Sea-level variability and change along the Norwegian coast between 2003 and 2018 from satellite altimetry, tide gauges, and hydrography

Mangini, Fabio; Chafik, Léon; Bonaduce, Antonio; Bertino, Laurent; Nilsen, Jan Even Ø.

doi:https://doi.org/10.5194/os-18-331-2022

Articles | Volume 18, issue 2

https://doi.org/10.5194/os-18-331-2022

© Author(s) 2022. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/os-18-331-2022

© Author(s) 2022. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 18, issue 2

Research article

|

18 Mar 2022

Research article |

| 18 Mar 2022

Sea-level variability and change along the Norwegian coast between 2003 and 2018 from satellite altimetry, tide gauges, and hydrography

Fabio Mangini, Léon Chafik, Antonio Bonaduce, Laurent Bertino, and Jan Even Ø. Nilsen

Download

Final revised paper (published on 18 Mar 2022)
Preprint (discussion started on 06 Jul 2021)

Interactive discussion

Status: closed

RC1:
'Comment on os-2021-55', Anonymous Referee #1, 27 Jul 2021

The manuscript is a very good contribution towards the use of coastal altimetry from ALES for local sea level trend and sea level budget assessments. For this reason, the scope of the article goes beyond the regional character of the study. While I strongly encourage the publication of this study, I must assign a major revision due to two main points that the authors will find specified in the attached document alongside other minor comments:

1) The insufficient description of the method to assess trends and in particular about the computation of the uncertainty, which raises the suspect that a fundamental issue such as serial correlation in geophysical time series is not being taken into account. The absence of such a description negatively affects the possible value of the results and the discussion about statistical significance

2) The several missing points and abundance of speculations in the discussion, in particular concerning the "sea level budget", which actually cannot be defined as such, since it does not provide any assessment of the components other than the steric ones. To solve this issue, the authors should either expand the discussion with the necessary analysis or reduce the aims and ambitions of the paper for what concerns Section 5

I think both the major and the minor issues highlighted in my review can be solved with a bit more work and possibly some additional analysis. Therefore I am willing to consider a revised version.

Citation: https://doi.org/10.5194/os-2021-55-RC1
- AC1: 'Reply on RC1', Fabio Mangini, 22 Oct 2021
  
  Dear sir,
  we would like to thank you for the comments on the manuscript. Attached here, you will find a PDF document with our answers. The document also contains the part of the manuscript that has been modified (with the line numbers to help you identify the changes in the manuscript). The line numbers refer to the manuscript without track changes.
  Kind regards,
  Fabio Mangini
  
  Citation: https://doi.org/10.5194/os-2021-55-AC1
RC2:
'Comment on os-2021-55', Anonymous Referee #2, 12 Aug 2021

The authors assess the quality of ALES-retracked altimetry data in terms of reproducing temporal variations of sea level along the Norwegian coast, specifically the annual cycle and the trend, by comparison with available tide gauge data. They then use the altimetry data in combination with observations from hydrographic stations (whith no tide gauge data there) to compare steric height variations to the total sea level variations.

Both issues of the paper, the quality assessment of the ALES-retracked altimetry as well as the analysis of sea level variations along the Norwegian coast are interesting for the reasons stated by the authors. However, both issues need more in-depth consideration. Also the methodology needs some additional descriptions.

It is not totally clear to me why the semi-annual cycle is fitted (eq. 1). It is never addressed anywhere later in the paper (or I have misunderstood). Is it included only to prevent, in case of data gaps, that the running annual mean reveals part of the seasonal signal, or aliasing issues? This has to be explained. If there is a signal with amplitudes comparable to that of the annual cycle, it would also be interesting to see, as additional part of the quality assessment, how, for this frequency, ALES-retracked altimetry compares to the tide gauges.

The authors state that sea level data from the tide gauges have been corrected for geoid height changes with respect to GIA. Is this correction applied to the altimetry data as well? If geoid height change from GIA is considered, then also instantaneous adaptation from changed loading caused by the Greenland ice melt should be included, which will probably outperform the GIA effect in most regions along the Norwegian coast (Siegismund et al, 2020). Why is, at the end, not the full sea level variation budget considered, but only the steric effect? The abstract states performance of the sea level budget. Consideration of the full budget would give the whole study more weight. If this is not intended the authors should explain the reasons to the reader.

