A newly reconciled dataset for identifying sea level rise and variability in Dublin Bay

Shoari Nejad, Amin; Parnell, Andrew C.; Greene, Alice; Thorne, Peter; Kelleher, Brian P.; Devoy, Robert J. N.; McCarthy, Gerard

doi:https://doi.org/10.5194/os-18-511-2022

Articles | Volume 18, issue 2

https://doi.org/10.5194/os-18-511-2022

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

https://doi.org/10.5194/os-18-511-2022

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

Articles | Volume 18, issue 2

Research article

|

22 Apr 2022

Research article |

| 22 Apr 2022

A newly reconciled dataset for identifying sea level rise and variability in Dublin Bay

Amin Shoari Nejad, Andrew C. Parnell, Alice Greene, Peter Thorne, Brian P. Kelleher, Robert J. N. Devoy, and Gerard McCarthy

Download

Final revised paper (published on 22 Apr 2022)
Preprint (discussion started on 11 Oct 2021)

Interactive discussion

Status: closed

RC1:
'Comment on os-2021-92', Philip Woodworth, 18 Oct 2021

18 October 2021

Comments on 'A newly reconciled data set for identifying sea level rise and

variability in Dublin Bay' by Nejad et al. (OSD)

This paper attempts to derive a long-term MSL record for Dublin Port, dealing with

technical problems in the record, validating the newly-obtained time series with

data from two other records in Ireland and from Newlyn and Brest. It is an interesting

exercise in trying to make sense of historical data, and I have no objection to its eventual

publication. However, I do have some major and minor comments I would like to see attended to.

One major comment is that there is no technical information or history provided on each gauge

(leaving aside Newlyn and Brest). Presumably Howth and Arklow are radar gauges? But also I would like to

see a table (perhaps not part of Table 1) which shows what type of gauge was

used at Dublin for which years. Presumably in the earlier years it was a float gauge, replaced

by a pressure sensor, or I would have thought a series of pressure sensors as they tend not

to last long.

The analysis of this paper hangs on the theory that the the pressure sensor(s) used in recent

years at Dublin was fairly reliable at low waters but imprecise for high waters. So, when was the

pressure sensor installed? If possible, what calibration method was used to level it in such that

low waters were accurate? And why were high waters less reliable? - the only thing I can think of

is that the wrong density (and maybe gravity) was assumed. If possible, what manufacturer was it? Was it

an absolute or differential sensor? There is none of this sort of documentary background in the paper

to support the authors' working theory that a major change took place in 1976. Also 1976 for me seems

rather early for a pressure sensor being used, which means the change point happened during the float

gauge era, and so would have had nothing to do with pressure sensors. Your theory at the moment hangs

entirely on the change point results (Fig A1) and would be much stronger with some additional evidence.

A second comment is that Table 1 is missing a line? From the text, I would think

it should look like:

1. Port Authority Annual 1938-1977 Annual Woodworth et al. (1991) HW+LW

2. ditto 1978-1988 ditto ditto

3. Port Authority Monthly 1987-2001 ditto

4. Harbourmaster ditto

5. NTGN ditto

6. Greene HW only

A third commment is that on line 1 of the abstract it says the new composite record for Dublin will be for 1938-2016,

but from Table 1 we see there is Dublin data to 2019. It looks like the 2016 constraint comes from

Newlyn and Brest data being only to 2016 (Table 3). That may have been the case when the first

draft of this paper was written but there are data to 2019 for both now so the new Dublin

record could go to 2019.

A fourth comment relates to the regressions at lines 116 and 140. I think the former is

fair enough (although see the additional reference mentioned below). The point is that

the MLW will contain a nodal component mostly due to changes in the tide (the roughly 3.7

percent of the M2 tide) and also a small nodal component in MSL. It will also have

a perigean component due to the tide. So ok.

But the latter is not so reasonable. MSL will have a nodal component (see

Woodworth, 2012) but it will be very small. So, instead of determining the true nodal

amount, your fit over 38 years will simply pick out noise from the much larger interannual variability

in the ocean circulation. In addition, there is no tidal basis for including a perigean

component in a parameterisation of MSL.

So I think this equation is not reasonable. Anyway, you don't discuss the determined

beta2-5. So I would also make a simple regression with beta0 and beta1 and check if those

parameters are similar to what you have here then use them instead. (If they are different then you have

a problem).

Minor comments:

22 - ... for them, while Camaro ... state ...

32 - given that

35 - We find that the datums of Dublin's ..

36 - and thus need further consideration. In addition,

38 - drop 'other'. Problems with datums are not atmospheric

45 - 2018, but Table 1 suggests 2019?

