the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Modelling ocean wave conditions at a shallow coast under scarce data availability – A case study at the western coast of the Mekong Delta, Vietnam
Roderick van der Linden
Dan Trinh Cong
Duong Hoang Thai Vu
Nguyet Minh Nguyen
Frank Seidel
Peter Oberle
Franz Nestmann
Andreas H. Fink
Abstract. Against the background of the rising sea level and land subsidence, protecting the progressively eroding coast along the Vietnam Mekong Delta becomes of tremendous importance. Within the presented work, design conditions for breakwaters were derived from offshore climate reanalysis data (ERA5), which were transferred to the nearshore by two numerical approaches, i.e. SwanOne and Delft3D, for different average and extreme wave and weather conditions. Within this process, design wave heights and periods at the nearshore could be determined for 10- to 100-year recurrence intervals. Both models thereby showed sufficient accuracy according to measurements in the field. Limitations must be made regarding the available spatio-temporal resolution, where reanalysis data showed a lack of short but high peak values compared to the observed measurements. Both numerical approaches showed their capabilities, where SwanOne offers a simple and fast calculation method, while it lacks of continuous effects like wind-generated swell or bottom friction. The Delft3D software on the other hand provides a more complete representation, not only of wave but also current dynamics, while it requires a much broader amount of input parameters and more complex boundary conditions. Within this study, the advantages and disadvantages of both models could be demonstrated, whereas for the final calculation of nearshore wave characteristics, only SwanOne was applicable based on the input parameters extracted from statistical analysis of long term ERA5 data.
- Preprint
(1713 KB) -
Supplement
(821 KB) - BibTeX
- EndNote
Moritz Zemann et al.
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2022-1447', Anonymous Referee #1, 02 Feb 2023
Comments and Suggestions for Authors
General review:
The manuscript of the research article entitled "Modelling ocean wave conditions at a shallow coast under scarce data availability-A case study at the western coast of the Mekong Delta, Vietnam" submitted as egusphere-2022-1477.
The main aim of this study is to determine extreme wave conditions to support the dimensioning of nearshore breakwaters.Therefore statistically analized long-term offshore data (wave & wind) from a climate reanalysis (ERA5) are verified by onsite measurements. Within the framework of a field campaign in July 2019, wave data were measured offshore and nearshore over a two week period. Within a modell approach, the third-generation wave model SWAN (Simulating Waves Nearshore) was applied to simulate wave propagation from offshore to nearshore using the 1D linear approach featured by SwanOne and the 2D spatial approach featured by Delft3D-WAVE. ERA5 data during the measurement campaign were thereby used as model input to validated with the on-site measurements. Afterwards statistical long-term data for average and extreme conditions from 40 year ERA5 time series were used as input to determine the site-specific design conditions. Besides, a comparison of SwanOne and Delft3D was performed for a eight day lasting storm event in 2000, to assess their overall applicability, reliability, and limitations under different conditions.
The authors have investigated an interesting topic in the Mekong Delta Coast, and indeed, the results are promising and valuable for design of detached breakwaters along the coast under scare data availability. The literature review and background are good, and the methodology, results, and conclusions are overall comprehensive and useful for coastal engineering activities in the delta.
Detailed comments:
Comment #1: The coastal bathymetry in the western coast of the MD is very shallow and characterised by gradually increasing slopes ranging only between 1:600 up to 1:1200; Since profile of selected transect 2, shallow but continuously increasing is much different from transect 1 where the sudden seafloor increase at the edge of the shelf (see Figure 2) so that some explanations or clarifications should be presented on SwanOne/ Delft3D computation resulting wave transmission from offshore to nearshore at location of existing detached breakwater installed (about 100-300m). Those clarifications will be more useful for engineering designs.
Comment #2: Continuing above mentioned, in section from line 300-324 and figure 6, there are different wave parameter computed and measured in transect 1 rather than that in transect 2 should be explaned.
