The paper provides a comprehensive analysis of three primary classifications for a catchment: meteorological, attributes, and response. By correlating this information, the paper identifies characteristic classes of flood responses. The main findings show that meteorological data has a much greater impact on flood response compared to land cover and catchment attributes. However, certain catchment attributes were also found to be correlated with the response.

Here are my main concerns about this paper:

1. The results don't contribute new knowledge about the streamflow-generating process. It's well known that streamflow is mainly controlled by factors such as precipitation, intensity, duration, and its distribution. A similar analysis using the rational method could yield the same results as presented in this paper.
2. While the classification found in the paper may have value for local or basin analysis, most of the results cannot be applied to other regions or countries. The attempt to connect with other countries in the discussion is qualitative and not valid for comparison without quantitative analysis. What is considered high or low, fast or slow in one country could be entirely different in another.
3. Throughout the paper, the authors mainly describe numerical findings that could be presented in a table. I believe that the value of research lies in the analysis, discussion, and implications of the findings. Additionally, many of the figures contain irrelevant information that doesn't help highlight the findings.

Minor comments:

Line 40. You refer many times in the text to behavior characteristics what I consider response types. When we talk about behavior, you are trying to characterize the catchment dynamic which is intrinsic to each catchment. In other words, you try to characterize the low filter function that transform input to outputs. I would suggest changing the word behavior for response which is a more precise word for what you are analyzing.

Line 77. The expression “solid data foundation” is a biased description of your research.

Line 94. This is not the right way to refer to information extracted from a webpage. Check the referring rules from the journal.

Line 109. How dense is the meteorological gauge network? How can we be sure that they are representative of the basin analyzed?

Figure 1. The gauge distribution is strongly biased to Yangtze and Huai Rivers. How can you develop an analysis by basin with this low density in the other basins?

Line 139. PCA is known to work well for linear factors. Did you check for non-linear relationships?

Line 163-168. You are presenting the same information as Table 2. You should summarize.

Line 173-178. You are presenting the same information as Table 2. You should summarize.

Table 2. Factors are hard to visualize. Add a bullet for each one.

Line 193-196. These lines should be at the beginning of the paragraph with a more detailed explanation of the method used.

Line 208. You should be more specific about how you got that. What are the values inside the table? Explain more.

Line 209. Typo. What is the value 33.2 or 33.3%?

Line 210. What clustering methods are you referring here?  

Line 226-254. You are just describing the data that could be summarized on an appendix table.

Figure 4. I would try 2 columns. Left: Flood event distribution. Right: Frequency histogram. Currently, it is too small to watch some differences in the distributions.

Line 268-283. You should add a discussion about your results. You are mainly describing information that could be in an appendix table.

Figure 5. This is too small. You could move this figure to the appendix and add a figure with a more informative visualization, maybe zoon in a small area. Maybe you should correlate with some of the PC factors in space, etc.

Line 288. How can you talk about class per basin if some of them have a few gauges?

Line 292. Why does the class 5 increase over time?

Section 4.4.1. you mainly describe the same information presented in the figure 7. You should add an analysis or discussion about the implication of your findings.

Figures 7 and 8. Do you need a big figure only to show almost non-significance in the factors?

Figure 9. A rainbow color scale is not recommended because it is very difficult to recognize visually what value is higher than others.

Line 354-361. What about the high collinearity between meteorological factors? If you have many factors representing the same, the relative importance decreases. I would try to group them for more general characteristics because you have many factors in the range r=0.15-0.21.

Line 362-392. You are just summarizing the results. Where is the analysis and discussion?

Figure 11. You should present only the figures that support your statements (4 maximum). Other figures could be in the appendix.

Heterogeneities in meteorological and underlying surface conditions usually result in remarkable spatial and temporal variabilities of flood events. It is very beneficial to investigate comprehensive variation characteristics of flood events and their formation mechanisms by clustering massive homogeneous events into some representative classes. This manuscript made an interesting contribution to understand meteorological and physio-geographical controls of flood event variabilities at class scale across China. Over a thousand flood events were selected from most of river basins in China. The sizes of flood events, meteorological and physio-geographical factors were impressive, and the investigation was convincing because multiple statistical analysis methods were adopted, including the hierarchical and partitional clustering methods, constrained rank analysis and Monte Carlo permutation test. This topic fits well with the scope of HESS, and the study is original. I think that some moderate revisions are required for this manuscript before publication.

Line 104, how to “assess” the potential meteorological and physio-geographical control factors of flood events?

Line 123, the T_bgn is calculated using the circular variable. Please explain the reason.

In the section of methods, many of flood behavior metrics or control factors were not independent. Why were they selected? Please clarify specifically.

Lines 142-147, 22 criteria were used to assess the classification performance and determine the best number of clusters. I agreed that it would be a robust way to select an optimal class number. However, most of the criteria were given as an abbreviation. Could you please give a detailed explanation about these criteria including full names, equations and units in the supplementary material?

Lines 285-297, the comparisons of flood events among different classes are largely based on percentages, but the flood event numbers at many stations were not the same. Please give the detailed introductions about the spatial and temporal distributions of flood event classes.

In Figure 1, the main river names should be replaced by the river basin names.

In Figure 5, the legend “Flood classes” should be changed to “Flood event classes”. Please remove shading from the stacked bars. That adds no information.

What are the means of 21 in Figure 5 and 0.46 in Figure 8?

In Figure 6, I suggested that the flood event numbers could be given for every year in all the basins.

In Figure 7, it should be changed to a single column of the five cases. The coefficients should be “correlation coefficients”.