50 - ... (LAT) which is xx cm below/above Poolbeg Datum. [I think that is about 200 mm?]

51 - this monthly data set is not in Table 1 (see above)? The second entry says monthly but that

is the second part of the PA Annual data set (Woodworth et al. 1991 data) and that is annual. Anyhow, can you check

all this?

53 - automatic comparison and checking of the ..

54 - it would be worth saying there is only 2 years overlap between this data and that of

the PA Annual.

57 - 2001-2009 but Table 1 says 2002-2009

60 - 2006-2017 but Table 1 says 2006-2019

62 - All NTGN data are defined relative to ODM and LAT. (presumably)

70 - each day. These ..

71 - metres. Data post-1998 had already been digitised ... intervals prior to the present

analysis, while for the post-2004 data the sampling frequency increased to 1 minute

intervals.

I'm sure this ok although what you are trying to explain are columns in some

excel sheet? Maybe check if it reads ok again. Also (line 72) you might explain that sometimes

you won't have two high waters when the second one falls into the adjacent day.

Table 1 - see my comments above. I think an additional column saying 'HW and LW' or

'HW only' would be useful

80 - There is found to be a systemmatic .. dataset presumably due .. [presumably because do

you know that for sure? i.e. did they round in one way?]

81 - shows monthly MHW from the Port ..

84 - datasets, each expressed with respect to LAT.

88-91 - should these lines be better in section 2 as they are describing adjustments made.

I didn't understand the sentence at lines 90-91 which seemed tountament to replacing one

with another.

prior to 2003 (Table 1) ..

92 - only tabulated MTL (i.e. the combination of MHW and MLW) is available for some years

instead of MSL. Therefore, MTL needs to be converted to MSL to be consistent throughout.

96-97 - I'd drop the sentence 'The recordings'

Figure 3 - see above. This can be extended to 2019?

101 - drop 'of the issue'

.. malfunction of the Dublin tide gauge ...

next line - the reader will want to know why the discrepancy is proportional to depth -

see above. I think you need an extra paragraph at least.

109 - sentence 'According'. Please write this more fully.

114 - ... an intercept, a term proportional to annual MLW

116 (Haigh et al., 2011; Woodworth, 2012). The two latter terms will primarily

contribute to the astronomical tidal component of MLW.

The reference is:

Woodworth, P.L. 2012. A note on the nodal tide in sea level records.

Journal of Coastal Research, 28, 316-323, doi:10.2112/JCOASTRES-D-11A-00023.1.

124 - and the original

129 - and exhibits a relatively .. to MSL. You then need a sentence to say how

you now interpret the information in Figure 3.

133 - what do you mean by atmospheric effects? If you mean simply the inverse barometer

then say so. If it is more complicated in trying to account for winds then say that

also.

... We first removed the inverse barometer effect on sea level at each site following ...

140 - see my comments about this equation above.

145 - has a mean

152 - earlier periods (i.e. pre-1953) are ..

153 - and are consistent

words 'and where lower'. This is obvious isn't it? I drop this.

157 - you mean Glacial Isostatic Adjustment (GIA) so reword:

.. are in regions of uplift due to Glacial Isostatic Adjustment (GIA) ...

159 .. greater consistency of the local sea level rates with the global

figure is to be expected (reference).

163 - It doesn't matter for this paper and don't change anything but I don't particularly

buy this. Ireland may be on the edge of the Atlantic but Dublin isn't particularly and

in fact Newlyn and Brest are closer to the gyre (see for example, Woodworth, P.L., Pouvreau, N.

and Woppelmann, G. 2010. The gyre-scale circulation of the North Atlantic and sea level at Brest.

Ocean Science, 6, 185-190, doi:www.ocean-sci.net/6/185/2010/.)

But one reference you could include is to the fact that interannual variations along the

European coastline are very similar, whether due to the gyre or long-shore processes or

whatever, see for example Figure 11 of:

Calafat, F.M., Chambers, D.P. and Tsimplis, M.N. 2012.

Mechanisms of decadal sea level variability in the eastern

North Atlantic and the Mediterranean Sea. Journal of Geophysical Research,

117, C09022, doi:10.1029/2012JC008285.

196 etc. - you should define statistical terms like N, Uni etc. for the general reader.

200 - the model output as such is not shown in Figure A1, simply the difference Dublin minus Newlyn

from which the model decides 1976 is a change point. And should that be difference and not absolute difference?

I agree something seems to happen around 1976, but it would be good to have an explanation

why - see above

References - you provide a doi for the first one but not others. Pugh et al. can be updated.

