The article “Influence of Storm Type on Compound Flood Hazard of a Mid-Latitude Coastal-Urban Environment” explores the differences between tropical cyclone (TC), extra-tropical cyclone (ETC) and non-cyclone (NC) as drivers of compound flooding to New York city, USA. The authors examine historic time-series of hourly rain and tide gauge data, using dependence and joint probability analysis methods, to explore the potential influence of storm type on near-simultaneous pluvial and storm surge flooding events. The study found that TCs dominate the most extreme pluvial/storm surge compound flood events, but ETCs are responsible for the majority of moderate and high frequency occurrences. There are important magnitude and lag differences depending on coastal location.

Strengths:

1. Aim and objectives: the aims and objectives of the study are clearly stated. The use of long time-series of gauged data is to be welcomed.
2. Discussion and conclusions: the focus on compound floods due to TC- and ETC-linked drivers is timely, given the rising frequency of extreme weather even within temperate zones, due to climate change.
3. Statement of limitations: it benefits this study that simplifications and assumptions are clearly stated. This text provides context and is a source of ideas for future research.

Minor issues:

1. Sea level rise trend: it was not completely clear how the trend of sea level rise has been removed from the 75-year time-series data measured at the tide gauges. Could the authors please expand on this?
2. Distinction between TCs, ETCs and NCs/convective storms: It was not clear to me how the authors categorized the different storm types in section 2. Was this pre-assigned to each storm by the National Hurricane Centre, or was a threshold (e.g. as defined by the Saffir-Simpson scale) applied afterwards? This is key information for anyone wishing to reproduce the study.
3. POT approach: could the authors please expand on the selection of the top-5 ranked rain/surge events each year? For context, it would be interesting to know how many of these events (out of all 75 x 5 events picked, per gauge), were categorized TCs, ETCs or NC/convective storms. A simple table would be enough. Only being able to capture a few TC events in the record, even with a long time-series, has to be recognized as an unavoidable limitation.
4. Return period assumptions: the data are not longer than 75-years, how do the authors defend the calculation of return periods in excess of this (e.g. in figure 8)?
5. Assumption of stationary storm surge over time: While it is stated that this first baseline assessment does simplify conditions, it could also be worth mentioning that recent research has identified that elsewhere, storm surge extremes are not in fact stationary, over similar time-scales (e.g. Calafat et al 2022, DOI: https://doi.org/10.1038/s41586-022-04426-5 )
6. Figures: the majority of figures would be difficult to read for those who print in B&W, or are color-blind. Would recommend a different color palette (colorbrewer2.org, for example, suggests great color combinations that overcome this problem). In addition, would suggest that Figure 1 would benefit from (a) a simpler (line drawing) background rather than satellite imagery; (b) an inset, or wider view, to illustrate the NYC location with more of the Long Island Sound and Atlantic visible (to better understand storm surge, and significance of storm orientation at each gauge location); and (c) to perhaps reconsider the color scheme of gauge location points for the reasons stated to above. Additionally, figure 8 would benefit from larger font in the x-, and y- axis labels.
7. Statement of relevance: the manuscript might benefit from a clearer description of the significance of the results of this study, which focuses on a relatively small urban watershed referencing a small number of gauges, to the current scientific knowledge of pluvial/coastal compound flooding. How do these findings contribute to the scientific conversation?

Overall:

Because of the use of hourly time-series data, this study provides useful insights into how lag time, magnitude, and orientation of storm-linked drivers all contribute to the state of flooding within an urban watershed of high economic value. The use of this more discrete data, creates a useful distinction between impacts in compound flooding due to TCs, ETCs, and convective storms. The study would benefit from clarifying some details of the methodology and results, as detailed above.

Technical corrections:

• L486 “in toto”?
• L114 - how long is the data collected at Battery gauge?
• L 143 – what is a “sewershed”?
• L165-L167. At a single gauge is this statement correct? This feature of storm surge is known due to onshore and offshore winds in different quadrants of the TC position; however usually one tide gauge records rising levels due to onshore winds, and a neighbor some km away would (hopefully be well-placed to) capture the negative surge due to offshore winds. Of course this effect changes with cyclone path/coastline orientation, and cyclone size.

This manuscript by Chen et al., presents an analysis on compound flood hazard for the New York City area. The analysis is focused on compound events of precipitation and storm-surge that are driven by different storm events classified as tropical cyclones (TC), extra-tropical cyclones(ETC) and neither events. Results are also presented for “all” events considered, to highlight the differences in return period of the hazard when frequency analysis does not consider event type. Results shown suggest that despite the fact that the frequency of compound surge and rain events is low, the compound risk associated toTC events need to be assessed separately to avoid underestimation of the risk. Analysis has been based on a long record of hourly rain and tide gauges. Carrying the analysis at an hourly scale offers clear advantages, with respect to past works focusing on daily, on the identification of “simultaneous” rain-surge events and investigation of lag of the peaks (from rain and surge) overall.

Overall, the manuscript is clearly written, and the discussion and conclusions are supported by the results presented. Furthermore, the analysis at hourly scale and the event-type investigation offers novel elements for this type of work. Most of my major concerns on the methodological framework have been acknowledged by the authors themselves in section 5.4 “Limitations and simplifications”, a fact that I appreciate because at the very least demonstrates that the authors understand and openly acknowledge the limitations of their approach and the complexity of the problem under study.

Below I list some additional comments (mostly minor) for the author’s consideration.

1. Thinking of estimation of lag or equivalently identification of “simultaneous” extremes of rainfall and surge, and considering that timing for the two variables is derived from different locations in space (tide gauges for surge and rain gauges for rainfall), there is some potential effect therefore on lag estimation. The authors somewhat refer to this effect in lines 394-395 and mention that this is further discussed in Section 5.3, but it is not discussed any further in that section. I think that elaborating further on this (and potential implications on the findings or methodology overall) is required.
2.  Line 46: “and frameworks”. Elaborate on what frameworks you refer to here, it is current statement is quite vague.
3. “Metro-scale rainfall”. The reason for estimating average over that scale and its incorporation in the overall analysis is not clearly explained.
4. L150: “we eliminate peaks that occur within 5-day windows”. I assume that you mean that only the max peak within a 5-day window was retained(?). Please clarify. 5 days is admittedly a long duration for small scale pluvial flood events.
5. What is the total (and per class) sample size of rainfall and NTR peaks? The exact numbers should be reported for the readers to appreciate the sample size involved in this analysis.
6. Figure 3 is not visually appealing (just the opinion of this reviewer), I wonder if you could improve how you convey the information in this figure.
7. Figure 6: Have you accounted for the bin width when you calculated the values? If not the y-axis should be labeled frequency instead of probability density.
8. Finally, considering that the compound flood hazard (i.e. total flood depth resulting from rainfall and surge) is not explicitly considered and thus conclusions on the influence of the storm types on the hazard cannot be directly derived without coupled simulations (as the author acknowledge), I would recommend modifying the title to “Influence of Storm Type on Compound Flood Drivers” or something along those lines.