Very high-resolution modelling of submesoscale turbulent patterns and processes in the Baltic Sea

Onken, Reiner; Baschek, Burkard; Angel-Benavides, Ingrid M.

doi:https://doi.org/10.5194/os-16-657-2020

Articles | Volume 16, issue 3

https://doi.org/10.5194/os-16-657-2020

© Author(s) 2020. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/os-16-657-2020

© Author(s) 2020. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 16, issue 3

Research article

|

26 May 2020

Research article |

| 26 May 2020

Very high-resolution modelling of submesoscale turbulent patterns and processes in the Baltic Sea

Reiner Onken, Burkard Baschek, and Ingrid M. Angel-Benavides

Download

Final revised paper (published on 26 May 2020)
Preprint (discussion started on 13 May 2019)

Interactive discussion

Status: closed

AC: Author comment | RC: Referee comment | SC: Short comment | EC: Editor comment

- Printer-friendly version

- Supplement

RC1: 'Referee Comment', Anonymous Referee #1, 14 Jun 2019
- AC1: 'os-2019-44', Reiner Onken, 24 Jun 2019
  - RC3: 'Reply of Reviewer #1 to Author's Comments os-2019-44-AC1-supplement.pdf', Anonymous Referee #1, 27 Jun 2019
    - AC2: 'Reply to Reviewer #1', Reiner Onken, 02 Aug 2019
RC2: 'Review', Anonymous Referee #2, 20 Jun 2019
- AC4: 'Reply to Reviewer#2', Reiner Onken, 05 Aug 2019
RC4: 'Review', Anonymous Referee #3, 18 Jul 2019
- AC3: 'Reply to Reviewer#3', Reiner Onken, 02 Aug 2019

Peer-review completion

AR: Author's response | RR: Referee report | ED: Editor decision

AR by Reiner Onken on behalf of the Authors (05 Aug 2019) Author's response Manuscript

ED: Referee Nomination & Report Request started (09 Aug 2019) by Markus Meier

RR by Anonymous Referee #1 (10 Aug 2019)

RR by Anonymous Referee #3 (27 Sep 2019)

Suggestions for revision or reasons for rejection

Onken et al. Very high-resolution modelling of submesoscale turbulent patterns
and processes in the Baltic Sea; Ocean Sci. Discuss., https://doi.org/10.5194/os-2019-44

Review 2

Major comments

I thank the authors for revising the manuscript and replying to my comments. The paper has been improved a lot. However, I still suggest to make some changes and I am not satisfied with some answers to my questions/comments and no action in those instances. Please find my arguments and some further comments below (page and line numbers below refer to the revised manuscript).

P1L8: Please delete the statement about the characteristic vertical speeds of 100 m day-1. I have at least three reasons for that. First, you define STPPs as submesoscale turbulent patterns and processes not mentioning that those also include internal waves. Secondly, you give some lower speed values in the Results and Discussion sections – you give 20 m day-1 and 15 m day-1, instead; this number 100 m day-1 is mentioned in Conclusions and Abstract only. Third, you suggested that the period of such internal waves is about 4 hours; thus even regarding internal waves the vertical speed should not be given using units meters per day (use m per hour or mm per s).

P1L19: Add the word “turbulent” when first mentioning STPPs.

P4L12: What do you mean by the near-adiabatic conditions in this context?

P6L7-13: The use and the impact of the cubic spline are still not clear to me. No criterion of choosing it is given except a subjective criterion of more likelihood of the submesoscale patterns to those expected. I am also not satisfied with the reply to my earlier comment on that. I do not understand how the more likely pattern with cyclonic shear zone narrower than anticyclonic shear zone appear by using the cubic spline if it was not the case when using linear interpolation. This effect could only occur at length scales less than 1 km if you interpolate a spatial field and the values in every fifth grid point of R100 are equal to the corresponding values in R500. Does it mean that you create these structures at the scales <1km while they were not present in the parent model?

