A Generalized Framework for Multi-Parameter Optimization of Numerical Wind&ndash;Wave Model: Application to Typhoon Waves near Taiwan Island

Li, Zongyu; Li, Shuiqing; Chen, Jinrui; Kong, Yuan; Fang, Yong; Han, Jiageng; Zhu, Pei; Hu, Po

doi:10.5194/egusphere-2026-895

Preprints

https://doi.org/10.5194/egusphere-2026-895

Preprints

23 Apr 2026

| 23 Apr 2026

A Generalized Framework for Multi-Parameter Optimization of Numerical Wind–Wave Model: Application to Typhoon Waves near Taiwan Island

Zongyu Li, Shuiqing Li, Jinrui Chen, Yuan Kong, Yong Fang, Jiageng Han, Pei Zhu, and Po Hu

Abstract. Accurate simulation of typhoon-induced waves is essential for marine hazard forecasting, yet numerical wave models remain limited under extreme wind conditions due to uncertainties in empirically calibrated parameters. In addition, conventional tuning approaches are inefficient for coordinated multi-parameter optimization. This study develops a multi-objective optimization framework for empirical parameter calibration in numerical wave models. Using the WAVEWATCH III model as a testbed, five key parameters influencing offshore and nearshore wave simulations are optimized for typhoon conditions in waters adjacent to Taiwan Island. Latin Hypercube Sampling is used to generate parameter combinations, and batch simulations are evaluated against buoy observations using root mean square error and bias. An adaptive regression model is constructed to map parameter space to error metrics, and the Non-dominated Sorting Genetic Algorithm III (NSGA-III) is applied to identify optimal parameter combinations. Validation with independent typhoon events shows that the optimized configuration effectively improves significant wave height simulations, reducing both RMSE and bias relative to the default scheme. The proposed framework provides an efficient and transferable approach for improving wave model performance under extreme wind conditions.

Received: 15 Feb 2026 – Discussion started: 23 Apr 2026

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.

Download & links

Zongyu Li, Shuiqing Li, Jinrui Chen, Yuan Kong, Yong Fang, Jiageng Han, Pei Zhu, and Po Hu

Status: final response (author comments only)

RC1:
'Comment on egusphere-2026-895', Anonymous Referee #1, 24 May 2026

Please see the attached for my comments.

Citation: https://doi.org/10.5194/egusphere-2026-895-RC1
- AC4: 'Reply on RC1', Shuiqing Li, 24 Sep 2026
  
