Analysing the 100year sea level record of Leixões, Portugal

Journal of Hydrology xxx (2013) xxx–xxx

Contents lists available at SciVerse ScienceDirect

Journal of Hydrology

journal homepage: www.elsevier .com/ locate / jhydrol

Analysing the 100 year sea level record of Leixões, Portugal

I.B. Araújo a,⇑, M.S. Bos a, L.C. Bastos a,b, M.M. Cardoso a

a CIIMAR/CIMAR, University of Porto, Rua dos Bragas 289, 4050-123 Porto, Portugalb Astronomical Observatory and DGAOT, Faculty of Sciences, University of Porto, Rua do campo Alegre 687, 4169-007 Porto, Portugal

a r t i c l e i n f o

Article history:Received 10 April 2012Received in revised form 9 November 2012Accepted 14 December 2012Available online xxxxThis manuscript was handled by Peter K.Kitanidis, Editor-in-Chief, with theassistance of Souheil M. Ezzedine, AssociateEditor

Keywords:Long-term sea levelTidesGauge

0022-1694/$ - see front matter � 2012 Elsevier B.V. Ahttp://dx.doi.org/10.1016/j.jhydrol.2012.12.019

⇑ Corresponding author. Tel.: +351 22 3401897; faxE-mail addresses: [email protected] (I.B. Araú

Bos), [email protected] (L.C. Bastos).

Please cite this article in press as: Araújo, I.B., etj.jhydrol.2012.12.019

s u m m a r y

A new data set from the tide gauge at Leixões (North–West Portugal) has recently been transferred fromits paper format into a digital time series of hourly sea level values. By measuring sea level variationssince 1890, this tide gauge station is one of the few in the world with over 100 yr of digitised hourlyrecords and the longest for the South West of Europe and Iberian Peninsula. This paper presents the pro-cedures adopted to recover the Leixões sea level data from paper chart records as well as the data qualitycontrol and data editing methodologies. The mean rate of sea level change between 1906 and 2008 is�0.70 ± 0.27 mm yr�1, which does not agree with the global mean sea level rise of 1–2 mm yr�1. No evi-dence for vertical land movement was found and Global Isostatic Adjustment influence on sea level, atthis location, can be neglected. It is likely that prevailing weather systems in the North Atlantic, especiallyin the winter, and local atmospheric pressure, influence sea levels at Leixões. A further contribution isfound from tides and surges. The evolution of the port cannot be ignored when trying to understandsea level change.

� 2012 Elsevier B.V. All rights reserved.

1. Introduction Until recently, the longest sea level record for South-West Eur-

The understanding of sea level changes has in the past decadeacquired renewed interest in view of climate change research.The use of historical sea levels to estimate local and global sea levelchanges has been invaluable (e.g. Holgate, 2007; Church andWhite, 2011). As Colosi and Munk (2006) have pointed out, thereis a large natural variability in sea level which makes it necessaryto have very long time-series to detect secular changes.

Sea level observations have a relatively long history dating backto the late 17th Century (Woodworth et al., 2011; Wöppelmannet al., 2006). Along the European Atlantic coast, examples of suchstations are Brest (Wöppelmann et al., 2006), Amsterdam (vanVeen, 1954), Liverpool (Woodworth, 1999), Stockholm (Ekman,1988) and Cadiz (Marcos et al., 2011), for which datasets date backto 1679, 1682, 1768, 1774 and 1880, respectively (c.f. Marcos et al.(2011) for a summary of long-term records).

Some of these also have the longest ‘continuous’ hourly read-ings registered through an automatic recording device composedof a drum attached to a float by a system of wires and counter-weights. The float-gauge system in a stilling well enables a tidalcurve to be drawn on a paper chart placed around the drum. It ap-peared in the 1st half of the 19th Century and was still the domi-nant tide gauge technology into the 20th Century.

ll rights reserved.

: +351 22 3390608.jo), [email protected] (M.S.

al. Analysing the 100 year sea le

ope has been that of Cascais (Antunes and Taborda, 2009), in Por-tugal (38� 41038.800N; 9�2505.400W), for which monthly mean sealevels exist since 1882 and hourly digitized levels only since1960. Other Spanish and Portuguese hourly sea level data, priorto the 60s, are likely to exist in paper format, nevertheless theirdigitisation has not been made.