It isn’t clearly stated how the optimal distances from and along the coast for spatial averaging the altimetry data are found. I guess, a set of distances is defined and the correlations are computed for each element? Please explain.

It would be nice to see a direct comparison of conventional altimetry and ALES-retracked altimetry for the Norwegian coast to see both, the improvements/changes in the observations dependend on location (and specifically distance from the coast) as well as the added value caused by the improvements when investigating spatio-temporal variations of sea level along the Norwegian coast. Direct comparison to Breili et al (2017) is not feasible due to different methodology and period. From the optimization of the distances for the spatial averaging of the altimetry data it seems that inclusion of data more offshore than 20 km, which has been the a-priori fixed maximum distance, could even improve the correlation with the tide gauge data. Has this been tested? That could mean, that conventional altimetry data could be of comparable quality because either the errors of the ALES-retracked altimetry data is still too high near the coast, or temporal variations are coherent for a wide stripe along the Norwegian coast and data away from the coast can do the job as well. So what is really the benefit from using the ALES-retracked data?

Minor issue:

In Figure 1 please mention, that the yellow diamonds are the hydrographic stations, and the red dots are the tide gauges.

Citation: https://doi.org/10.5194/os-2021-55-RC2
- AC2: 'Reply on RC2', Fabio Mangini, 22 Oct 2021
  
  Dear sir,
  we would like to thank you for the comments on the manuscript. Attached here, you will find a PDF document with our answers. The document also contains the part of the manuscript that has been modified (with the line numbers to help you identify the changes in the manuscript). The line numbers refer to the manuscript without track changes.
  Kind regards,
  Fabio Mangini
  
  Citation: https://doi.org/10.5194/os-2021-55-AC2

Peer review completion

AR: Author's response | RR: Referee report | ED: Editor decision | EF: Editorial file upload

AR by Fabio Mangini on behalf of the Authors (22 Oct 2021) Author's response Author's tracked changes Manuscript

ED: Referee Nomination & Report Request started (27 Oct 2021) by Joanne Williams

RR by Anonymous Referee #1 (03 Nov 2021)

Suggestions for revision or reasons for rejection

Document of reference: revised_manuscript_with_track_changes.pdf

I would like to thank the authors for providing a satisfying revisions. I have only a couple of minor points left.

Concerning my two former major comments, I think they are mostly answered. It is good that the authors are now transparent in describing the way in which they define the effective degrees of freedom and the uncertainties, although I must say I am not fully familiar with it.
I am still a bit puzzled by the quality of the fitting of the variogram in some stations compared to others, for example in Figure A2 the fitting of Honnigsvag and Vardo (last two locations) showing the same lags as the fitting of Stavanger, Bergen, Maloy, Alesund and Kristiansund in the second line. But then, compared for example to the much stronger autocorrelation seen in the Appendix B, probably this suboptimal fitting is not that problematic and does not modified the conclusions very much, as the authors say. I’d appreciate a comment of the authors for my own knowledge about that, but I am not asking a revision on this point.
What I do ask, instead, is to be clear about what references have been used to apply this methodology (or methodologies) concerning Appendix A. No references are cited right now. And, secondly, I think the caption of the figures is wrong. For example in Figure A1 you write: “variogram of the difference between the detrended and deseasoned SLA estimated from the ALES-retracker satellite altimetry (empty circles)”. But this is not a variogram of the differences, it’s a variogram of the the detrended and deseasoned SLA, isn’t it? The only variogram “of the differences” is in Figure A3. Please double check.

Concerning my previous comment “Nevertheless I don't get one point: why such strict thresholds and not more simply iterating the two distances for each tide gauge selecting each time the distances that yield the best correlation?”, the added paragraph looks still confusing to me.
So, let’s take the distance “along coast”. You write “we test different combinations...between 20 and 200 km”. Which different combinations? Every 20 km? Every 50?
Then you write: “we have performed a sensitivity test on the distance from the tide gauge allowing it to range between 15 and 400 km: as before, we have found little difference in the final results.”
Then you write: “Indeed, the maximum values of the linear correlation coefficients occur for distances along the coast that range between 140 and 200 km, with them being 200 km at 13 out of 22 tide gauges.”. So, if you get the best correlation using 200 km (the maximum value of your threshold), does the correlation get better if you use, say, 300 km?
What I mean is that all these paragraphs put together are confusing for the reader. I appreciate you want to keep your methodology, but could you write it in a coincise way in one single paragraph, maybe with a formula that helps the reader to reproduce the procedure?

