table 1: The Regime column is in French.
9. Lines 112-115: For which typical catchment size are these comments relevant?
10. Lines 121-124: Sorry but it was unclear for me whether point simulations are produced by the GWEX generator and then averaged at the catchment scale to feed the hydrological model.
11. Line 123: The disaggregation process was unclear for me.
12. Line 133: Is there some bias in the spatial consistency of extreme events?
13. Line 144: At the hourly time step, the 50-80% quantiles are not very high.
14. Table 2: I did not understand why the snow correction factor (SFCF) can be below 1. Are there actually cases where measurements overestimate snow and should be corrected by SFCF below 1?
15. Line 186: I was wondering whether the cross-correlation is not excessively influenced by a few extreme events. Should not the cross-correlation be calculated on transformed rainfall or streamflow (for example with square-root) to limit the influence of a few very large events?
16. Lines 278-279: Is this really surprising or was it expected?
17. Section 4.2: I found that the authors could further discuss the implications of their work for classical approaches of extreme flood estimation methods. I think they could discuss how their results corroborate (or not) past findings from comparative studies of flood estimation methods. I was thinking for example about the Extraflo project (Lang et al., 2014, https://doi.org/10.1051/lhb/2014010) in which a large range of flood estimation methods were compared, in gauged and ungauged conditions, using statistical approaches or methods based on continuous or event-based hydrological modelling. Other studies also attempted to compare methods (e.g. Okoli et al., 2019, https://doi.org/10.2166/nh.2019.188, and references therein).