The Leixões Port authority recently brought to our attention thatthere were additional long-term sea level records that can be addedto that of Cascais, namely Cantareira in the Douro River (41�804700N;8�400000W) and Leixões (41�1101200N; 8�4201700W), approximately5 km North of Cantareira (Fig. 1). This led to the digitization of theLeixões sea level observations that will be discussed herein. The firstrecords date back to 1890, making this series the longest hourly-dig-itized record in the country and Iberian Peninsula. The objective ofthis paper is to present this data set and to investigate if it can beused to study secular sea level changes in the region.

This paper starts by giving an overview of the Port of Leixões, itshistory and its tide gauges. This is followed by a description of thedataset that was digitized and the methodology used in the recoveryof the records, which includes the data quality check (editing). Final-ly, we present the results and first interpretation of this record.

2. Historical background

2.1. Evolution of the Leixões Port

The Port of Leixões is located on the northern Portuguese coastapproximately 5 km north of the Douro River mouth. The port

vel record of Leixões, Portugal. J. Hydrol. (2013), http://dx.doi.org/10.1016/

Fig. 1. Location and diagram of the Port of Leixões and its tide gauge station (shown by the black circle). Dates are given for the main construction works described inSection 2.

2 I.B. Araújo et al. / Journal of Hydrology xxx (2013) xxx–xxx

extends from Leça da Palmeira (North) to Matosinhos (South), cov-ering an area of 180 ha, see Fig. 1.

The construction of the port started in 1884 and evolvedthroughout the following Century as a result of major land recla-mation that substantially altered the original landscape. Sea levelshave been measured from tide gauges at Leixões since 1890 withvery few interruptions despite several severe political and financialepisodes throughout those years. The history of this port (withnumerous anthropogenic influences) should therefore be consid-ered alongside to its sea level records.

Construction of the artificial port started in July 1884 with theNorth (1579 m) Breakwater and the South (1145 m) Breakwaterbuilt with 50 ton granite block foundations on existing beachrocks. In 1890 a low-rising breakwater added a few metres to theNorth breakwater. These structures were mostly concluded in1892. Construction of a commercial harbour started later in 1914with a docking berth on the South Breakwater. However, workswere put on hold until 1932 as a result of national political unset-tlement and the financial constrictions of the great recession.When construction picked up in 1932, increase in shipping activityand mooring problems (conditioned by increased intensity ofwaves) called for new expansion of the Port. This occurred inlandinto the Leça estuary (1932–1940) with Dock 1 and the construc-tion of an extension to the North Breakwater (1932–1945) to re-solve the wave effects within the harbour. From 1956 to the 60sfurther expansion took place inland with Dock 2 (0.5 km). Dam-ming at the end of Dock 1 was necessary to avoid salt intrusionto the (substantial) excavated grounds up the Leça River and adja-cent area. This involved the diversion of the course of the Leça Riv-er into the harbour, outside Dock 1.

Further expansion included a fishing harbour (1965–1968), aterminal for tankers and the raising of the height of the breakwaterextension to above submersion level during the 60s. The NorthContainer Terminal (1974–1979) and the extension to Dock 2(1974–1983), i.e. Dock 4, followed. At the end of the 80s the break-water was increased and in the 90s the Marina and South Con-tainer Terminal were built (Fig. 1). A summary of all the majorconstructions in the harbour and change of instrumentation is gi-ven in Table 1 and Fig. 1.

2.2. Description of the Leixões tide gauges and datum

The continuous water level measurements at Leixões have been,and are still made using a float gauge in a stilling well. The oldest

Please cite this article in press as: Araújo, I.B., et al. Analysing the 100 year sea lej.jhydrol.2012.12.019

record found dates back to 1890. The exact location of the gauge upto the early 1900s is questionable as the only information availableis from the header on the 1890 record, which mentions ‘‘LeixõesNorth Breakwater’’ whilst headers in the 1892/3 records mention‘‘Leixões South Breakwater’’.