Comment #3: Reconsider explanation on differences in Hs and Ts between transect 1 and transect 2 (section from line 372 – 380 and figure 7) it is possible mainly due to different bathymetry profiles but not different models (Swan/ Delft3D)?
Comment #4: Regarding term of “return levels of the maximum individual wave height“ presented in line 180 and figure 3, an clarification should be made for that term is based on what (design criteria or wave parameter specification); In case a new design criteria should be developed as max wave height level, the figure 3 should be refined tecnically for not ”example“, otherwise that figure should be removed to avoid reading confusion.
Comment #5: Minor comments:
- There is no comparison between new recommended method and National design criteria of sea dyke (MARD, 2012) on specification of “design wave parameters”, TCVN points must be removed from figure 2 and para in line 401-402 should be reconsidered.
- Please check the information (line 88-89), long-term wave data are not available from national stations at Phu Quoc island;
- Some recommendations for next studies should be raised in the “conclusion” section.
Citation: https://doi.org/10.5194/egusphere-2022-1447-RC1 -
AC1: 'Reply on RC1', Moritz Zemann, 01 Apr 2023
Response to reviewer 1
Dear reviewer, thanks a lot for your valuable comments. We considered them carefully and will take them into account for the preparation of the revised manuscript.
Comment
Author response
Comment #1: The coastal bathymetry in the western coast of the MD is very shallow and characterized by gradually increasing slopes ranging only between 1:600 up to 1:1200;
Since profile of selected transect 2, shallow but continuously increasing is much different from transect 1 where the sudden seafloor increase at the edge of the shelf (see Figure 2) so that some explanations or clarifications should be presented on SwanOne/Delft3D computation resulting wave transmission from offshore to nearshore at location of existing detached breakwater installed (about 100-300m). Those clarifications will be more useful for engineering designs.According to your comment, the description of the two transects differences at line 128 ff. was supplemented with the following text:
"Transect 2 shown an unconventional beach profile as the coastal plain extends 15 km to the ocean just to be continued with a sudden drop to the ocean bed. This type of cross profile represents the beach morphological around Ca Mau cape. While transect 1 shows a more common cross profile which could be found at the rest of the west coast. As SWAN uses the method of stochastic wave fields, a wave running along transect 2 will face a strong transformation when reaching the sudden drop and therefore most likely will break and dissipate its energy. In contrast the shallow and slowly increasing bathymetry at transect 1 might not cause enough interaction to force an incoming wave to break and change its energy spectrum.”
Comment #2: Continuing above mentioned, in section from line 300-324 and figure 6, there are different wave parameter computed and measured in transect 1 rather than that in transect 2 should be explained.
Dear reviewer, we calculated the significant wave height and the peak period for both transects where we compared for each the two onsite measurements (red), the Input and result of the SWANOne model (blue) and the input and results of the deltf3d model (green) while the dashed line always gives the offshore location and the solid line the onshore location. However, we will change the text to make this parallel approach more visible.
Comment #3: Reconsider explanation on differences in Hs and Ts between transect 1 and transect 2 (section from line 372 – 380 and figure 7) it is possible mainly due to different bathymetry profiles but not different models (Swan/ Delft3D)?
According to your comment, we had some thought about this as well. However, if the differences would originate from the different models, then the blue crosses (SWAN) would not match the green solid line (Delft3D) which means that both models show rather similar results for the different locations. As a consequence, the difference in periods and wave height must be related to boundary conditions at each transect, and here furthermost to the different bathymetries. We will add this explanation to the text as well.