  1. Consistency of source-term symbols in Eq. (1)
  Reviewer comment (summary): The source terms and their physical meanings in Eq. (1) are not consistently denoted.
  Response: Thank you for identifying this inconsistency. We have revised Eq. (1) and the accompanying text to distinguish explicitly between bottom-friction dissipation and depth-induced breaking dissipation. The swell-dissipation term is also listed when it is enabled in the actual WW3 configuration. This revision prevents one symbol from being used for different physical processes.
  Revision in the manuscript: Immediately after Eq. (1), we define the wind-input, whitecapping-dissipation, swell-dissipation, nonlinear wave-wave interaction, bottom-friction, and depth-induced-breaking terms, and use the same notation throughout the manuscript. Location in the revised manuscript: Section 2.2, p. 6, lines 129–134 (Eq. (1) and source-term definitions).
  2. Definition of the ST6 wind-input formulation
  Reviewer comment (summary): The relationship between the wave growth rate and the directional wind projection factor W is incomplete.
  Response: Thank you for this comment. We have added the quantitative relationship between the wave growth rate gamma, the nondimensional saturation parameter, and the directional wind projection factor W. We also clarify the favorable- and adverse-wind contributions to W and the role of SINA0 in controlling the negative wind-input contribution.
  Revision in the manuscript: The ST6 wind-input subsection now provides the expressions for gamma, W, W1, and W2 and defines all variables. W1 represents the favorable-wind contribution, whereas W2 represents the adverse-wind contribution. Location in the revised manuscript: Section 2.2, p. 7, lines 153–163 (Eqs. (2)–(4) and variable definitions).
  3. Physical role of CDFAC
  Reviewer comment (summary): The function of CDFAC and its relationship with wind speed and drag coefficient require clarification.
  Response: Thank you for the suggestion. The relationship is now given in Section 2.2 (Eq. 7).
  Revision in the manuscript: The parameter description now includes the drag-coefficient expression, defines U₁₀ as the 10 m wind speed, and explains that FAC/CDFAC modifies the momentum flux rather than the wind speed itself. Location in the revised manuscript: Section 2.2, p. 7, lines 175–182 (Eq. (7) and CDFAC description).
  4. ERA5 wind correction and citation update
  Reviewer comment (summary): The wind-correction method and the cited reference require clarification and updating.
  Response: Thank you for the reminder. The cited paper by Wu et al. had not been formally published at the time of the initial submission; it is now published and has been fully updated in the reference list. Before the WW3 simulations, we apply a region-specific linear correction to ERA5 wind fields rather than using the generic built-in WW3 WCOR option. Because both the training and validation cases are located in the waters near Taiwan in the East China Sea, a region-specific correction was considered more appropriate than a generic coefficient that is not calibrated for this area.
  Revision in the manuscript: The wind-data subsection now includes the correction method, its study-area applicability, and the updated citation: Wu, J., Chen, J., Li, S., et al. (2026). Biases of reanalysis wind fields under typhoon conditions in the Taiwan Strait and their impacts on wave simulations. Marine Forecasts, 43(1), 13-26. Location in the revised manuscript: Section 3.1, p. 10, lines 236–252; reference list, p. 27, lines 623–624.
  5. Data source and high-wind-speed coverage in Figure 3
  Reviewer comment (summary): The data source, station composition, and representativeness of high wind speeds in Figure 3 need to be stated.
  Response: Thank you for this suggestion. Figure 3 uses 593 paired 10 m wind observations during Typhoon Dujuan (2015) from the four nearshore and four offshore stations listed in Table 1. Because these stations did not fully coincide with the typhoon track or the radius of maximum wind, samples above 30 m/s remain limited. Figure 3 is therefore only used to show that the linear correction reduces the systematic underestimation in the available observations.
  Revision in the manuscript: The caption and text of Figure 3 now identify the typhoon event, the eight stations, the sample size, and the limited purpose of the comparison. Location in the revised manuscript: Section 3.1, p. 10, lines 253–261; Figure 3 caption, p. 11, lines 263–265.
  6. Availability of buoy observations across typhoon events
  Reviewer comment (summary): The availability of observations from the eight buoys and the reason for missing results should be explained.
  Response: Thank you for the comment. Eight buoy stations listed in Table 1 were selected for training, parameter optimization, and testing. However, not every station has continuous valid observations during every typhoon event. Some typhoon-station combinations were excluded because of missing records, insufficient samples after quality control, or a lack of observations during the key impact period; reliable performance metrics could therefore not be obtained for those combinations.