P6L31-P7L4: I still do not understand this statement about salinity as the ideal parameter. You just have to accept the discrepancy in the spatial resolution of your model and atmospheric forcing that could cause some peculiarities in your model results at small spatial scales. Surface temperature and salinity both determine surface water density and, thus, vertical stratification, which, in turn, influences vertical mixing. The latter creates variability in both salinity and temperature distribution (vertically and laterally).

P7L13-14: The use of the cumulative average TKER500 is still unclear.

P8L29-31: Also here, a subjective criterion is mentioned to justify the choice of the mixing scheme. Why do you state that the dumping of STPPs should be minimal?

P9L1-2: I think you should not directly compare your numerical experiment with the observations.

P9L11: Earlier, you stated that salinity is an ideal parameter, but here you picked up the temperature distribution.

P9L20-21: Please, correct it. What does it mean that frequency grows with the frequency interval? I think the figure caption of Fig. 6 should be corrected as well (see below).

P11Section4.2.3: Is the length scale of 1 km popping up here and there something characteristic for this study or has it more general meaning? Why different processes and structures like internal waves and fronts exhibit this length scale?

P12L1-8 and Fig 10: I would like to see the entire depth range of this figure. I guess, only 3-4 grid points are in the depth interval of 0-10 m while along the horizontal axis the number of grid points is two orders of magnitude larger. Could the figure be presented without interpolation in vertical (as colored boxes)?

P12L9: It should be Section 4.4.3.

P13L23 and Fig. 11: Why different examples are drawn from different depths (here at 2 m depth, you also use 3 m, 5 m and the base of the first layer)?

P14L21-22: Here you differentiate between the terms STPPs and internal waves.

P20L3-16: Please remember that you compare the model results without atmospheric forcing and observations (in some other part of the ocean). In addition, you compare salinity at 15 depth with SST (since in your run, the cyclone is hardly visible in SST). Is this approach acceptable?

P21L31-32: Is it something general or related to the Baltic Sea conditions (why 40 m)?

Tables

Table 1: You answered my previous comment that you have only 10 vertical layers in your model. I do not know any other modeling study having so small number of vertical layers if the aim was to model submesoscale patterns. Is this characteristic ratio of vertical to horizontal step of 1:20 a good choice? Referring to HBM, which also has similar characteristic layer thickness, is not relevant since they do not have an aim to permit submesoscale features in HBM.

Figures

Fig. 6: Please, explain what is on the vertical axis. Are the values the number of occasions with the gradient less than the value on the x-axis? I guess it is not the frequency.

Fig. 10: Could it be possible to show the entire depth range? How many grid points you have in vertical in the depth interval of 0-10 m?

Hide

RR by Anonymous Referee #4 (07 Oct 2019)

Suggestions for revision or reasons for rejection

Referee report of the revised manuscript “Very high-resolution modelling of submesoscale turbulent patterns and processes in the Baltic Sea” (version of 2 Aug 2019) by Reiner Onken, Burkard Baschek, and Ingrid M. Angel-Benavides, submitted for consideration to Ocean Science.

---

General comments

This manuscript describes a modelling study of submesoscale dynamics in the southern Baltic Sea. The topic of the paper is interesting. While there are studies of submesoscale dynamics from the Baltic area, more are surely needed. The authors have clearly dedicated considerable effort to their investigation and I appreciate the motivation of this manuscript. I did not review the original manuscript, and therefore will review the revised manuscript in view of the interactive public discussion available online. Overall, I see that several noteworthy improvements have been made to the manuscript during the peer-review process. However, after reading the revised manuscript, I have some remaining concerns that I would like to be addressed before publication.

First, I must admit that I am still somewhat unclear about the objectives of the manuscript. Upon my first reading, I got the impression that the purpose of the manuscript was to attempt realistic modelling of the submesoscale phenomena to aid “the interpretation of observations that were taken during the “Expedition Clockwork Ocean”, as is stated in the abstract. Upon closer inspection it seems that, while the configurations are mostly realistic, there are aspects that could be described simplifications or idealisations, and that connection to “Expedition Clockwork Ocean” is minimal.

