the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
UPSurgeML (v1.0) – A Machine Learning Workflow for Probabilistic Ensemble Forecast of Tropical Cyclone Storm Surge
Abstract. UPSurgeML is an open-source software package for probabilistic storm surge analysis. The code is written in Python, C, and Fortran, and the codebase complies with the National Centers for Environmental Prediction’s (NCEP) guidelines, making it suitable for deployment on High Performance Computers (HPC) like the Weather and Climate Operational Supercomputing System (WCOSS). By utilizing an unstructured mesh, UPSurgeML refines the prediction of water elevations in complex coastal regions. It provides both deterministic and probabilistic storm surge forecasts driven by a small ensemble of hurricane forecasts. UPSurgeML uses parametric hurricane wind and pressure fields, and astronomical tide forcing with a hydrodynamic ocean model to simulate water elevations. Unlike the current Probabilistic Surge (PSurge) model utilized by the National Hurricane Center (NHC), UPSurgeML employs a machine learning surrogate modeling framework. This approach enables a smaller (order of magnitude reduction) hydrodynamic ensemble size that would otherwise be necessary. Consequently, UPSurgeML provides high-resolution water elevation probability fields and exceedance levels with reasonably low computational cost, and has demonstrated its usability as experimental guidance during the hurricane seasons.
- Preprint
(2245 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 18 Dec 2026)
-
CEC1: 'Comment on egusphere-2026-4586 - No compliance with the policy of the journal', Juan Antonio Añel, 23 Sep 2026
reply
-
AC1: 'Reply on CEC1', Fariborz Daneshvar, 28 Sep 2026
reply
Dear Dr. Añel,
Thank you for your prompt review of the manuscript and for guiding us on the journal's Code and Data Policy. To address your specific comments:
-
Third-Party Codes and Data: I published a newer version of code and data on Zenodo (https://zenodo.org/records/23022229) and added the exact versions of SCHISM, PaHM, and StormEvents code and data that were used for this manuscript. These are now explicitly listed in the revised Code and Data Availability section.
-
Machine Learning Training Data: All datasets (including training data) used to develop UPSurgeML are fully included in the primary (https://doi.org/10.5281/zenodo.20801237) and revised Zenodo repositories (https://zenodo.org/records/23022229). This is now explicitly explained in the revised Code and Data Availability section.
Below is the revised Code and Data Availability Statement that incorporates these updates:
"""
Code and Data Availability Statement:
The specific version of the UPSurgeML model used in this manuscript, along with all associated training data, input/output files, and scripts used to generate plots, are permanently archived on Zenodo at https://zenodo.org/records/23022229 (Daneshvar et al., 2026). For active development, the latest version of UPSurgeML is maintained on the NOAA Office of Coast Survey (OCS) GitHub repository (https://github.com/noaa-ocs-modeling/UPSurgeML).
This work relies on third-party models and datasets including SCHISM, Parametric Hurricane Model (PaHM), and StormEvents. The exact source code versions of SCHISM (@7fc47d1), PaHM (@2043c99), and StormEvents (v2.3.7) used in this study have been permanently archived on Zenodo at https://zenodo.org/records/23022229 (Daneshvar et al., 2026) to ensure replicability.
"""Please let us know if these updates satisfy the journal's policy so that our manuscript may proceed to the discussion and peer-review phase. We are happy to make any further adjustments if necessary.
Sincerely,
Fariborz Daneshvar, Ph.D.
Senior Physical Scientist | Ocean Associates Inc
Storm Surge Modeling Team
Coastal Marine Modeling Branch
NOAA National Ocean Service | Office of Coast SurveyCitation: https://doi.org/10.5194/egusphere-2026-4586-AC1 -
CEC2: 'Reply on AC1', Juan Antonio Añel, 29 Sep 2026
reply
Dear authors,
Thanks for addressing this issue. I have checked the repositories and we can consider now the current version of your manuscript in compliance with the code policy of the journal.
Juan A. Añel
Geosci. Model Dev. Executive Editor
Citation: https://doi.org/10.5194/egusphere-2026-4586-CEC2
-
-
AC1: 'Reply on CEC1', Fariborz Daneshvar, 28 Sep 2026
reply
-
RC1: 'Comment on egusphere-2026-4586', Anonymous Referee #1, 10 Nov 2026
reply
Review Comments for: “UPSurgeML (v1.0) – A Machine Learning Workflow for Probabilistic Ensemble Forecast of Tropical Cyclone Storm Surge”
In this manuscript, the authors present and evaluate the use of a new machine learning probabilistic ensemble forecast system, called UPSurgeML. They utilized the UPSurgeML model to evaluate three different perturbations of Hurricanes Ike, Sandy and Idalia. The methodology is presented along with the evaluation of these three hurricanes for the 50% median changes and 10% exceedance with respect to the maximum water elevations. Currently, I recommend that this manuscript undergo major revisions. The authors should consider the following points when revising the manuscript:
- The authors should consider adding a significant amount of the background information for the methods utilized in this work. For example, they are using a machine learning technique discussed in the Pringle 2023 study; however, they never provide summary details of the methodology and results. Currently, most of the references to background information utilized in this study is only one sentence in the manuscript.