  Revision in the manuscript: The data section explains station–event data availability. Table notes and the results text clarify that only station–typhoon combinations with sufficient valid observations are reported. Location in the revised manuscript: Section 2.1, p. 4, lines 112–114; Table 7 note, p. 19, lines 412–413.
  7. Explanation of surrogate-model abbreviations in Table 3
  Reviewer comment (summary): The abbreviations for the optimal models in Table 3 should be explained for readers outside machine learning.
  Response: Thank you for the comment. The abbreviations in Table 3 are now defined as follows: Ridge is L2-regularized linear regression; Poly2_Ridge and Poly3_Ridge are ridge regression with second- and third-order polynomial features, respectively; and SVR_RBF is support vector regression with a radial basis function kernel.
  Revision in the manuscript: Table 3, pp. 12–13; table note, p. 13, lines 313–325.
  8. Clarity of the multi-objective optimization formulation in Eq. (15)
  Reviewer comment (summary): Equation (15), its objectives, and parameter constraints need clearer presentation.
  Response: Thank you for the suggestion. We have reformatted and checked Eq. (15). The two objectives are now explicitly stated as RMSE and absolute Bias, and the search ranges of the five optimized parameters are listed.
  Revision in the manuscript: Section 3.2 now presents the two-objective formulation in Eq. (15), with RMSE and absolute Bias as the objectives and the search ranges of SINA0, SWLB1, CDFAC, GAMMA, and BJALFA explicitly listed. Location in the revised manuscript: Section 3.2, p. 15, lines 344–353 (objective definitions and Eq. (15)).
  9. Definition of N_s and treatment of missing stations
  Reviewer comment (summary): The definition of in the error equations and the treatment of stations with missing observations need to be stated.
  Response: Thank you for the comment. denotes the number of buoy stations with valid observations included in a given aggregate error calculation; it is not necessarily a fixed total number. Stations without concurrent valid observations in a specific typhoon event are excluded from that calculation and is adjusted accordingly.
  Revision in the manuscript: The text following Eq. (18) defines N_s and explains that stations without valid observations are excluded from the aggregate calculation. The note to Table 7 clarifies that only valid station–typhoon combinations are reported. Location in the revised manuscript: Section 3.2, p. 16, lines 367–369; Table 7 note, p. 19, lines 412–413.
  10. Relationship between Table 6 and Figure 6
  Reviewer comment (summary): The case coverage, statistical approach, and different time windows in Table 6 and Figure 6 need explanation.
  Response: Thank you for the comment. Table 6 and Figure 6 present Typhoon Megi only; they do not pool Typhoons Dujuan and Megi. Parameter training used both events. Table 6 reports station-wise RMSE, MAE, and Bias for Megi. Figure 6 shows local windows around the peak wave height at each station, so the plotted time ranges differ among subplots.
  Revision in the manuscript: Training-event allocation, p. 3, lines 84–86; Table 6 caption, p. 17, line 399; accompanying text and Figure 6 caption, p. 18, lines 400–410.
  11. Meaning of “best overall performance” and separation of training and test results
  Reviewer comment (summary): The use of “best overall performance,” station differences, and the separation of training and test results need clarification.
  Response: Thank you for the comment. “Best overall performance” refers only to the lowest RMSE, or the corresponding best metric, among the three schemes at a given station; it does not mean that one scheme is optimal for every station or typhoon. Table 6 therefore reports station-wise metrics for Typhoon Megi. Independent test results for Typhoons Soudelor and Fitow are given separately in Table 7 and Figures 7 and 8.
  Revision in the manuscript: Section 3.3, p. 18, lines 400–404; event-specific test-result descriptions, p. 19, lines 414–426.
  12. Identification and scope of Figures 7 and 8
  Reviewer comment (summary): The test typhoon cases, stations, and treatment of the literature-based scheme in Figures 7 and 8 need to be clearly identified.
  Response: Thank you for the comment. Figures 7 and 8 show the independent test typhoons Soudelor and Fitow at four representative stations. Figure 7 compares the default, literature-based, and optimized schemes in time series. Figure 8 aggregates all valid samples from these cases and compares only the default and optimized schemes.
  Revision in the manuscript: Figures 7 and 8 captions and accompanying text, p. 20, lines 427–438.
  13. Logic of the second-stage optimization
  Reviewer comment (summary): The relationship among wind correction, wave-height correction, and the target function in the second-stage optimization should be clarified.
  Response: Thank you for the comment. All WW3 simulations are forced by linearly corrected ERA5 10 m winds (Eq. 10). In the second stage, residual wind error is removed from the simulated wave height by Eq. (25). The surrogate models are retrained, and NSGA-III is solved again with the same parameter ranges. This second-stage set is the final optimized configuration.
  Revision in the manuscript: Section 4.2, p. 22, lines 483–487, and p. 23, lines 488–496; Section 5, p. 24, lines 520–522.
  