Despite not being able to specify the exact location of the tidegauge on both breakwaters, following historical accounts of theevolution of the port (Table 1) we estimate that between 1890and 1892 the gauge would have been sheltered within the NorthBreakwater inner port. Reports of constant repair and buildingworks on the North Breakwater, as a result of regular breachingby violent storms (in 1887, 1888, 1892), is a plausible explanationfor the transfer of the tide gauge from the North Breakwater to theSouth Breakwater (registered in the 1892 record header). Thisassumption is based on a clear definition of the South Breakwateras stated on port plans which should not be confused with the In-ner South Pier (also referred to as docking pier or tide gauge pier)that later lodged the gauge and currently forms part of the NorthContainer Terminal (seen in Fig. 1).

The location of the gauge between 1894 and 1916 is also uncer-tain. We have been able to confirm the existence of a tide gaugehut at the end of the inner south pier from a 1917 photo of a shipcollision with the Inner South Pier. Therefore, it seems reasonableto assume that the tide gauge hut has remained on the Inner SouthPier since 1917.

The coordinates of the present tide gauge station (at the North-west end of the North Container Terminal) are 41�1101200N and8�4201700W (black circle in the left panel of Fig. 1). It consists of afloat gauge in a stilling well connected to an Ott RV20 paper chartrecorder and an Ott Thales digital logger, see Fig. 2.

In 1956 the Portuguese Hydrographic Institute (IH) took overresponsibility for the maintenance of the gauge and for data acqui-sition, previously overseen by the Geographic and Cadastral Insti-tute (IGC; note that IGC was established in 1926 and changingname to IPCC (Instituto Português de Cartografia e Cadastro) in1994 and to IGP (Instituto Geográfico Português) in 2002). Duringthis transition (1956–1958) an Ott V (Bosum MKV) tide gauge wasinstalled remaining in use until January 1983 (based upon an in-voice letter). No reference was found for the gauge used from1983 to 1992. However, an Ott MTG was installed in 1992, whichwas changed for the current gauge in 2004.

IH has a dataset of hourly sea levels extracted from the paperrecords. From 1956 until 2000 those hourly levels are referencedto the local datum, i.e. Zero Hidrográfico de Leixões (ZHL described

vel record of Leixões, Portugal. J. Hydrol. (2013), http://dx.doi.org/10.1016/

Table 1Relevant milestones in the history of the Port and its tide gauge.

Dates Events

1884–1892 Start of Port with construction of the north and south breakwater1914–1931 Commercial harbour1932–1940 Dock 11932–1942 Extension of the north breakwater1956–1975 Dock 21965–1968 Fishing harbour1974–1979 Container Terminal1974–1983 Dock 41956/7–1983 OTT V gauge1992–2004 OTT MTG gauge2004- OTT RV gauge/thales digital recorder/stilling well changes

I.B. Araújo et al. / Journal of Hydrology xxx (2013) xxx–xxx 3

below). From 2000, both analogue readings and higher frequency6-min digital readings are referenced to the hydrographic zero da-tum or national chart datum (ZH, described below). The main pur-pose of acquiring these sea level data has been for the prediction oftides in the harbour (Reis, 2005).

Ancillary data (metadata), essential in the analysis of sea levelrecords, usually include levelling and gauge maintenance informa-tion as well as other relevant details. Leixões gauge metadata priorto 1956 was handed over to IH in 1956 (Gonçalo Crisótomo (IGEO),personal communication) and IH has continued to compile infor-mation which is held in the form of loose records. Future interpre-tation of Leixões data will benefit once a clear compilation of allexisting documents held by IH that report on gauge location, gaugerecorders, maintenance and levelling, as well as other relevantinformation, are made available.

Leixões tide gauge station is part of the national tide gauge net-work. It is also listed as station 791 on the Permanent Service forMean Sea Level (PSMSL) database (Woodworth and Player, 2003)from which annual and monthly mean sea level (for 1956–1995)can be retrieved with reference to the Hydrographic Zero (ZH).