- Were Penny (2023) and persistent radius of maximum wind options also tested with the central pressure options presented in Chavas (2025) and Courtney and Knaff (2009)? Why do the UPSurgeML framework options presented in Table 1 include the non-recommended combinations (e.g., persistent Holland B)? What is the benefit to evaluating these different methods?
- The authors should update their manuscript to be more of a journal article instead of a technical report. Currently this article reads more like a technical report with the addition of the information on running the program. For example, the sentence on page 12 lines 10-13 reads like information that would be included in a user guide than a scientific article. Additionally, this is true of page 12 line 17 and page 13 line 7. The authors might consider including this information as supporting documentation within the appendix or perhaps combined into a single figure with the full workflow for the Sandy test case.
- The authors should consider moving Figure 1 later in the manuscript after they have provided the descriptions of the different models. Currently, the figure has included many abbreviations that have not been defined. For example, PAHM is included in the figure; however, this is not defined until page 8 in the Simulation section.
- Figure 3 shows the trajectory and intensity for the three hurricanes evaluated; however, it is not easy to determine which hurricane each track is as the variation in the lines does not appear in the figure. Additionally, it is hard to evaluate the official track in Figure 4. The authors should consider updating this figure with a different color for this track. The authors should spell out “official” in legend.
- The authors have chosen three hurricanes as their testcases; however, they did not provide the reason these three storms were chosen. The authors should include additional information on these storms and why they are being used to evaluate the UPSurgeML framework. Additionally, on page 10, the authors state that “Hurricane Ike was one of the largest storms in the Gulf Coast region” – the authors should include a reference for this information.
- The authors might consider comparing their results to PSurge to show what additions the UPSurgeML framework provides as compared to the previous methodology.
- The authors indicate that UPSurgeML framework was tested during the 2025 Atlantic hurricane season. What were the results from the hurricane season?
- Lastly, there are some issues with grammatical errors in the manuscript. The authors should carefully read through the manuscript again.
- On page 6 (line 6): “strom” needs to be “storm”.
- On page 6 (lines 11-12): In the previous sentence, two options are presented for retrieval. But in this location, the authors introduce a third option. The authors should consider revising the text for clarity on the different options.
- On page 6 (lines 17-23): The authors should consider adding more detail for each of the listed methods for missing central pressure and radius to maximum winds. Currently, the authors only provide references with no description of the actual methods.
- On page 7 (lines 3-4): The authors mention the KL-PC surrogate framework here; however, they do not define KL-PC until page 9 (line 16).
- On page 8 (lines 17-18): Consider changing “using 240 seconds (four minutes) timesteps” to “using 240-seconds (4-minute) timestep”.
- On page 10 (line 9): remove “the” before landfall.
- On page 10 (line 14): add “third most” for devastating storm.
- On page 11 (line 11): “dev” should be included as “development”. The authors should search through the document for instances of "dev"..
- Figure 5: Are the squares also color-coded by water elevation or are they just black outlines?
Citation: https://doi.org/10.5194/egusphere-2026-4586-RC1
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 460 | 197 | 60 | 717 | 63 | 57 |
- HTML: 460
- PDF: 197
- XML: 60
- Total: 717
- BibTeX: 63
- EndNote: 57
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Dear authors,
Unfortunately, after checking your manuscript, it has come to our attention that it does not comply with our "Code and Data Policy".
https://www.geoscientific-model-development.net/policies/code_and_data_policy.html
The Code and Data Availability section of your manuscript does not contain the repositories for the code and data used in your manuscript. This includes codes such as SCHISM, PaHM, StormEvents, etc. which are cited in the text and references, but not mentioned in the Code and Data Availabilty section. Moreover, they such codes are stored in sites that do not comply with the policy of the journal. It is specially striking that you cite several Git sites -- or even published papers, no repositories--- when the policy of the journal makes clear that Git sites are not acceptable.
Also, it is not clear if the training data that you have used is available in the Zenodo repository that you provide, as the training process for the system presented is not explained in your submitted manuscript. For machine learning works submitted to the journal we request that all the training and output data for the presented models are properly stored in repositories acceptable according to the policy of the journal.
The GMD review and publication process depends on reviewers and community commentators being able to access, during the discussion phase, the code and data on which a manuscript depends, and on ensuring the provenance of replicability of the published papers for years after their publication. Due to the lack of the compliance mentioned above, your manuscript should not be accepted for Discussions and peer review in the journal. Please, therefore, publish your code and data in one of the appropriate repositories and reply to this comment with the relevant information (link and a permanent identifier for it (e.g. DOI) -- a new version of the Code and Data Availability section) as soon as possible. We cannot have manuscripts under discussion that do not comply with our policy.
Later, if the Topical Editor decides to continue with the review or publication process of your manuscript and you are requested to upload a new version of it, then The 'Code and Data Availability’ section of your manuscript must also be modified to cite the new repository locations, and corresponding references added to the bibliography.
I must note that if you do not fix this problem, we cannot continue with the peer-review process or accept your manuscript for publication in GMD.
Juan A. Añel
Geosci. Model Dev. Executive Editor