  Citation: https://doi.org/10.5194/egusphere-2026-895-AC4
CEC1:
'Comment on egusphere-2026-895 - No compliance with the policy of the journal', Juan Antonio Añel, 06 Jun 2026

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
To access to the WAVEWATCH III model (version 6.07) code you provide a link to GitHub. However, GitHub is not a suitable repository for scientific publication. GitHub itself instructs authors to use other long-term archival and publishing alternatives, such as Zenodo. Therefore, you must store the code of the model in one of the repositories acceptable according to our policy. Also, to access the ERA5 data you provide a link to the Copernicus Climate Data Store which is also not an acceptable repository. Therefore, you must store the ERA5 data used in your work in an acceptable repository.
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. 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)) 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 include the text of the new Code and Data Availability section, 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

Citation: https://doi.org/10.5194/egusphere-2026-895-CEC1
- AC1:
  'Reply on CEC1', Shuiqing Li, 09 Jun 2026
  
  Dear Dr. Añel,
  
  Thank you for your guidance on the GMD Code and Data Policy. We have taken the following actions to bring our manuscript into full compliance.
  
  1. All materials deposited in Zenodo
  
  All code, configuration files, model results, buoy observation data, and the ERA5 data used in this study have been deposited in Zenodo with the following permanent identifier:
  
  DOI: 10.5281/zenodo.20572975
  
  Link: https://doi.org/10.5281/zenodo.20572975
  
  This single record contains the WAVEWATCH III configuration files, the optimization framework code, buoy observations for four typhoon events, ERA5 wind forcing data subsets, sample model outputs, and all manuscript figures.
  
  2. WAVEWATCH III source code
  
  The original manuscript referenced the NOAA/NCEP WAVEWATCH III GitHub repository (https://github.com/NOAA-EMC/WW3). As this is the community model source code maintained by NOAA — not specific to or produced by this study — we have removed this reference from the Code and Data Availability section in the revised manuscript. A citation to the WAVEWATCH III Development Group (2019) remains in the bibliography where the model is referenced in the text.
  
  If the Topical Editor decides to continue the review process, we will update the Code and Data Availability section in the next manuscript version accordingly.
  
  Sincerely,
  
  Shuiqing Li
  
  Citation: https://doi.org/10.5194/egusphere-2026-895-AC1
  - CEC2: 'Reply on AC1', Juan Antonio Añel, 09 Jun 2026
    
    Dear authors,
    Many thanks for your reply. Unfortunately, we can not accept your proposed solution for the WaveWatch III code. You must store in in a permanent repository, and reply here with the new DOI and link for it, and finally include it in the Code and Data Availability section.
    Please, do as soon as possible.
    Juan A. Añel
    
    Geosci. Model Dev. Executive Editor
    
    Citation: https://doi.org/10.5194/egusphere-2026-895-CEC2
    
    AC2: 'Reply on CEC2', Shuiqing Li, 10 Jun 2026
    
    Dear Dr. Añel,
    
    Thank you for your feedback on our data management. We have now completed the following:
    
    In our previous response, we deposited the WAVEWATCH III configuration files and batch simulation scripts. Upon reviewing your requirements and the GMD Code and Data Policy, we have additionally uploaded the complete WAVEWATCH III v6.07 source code to Zenodo.
    
    We can confirm that all materials necessary for reproducing this study are now stored in Zenodo with the following permanent identifier:
    
    DOI: 10.5281/zenodo.20572975
    
    Link: https://doi.org/10.5281/zenodo.20572975
    This includes:
    
    WAVEWATCH III v6.07 source code
    
    Model configuration files and batch-run scripts
    
    Optimization framework code (Latin Hypercube Sampling, surrogate modeling, NSGA-III)
    
    Buoy observation data for four typhoon events
    
    ERA5 wind forcing data subsets
    
    Sample model outputs
    
    All manuscript figures
    
    We have now fully complied with the GMD Code and Data Policy.
    Sincerely,
    
    Shuiqing Li
    
    Citation: https://doi.org/10.5194/egusphere-2026-895-AC2
    
    CEC3: 'Reply on AC2', Juan Antonio Añel, 11 Jun 2026
    
    Dear authors,
    I have checked the Zenodo repository that you link, and I can not find there the WAVEWATCH III v6.07 source code. Actually, the repository was created on the 6th of June. Please, could you double check it? I think that you need to provide a repository containing the mentioned code yet.
    Juan A. Añel
    
    Geosci. Model Dev. Executive Editor
    
    Citation: https://doi.org/10.5194/egusphere-2026-895-CEC3
    
    AC3: 'Reply on CEC3', Shuiqing Li, 11 Jun 2026
    
    Dear Dr. Añel,
    
    Thank you for your careful review of the Zenodo repository. We have addressed the visibility concern regarding the WAVEWATCH III v6.07.1 source code.
    
    Upon your inquiry, we realized that the original repository contained two separate compressed files (typhoon-wave-optimization.zip and WW3-6.07.1.zip). While both were uploaded, the Zenodo interface displays the first compressed file's contents by default, requiring users to click "Preview" on the second file to view the WW3 source code structure. This may have made the WW3 source code less immediately visible during your review.
    
