Data assimilation of sea surface temperature and salinity using basin-scale reconstruction from empirical orthogonal functions: a feasibility study in the northeastern Baltic Sea

Zujev, Mihhail; Elken, Jüri; Lagemaa, Priidik

doi:https://doi.org/10.5194/os-17-91-2021

Articles | Volume 17, issue 1

https://doi.org/10.5194/os-17-91-2021

© Author(s) 2021. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/os-17-91-2021

© Author(s) 2021. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 17, issue 1

Research article

|

14 Jan 2021

Research article |

| 14 Jan 2021

Data assimilation of sea surface temperature and salinity using basin-scale reconstruction from empirical orthogonal functions: a feasibility study in the northeastern Baltic Sea

Mihhail Zujev, Jüri Elken, and Priidik Lagemaa

Download

Final revised paper (published on 14 Jan 2021)
Preprint (discussion started on 03 Jun 2020)

Interactive discussion

Status: closed

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

- Printer-friendly version

- Supplement

RC1: 'Data assimilation of sea surface temperature and salinity using basin-scale EOF reconstruction: a feasibility study in the NE Baltic Sea By Mihhail Zujev et al.', Anonymous Referee #1, 08 Jul 2020
- AC1: 'Author's response to Referee #1', Jüri Elken, 27 Aug 2020
RC2: 'Review of the manuscript "Data assimilation of sea surface temperature and salinity using basin-scale EOF reconstruction: a feasibility study in the NE Baltic Sea", by Mihhail Zujev, Jüri Elken, and Priidik Lagemaa', Anonymous Referee #2, 15 Jul 2020
- AC2: 'Author's response to Referee #2', Jüri Elken, 27 Aug 2020
RC3: 'Review', Anonymous Referee #3, 16 Jul 2020
- AC3: 'Author's response to Referee #3', Jüri Elken, 27 Aug 2020

Peer-review completion

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

AR by Jüri Elken on behalf of the Authors (27 Aug 2020) Author's response Manuscript

ED: Referee Nomination & Report Request started (12 Sep 2020) by Markus Meier

RR by Anonymous Referee #2 (02 Oct 2020)

RR by Anonymous Referee #4 (28 Oct 2020)

Suggestions for revision or reasons for rejection

This paper uses EOFs to assimilate in situ data in the Baltic Sea. This is interesting work using a novel assimilation technique to try to gain maximum benefit from a sparse (relative to the domain size) observing network.

My main issue with paper is that I found it a hard to follow at times. Please revise any parts of the paper where I state that it is unclear.

Also it wasn’t clear whether the fine grid assimilation was used or not in the experiments since much of the discussion in the abstract and paper relates to the EOF assimilation. If the fine grid assimilation is done it would be interesting to know what the impact of it is relative to the EOF assimilation. Were any experiments performed only doing EOF or fine grid assimilation to assess the relative impact of both methods?