Comment #4: Regarding term of “return levels of the maximum individual wave height“ presented in line 180 and figure 3, an clarification should be made for that term is based
on what (design criteria or wave parameter specification); In case a new design criteria should be developed as max wave height level, the figure 3 should be refined technically for
not ”example“, otherwise that figure should be removed to avoid reading confusion.Dear Reviewer, I think there might be a misunderstanding here: The maximum individual wave height is actually referring to the data from the ERA5 offshore position at transect 1, which were used to calculate the return levels in Figure 3. The return levels for 10, 20, 30, 50 and 100 year return periods were then transferred to the nearshore by SwanOne to determine design wave heights for the breakwater. So these values are not design criteria but the input wave parameter specification. As we calculated these return levels for different wave heights (Hs, HMax) and as well for the wave period at two positions, the figure was labeled an example since we show only one out of the calculation to illustrate the approach. According to your comment we will adapt the text and figure caption to make this more clear to the reader.
There is no comparison between new recommended method and National design criteria of sea dike (MARD, 2012) on specification of “design wave parameters”, TCVN points must be removed from figure 2 and para in line 401-402 should be reconsidered.
According to your suggestion, we removed the points from the figure. However, the idea of designing the coastal protection according to the number of people living in the protected area is a general Vietnamese concept which is mentioned in the dyke protection regulations and therefore should apply for any protection measure. In the case of our study we only make use here of the recommended return period and therefore decided to keep this citation inside our text. However, the TCVN points were removed according to your suggestion.
Please check the information (line 88-89), long-term wave data are not available from national stations at Phu Quoc island;
Thanks for this hint, according to your comment, we changed the name of the station to Tho Chu which is the closest station to our area of interest featuring long term data. Phu Quoc indeed offers some long-term time series but is out of operation since 2010.
Some recommendations for next studies should be raised in the “conclusion” section.
According to your suggestion, we will add some suggestions for further studies to our conclusion section.
-
RC2: 'Comment on egusphere-2022-1447', Anonymous Referee #2, 06 Feb 2023
General comment:
The work by Zemann et al compares the performances of the 1D and 2D SWAN model at two locations along the Vietnamese coast. The research topic may be of interest for the coastal engineering community; however, there are two major flaws that point to a rejection of the manuscript in its present form:
- When calibrating and validating a model, both error and correlation metrics should be computed based on modeled and observed data. The use of indexes such as NRMSE, BIAS, Pearson’s rho etc. allows to assess which model setting works better during calibration, and/or which model provides a better fit with respect to measured data. Here, it seems that the Authors only evaluated the goodness of fit of the curves in Figures 4 and 6 at naked eye, which of course does not suffice and cannot therefore be accepted.
- English grammar is very hard to follow. I am no native speaker, but I found several typos/weird sentence constructions throughout the text. See below for a few examples (but please note the list is not exhaustive):
- Line 29: please review the wording. In the present from it seems like there are two coastal stretches in the whole country, which is obviously not the case. Also, please do not use capital letters after colons.
- Lines 31-32: the two sentences are not related to one another. Therefore, you should remove “However”. Actually, the whole sentence about Vietnam’s economy is not really well suited there.
- Line 59: ofthen -> often. Same at line 90.
- Line 73: depending -> dependent (or “tidal ranges show different intensities depending on” or something like that).
- Line 92: “which approximately equals to 55 km for the 93 MD are.” This sentence makes no sense to me.
- Line 98: modell -> modeling. Please also add the appropriate reference to the SWAN model.
- Line 152: The paragraph is unclear. Please rephrase.
- Line 247: where -> were. See also at line 256.
- Line 250: what does “unconfident” mean?
- Line 259: preiods -> periods
- Line 264: there is a warning message due to a missing link. More attention should be put on the text when submitting a manuscript.
- Line 268: later -> latter.
Other comments:
If I understood correctly, the main advantage of using a 1D approach is that it only needs wave data defined at a single location to be used as boundary conditions. But then, why not to use a 2D grid assuming the same wave data as homogeneous along the grid sides?
Line 86: a range cannot be defined by a single value (e.g., 5.5 s and 3.5 s).
Line 174 on: much more details should be provided when commenting on the Extreme Value Analysis. Why were those distributions selected? It was performed any GOF test? Why do Authors rely on Blocks Maxima approach instead of e.g., a Peak Over Threshold approach?