Citation: https://doi.org/10.5194/os-2021-92-RC1
- AC1: 'Reply on RC1', Amin Shoari Nejad, 29 Oct 2021
  
  Dear Prof. Woodworth,
  Many thanks for reviewing the paper. Your reviews are very constructive and appreciated. Below I will try explaining the issues you raised:
  1^st major comment: One major comment is that there is no technical information or history provided on each gauge …
  We will add Table 1 in the supplement file (attached here) including the information on the tide gauges installed at Dublin port. However, please note that we do not claim with certainty that the tide gauge at Dublin port is faulty or has been faulty since 1976. We report the results of the comparison of monthly MLW, MSL, and MHW of Dublin Port, Howth Harbour and Arklow in Figure 3 and Table 3 of the paper. According to the comparison, it seems clear that the MLW time series are in higher agreement with each others compared to MSL and MHW. We have not investigated the cause behind this lack of agreement, and we only suggest that drift in the pressure sensor is a possibility worthy of further investigation. Regardless of what caused the apparent divergence in the high waters, we have shown that it is possible to re-create the MSL using MLW to correct for the bias (Figure 5-6 suggest that the output of our model is more aligned with MSL in Arklow, Howth Harbour, Brest and Newlyn compared to the original MSL data). To build the regression model we needed to validate the training MSL data and for that we compared the MSL of Dublin Port to those of Newlyn and Brest in Figure 4 and, based on this comparison, we decided to choose 1938-1976 as the period for training the regression model. The change point model gives additional support for choosing the year 1976 as the end point for the training period of the regression model and is not introduced for any other purposes. Hence, we neither claim that the drift has started in 1976 nor that the data during 1976-2003 (approximately) are incorrect. In fact, according to Figure 5, the original MSL and the modelled MSL are in good agreement in that period which suggests that a possible drift in high waters was not significant during that period. Otherwise it would have been corrected by the model as can be seen in the period 2003-2016.
  2nd major comment: A second comment is that Table 1 is missing a line? ...
  Thank you for pointing out to the issue with Table 1. We will replace it with Table 2 in the supplement file.
  3rd major comment: A third comment is that on line 1 of the abstract it says the new composite record for Dublin will be for 1938-2016, but from Table 1 we see there is Dublin data to 2019 …
  We excluded data post 2016 because the data exhibit very large variations and cannot be validated against Howth Harbour and Arklow even for MLW (shown in Figure 1 of the supplement file). We will clarify this issue in our revised manuscript.
  4^th major comment: A fourth comment relates to the regressions at lines 116 and 140 …
  We ran the model both with the perigean correction and without it, and there is no significant difference between the two, except for 1953-2016. However, the SLR rate has slightly reduced uncertainty when including the perigean terms. We will remove the perigean following your comment as the model is simpler and the results are nearly identical to the current estimates (results in Table 3 of the supplement file).
  Minor comments:
  We will fix the typos and ensure the text is more consistent and clearer throughout the paper.
  Specific responses to your inquires in the minor comments:
  Line 80 - There is found to be a systematic .. dataset presumably due .. [presumably because do you know that for sure? i.e. did they round in one way?]
  There was an overlap of 5 years between the lower vertical resolution data and higher resolution data. We calculated the mean high water and mean low waters for each and computed the offset. Hence we were able to be sure of the difference.
  Line 200- the model output as such is not shown in Figure A1, simply the difference Dublin minus Newlyn from which the model decides 1976 is a change point. And should that be difference and not absolute difference? ...
  We use the absolute difference because we are not interested in whether the changes are positive or negative but rather are looking for any changes in the MSL difference between sites. A model which uses the raw rather than absolute difference suggests multiple change-points throughout the time period, and would require a multiple change-point model. Such a model seems unnecessarily complex for this situation.
  References:
  Murphy, J., Sutton, G., O’Mahony, C., and Woodworth, P.: Scoping Study to Assess the Status of Irelands Tide Gauge Infrastructure and Outline Current and Future Requirements, Tech. rep., 2003.
  
  Citation: https://doi.org/10.5194/os-2021-92-AC1
RC2:
'Comment on os-2021-92', Sarah Bradley, 21 Dec 2021

Please see attached document

Citation: https://doi.org/10.5194/os-2021-92-RC2
- AC2: 'Reply on RC2', Amin Shoari Nejad, 21 Jan 2022
  
  Dear Dr. Bradley,
  Many thanks for reviewing the paper. Your comments are appreciated. Below we go through and respond to the queries. We look forward to hearing from you shortly.
  Kind regards,
  Amin Shoari Nejad on behalf of all authors.
  