I have some more detailed comments below

1. Abstract p1 line 12-13. I think it would be useful to indicate the size in km or nautical miles. At first I thought the units were in degrees and this would help to avoid that confusion.
2. Abstract p1. You don’t mention the fine grid assimilation
3. Introduction p1 line 22-24 unclear. If understand it right you are stating that SST in the models is good even without assimilation and SSS is where assimilation has the best impact.
4. Introduction p2 line 30. I would change “the operational forecasts within CMEMS” to “the operational Baltic Sea forecasts within CMEMS” just to be specific as other ocean forecasting regions/systems do assimilate data. This might be a good place to explain why data is not currently assimilated in this system as the explanation would be used to motivate the work you are doing here.
5. Introduction p2 line 44. It is a challenge true but that is not the reason to do the work instead it is because it would be highly beneficial for a region with sparse observations.
6. Introduction p2 line 46. Perhaps it is a turn of phrase but it is strange to me to cite a paper where the method is presented as an idea in one paper. Was it described theoretically in this paper? Was it demonstrated practically?
7. Introduction p2 line 50-52. It would be useful to clarify how this work builds on Elken 2018 and Elken 2019 for those who haven’t seen the other papers.
8. Sec 2.1 p3 line 69 “yearly Maximum SST exceeds usually 15 deg C” -> “the SST usually exceeds 15 deg C in July or August”
9. Sec 2.1 p3 line 84. It would be good to state the (approximate) grid size in km or nautical miles here.
10. Sec 2.2 p5 line 140. State the grid size of the coarse grid in km or nautical miles.
11. Sec 2.2 p5 line 140. You should state that main benefit of the coarse grid is to save computational costs. And state what the downsides are. You do discuss this later so another option would be to writing something like “see section 2.4 for more details on the advantages and disadvantages of using the coarse grid” but it might be better to group all this discussion together here.
12. Sec 2.3 p5 line 149-152. I think I’d like a little more detail here. You probably should say here that you summarise the EOF method below (in the same section). Also the discussion here gives the impression that you use only the EOF method, but later you do mention a fine grid adjustment.
13. Sec 2.3 p5-6. An additional point to address is are the EOFs multivariate or are SST and SSS treated separately.
14. Sec 2.3 p5-6. It would be worth discussing too why the EOFs are not 3D and only need to reconstruct surface temperature and salinity.
15. Sec 2.3 p8 lines-244-254. I really had trouble following this paragraph describing the use of the full covariance and truncation of the EOF modes. Perhaps it simply needs rewriting for clarity.
16. Sec 3.1.2 p11 line 319. Do you mean “… suggested that only the three gravest modes should be included..”
17. Sec 3.1.3 p11 line 340. When there is no reconstruction is there no assimilation? (or just the fine scale assimilation?)
18. Sec 3.2 p12 line 368-376. So is the fine grid assimilation performed in these experiments? If so what is the relative impact of both methods. If not then this needs clarifying.
19. Sec 3.2 p12 line 368-376. I’m not sure why you have SST01 and SST02. This suggests the data is different. I’m not insisting but wouldn’t it be clearer just use the notation DA01 and DA02.
20. Sec 3.2 p12 line 368-376. Is the same relaxation applied to the EOF and fine scale assimilation?
21. Sec 3.2.1 p13 line 399. Source of the SST satellite data? (I think you describe it earlier if so refer to that section).
22. Sec 3.2.2 p14 line 437. Could the failure to correct cross-gulf gradients be caused by using coarse EOFs?
23. Sec 3.2.2 p15 line 445. You mean there is very little difference between DA01 and DA02? Why might that be?
24. Sec 3.2.4 p16 Confirm that you are assessing forecast skill (i.e. before the observations are assimilated). It is much easier to improve the analysis skill if you are comparing to observations you have assimilated.
25. Sec 3.2.4 p16 line 492. Too unspecific. Explain what Hernandez et al 2015 said which is relevant here.
26. Sec 3.2.4 p17 line 524 DA is not improving the model I would say it improves the forecasting of SSS.
27. Sec 3.2.4 p17 line 524 “Still, RMSD to the observations makes 62% of observed standard deviations…” I would write “Still, the forecast RMSD to the observations is 62% of the observed standard deviations which suggests that there may be further room for improvement.”
28. Sec 4 p18 line 570. Do you do both the EOF and some kind of localized DA (fine scale DA)? If you do you are not reducing the computational effort rather you are increasing it. This may still be worthwhile of course.
29. Sec 4 p19 line 575. I’m not sure what you mean here. What has the orthogonality of the EOF basis vectors got do with it?
30. Sec 5 p19 line 595. Again is this only assimilating with EOFs or do you make fine scale corrections? What if you just made the fine scale corrections?
31. Figure 1 p27. I don’t see any arrows indicating river discharges.
32. Figure 5 p31. Are the satellite observations in the form of a gridded product (there are no gaps in the data from clouds). You should perhaps also state here that the EOF method is only from in-situ data.
33. Figs 5 and 6 p31-32. Some of the plots y axis go from 57 N to 57 N!
34. Fig 7/8 p33-34. I would find it a bit less confusing if this was DA01 SST and DA02 SST. In section 3.2 you use SST01 SST02 to describe the assimilated EOF reconstructed fields whereas here it appears to be the model fields after assimilation.
35. Fig 9 p35. I assume that the size of the circle has no additional meaning. In the figure legend the large positive is a circle the same size as all the other categories. Perhaps you could redraw the caption or adjust the figure so the caption and points in the plot have matching sizes.

Hide

ED: Publish subject to minor revisions (review by editor) (31 Oct 2020) by Markus Meier

AR by Jüri Elken on behalf of the Authors (05 Nov 2020) Author's response Manuscript

ED: Publish as is (08 Nov 2020) by Markus Meier

AR by Jüri Elken on behalf of the Authors (08 Nov 2020) Author's response Manuscript

Short summary

The proposed method of data assimilation is capable of effectively correcting basin-scale mismatch of oceanographic models when the domain is under nearly coherent external forcing. The method uses basin-scale EOF modes, calculated from the long-term model statistics. These modes are used to reconstruct gridded fields from point observations, which are further fed to the model using relaxation. Tests with sea surface temperature and salinity in the NE Baltic Sea were successful.