Next Article in Journal
Marine Boundary Layer Cloud Boundaries and Phase Estimation Using Airborne Radar and In Situ Measurements During the SOCRATES Campaign over Southern Ocean
Next Article in Special Issue
Collaborative Station Learning for Rainfall Forecasting
Previous Article in Journal
Characterization of VOC Emissions Based on Oil Depots Source Profiles Observations and Influence of Ozone Numerical Simulation
Previous Article in Special Issue
Determining Wind Shear Threshold by Using Historical Sounding Data in Experimental Area
 
 
Article
Peer-Review Record

Application of Hybrid Data Assimilation Methods for Mesoscale Eddy Simulation and Prediction in the South China Sea

Atmosphere 2025, 16(10), 1193; https://doi.org/10.3390/atmos16101193
by Yuewen Shan 1, Wentao Jia 1,2,*, Yan Chen 1 and Meng Shen 1
Reviewer 1:
Reviewer 2:
Reviewer 3:
Reviewer 4: Anonymous
Atmosphere 2025, 16(10), 1193; https://doi.org/10.3390/atmos16101193
Submission received: 2 September 2025 / Revised: 29 September 2025 / Accepted: 9 October 2025 / Published: 16 October 2025

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This is a well-structured and valuable study that provides a practical comparative analysis of novel hybrid data assimilation methods in a realistic application to mesoscale eddies in the South China Sea. The results are compelling, clearly demonstrating the advantages of the hybrid methods, particularly IEWVPS for overall assimilation and EnKF-based methods for computational efficiency in forecasting. The paper is recommended for acceptance after proofreading for minor grammatical errors.

Author Response

General comments:  This is a well-structured and valuable study that provides a practical comparative analysis of novel hybrid data assimilationmethods in a realistic application to mesoscale eddies in the South China Sea. The results are compelling, clearly demonstratingthe advantages of the hvbrid methods, particulanly lEWVPs for overall assimilation and EnKF-based methods for computationalefficiency in forecasting. The paper is recommended for acceptance after proofreading for minor qrammatical errors.

 

Response :  Thanks for your suggestion. Since English is not my native language, there is still much room for improvement in my English writing skills. Of course, I am well aware that the English writing standards have a significant impact on the readability of the paper and the readers' experience. Therefore, I will carefully review the grammar errors in the paper and make the necessary corrections. Thank you!

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Please, make the application of DA methods more readable, in particular I recommend to write the algorithm where step by step it will be presented

Comments for author File: Comments.pdf

Author Response

1. The detailed description of the DA algorithm(s) step by step, not only formulas including theirparallelization scheme.

Response :  Thank you for your suggestion to provide more details on the DA algorithm and the parallelization scheme. This requirement is crucial for enhancing the reproducibility of the method. We have supplemented the DA algorithm with a step-by-step detailed description, covering the entire process from initialization, prediction to analysis. The above content will be presented in the form of twos charts and included in Appendix B.

Other responses are reflected in the attachment; please refer to it for review.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

After reviewing this manuscript, I believe it presents an application and comparison study of previously published methods. I only have two simple questions for authors: 

1. Which method is the "Best" you preferred? And good for real application?
2. For the long-term test, how is the performance beyond one month?

Author Response

General comments:  After reviewing this manuscript, l believe it presents an application and comparison study of previously published methods. l only have two simple questions for authors:

1.Which method is the "Best" you preferred? And good for real application?

Response:  This is a very interesting question. From the current development status of marine numerical models in various countries around the world, it can be seen that the stability and computational complexity of 4DVAR and EnKF are significantly superior to those of the PF method. These two methods each have their own advantages. 4DVAR is more suitable for regional models, while EnKF is better suited for global models. However, in the future, as the resolution of numerical models continues to increase, especially for the in-depth study of sub-mesoscale ocean processes, the advantages of the PF method will gradually become apparent. I will incorporate the relevant explanations in the "Conclusion and discussion" section.

 

2. For the long-term test, how is the performance beyond one month?

Response:  Thanks for your suggestion. In this paper, we conducted assimilation experiments and forecast experiments on mesoscale vortices respectively. Firstly, we conducted a 1-month assimilation experiment using the ROMS model in the northern part of the South China Sea to evaluate the effectiveness of different methods. Among them, the hybrid data assimilation methods performed better than the other methods. In the forecast experiment, we only conducted the experiment for 15 days. Because when the forecast period exceeds half a month, the error increases sharply and the mesoscale phenomena become less obvious. This can be understood as exceeding the limit of "predictability". At this point, comparisons among different methods have lost their significance.

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments and Suggestions for Authors

The reviewer considers the topic of this manuscript to be of interest to an international readership. However, several comments are provided to enhance the quality of the manuscript:

  1. In the Abstract, the results should be presented quantitatively rather than qualitatively.
  2. A citation relevant to hydrodynamic modeling and ensemble Kalman filtering should be included, for example: https://www.mdpi.com/2073-4441/16/23/3530
  3. The novelty of this study should be explicitly described in the Introduction section.
  4. Equation numbering should avoid the use of periods; for instance, “(1.)” should be revised to “(1).”
  5. In Section 2.3 (ROMS Configurations), a brief description of the ROMS model would be helpful.
  6. A time step used in the ROMS model should be stated.
  7. Figure A1 should be cited within the main text.
  8. Section 6 (Conclusion and Discussion) should be divided into two separate sections: Discussion and Conclusion. The Discussion should address the study’s limitations and advantages.

Author Response

General comments:  The reviewer considers the topic of this manuscript to be of interest to an international readership. However, several comments are provided to enhance the quality of the manuscript:

1.In the Abstract, the results should be presented quantitatively rather than qualitatively.

Response:  We highly agree that the suggestion of "quantitative presentation of results" you proposed can more intuitively highlight the research value. Based on the core data from sections such as Section 3.2 "RMSE Statistics" and Section 5 "Computational Cost" of the paper, we will supplement key quantitative indicators in the abstract: for instance, we will clearly state that "after data assimilation, compared with the experiment without data assimilation, the Root Mean Square Error (RMSE) of the simulations of mesoscale eddy Sea Surface Height (SSH) in the northern South China Sea by the Ensemble Kalman Filter (EnKF), Localized Weighted Ensemble Kalman Filter (LWEnKF), 4-Dimensional Variational Data Assimilation with Incremental Strong Constraint (IS4DVar), and IEWVPS decreased by 55%, 65%, 65%, and 80% respectively; while the RMSE of their Sea Surface Temperature (SST) simulations decreased by 77%, 78%, 74%, and 82% respectively". Meanwhile, we will also supplement key data on computational cost – "the IEWVPS takes approximately 907 seconds for a single assimilation cycle, whereas the LWEnKF only takes 24 seconds, and its assimilation accuracy in the later stage can approach that of the IEWVPS" – thereby making the effectiveness and practicality of the research results more clearly verifiable.

Other responses are reflected in the attachment; please refer to it for review.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The authors clearly and adequately addressed the reviewer's comments. I recommend this paper to be published as it is.

Reviewer 2 Report

Comments and Suggestions for Authors

I agree with your text my remained my remark about the algorithm's presentation

Comments for author File: Comments.pdf

Reviewer 4 Report

Comments and Suggestions for Authors

The authors should revised the manuscript according to my previous comment. I recommend that the revised manuscript can be accepted for publication. 

Back to TopTop