Please find my comments in the attached .pdf. The overall method, data, and evaluation technique has the potential to provide a valuable contribution to predicting variability in channel capacity through residuals of the average stage-discharge curve. Inquiry into relevant scientific questions are presented. However, the current interpretation and analysis makes assumptions that may not be valid, lacks clarity, and requires more direct links between cause and effect than are stated within the article. Therefore, the article requires major revisions, including specificity of research aims, results interpretation, consideration of applied terminology, and acknowledgement of additional limitations. For instance, the first aim of the paper is to map the spatial variability of geomorphic response to extreme storm events, but the authors fail to acknowledge or address spatial correlation and bias in the stream gaging network. The definition of extreme in this article is unclear and it is unknown to what extent the included storms are extreme or quite frequent. The second aim is to understand the impacts of these storms on the stage-discharge relationships at gaged sites as a proxy for changes in flood hazard. However, this makes the assumptions that the storms alone are responsible for any observed changes in the residuals. While possible, other geomorphically significant events could have occurred that are unaccounted for. Further, the authors include other metrics in addition to the storms for predicting residuals, which makes it difficult to separate the impact of other drivers from the storms. For these reasons among others, I suggest major revisions prior to reconsideration for publication. More detailed comments are provided in the attached document.

The proposed study is relevant and novel for the field. Thus, it should be worthy of publishing in this journal. However, there are some details that needs to be revised before accepting it for publishing.  Below you will find a list of my main points.

1. The articles needs desperately a discussion section separated from the results. Right now, everything is cramped into a single section that makes difficult to take the main points out. 

2. I would also add as a minimum a paragraph in the conclusion section (if not its own standalone section) about the framework limitations. This is crucial in any prediction framework so readers are aware of it when taking decision. 

3. The introduction (and other sections) have too many small paragraphs (1-3 sentences) that disrupts the reading of the manuscript. I will suggest to combine paragraphs that convey the same message. 

4. Did the authors had a minimum years of data treshold when selecting the gauges? It seems imperative to have one, since a gauge with 5 years of data will yield very different results than one with 50 years of data. Also, what are the general statistics of the gauge data? For example, what is the mean length of record, amount of cross sections measurements, etc.

Specific Comments:

 Line 44:  is not clear the statement. Since during a flood event, flow within the channel can change due to external factors, stormwater discharge o compounding flood at coastal estuaries.

Line 64: the “secondary channel” that is referring in Figure 1 should be highlighted in the figure itself to help the reader understand the point. 

Figure 1: The figure needs a north arrow and scale bar. I also strongly suggest the authors to use a GIS platform to enhance the quality of the figure. The figure also needs a location map.

Line 93 and 95: both sentences start with “Despite some limitations …” Please rephrase.

Line 106: I will summarize all the gauges selected with their corresponding ID in a text file (or any other format file)  and upload it to a repository for easy sharing. Then, the reader could see exactly which stations were selected. This helps the open data statement in the research community.

Line 110: did the authors downloaded also discharge values from the NWS or it was just flood stages as it is mentioned in this line? If the cross section data (width and depth of the river) were obtained from the USGS, why not also use their created flow-stage curves. My biggest question is from where the authors obtained the discharge values for the creation of their rating curves, since NWS only provides stage level whereas USGS provides both stage and discharge in most gauges.

Figure 2: I do not support have several lines of text if the figure caption just to describe the different climate regions in the map. Also, the authors also explain the abbreviation in the results section when talking about it. Thus, I strongly recommend having a nomenclature section that summarizes all of these, including the variables from table 1. Then, the reader can easily find it.

Line 161: the statement of the reason for change in capacity (deposition) has been already mentioned in Line 159. Please rephrase or remove.

Line 177: why that the authors only focused on a very narrow range of years for their storm event? This seems like a big limitation, especially since the latest year of the record was a decade ago. The authors needs to justify their selection as a minimum.

Table 1: there are some variables that have their “unit” column empty. For example, Peak , Q2, etc. This might be a typo since if the variable does not has unit the authors specify with a dimensionless or N/A. Also, the table is too long for a peer-review article. I strongly suggest dividing the table into three separate ones, one for each variable type. There are also some variables like BFI_AVE that their description is a quarter of the page due to being squeezed in the small column width. I would suggest the authors to place the long variable description as a footer in the table or in the appendix as part of the nomenclature section.

This paper focuses on predicting changes in flood hazard, primarily driven by 'major storm events.' The study analyzed flood hazard changes for 3,101 gauges across the Continental United States (CONUS) using a machine learning model (self-organized maps) that incorporated 38 explanatory variables, including atmospheric, hydrologic, and geomorphologic factors. The findings are highly relevant and could serve as a valuable reference for understanding channel capacity in relation to storm events. However, I believe that the authors should enhance the manuscript's overall structure and clarify certain technical aspects. Therefore, I recommend major revisions. I am optimistic that by addressing these comments, the authors can enhance this already interesting work.

Major Comments:

1. The authors state in their paper objectives and title that they are exploring the impact of major storm events on flood hazard through changes in channel capacity. However, the explanatory variables used in their machine learning models include not only storm-related properties but also hydrologic and geomorphologic variables. It's unclear how the authors discern from the ML models that the changes in channel capacity are primarily attributed to storm properties and not influenced by other factors. Additionally, the analysis is focused on the dataset containing "major storm events," which implies that changes in channel capacity are associated in the ML to major events. What about changes in channel capacity during non-storm events? In other words, can the ML capture changes in channel capacity without storm property variables (variables as described by Falcone, 2011)?
2. The overall organization of the introduction and methods sections lacks necessary details. To better understand the techniques used to estimate the residuals and the various simplifications (such as manual filtering of outliers) required for the methodology, I had to refer to Slater et al., (2015). For example, in Figure 3, panel b, the authors mention outliers but do not clearly identify them or provide specific details.
3. The results and discussion sections need to be reorganized. It is recommended to create a separate section for the discussion. Additionally, for improved readability, it is advisable to create a single section dedicated to limitations and future work.

Minor Comments:

• Figure 3: It would be helpful to include a time series with streamflow data to illustrate the magnitude of the April 2007 flood. Panels c and d are confusing since they may give the impression that there's only one change in flood capacity per gauge, which might not be the case.
• Line [15]: Please clarify what you mean by "geomorphologic impacts."
• Consider adding a schematic figure that explains the core concept of conveyance capacity before and after a storm event. Real data examples would be beneficial in illustrating this concept. Slater et al., 2015, offers a useful example in this regard.
• Figure 9: Please provide information on how the 95% confidence bound of the current stage-discharge relationship was calculated.