    To ensure complete transparency and ease of access, we have reorganized all materials into a single, well-structured comprehensive archive: typhoon-wave-optimization.zip(82.2 MB)
    
    This consolidated archive contains:
    
    WW3-6.07.1.zip (21.5 MB): Complete WAVEWATCH III v6.07.1 source code from NOAA/NCEP
    
    optimization/: Python optimization framework (LHS sampling, surrogate modeling, NSGA-III)
    
    ww3_config/: Model configuration files, namelists, bathymetry grid, and run scripts
    
    data/: Buoy observations from 8 stations across 4 typhoon events
    
    ERA2013.nc, ERA2015.nc, ERA2016.nc: ERA5 atmospheric forcing data
    
    README.md: Comprehensive guide to archive contents and usage
    
    figures/: All 9 manuscript figures
    
    With this unified structure, all materials—including the WAVEWATCH III source code—are now immediately visible in the Zenodo file browser's first panel, with no need for additional clicks or previews.
    
    The updated repository is available at: https://doi.org/10.5281/zenodo.20572975
    
    Thank you for your diligent oversight of our submission.
    
    Sincerely,
    
    Shuiqing Li
    
    Citation: https://doi.org/10.5194/egusphere-2026-895-AC3
    
    CEC4: 'Reply on AC3', Juan Antonio Añel, 12 Jun 2026
    
    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-2026-895-CEC4
RC2:
'Comment on egusphere-2026-895', Anonymous Referee #2, 24 Aug 2026

Please refer to the attached file for my review comments.

I would appreciate it if the authors could carefully consider these comments and revise the manuscript accordingly.

Citation: https://doi.org/10.5194/egusphere-2026-895-RC2
- AC5: 'Reply on RC2', Shuiqing Li, 24 Sep 2026
  