  1: Section 4: Prior to calculating the MSL rates line 140 states “ yearly MSL from Brest and Newlyn .. removed atmospheric effects following”. If to use the MSL from Brest +Newlyn the atmospheric effects need to be removed; why was this not removed in Section 3 when estimating the Mean sea level for Dublin Port.
  
  Regarding your question as to why we didn’t remove the atmospheric effects when modelling MSL using MLW in Section 3 of the paper, the answer is that our atmospheric dataset for Dublin port is limited to 1948 onwards whereas we wanted to train the model on the longer (and therefore more robust) data set from 1938 to 1977. If we were to first remove the atmospheric effects and then fit the model, we would have to remove 10 years of data (1938-1947). We did attempt this but found that it increases the uncertainty of the model predictions substantially. In fact, we tried two approaches: 1) removing the atmospheric effects from both MSL and MLW and then fitting the model on the data during 1948-1977, and 2) training the model for the period 1938-1977 without removing the atmospheric effects for re-creating MSL using MLW, and then removing atmospheric effects from MSL to calculate the SLR rates. We concluded that the second approach is superior because the extra training data resulted in lower uncertainty associated with the predicted values of the model. We will add a sentence to the paper clarifying why we used this period.
  
  2: Section 5 and Section 6: Unless the journal requires both discussion and conclusion section: these could be combined as “Discussion +Conclusions” In the introduction the authors refer to the difference between satellite observations (line, 23 “ sea level rising at a rate of 2-3 mm/yr), Dublin City council rate (line 30: reports a 6-7 mm/yr SLR between years 2000-2016) and the rate from previous published studies. Could the authors refer to these differences in the discussion or conclusions. Why are the satellite records different, for example?
  
  With regard to the question of why there is a difference between the reported satellite measurements and Dublin council rates mentioned in the introduction, the former is the open ocean rate near Ireland and the latter is specific to Dublin port which is subject to local factors. Differences are expected between the satellite observations and local tidal sites in studies such as these. Our final rates of change contextualise the rates of rise reported by DCC and previous authors. We will add a footnote to this paragraph so that future readers are not confused by this discrepancy.
  
  Minor questions:
  
  1:Section 1: Line 34-40 “find problems with the MHW measurements which indicate a drift over time” From reading section 3, I assume this “drift correction” was calculated using the Bayesian multivariate linear regression. I think it would to mention this.
  
  That is correct, the Bayesian model is used to correct for the possible drift. Thanks for your suggestion, we will clarify this in the introduction as it is done in the latter sections.
  
  2: Data collection for Dublin Port: Line 45 - 66. We complied MHW... and, where available, mean sea level for Dublin Port. Do all five datasets have data for MHW, MLW, MTL and MSL? If not, could you add to table 1 which do/do not. From Figure 1, I assume they all provide MHW, but which ones do not record MSL?
  
  Thank you for your comment on Table 1. We will add a new column to that table including the available variables (MLW – MSL – MHW – MTL) for each dataset.
  
  3: Section 4: Perhaps rename “rates of sea level rise at Dublin Port and nearby tide-gauges”. Just to make it clearer the differences of this section to Section 3.
  
  Thank you for your suggestion. We will change the title of Section 4 to “Rates of sea level rise at Dublin Port and nearby tide-gauges” to make the distinction.
  
  Specific comments about figures
  
  Comments on Table 1:
  
  Regarding Table 1, we will add a datum column following your suggestion. However, we think combining Table 1 and Table 2 could cause some confusion because the two tables have different purposes. The first one is describing the various data sources used to create a reconciled dataset for Dublin port whereas the second one is introducing the other datasets from other locations that we used to validate the reconciled Dublin port dataset.
  
  About the inconsistency in dates specified in the text and the table, we apologize for the problem. There are some typos (e.g. 1978 should be 1987) that need to be fixed and some clarifications need to be made regarding why we used the data until 2017 while our dataset ends in 2019. We will fix these problems and make sure the text and the tables are consistent.
  
  2: Figure1: The length of the records shown do not correspond to the length of the records in Table.1. For example: Port Authority (blue) ~ 1968 to 2011? This may be a plotting problem or the resolution of the figure.
  
  Thank you for pointing out to the problem with Figure 1, the labels need a swap between the Greene dataset and the Port Authority dataset. We will fix the table in our revised manuscript.
  
  3: Line 89: “we now use our newly merged dataset (..) MSL and MHW from 1938 and 2019. As you refer to the period 1938 - 2019, why is Figure1 showing the MHW from 1968 to 2019?
  