I wish the authors would once more revisit their abstract and introduction to reconsider how they state their aims and hypothesis in the light of their main objective. This would make it possible for the reader to better assess the impact of some of the methodological choices made in the manuscript. As it currently stands, it seems the stated objective does not fully correspond the actual research.

Second, there seems to be very limited possibilities for the validation of the two ROMS configurations. I would hope that this was better reflected in the manuscript, and that the associated uncertainties are clearly stated. Overall, the limitations of the study are not that clearly articulated, especially in the abstract. All modelling has its limitations, and this is often OK, as long as those limitations are acknowledged in a sufficient manner.

Third, I must also say that I found some of the methodological choices (e.g. resolution, run length) surprising and I had difficulty understanding the rationale behind them. I feel that these choices are not always adequately justified in the text. I hope that the authors would either a) justify these methodological choices in more detail to relieve potential concerns or b) consider if the objective of the paper would be better served if the modelling configurations were once more revised and models rerun. I guess my main concern here is that as there are some limitations with respect to validation, then it is even more important that modelling methodology is well thought out and clearly articulated.

Based on my reading of the manuscript, I think the manuscript can eventually be suitable for publication, if these most major problems can be resolved and limitations of the work are clearly enough articulated. Please see the specific comments for more details on these issues.

---

Specific comments

p1,4,6,8,22,29: What is the purpose of the frequent mentions of “Expedition Clockwork Ocean”? As far as I can tell, the only relation to this paper seems to be that the modelling period was chosen to coincide with the experiment, but the results of the experiment are not referenced. Wouldn’t it be sufficient to briefly mention this experiment once as the motivation for the modelling domain and time span, e.g. in section 2.1? Currently, it leads the reader to (incorrectly) assume that the results of the experiment would be discussed.

p1l3: “The modeled STPPs provide an aid for the interpretation of observations that were taken during the “Expedition Clockwork Ocean” in the same region in June 2016.” One can understand this sentence so that “June 2016” only applies to the observations and assume the model run is longer. It would be important to clearly state in the abstract that the model was run for June 2016 only. Please consider modifying the text accordingly.

p1l6: “The comparison of the results with satellite images shows good agreement.” I would disagree with this characterisation. Yes, the comparison presented shows fair agreement, but the validation is far from complete, somewhat subjective, and, it seems, quite difficult to do. Please consider reformulating so that these difficulties of validation are reflected in the abstract.

p1l7: Please insert “Both ROMS configurations used 10 sigma levels in the vertical.” or something similar to the abstract. The number of levels is important to mention here, as it is smaller than what most modellers in the area would likely expect nowadays.

p1l21: “In this article, high-resolution modeling is used to understand and interpret the observed features.” It seems to me that the observations from “Expedition Clockwork Ocean” are not used in any significant degree?

p5l3: Section “2.2 HBM”. “the model domain is separated into 25 depth levels” This, I think, is misleading. 25 is not the number of levels in the native HBM model grid, but rather just the number of levels provided in the CMEMS product. The native grid is much finer (as indicated in the documentation of the product available from the CMEMS website, http://resources.marine.copernicus.eu/documents/PUM/CMEMS-BAL-PUM-003-006.pdf, p13). This difference should be clearly indicated in the manuscript, especially since your vertical resolution is much coarser.

You say that you have used the HBM model results. I think it would be more accurate to say that here that you are using the CMEMS Baltic Sea forecast product, which is currently based on the HBM model. This is significant for a number of reasons. Most notably, the output can have differences with the native model output like the one with vertical resolution. Furthermore, the output files you are using can in principle originate from different forecasting centres, if there have been operational issues, or even from different model versions, if your time span is long. For these reasons, I recommend you e.g. consider changing the title of this subsection to “CMEMS Baltic Sea forecast product” or something similar, and revise the text accordingly

Also, the CMEMS product is updated every now and then, resolutions and other specifications change. For this reason, it would be good to clearly specify which version of the product you used, and to link/reference the correct version of the user guide for the product, and possibly also the corresponding validation documentation. For this very same reason, it would be helpful for the reader if the main characteristics of the product were clearly reiterated in the text. Most notably, I would have hoped to see some reference to the atmospheric forcing source and resolution here.