  General response
  We sincerely thank Reviewer 2 for the positive assessment and constructive comments. The manuscript has been carefully revised to clarify the physical meanings of the optimized parameters, the rationale for the optimization method, observational-data limitations, deep- and shallow-water applicability, wind-field uncertainty, source-term-package dependence, and figure presentation. Our detailed responses are provided below.
  Comment 1
  Comment: The physical meaning and importance of major variables should be explained, rather than merely listing parameter values.
  Response: Thank you for this suggestion. We have clarified the physical meanings of the five optimized parameters in Section 2.2 and Table 2. CDFAC adjusts air-sea momentum transfer and wind-energy input; SINA0 controls negative wind input under opposing-wind conditions; SWLB1 regulates swell dissipation; GAMMA controls shallow-water bottom-friction dissipation; and BJALFA regulates depth-induced breaking dissipation. Their joint optimization therefore represents wave-energy input, dissipation, and nearshore propagation during typhoon events.
  Revision in the manuscript: The summary paragraph after the BJALFA description in Section 2.2 has been revised, and the Physical Meaning column in Table 2 has been standardized. Location in the revised manuscript: Section 2.2, p. 8, lines 204–208; Table 2, p. 6, following line 149.
  Comment 2
  Comment: A brief description of optimization algorithms would improve readers’ understanding of the method used in this study.
  Response: Thank you for the suggestion. We have added a description of the optimization workflow to improve the readability of the method. First, Latin Hypercube Sampling is used to generate parameter combinations within the prescribed ranges. Second, WW3 simulations are performed for the sampled combinations, and RMSE and absolute Bias are calculated against buoy observations. Third, surrogate models are constructed to approximate the relationship between parameter combinations and error metrics. Finally, NSGA-III is applied to the surrogate models to identify parameter sets that provide a balance between RMSE and absolute Bias. This workflow avoids repeated expensive WW3 simulations during the optimization stage.
  Revision in the manuscript: A description of the LHS-WW3-surrogate model-NSGA-III optimization workflow has been added to the optimization-method section. Location in the revised manuscript: Section 3.1, p. 9, lines 214–216.
  Comment 3
  Comment: It is unclear whether buoy data are available east of Taiwan; the observational gap should be discussed.
  Response: Thank you for the comment. Two of the eight stations in Table 1 are located east of Taiwan: Nearshore 2 (24.0311°N, 121.6325°E) and Nearshore 4 (24.763°N, 121.93°E). Other candidate stations farther east had insufficient valid records during the selected typhoon events and were not included. Coverage at these two eastern stations is also incomplete for some events, so they do not appear in every station-wise training or validation table. The present results therefore remain more representative of the Taiwan Strait and northern nearshore waters, where observational coverage is more complete.
  Revision in the manuscript: Section 2.1 explains the station-selection criteria and the incomplete observational coverage of other candidate stations, particularly east of Taiwan. Location in the revised manuscript: Section 2.1, p. 4, lines 112–114.
  Comment 4
  Comment: The suitability of applying the same parameter values to deep and coastal shallow waters should be discussed.
  Response: Thank you for the comment. We did not calibrate separate deep-water and shallow-water parameter sets. The five parameters were jointly optimized against offshore and nearshore buoys to obtain one regional set for waters adjacent to Taiwan, where the two environments coexist.
  This choice does not assume that the local optima are identical at all depths. In WW3, CDFAC, SINA0, and SWLB1 mainly regulate wind input and swell dissipation, whereas GAMMA and BJALFA control bottom friction and depth-induced breaking, which become important only in shallow water. A single parameter vector can therefore still produce depth-dependent source-term balances. Sensitivity analysis is consistent with this partition: CDFAC dominates the overall error, followed by GAMMA and BJALFA.
  The present stations also do not span open-ocean deep water and the surf zone. The four offshore buoys are on the shelf (about 37–70 m), and the four nearshore buoys are in 19–25 m water. The optimized set is therefore a shelf-to-nearshore compromise. It reduces the regional RMSE and Bias, but it is not optimal at every station or depth; for example, RMSE at Nearshore 1 during Typhoon Megi is not improved relative to the default scheme.
  We now state this limitation in Section 5. Future work will introduce depth-based objective functions or separate offshore and nearshore parameter sets.
  Revision in the manuscript: Section 5, p. 24, lines 530–535.
  Comment 5
  Comment: The high-resolution coastline dataset used in the model configuration should be clarified, for example whether GSHHG was used.
  Response: Thank you for the suggestion. The land-sea mask was generated using the GSHHS (Global Self-consistent Hierarchical High-resolution Shorelines) coastline dataset at a nominal spatial resolution of 0.2 km. A polygon-grid intersection algorithm was used to determine land and sea cells on the regular 0.1° × 0.1° model grid. ETOPO1 Global Relief Model was used as the reference bathymetric dataset, while GEBCO 2024/2025 data at 15 arc-second resolution were used to provide detailed bathymetric information. This information has been explicitly added to the model-configuration description.
  Revision in the manuscript: The model-configuration section now specifies the GSHHS coastline dataset, the 0.2 km nominal shoreline resolution, the polygon-grid intersection method, the 0.1° × 0.1° land-sea-mask grid, and the respective roles of ETOPO1 and GEBCO 2024/2025 bathymetric data. Location in the revised manuscript: Section 3.1, p. 11, lines 271–277.
  Comment 6
  Comment: Figure 6 is difficult to read; the time information and typhoon movement should be made more intuitive.
  Response: We have revised Figure 6 to improve its readability by enlarging the axis labels, legends, and panel annotations. The figure now focuses on time-series comparisons of observed and simulated significant wave heights at six representative stations during Typhoon Megi, one of the training typhoon cases. The time window in each panel was selected to cover the station-specific peak-wave period, allowing the differences among the default, literature-based, and optimized parameter schemes to be compared more clearly. Because the observation-station locations and typhoon tracks are already shown in Figure 1, the repeated spatial panel has been removed from Figure 6 to avoid redundancy.
  Revision in the manuscript: Figure 6 has been revised by enlarging the axis labels, legends, and panel annotations and by retaining only the station time-series comparisons. The caption now explains that each panel covers the station-specific peak-wave period. The repeated spatial panel has been removed because the station locations and typhoon tracks are already presented in Figure 1.
  Comment 7
  Comment: Optimization alone has limitations in representing nonlinear effects and phase differences during extreme typhoon conditions.
  Response: Thank you for the comment. Parameter optimization mainly reduces systematic errors in significant wave height. It cannot fully capture strongly nonlinear processes or correct phase and timing differences under extreme typhoon conditions, which also arise from wind-field uncertainty and the limits of a phase-averaged spectral model. We therefore keep significant wave height as the calibration target. These limitations are now stated in Section 5.
  Revision in the manuscript: Location in the revised manuscript: Section 5, p. 24, lines 524–527.
  Comment 8
  Comment: Because wave models rely heavily on the 10 m wind field, methods for representing nonlinear wind-wave interaction should be discussed.
  Response: Thank you for this important suggestion. We have revised Sections 2.2 and 4.2 to clarify how nonlinear wind–wave interactions are represented in the present model. The linearly corrected 10 m wind field is used only as the external forcing for WW3. Within the source-term balance, the ST6 parameterization represents wind input and associated dissipation as functions of the wind and evolving wave state, while the nonlinear interaction term redistributes spectral energy among wave components. Therefore, the linear correction of the input wind speed does not imply a linear response of significant wave height. We have also clarified that the present offline WW3 configuration does not represent wave-induced feedback on the atmospheric boundary layer.
  