  Figure 1 is trying to illustrate the consistency among the monthly values of different datasets used for the reconciliation (we don’t have monthly data pre-1968), whereas 1938-2019 is the duration of the reconciled dataset with annual resolution. We will add a sentence to clarify this in the next version of the paper.
  
  4: Table 2 and Figure 3: Can you clarify why the time axis is only to 2017 (but the records extend to 2019?)
  
  We decided to remove the data post 2016 because of its poor quality. This is mentioned in our previously submitted paper (the current one is the revised version) and we will further clarify this in our revised manuscript.
  
  Citation: https://doi.org/10.5194/os-2021-92-AC2

Peer review completion

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

AR by Amin Shoari Nejad on behalf of the Authors (14 Feb 2022) Author's response Author's tracked changes Manuscript

ED: Publish subject to minor revisions (review by editor) (24 Feb 2022) by John M. Huthnance

Dear authors.
Thank-you for your revised manuscript. I think that some “minor revision” is still needed. Some of this is related to your rather brief response to Reviewer 1; some of the points not addressed in the response are not treated in the revised manuscript either. There still seem to be some unexplained discrepancies between dates. Please revise taking into account the comments below; these include some “editorial” comments seeking clarity for the reader.
Yours sincerely
John Huthnance (editor)

Specific comments.
Line 7. I think CI should be defined as “confidence interval”.
Line 20. Better “increase” –> “rise”
Line 48. You refer to “Annual high and low water . . . for the period 1938 to 2001” but lines 49 to 51 only mention the datum for data up to 1988 and line 54 refers to only “two overlapping years”. Why ignore the annual data for 1989-2001?
Line 74. “The second-high tide”. As Reviewer 1 said, on some days there is only one high tide. You should say what procedure allowed for this.
Table 1. I miss the definition of MTL. Please also give an explicit definition of mean sea level (MSL) so that the difference between MTL and MSL is clear.
Line 78. Elsewhere (line 62, Table 1) the NTGN dataset only begins in 2007 giving only three years overlap. Please be consistent.
Lines 80-81. “. . dataset presumably due . .”? (c.f. Reviewer 1, unless you are absolutely sure of the reason).
Line 82. “. . no overlaps . .” but Table 1 shows Port Authority monthly completely covering Harbourmaster. Likewise Figure 1: why does the Port Authority stop at 2001 or 2002 whereas in Table 1 it continues to 2017?
Table 2. I think this should include the tide gauge type (radar?) as implicitly asked for by Reviewer 1. For the Dublin gauges, I think some of the detail sought by Reviewer is still lacking: “calibration method used to level it in such that low waters were accurate?” “why were high waters less reliable?” “density (and maybe gravity) assumed”, “manufacturer?” “absolute or differential sensor?”
Line 99. “. . malfunction of the Dublin tide gauge . .” (the reader needs to know which tide gauge)
Line 104. Omit “to”
Lines 120 and two lines before 145. “mean process” –> “statistical mean” or “mean of the process” or simply “mean”?

Lines 137-138. “Due to the lack of atmospheric data for Dublin Port before 1948, we removed the atmospheric effects after correcting for MSL.” This sentence is very unclear. How do you remove atmospheric effects without atmospheric data? What does “correcting for MSL” mean (is it a reference to some procedure in section 3)? Please be explicit.
Why do figure 6 and Table 5 only start in 1953? Line 137 refers to atmospheric data from 1948.
Four lines after 140. Omit “, and a harmonic function with periods of 18.6 years and 4.4 years”

Lines 160-161. “regions of relative sea level fall . . . which will reduce relative sea level rise” is rather confusing. Better “regions of uplift due to . . . (GIA) which reduces relative sea level rise”.
Line 163. Supply the “reference”!
Line 175. “. . as the more . .”
Sections 5 and 6. These can be merged as a Discussion and Conclusions section if wanted.

Lines 210-211. This sentence does not read correctly. What “is carried out . .”?

Hide

AR by Amin Shoari Nejad on behalf of the Authors (12 Mar 2022) Author's response Author's tracked changes Manuscript

ED: Publish subject to technical corrections (18 Mar 2022) by John M. Huthnance

AR by Amin Shoari Nejad on behalf of the Authors (24 Mar 2022) Author's response Manuscript

Short summary

We have collated multiple sources of tide gauge data for Dublin Port, and subsequently corrected them for bias. We have then shown that these corrected mean sea level measurements agree with nearby tide gauges to a far higher degree than the raw data. A longer-term comparison with Brest and Newlyn also indicates overall agreement. Our final adjusted dataset estimated the rate of sea level rise to be 1.1 mm/yr between 1953 and 2016 and 7 mm/yr between 1997 and 2016 at Dublin Port.