What was the reason for selecting this forecasting product as the boundary condition for your ROMS configurations, other than its easy availability? Did you consider other options? It seems to me there were suboptimal features in the CMEMS product for your particular application (such as the poor vertical resolution in the product when compared to the native grid), and perhaps a more suitable source for boundary data could have been identified. For example, have you contacted the CMEMS service desk or the authors of the HBM model to see if files more suited for your application could be provided for research purposes? (E.g. more vertical levels.) Or, perhaps the CMEMS reanalysis product for the Baltic Sea would have been a better choice? (As the reanalysis product is also from time to time updated, it is of course possible it did not cover your investigation time at the time you made the model runs. It does cover it now, however.)

p5l23 “In addition, the bathymetry of HBM is not contained in the output provided by CMEMS.” Did you ask if the CMEMS service desk could provide the bathymetry file for you? If so, what was their response? I find it surprising if this information was not made available for you, especially if you requested for it for scientific purposes.

p6l20 What was the reason to use DWD forcing? How does the resolution of this dataset compare to the resolution of the atmospheric forcing data used in the HBM model? As the atmospheric forcing in ROMS differs from HBM, I would appreciate if this choice was justified in the text.

p8l1 “On the whole, the apparent similarities between the observations and the R500 simulations make the model output trustworthy.” I would disagree with this statement. Yes, the comparison presented shows fair agreement, but the validation is far from complete and somewhat subjective (as evidenced by the previous round of referee comments, e.g. RC3). I think the best you can say based on this validation is that it builds some basic confidence that your model configuration may produce realistic results. Please consider reformulating this accordingly.

p8l31, and p28 Tables 1 and 2. These tables indicate that the number of grid points in both ROMS configurations is, by today’s standards, manageable (apprx. 400 times 1000 times 10). The time span of the run was quite short for both of the ROMS configurations, under one month. Could you please elaborate how you reconcile this with the claim that “R100 is rather expensive in terms of computer resources” on page 8? From my experience I find this claim rather surprising. In fact, your response to reviewer #2 in the first round (AC4) seems to indicate that the need for CPU time for the simulation was in reality quite low, only 2.5 days. I do not understand why the CPU time was a limiting factor, why model was not run for a longer time period, or why you didn’t increase the number of vertical levels.

p14l27 If these video clips have been made by the authors, would it be possible to place them in a reliable repository receiving a DOI (as suggested by this journal, see https://www.ocean-science.net/for_authors/submit_your_manuscript.html)? This would ensure that readers are able to access them also years from now.

p21l6 und -> and

p29 Fig.1: Domain of the nested R100 model configuration could be indicated in this Figure. The reference to “Expedition Clockwork Ocean” does not seem to be as useful.

Overall: Please consider using e.g. NM or nmi for “nautical mile”, as suggested by the the International Committee for Weights and Measures (ref https://www.bipm.org/utils/common/pdf/si_brochure_8_en.pdf, p127)

Finally, please consider if the colour maps you have chosen for the Figures are optimal for your purposes (ref e.g. https://doi.org/10.1038/519291d).

Hide

ED: Reconsider after major revisions (10 Oct 2019) by Markus Meier

AR by Reiner Onken on behalf of the Authors (12 Nov 2019) Manuscript

ED: Referee Nomination & Report Request started (22 Nov 2019) by Markus Meier

RR by Anonymous Referee #4 (27 Nov 2019)

Suggestions for revision or reasons for rejection

Referee report of the re-revised manuscript “Very high-resolution modelling of submesoscale turbulent patterns and processes in the Baltic Sea” (version of 12 Nov 2019) by Reiner Onken, Burkard Baschek, and Ingrid M. Angel-Benavides, submitted for consideration to Ocean Science.