At Leixões ZH is 2.00 m below the national levelling datum(Cascais Helmert 38). The latter datum was determined in 1938by the Mean Sea Level (MSL) estimated from 56 yr (1882–1938)of sea level measurements from the Cascais tide gauge.

The tide gauge benchmark (TGBM) at Leixões consists of around bronze horizontal plate (letter mark 2BT) emplaced to the

Fig. 2. Leixões tide gauge at the end of the Container Terminal pier. The tide gauge shelwires that connect to the stilling well through an opening on the floor below the record

Please cite this article in press as: Araújo, I.B., et al. Analysing the 100 year sea lej.jhydrol.2012.12.019

northern side of the tide gauge hut (Fig. 3). It was installed by IH(1956–1957) substituting the previous existing N1L plate, usedby IGP. Sea level observations prior to 1956 were referred to datumestablished by IGP. To the best of our knowledge, benchmarks Ot1,Ot2 and N1L were used by IGC while M6 and 2BT, amongst others,are used by IH (confirmed location of some of the benchmarks usedby IH are shown in Fig. 3).

From 1956 till 2000, sea levels have been measured to the localchart datum known as the Leixões Hydrographic Zero, ZHL. How-ever, after the installation of the digital gauge both recorders wereset to measure with reference to national chart datum, i.e., ZH. Therelationship between ZHL, ZH and the national levelling datum(NP), are given in Fig. 3. The local ZHL datum is 0.326 m aboveZH and 1.674 m below NP.

Global Positioning System (GPS) equipment was installed by theFaculty of Science of the University of Porto in September 2008 buta longer dataset is needed to derive reliable estimates of verticalstability of the gauge.

Sousa et al. (2011) used satellite based InSAR (InterforometricSynthetic Aperture Radar) and Multi-Temporal Techniques (MTI)to evaluate the stability of the Port of Leixões area. The methodol-ogy chosen was based on the Stanford Method for Persistent Scat-terers/Multi-Temporal InSAR (StaMPS/MTI), which combines bothPS and SB methods allowing the identification of scatters that dom-inate the scattering from the resolution cell (PS) and Slowly-Decorrelating Filtered Phase (SDFP) pixels, identified due to the

ter harbours a digital (Thales) and a paper recorder. Both are linked to the float byers. The PVC tube inside the stilling well minimises any breaching of the walls.

vel record of Leixões, Portugal. J. Hydrol. (2013), http://dx.doi.org/10.1016/

Fig. 3. Reference levels and datum for Leixões tide gauge station.

4 I.B. Araújo et al. / Journal of Hydrology xxx (2013) xxx–xxx

use of interferograms formed only between images separated byshort time and space baselines (SBs). This was applied to 47 ERS-1/2 scenes and 19 Envisat scenes covering the Port of Leixõesand, therefore the tide gauge location, between 1992 and 2007. Re-sults estimated rates of 0.35 mm yr�1 (with coherence 0.75) and0.28 mm yr�1 (with 0.71 coherence) in ERS and ASAR, respectively.

Moreover, we have applied a linear fit to the Sousa et al. (2011)MTI results, from which a trend was estimated using ordinaryleast-squares and standard errors. An offset in the estimation pro-cess was included to allow for the difference in height caused byusing two different satellites (ERS-1/2 before 2002 and Envisatafter 2002). An insignificant 0.20 ± 0.24 mm yr�1 was found (offset:0.60 ± 2.16 mm; nominal bias: �0.74 ± 0.69 mm) for the tide gaugelocation suggesting stability of the area of land around the station.

3. Data and methodology

3.1. Sea level dataset digitisation

The gauge records from Leixões are fairly continuous, datingback to 1890 and with main discontinuities occurring within thefirst three decades. From 2004 onwards the quality of the analoguerecord declines as a result of several breakdowns and malfunctionsof the logger. Nevertheless, the available digital recordings post2004 are able to ensure data continuity and have been used forthe 2004–2008 period.