Concerning my previous comment related to the sentence “These results suggests that the detrended and deseasoned SLA in the south vary over smaller spatial scales compared to the north.” I do not find the answer fully satisfactory. You have added a general comment, (which I comment later), but you have not answered on whether you think that the location of the tide gauge plays a role. Please look for example at the values shown in Figure 5, concerning the tide gauge 12. Why is the standard deviation of the correlation coefficients and of the RMSD very high at that point? Is it anything to do with the fact that this tide gauge is located in a fjord or not? In other words: do you think that the spatial scales of the detrended and deseasoned SLA are fully independent of the conditions in which the tide gauge is located? (open sea, vs fjord, vs ….)

Line 738: “a few thousands of kms”...you mean a few hundreds?

Lines 741-742 “we reduce the noise in the SLA from altimetry which might result, for example, from the rough topography of Norway”. I don’t see why the rough topography of the land would create noise in a sea level dataset? Please clarify, but I tend to believe that the reason why the “noise” is reduced is that by averaging over long distances, expecially along a coherent coast, you include more altimetry tracks and therefore you improve the temporal sampling from altimetry. This has been demonstrated, for example, with the concept of the “Zone of Influence” in Oelsmann et al., 2021 (https://doi.org/10.5194/os-17-35-2021)

Appendix C: this is a useful comparison, but please make sure to write in the lines that these results are not only influenced by the retracker of choice, but also by the set of geophysical corrections applied to the retracker, which surely have differences between OpenADB and your source of “conventional altimetry”. I appreciate the fact that this addition is in an Appendix and does not need to be the core of the article, since the use of ALES and its advantages with respect to conventional retracking have been validated in several other papers nowadays.

Hide

RR by Anonymous Referee #3 (14 Dec 2021)

Suggestions for revision or reasons for rejection

This paper had gone through one round of review when the editor sent it to me for a third opinion. The paper investigates sea-level variability and trends along the Norwegian coastlines from tide gauge observations and altimetry data. They also use hydrographic data to estimate the steric contribution to the observed sea-level changes. The revised paper does not have any serious flaws which would prevent its publication, the use of statistics is overall appropriate, and the presentation of the results meets the required standards. That said, I think that the paper could be strengthened. To this end, I have several questions of clarification as well as some corrections and suggestions to make, as outlined below.

Line 24. A hyphen is not required here. Generally, the term “sea level” only needs to be hyphenated when it is used as an adjective before a noun.

Line 25. I suggest: “Accurate estimation and attribution of…”

Lines 24-29. I think that the argument for this study could be made more convincingly in this first introductory paragraph by emphasizing the societal impacts of understanding local sea level. For example, the authors start the introduction by saying that “Sea level is considered a key indicator to monitor the earth’s energy imbalance and climate change”, but this point is not even relevant to the work presented here because it is global mean sea level that is a key climate indicator rather than local sea level. I would say that the motivation for this study is that changes in local sea level can considerably differ from global mean sea level, and hence understanding such local changes is crucial to the implementation of coastal adaptation plans.

Lines 52-53. Here I would add a clarification along the lines of “typical of the open ocean for distances to the coast of about 3 km”.

Line 65. You do not really assess the “sea-level budget”. I would suggest you reword this sentence.

Lines 104-105. Could you please specify here exactly which tidal corrections have been applied? This is important because, while tide gauges only sense the ocean tide and the ocean pole tide, altimeter data are also influenced by the solid earth tide, the load tide, the solid earth pole tide, etc.

Section 2.1. Altimetry measurements are influenced by GIA, albeit differently from tide gauge data. Yet there is no mention of any GIA corrections applied to the altimetry data, even though tide gauge records appear to have been corrected for GIA (both crustal uplift and gravity contributions).

Lines 145-147. “lesser extent”. In some regions, the gravity GIA contribution is larger than the VLM induced by GIA. Perhaps, you could clarify that in Norway VLM is the dominant GIA contribution.

