This paper presents a well-structured study with a solid methodological approach to assessing the spatial correlation of potential compound flooding along the U.S. coastlines. The work represents an appreciable scientific contribution by addressing a relevant and previously underexplored aspect of compound flooding.

I recommend acceptance subject to technical corrections. In particular, there is a typographical error in line 138, where “Santa Monita” should be corrected to “Santa Monica.” There may be other similar typographical issues worth checking.

No further technical issues were identified.

The study proposes a statistical framework for evaluating the spatial correlation of compound flooding, demonstrating its use for some coastal locations in the USA. While the methodology appears potentially valuable for flood risk assessments in coastal cities, the manuscript currently lacks sufficient detail to fully convey its approach and practical implementation. This limits the reader’s ability to understand and apply the proposed framework. Furthermore, the accompanying scripts intended to reproduce the analysis are incomplete.

I offer the following comments to help improve the article.

1. L108: Why did the authors use daily data and not instantaneous (e.g., hourly or 15-minute) data?
2. L120-139: Instead of filling up data gaps based on information from nearby stations, would have been possible to perform all the analyses only considering time periods and storms unaffected by these gaps? Please discuss pros and cons of the two alternative approaches.
3. L161-186, as well as Section 2.4, provide a description of critical methodological steps of this work; however, they lack a rigorous mathematical framework. The authors should include suitable mathematical formulations, possibly using subscripts for referring to generic locations or events, when explaining the proposed approach.
4. While addressing the previous comment, the authors should clarify if and how they account for mixed-population effects within each water-level or flow time series. In several parts of the manuscript (e.g., L288, L355, L368, L415), the authors observe that there are different possible generating mechanisms for these extremes, such as tropical cyclones and extra tropical cyclones affecting storm surges, or snowmelt and convective events affecting riverine flooding. Could the coexistence of different parent distributions limit the applicability of the proposed methodology? The authors should discuss these aspects in detail.
5. L402: it is expected that larger storms will have a greater spatial footprint and affect a larger number of stations. The authors should therefore rephrase the sentence at L402. For example: “However, these relative occurrence rates become significantly higher with increasing thresholds, reflecting the fact that bigger storms can affect more locations”
6. The code provided to reproduce the analyses (assets for the review process) apparently lacks several parts, including a routine for automated threshold selection (L154), and subroutines that are called within the provided scripts. Those subroutines contain functions that are used in the shared scripts; without those files, it is impossible to run the provided codes. E.g., in the R script “Cal_dependence_autmoated_thres.R”, calls to the following missing scripts are included: “transFun.HT04.R”, “predict.mex.conditioned.R”, “u2gpd.R”, “revTransform.R”, “Migpd_Fit.R”, “mexTransform.R”.

Minor comments

1. L49-50: add references supporting this statement regarding the effects of TCs and ETCs on extreme and moderate, more frequent events, respectively.
2. Fig 2b, L293-296: How were the 5th-95th-percentile confidence intervals obtained?
3. Fig 4 is unclear. Does the y-axis show the percentage of events affecting 1, 2, …, etc. distinct locations over the 10,000-simulation period? If yes, please consider using a more informative label for the y-axis.