The 1890–2004 paper records were scanned into a digital imageformat that insured the legibility of all the details on the paper.Hourly sea levels were extracted from an image display of the dig-ital paper record using a custom-made software program for thistask. A few adjustments were made according to hand writtennotes on the records, most related to Daylight Saving Time (DST)shifts. Before the 1930s the data seem to have suffered from clockproblems or bad clock adjustments when the paper rolls werechanged each week because the records show clearly a tidal resid-ual that varies from week to week after the predicted tides have

Please cite this article in press as: Araújo, I.B., et al. Analysing the 100 year sea lej.jhydrol.2012.12.019

been removed. For that reason it was not always possible to verifyif summer time or wintertime was used.

The hourly data were stored in a database following a few adjust-ments in time according to hand written notes on the records. Thisprovided a continuous time set of hourly relative sea levels.

3.2. Data quality and editing

The malfunctioning of a conventional stilling well tide gaugecan cause errors as a result of calibration, datum control, slidingof the pulley cables, obstructions in the well, and especially withthe older mechanical clockwork timing (IOC, 1985; Pugh, 1987).Nevertheless, if well maintained it has proven to be a reliableway of recording sea level.

The repeatability of the operator of the software used to obtainthe hourly levels from the scanned paper rolls was tested by com-paring a year of extracted hourly values to a second independentextraction of the same year. This process was also done with otheryears, with 20yr intervals, and the results showed that extractedhourly levels are reproducible with a standard deviation of 1.03 cm.

The repetitive and time-consuming task of selecting hourly val-ues can be another source of operator error. Any detected outlierswere corrected by comparing each year of hourly digitised sea levelobservations with a year of hourly sea levels predictions obtainedfrom the harmonic analysis of a selected year of data unaffected byDST adjustments.

After the previous procedures, each year of data (withP300 days) underwent robust editing following the method ofAraújo and Pugh (2008). Non-tidal residual values, obtained aftera harmonic analysis on each individual calendar year of data, werescrutinised in 10-day blocks to identify measurement errors abovea threshold level of two standard deviations of the residual vari-ability (below this threshold apparent errors were tolerated). Onlythose signals that were clearly errors in the measurements wereadjusted by substituting tidal-residual values by correspondinglow-pass filtered values. A second harmonic tidal analysis was thenperformed on the adjusted observed values.

vel record of Leixões, Portugal. J. Hydrol. (2013), http://dx.doi.org/10.1016/

Fig. 4. Observed sea level standard deviation (SL STD) with a 18.6 nodal cycle fit(top); and corresponding fit residual series (bottom).

I.B. Araújo et al. / Journal of Hydrology xxx (2013) xxx–xxx 5

The annual and monthly MSLs obtained after the data qualitycheck and editing described herein are made available with thispaper and also via the PSMSL database.

3.3. Annual mean sea level and adjustment for air pressure

The whole data set of hourly sea level data was divided into cal-endar years that were analysed with the TASK-2000 tidal analysissoftware (Bell et al., 2000). Annual MSL is one of the results ob-tained from this analysis. However, a realistic estimate of the er-rors in an annual mean sea level calculation is necessary todetermine if the changes in MSL from year to year are significantlydifferent from zero or if they fall within the measurement noise.The TASK-2000 software ignores the temporal correlation that ex-ists in the data although this dependency needs to be taken into ac-count to obtain a realistically estimated error for the computedannual mean values.

The properties of this temporal correlation were computed fromtidal residuals using a simple first order auto-regressive noisemodel (Brockwell and Davis, 2002) and resulted in a mean errorof 9 mm which is 6–9 times increase of the uncertainty of the esti-mated annual mean sea level compared to the standard error.

MSL response to atmospheric pressure requires a long-termMean Sea Level Pressure (MSLP) record to cover the length andtime of the sea level record. Version 2 of the 20th Century Reanal-ysis data provided by the NOAA/OAR/ESRL PSD, Boulder, Colorado,USA, at http://www.esrl.noaa.gov/psd/ contains global weatherconditions from the year 1891 to 2008 (Compo et al., 2006,2011; Whitaker et al., 2004). Monthly pressure reduced to MSLPvalues were extracted from the 2� � 2� global grid for the gridpoint closest to Leixões. MSLP from 1906 to 2008 shows no trend.Correlation between detrended annual MSL and detrended MSLPresulted �16.32 ± 7.9 mm mbar�1 trend, within 1 standard errorfrom the theoretical inverted barometer effect (�10 mm mbar�1).The difference is explained by the effect of winds directly corre-lated with atmospheric pressure (c.f. Mathers and Woodworth,2001). This observed relation was used to remove the influenceof the atmospheric pressure from the annual MSL.