I would like to thank the authors for their reply and the changes implemented to the manuscript. Unfortunately, some points in their reply have left me uneasy. My most major concern about this manuscript was and still is the modelling methodology. I was surprised by some of their choices and asked the authors to better justify them or explain the process which lead to those choices. For some issues they did so, for which I thank them. But for other chose not to do so, and for some issues I did not find their justifications convincing.

As sound methodology is one of the most important aspects of any scientific study, it is of utmost importance that modellers do their due diligence when designing modelling experiments. I am not yet convinced this has been the case in this study. Therefore I have no choice but to recommend another major revision. I hope the authors would take some time to think what can be considered a “reasonable effort” on their part to make sure they can justify their choices adequately, and whether they feel confident that their modelling configuration is well enough documented in the manuscript, along with any potential weaknesses. I emphasize that I do not claim that there necessarily are any fundamental flaws in the methodology, but rather that their replies (and in some cases the lack thereof) have left me with insufficient information to determine if this is so.

Please find my specific comments below. Numerals in parenthesis refer to the replies made by the authors in response to my original review. At this time I refrain from making what I would consider minor comments so they do not shadow these more major issues.

In their reply, the authors do not address in any way my three main comments presented under “General comments”. I was truly surprised by this decision. These were the questions I was most looking to be answered, but there are no answers. Because of this I feel I lack information to make a fully informed review. Their other replies do not really address these comments fully. Regarding my first major comment asking the paper’s objectives to be clarified, I see that there have been some edits to the introduction. On one hand I feel this comment could have been addressed more carefully. However, if the authors find their manuscript now accessible enough and do not want to address this in more depth, I have no further comments on this issue.

Regarding my second and third point about documenting the limitations of the study and justifying their methodology, I see improvement and that needs to be acknowledged. However, I feel it would have been possible for the authors to be much clearer on this issue. For example, regarding their reply (8), I don’t think that “worrying the readers” is an acceptable reason to omit information that would help the readers make their own conclusions about the study. This approach of omitting information is deeply worrying and leaves me to wonder what else has been omitted from the manuscript because the readers would find it upsetting.

Also regarding reply (8), I would again ask the authors to insert into the text the version number of the CMEMS product they used. This version number is identified on the cover of the product user manual (and in the following change record) as e.g. “4.0” or “3.0”. The issue of the user manual is useful information, but it is not the same thing as the version of the product.

In their reply (9), the authors indicate they have not really done any kind of survey of available boundary condition products before settling on the CMEMS product, or indeed even considered other options. They also say they “cannot remember” if the reanalysis product was available at that time. In my opinion, “due diligence” or “reasonable care” would have required at least rudimentary survey of available state-of-the-art options for boundary data, along with their main features. I sincerely hope this reply was some kind of misunderstanding, perhaps due to linguistic challenges.

In their reply (10) they state that they did not do any effort to obtain the bathymetric data of their boundary data. In fact, they still have not done so after being asked about this by myself and earlier by another referee. Rather, in their reply they just state their assumption on what this file “probably” is. Yet they still cite the lack of this file as a rationale for their nesting procedures in the manuscript. In my opinion, in this case a reasonable requirement would have been at least one email asking for this file from the service desk. If they do not provide it after asking, fine. If the file is not useful for whatever reason, then that’s fine too. But I do not understand why no effort was taken to obtain it, especially after the comments made by one of the other reviewers.

If we take replies (8), (9) and (10) together, they leave the impression that the authors have decided to use the CMEMS product as their boundary data without much consideration for its features, and now do not want to explain their choice or the main features of the dataset in the paper. This impression may very well be unintended, but I would kindly suggest to the authors to reconsider if it would in fact be better to just briefly explain these things to the readers to avoid such misunderstandings.