Lines 168-176. Could you please specify the temporal resolution of the hydrographic data?

Line 195. While there is nothing wrong with this nonlinear regression model, the model can be easily written in linear form, which greatly simplifies the estimation of the regression coefficients and their uncertainty: z(t) = a + b*t + c*cos(wa*t) + d*sin(wa*t) + e*cos(was*t) + f*sin(was*t), where “wa” and “was” are the annual and semiannual angular frequencies. I am not suggesting the authors redo the analysis with the linear model, but I thought it was worth pointing this option out.

Line 203. “depend”.

Line 212. This equation appears three times in the manuscript. I would suggest you write this equation only once, possibly in the Appendix, number it, and then refer to it whenever you use it.

Line 268. As written, Equation (3) is incorrect. In particular, it is missing a minus sign. Assuming a standard definition for the haline contraction coefficient (beta = 1/rho * drho/dS), the equation as it stands implies that sea level increases as salinity increases, which is not correct. I assume that this is just a typo and that the halosteric calculations have been done properly, otherwise the authors need to adjust their results.

Lines 281-284. I do not understand what the authors mean here. Could you please clarify?

Line 395. The term used in Benveniste et al. (2020) is “fractional differences”.

Lines 395-399. It is not clear how the authors compute the FDs. In Benveniste et al. (2020), “tau” is defined as the trend of the time series of sea-level differences. That is, Benveniste et al. (2020) subtract the altimetry time series from the tide gauge time series, and then compute the trend of the residual time series. Here, the authors define “tau” as “the linear trend difference between altimetry and each tide gauge”, which seems to imply that “tau” is the difference of the two trends from the tide gauge and altimetry. If this is how you do it, what value do you use for the standard error (SE) in the FD equation? The whole point of defining “tau” as the trend of the residual time series as in Benveniste et al. (2020) is that the value of SE is straightforward to calculate. When you define “tau” as the difference of two trends, as opposed to the trend of sea-level differences, then you need to make assumptions about the dependence between SE_TGs and SE_altimetry. Please clarify.

Lines 399-400. From Fig. 8, the confidence intervals overlap significantly, even at Tregde, Måløy, and Bergen, so I am surprised that FDs are > 1 at those stations.

Section 5.2. This section is called “Steric contribution to the sea-level trend”. However, the authors do not really place the steric contribution in the context of the sea-level trends. They simply say that the steric trends lie within a particular range and then refer to Fig. 11. I would expect the authors to say what fraction of the sea-level trends is explained by the steric contribution, distinguishing between stations. This Section could benefit from a more focused discussion of the results.

Line 637 (Appendix A). Strictly speaking, this is the semi-variogram.

Lines 649-651. How do estimates of the degrees of freedom (dof) computed with this approach compare to estimates computed using the equation: dof * (1 – r)/(1 + r), where ‘r’ is the autocorrelation? Providing such a comparison would give the reader additional confidence in your uncertainty estimates.

Abstract. The authors write “we find that the sea-level annual cycle is more affected by variations in temperature than in salinity, and that both temperature and salinity give a comparable contribution to the sea-level change along the entire Norwegian coast”. I find this sentence a bit confusing. Seasonal variations are “sea-level changes”, so what do you mean by “contribution to sea-level change”? Could you please rewrite this important sentence to improve clarity?

Hide

ED: Publish subject to minor revisions (review by editor) (03 Jan 2022) by Joanne Williams

AR by Fabio Mangini on behalf of the Authors (16 Jan 2022) Author's response

EF by Manal Becker (18 Jan 2022) Author's tracked changes

EF by Manal Becker (18 Jan 2022) Manuscript

ED: Publish subject to technical corrections (07 Feb 2022) by Mario Hoppema

AR by Fabio Mangini on behalf of the Authors (14 Feb 2022) Manuscript

Short summary

We validate the recent ALES-reprocessed coastal satellite altimetry dataset along the Norwegian coast between 2003 and 2018. We find that coastal altimetry and conventional altimetry products perform similarly along the Norwegian coast. However, the agreement with tide gauges slightly increases in terms of trends when we use the ALES coastal altimetry data. We then use the ALES dataset and hydrographic stations to explore the steric contribution to the Norwegian sea-level anomaly.