This paper studies drivers of flooding and flood change in the Geul River catchment (Netherlands). Better understanding the drivers of flood change is a very topical subject and this paper produces a useful contribution on this topic. It especially stands out by providing a more in-depth insight than many large-sample studies (involving many catchments at once) have managed to provide on this topic, while still providing methodologies and insights that can be more widely adopted in understanding drivers of flood and flood change. However, at the same time, the paper seems to suffer from one major issue. I personally recommend the publication of this after this can be addressed meaningfully

Major comments

The consideration of antecedent wetness as a flood driver relies on a threshold API value (exceeding 1). This API index is based on antecedent precipitation and does not take any evaporative processes into account. The latter seems somewhat problematic as soil wetness in this region tends rho be very seasonal (as ET is low in winter and high in summer) which very likely causes the strong seasonality in maximum flow and flood events (see e.g. Figure 3) but which is not visible in any of the considered flood drivers. Therefore it seems that the importance of soil wetness does not reflect soil wetness in this paper, but reflects relative wetness compared to what is normal for that part of the season (which is not relevant to the study?). This problem likely causes a strong bias in all results and thus the overall conclusions

Minor comments

• It seems like the statement “Results suggest that extreme 24-hour precipitation cannot solely lead to floods.” is unlikely but not physically impossible. Therefore, I recommend rephrasing “cannot”.
• L15: “Unprecedented precipitation” seems like a bold statement when it’s not specified for example since the observational record started, or some clause that determines the period over which we talk.
• L33: this statement could, in addition, be supported by some other publications that show the importance of antecedent wetness in other places.
• Fig 2. Check the label of “Feb”.
• L144: “all-4day” misspelled?
• I’d recommend (but maybe this is just personal taste you can ignore) to start the results paragraph with a sentence that summarizes the result. This would make it easier for a reader to focus on when reading the details in the figure that follows. This essentially applies to each new paragraph in the results.

The manuscript “Flood drivers and trends: a case study of the Geul River Catchment over the past half century” by Tsiokanos et al., analyses the long-term temporal variability of flood drivers for the Geul river catchment. The study adopts an interesting multi-temporal approach to analyze temporal trends of floods and their drivers and finds that 1-day extreme precipitation alone does not explain flood changes, rather heavy prolonged rain, and wet initial conditions. The manuscript is well written, and the analyses and results are presented in a convincing way. Please find my comments below:

Major comment:

1. The aim of the study should be clarified. Is it to develop a methodology (L61-64) or to understand flood trends and their drivers in the catchment (L69-72)? These lines appear quite disconnected in the introduction. Furthermore, is the multi-temporal trend approach new (L7, L61-64), or it was proposed by Hannaford et al. (2021) and Murphy et al. (2020) as stated in line 52-53? Please clarify.

Specific comments:

1. “critical precipitation” terminology. In several parts of the manuscript (abstract, introduction and discussion) the authors draw conclusions on the “critical precipitation (precipitation that leads to floods)”. It is not fully clear to what of the analyzed precipitation indices they refer to. Please clarify.
2. L145-148: It is not clear if these lines describe an extra criterion used. How do you practically ensure that P_MD is higher than P₉₉ ? What do you do when this is not the case (L148)?
3. L158: how is FE defined?
4. L207: “Trends in P_kD are based on the. annual maximum values”. What does it mean? Do you refer to annual maximum discharges and the fact that P_kD is calculated using k days hat preceding flood events? Please clarify.
5. L218-220: Why are different assumptions used for the MK test for precipitation and discharge trends? Why do you account for autocorrelation in annual maximum discharge series? Annual maximum values are typically considered uncorrelated by construction as they belong to different blocks/years.
6. L221: What t do you consider in the analyses?
7. Table 2: Last column. Shouldn’t it be “Reverse relative frequency?
8. L304-308: these lines were not fully clear to me.

The authors present an event-based analysis of flood drivers on a 344 km² catchment using 50 years of concurrent daily rainfall and continuous streamflow data. The main conclusions are that heavy 4-day precipitation is the primary high flow driver and this, when combined with wet antecedent conditions, provides a stronger indication of flood likelihood than extreme daily precipitation alone.

Overall, I think the evidence presented provides reasonable support for the conclusions drawn, but this evidence could be strengthed and clarified. The authors have selected an interesting topic and a very worthwhile case study. I have three major comments and some minor ones, as detailed below.

Major comments:

1. The floods considered are based on a small number of factors (daily and 4-day precipitation occurrences occurring in the highest 1% and 5% of wet-day events, API, and various joint combinations). While in concept these are reasonable surrogates for the underlying flood processes of most relevance, it is a little surprising that no attempt appears to have been made to select factors of specific relevance to the catchment. For example, rather than adopt an arbitrary API, the decay factors of an API function could be fitted to the selected flood maxima and then used in the event-by-event analysis. Alternatively, a simple daily soil-moisture accounting function could be derived that implicitly allows for the influence of rainfall sequencing and evaporation; even without fitting to any observed data such an approach would appear to have greater efficacy than the adopted indicator of wetness. Similarly, a simple correlation analysis could be used to justify the number of days adopted for the multi-day precipitation index, as at present no discussion is provided to justify the “critical” duration adopted. Such analyses would strengthen the physical reasoning used to assess the relative importance of the different flood drivers and may reveal greater insights about the nature of the interactions involved.
2. Most of the analyses focus on the sample of events where it is known that conditions have resulted in floods. However, concentrating on the sample of 870 multi-day precipitation events (noted in Table 2) and examining the moderating factors which led to 50 annual maxima events should provide more insight about the processes leading to floods than does focusing on the much smaller sample of known flood maxima. For example, the analysis of these 870 events using similar diagnostics to that used in Fig 5 would make it clearer what combinations of factors lead to major flooding and which don’t. It may be found that the combinations of conditions that are associated with floods may in some (or many) cases not lead to flooding, and this may highlight the influence of an additional factor that has not been considered. The “reverse” analysis described in the paper thus needs more focus and attention.
3. The results are consistent with physical reasoning though in places I had to work quite hard to follow the logic of the narrative and the specific details of the results. It would thus be useful if the authors tightened up the narrative and provided additional discussion. For example:
   1. the information presented in Table 3 needs further explanation as the supporting discussion on this was not particularly helpful.
   2. While the information presented in Figure 4 is broadly clear, I do not understand how the relative frequencies are calculated and why selected combinations of them don’t add up to 100%.
   3. Fig 5 provides is a useful analysis as it differentiates between floods of different magnitude, yet it is not entirely clear what the different symbols are in Figure 5 denote - they appear to differ from the indicators listed in Table 1? It would perhaps be useful to examine such correlations for all selected indicators, allowing for timing lags as needed?
   4. Further efforts should be made to strengthen the narrative thread throughout the paper as in many places I found myself going back and forth within the current and previous paragraphs to make sure I was following the intended logic. For example in Section 2.3.3 the discussion around the logic of the selected indicators commences before they are clearly defined two paragraphs later.

Minor comments:

• Figure 2(b) – x-axis label is incorrect (it is not a rate, but rather the proportion of time that the given flows are exceeded)
• Line 219-220 – why is it the serial correlation of the precipitation time series assumed and not simply calculated?
• Line 249-250 – the justification for the last sentence of this paragraph is not clear
• Line 256 – should 86% be 83.7%?
• Line 397 – clearer justification is required for the 3^rd sentence in this para regarding the cause for the rise in severe precipitation
• There are numerous small errors with the use of prepositions and other minor grammatical problems, and these should be reviewed and corrected.