Based on my experience from the ocean modelling community, it is not accurate at all to call these model configurations “expensive in terms of computer resources” in the year 2019. Their reply (13) justifies this by claiming that this is a matter of opinion, but frankly, the given information seems to indicate that this model is not computationally expensive when compared to other commonly used configurations in the area. This can be e.g. regional CMEMS configurations, or recently published coastal modelling configurations from the Baltic Sea basin. It is also run on a relatively modest computational system. The test phase of 3 months seems very much normal for this kind of a study, not exceptionally long at all. I ask the authors to remove this statement from the manuscript.

Hide

RR by Anonymous Referee #3 (02 Jan 2020)

Suggestions for revision or reasons for rejection

Onken et al. Very high-resolution modelling of submesoscale turbulent patterns
and processes in the Baltic Sea; Ocean Sci. Discuss., https://doi.org/10.5194/os-2019-44

Review 3

Final comments

I am not satisfied with the response of the authors to some of my previous comments and questions. The major problems are related to the authors’ concept that the internal waves are part of submesoscale processes and the set-up of models – too low number of vertical layers, nesting of models with and without atmospheric forcing, different turbulent parametrization, etc.

The authors refer to a paper by McWilliams (2016) to support their approach that internal waves are also among the submesoscale turbulent patterns and processes. At the same time, Williams (2016) clearly distinguishes between the internal gravity waves (IGW) and submesoscale currents (SMC). The scales of SMC and IGW are coinciding, but they are two separate branches of the flow of energy from mesoscale processes to microscale dissipation. Although in the term “submesoscale processes”, the scale is mentioned, it is more than that. See, e.g., a definition by McWilliams (2019): “Besides their identifying scales, submesoscale currents are distinctive in their flow patterns, their essential dynamical processes, and their consequences for transport, mixing, and dissipation in the general circulation”. I think it is misleading to define internal waves and submesoscale processes as submesoscale turbulent patterns and processes. It also causes a misleading statement that submesoscale processes are characterized by strong vertical speeds of O(100 m day-1).

Since some questionable assumptions were used and clearly different settings were applied to the nested models, readers still might wonder what results are of general value and what results are related to the approach. For instance, the authors state that the submesoscale features appear preferably in the shallow areas with water depths less than 40 m. Is it a natural phenomenon or is it related to the model resolution? One could argue that in the deeper areas, the vertical resolution of the model is too coarse to simulate submesoscale processes.

There are also smaller issues, which could be omitted to allow to shorten and focus the manuscript better. For instance, it is not relevant to discuss the long-term precipitation and evaporation to justify why salinity is an ideal parameter (passive tracer), etc.

On the other hand, the authors have conducted a thorough analysis of their results using a series of parameters and their distributions related to certain processes, patterns or mechanisms and compared the findings with other papers/authors. Thus, I think the paper is worth to be published to intensify the discussions on the nature and features of submesoscale processes. I leave the decision to the Editor(s).

McWilliams J.C. 2016. Submesoscale currents in the ocean. Proc. R. Soc. A 472: 20160117. https://doi.org/10.1098/rspa.2016.0117.

McWilliams J.C. 2019. A survey of submesoscale currents. Geosci. Lett. 6:3. https://doi.org/10.1186/s40562-019-0133-3.

Hide

ED: Reconsider after major revisions (24 Jan 2020) by Markus Meier

AR by Reiner Onken on behalf of the Authors (06 Mar 2020) Author's response Manuscript

ED: Publish subject to technical corrections (28 Mar 2020) by Markus Meier

AR by Reiner Onken on behalf of the Authors (23 Apr 2020) Manuscript

Short summary

In order to provide an aid for the interpretation of observations of the formation, structure, and dynamics of submesoscale patterns, a numerical model is applied in a double-offline-nested setup to a sub-area of the Baltic Sea. A model with 500 m horizontal resolution is nested into an existing operational model in order to create a realistic mesoscale environment. Turbulent patterns with horizontal scales < 1 km are resolved by a second nest with 100 m resolution.