Line 183: why is the water level only relevant for the 1D model?
Line 207: I do not know about WAM, but I am sure WavewatchIII also allows for wave breaking and triad wave-wave interaction.
Line 222: while I do not argue the suitability of a stationary approach, I would not say that the computational domain is small (indeed it is order of kilometers).
Line 396: Based on the results, I do not understand why the 1D model would be particularly suited for propagating extreme waves.
Citation: https://doi.org/10.5194/egusphere-2022-1447-RC2 -
AC2: 'Reply on RC2', Moritz Zemann, 01 Apr 2023
Response to reviewer 2
Dear reviewer, thanks a lot for your valuable comments. We considered them carefully and will take them into account for the preparation of the revised manuscript.
Comment
Author response
When calibrating and validating a model, both error and correlation metrics should be computed based on modeled and observed data. The use of indexes such as NRMSE, BIAS, Pearson’s rho etc. allows to assess which model setting works better during calibration, and/or which model provides a better fit with respect to measured data.
Here, it seems that the Authors only evaluated the goodness of fit of the curves in Figures 4 and 6 at naked eye, which of course does not suffice and cannot therefore be
accepted.Dear reviwer, according to your comment we calculated the suggested values for NRMSE, Pearsons r and BIAS. NRMSE for wave height and wave period ranged below 0,5 which is typically considered as acceptable, with some are even below 0,2. BIAS for wave heights ranges between +0,07 and -0,18. Pearsons r shows a positive correlation for all wave heights and almost all wave lengths; mostly in the range between 0,3 - 0,9. Altogether we consider this as acceptable level to follow the suggested approach of our manuscript. For all calculations, transect 2 thereby showed a better agreement between measured and modelled values than transect 1. The numbers will be provided and explained in the revised version of the manuscript. For the 2d model, there was already a calibration process, improving the RMSE stepwise, which was given in the supplementary materials 7. However, for the 1D model we have to point out that the model can only be validated as SWAN1, besides the input of wave conditions, wind conditions and bathymetry, gives no further options for optimization, e.g. arranging the bottom roughness or adding tidal effects.
English grammar is very hard to follow. I am no native speaker, but I found several typos/weird sentence constructions throughout the text. See below for a few examples (but please note the list is not exhaustive):
Thanks for your recommendation. Before the resubmission of the revised version, we will perform a professional proofreading of the whole manuscript.
Line 29: please review the wording. In the present from it seems like there are two coastal stretches in the whole country, which is obviously not the case. Also, please do not use capital letters after colons.
The text was changed according to your recommendation and the whole text was checked for capital letters after colons!
Lines 31-32: the two sentences are not related to one another. Therefore, you should remove “However”. Actually, the whole sentence about Vietnam’s economy is not really well suited there.
The text was changed according to the recommendation and the sentence about the economy was skipped
Line 59: ofthen -> often. Same at line 90.
The text was changed in the revised manuscript.
Line 73: depending -> dependent (or “tidal ranges show different intensities depending on” or something like that).
The text was changed in the revised manuscript.
Line 92: “which approximately equals to 55 km for the 93 MD are.” This sentence makes no sense to me.
According to your suggestion, the text was changed: "More accurate climate reanalyzes like ERA5 (ECMWF Reanalysis 5) do not consider the relevant areas close to the coast as they typically start with around 10 km distance and comprise a grid resolution of 0.5°, which approximately equals to 55 km² for the MD."
Line 98: modell -> modeling. Please also add the appropriate reference to the SWAN model.
We added an appropriate reference (The Swan team, 2023) to the text.
Line 152: The paragraph is unclear. Please rephrase.
The text was changed to a more clear statement: "To compare these measurements with the hourly available wave heights and periods from ERA5, an on hour average was calculated over the same timesteps from the measured data and later used for verification. "
Line 247: where -> were. See also at line 256.