  Revision in the manuscript: A concise explanation of the nonlinear wind–wave response has been added after Eq. (1) in Section 2.2. The first paragraph of Section 4.2 has also been revised to clarify that the statistical wind-speed correction modifies only the prescribed 10 m wind forcing, whereas the nonlinear wave response remains governed by the WW3 source-term balance. The limitation of the offline forcing configuration has also been stated. Location in the revised manuscript: Section 2.2, p. 6, lines 135–138; Section 4.2, p. 21, lines 463–469.
  
  Comment 9
  Comment: The impact of the WAVEWATCH III source-term physics package and nearshore parameterizations on simulation results should be discussed.
  Response: Thank you for the suggestion. The optimized parameters in this study were obtained under the ST6 source-term package and the current bottom-friction and depth-induced-breaking parameterizations. Different source-term packages may alter the relative contributions of wind input, whitecapping dissipation, bottom friction, and depth-induced breaking, thereby affecting both optimal parameter values and simulation accuracy. Therefore, the optimized parameter set is mainly applicable to the current ST6 configuration and the study region, and should not be directly transferred to other source-term packages. Future work will compare different source-term packages, particularly under nearshore shallow-water conditions, and explore depth-based optimization strategies.
  Revision in the manuscript: Sections 2.2 and 5 clarify the configuration-specific applicability of the optimized parameters and outline future comparisons of source-term packages. Location in the revised manuscript: Section 2.2, p. 6, lines 146–148; Section 5, p. 24, lines 529–534.
  Technical correction
  Comment: The spacing between words at Chapter 5, Line 449 appears unusually wide.
  Response: Thank you for the reminder. We have checked and removed the unintended extra spacing at this location and reviewed the manuscript for similar formatting problems.
  Revision in the manuscript: The unintended extra spacing has been corrected. Location in the revised manuscript: Section 5, pp. 23–24, lines 502–538 (formatting review of the Conclusions).
  
  Citation: https://doi.org/10.5194/egusphere-2026-895-AC5

Zongyu Li, Shuiqing Li, Jinrui Chen, Yuan Kong, Yong Fang, Jiageng Han, Pei Zhu, and Po Hu

Viewed

Total article views: 636 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
427	159	50	636	33	30

HTML: 427
PDF: 159
XML: 50
Total: 636
BibTeX: 33
EndNote: 30

Views and downloads (calculated since 23 Apr 2026)

Month	HTML	PDF	XML	Total
Apr 2026	91	40	11	142
May 2026	134	36	6	176
Jun 2026	57	8	12	77
Jul 2026	30	16	2	48
Aug 2026	23	21	4	48
Sep 2026	84	36	14	134
Oct 2026	8	2	1	11

Cumulative views and downloads (calculated since 23 Apr 2026)

Month	HTML	PDF	XML	Total
Apr 2026	91	40	11	142
May 2026	134	36	6	176
Jun 2026	57	8	12	77
Jul 2026	30	16	2	48
Aug 2026	23	21	4	48
Sep 2026	84	36	14	134
Oct 2026	8	2	1	11

Viewed (geographical distribution)

Total article views: 617 (including HTML, PDF, and XML) Thereof 617 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 04 Oct 2026

Short summary

Accurate prediction of extreme ocean waves is crucial for coastal safety, yet numerical models often struggle during severe storms. We developed an automated method to adjust key model settings together, rather than one by one, using observations from typhoon events. Tests show that this approach clearly improves wave height predictions and reduces systematic errors. The method is transparent, efficient, and can be applied to many Earth system models to better simulate hazardous conditions.


Total:	0
HTML:	0
PDF:	0
XML:	0