The text was changed in the revised manuscript.
Line 250: what does “unconfident” mean?
“unconfident” was changed to “unrealistic”
Line 259: preiods -> periods
The text was changed in the revised manuscript.
Line 264: there is a warning message due to a missing link. More attention should be put on the text when submitting a manuscript.
Thanks for this hint, you are totally right. We apologize for not checking the manuscript with the appropriate accuracy before we submitted it to the journal. The missing link was added!
Line 268: later -> latter.
The text was changed in the revised manuscript.
If I understood correctly, the main advantage of using a 1D approach is that it only needs wave data defined at a single location to be used as boundary conditions. But then, why not to use a 2D grid assuming the same wave data as homogeneous along the grid sides?
The goal of our research is actually to verify whether 1D could be used instead of 2D, specifically for the case of the Mekong Delta. In this case using 1D would be much more economic in terms of the necessary input data, pre-processing (setup of the model), computational time, and post-processing compared to 2D. In fact you do not only need the wave data equally as well the bathymetry, wind and current data, tidal changes etc.. Each of them is needed not only for a single transect but for the whole area. We will add some explanations about this to the revised manuscript to highlight the economical benefit of 1D compared to 2D.
Line 86: a range cannot be defined by a single value (e.g., 5.5 s and 3.5 s).
The text was changed: "Average wave periods followed a similar pattern with longer periods corresponding to greater wave heights, with an average of 5.5 s for the eastern coast during the northeast monsoon, and an average 3.5 s at the western coast during the southwest monsoon (Marchesiello et al. 2017). "
Line 174 on: much more details should be provided when commenting on the Extreme Value Analysis. Why were those distributions selected? It was performed any GOF test?
Why do Authors rely on Blocks Maxima approach instead of e.g., a Peak Over Threshold approach?We are aware that there are different approaches for estimating return levels like using POT instead of block maxima as mentioned in your comment. For the revised version of our manuscript, we will carefully assess if it is meaningful to use a different approach than the one which is presently used for the different variables. At least we will more carefully justify our choice of distributions backed up by GOF tests and provide more references to previous studies.
Line 183: why is the water level only relevant for the 1D model?
The text was changed and includes now both models!
Line 207: I do not know about WAM, but I am sure WavewatchIII also allows for wave breaking and triad wave-wave interaction.
Thanks for your comment. Your totally right here! According to your hint, we remove the Wavewatch III citation from our manuscript. For WAM we doublechecked that it does not include breaking and triad wave-wave interaction in their propagation equation and therefore remain that one.
Line 222: while I do not argue the suitability of a stationary approach, I would not say
that the computational domain is small (indeed it is order of kilometers).According to your comment, we changed the text to the following, trying to avoid further misunderstandings of "small":
“Additionally, the simulation is only calculated in a stationary mode. This is reasonable for applications where wave travel through computational domain within simulation period (Delft WAVE user manual, p49). Moreover, in this case bathymetry contour lines stay relatively parallel to each other and the coast, and stationarity assumptions of instantaneously reacting waves to the wind field fluctuation are acceptable (Rogers et al. 2007).”
Line 396: Based on the results, I do not understand why the 1D model would be particularly suited for propagating extreme waves.
You are right, in fact both models are suitable to propagate extreme waves in an appropriate way. However, the presented approach based on scare data availability in terms of measurements and therefore follows the idea of using extreme wave conditions derived from statistical analysis of ERA5 as input. Such a simplified approach would not be possible for Delft 3D or at least would take much more time in generating all necessary boundary conditions, as much more parameters at multiple locations are needed to set up the model. The text was adapted here to clarify this detail.
Moritz Zemann et al.
Moritz Zemann et al.
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 317 | 95 | 16 | 428 | 30 | 3 | 7 |
- HTML: 317
- PDF: 95
- XML: 16
- Total: 428
- Supplement: 30
- BibTeX: 3
- EndNote: 7
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1