4. Results

4.1. Observed sea level

The standard deviation in the observed sea level variations, foreach year of the observed (hourly) sea levels, Fig. 4, is clearly af-fected by the 18.6yr nodal cycle with an amplitude of 3.2%(25.9 mm). This is within the ±3.7% nodal modulation of the semi-diurnal Equilibrium Tide.

A regression model using least squares was fit to the observedsea levels (Fig. 4). A 0.02 mm yr�1 decrease is found in the standarddeviation when the nodal cycle is removed. The fit residuals (level-fit) are given in Fig. 4. Araújo (2006) applied an identical fit to theobserved sea level standard deviation from stations along the Eng-lish Channel, France, Spain and Portugal. Negative trends werefound for Vigo (�0.12 mm yr�1) and Brest (�0.04 mm yr�1), con-sistent with increase in M2 and S2 tidal amplitudes. Positive trendswhere found for Newlyn, Santander and Cascais.

4.2. MSL

The estimated annual MSL values together with their correctederror bars are shown in Fig. 5. The years earlier than 1906 havebeen discarded because within each year there was less that 82%of days with observations (<300 days). A 5 yr moving average isalso shown to visualise the long term changes in MSL more clearly.

Please cite this article in press as: Araújo, I.B., et al. Analysing the 100 year sea lej.jhydrol.2012.12.019

Relative MSL over the period of 1906–2008 has decreased atan average rate of 0.70 mm yr�1 ±0.27 mm yr�1. This result con-tradicts both global and regional MSL estimates. Negative ratesin MSL have only been reported in Northern Europe (Scandina-via) where Glacial Isotactic Adjustment (GIA) are responsiblefor the vertical component of land movement which in turn af-fect MSL.

The ICE-5G (VM2 L90) model (version 1.3) for PSMSL Leixõestide gauge site (http://www.atmosp.physics.utoronto.ca/~peltier/data.php, 13th August 2012) predicts an average present-day rateof vertical motion of the solid earth due to Glacial Isostatic Adjust-ment (GIA) of �0.23 mm yr�1, whilst the predicted present-dayaverage rate of sea-level rise due to GIA is �0.20 mm yr�1 (Peltier,2004). The GIA influence on sea level is within the standard error ofthe estimated MSL trend for this site and has, therefore, beenneglected.

Long-term trend estimates are known to be affected by the var-iability in the decadal estimates. Holgate (2007) found differentdecadal rates of change during the 50s with highest negative ratescentred in 1964.

Leixões averaged annual MSL and standard deviation estimatedover 10 yr periods confirm that higher decadal MSL averages arefound prior 1970, with the exception of a low decadal average in1950s (Fig. 5). MSL trends have therefore been estimatedconsidering:

(i) the 50 yr of data prior and post IH responsibility for thegauge, i.e. 1906–1955 and 1956–2008;

(ii) the period in which the decadal average MSL and standarddeviation decreased, i.e. 1970–2003/08;

(iii) the 1956–2003 period, as to avoid any bias introduced bythe 0.326 m levelling adjustment of the digitized levels (ZHto ZHL) and the changes to the stilling well, both includedin the 2004–2008 data.

No statistical significant trends have been found.Whilst changes in sea levels can occur over different time

scale (e.g. decadal) coherence is expected between neighboringtide gauge stations at which the regional dynamics (steric andmeteorological) is expected to be similar. Leixões does not showcorrelation with neighboring Lagos and Cascais stations in Por-tugal, nor to Northern Iberian stations (Vigo, Coruna,Santander).

vel record of Leixões, Portugal. J. Hydrol. (2013), http://dx.doi.org/10.1016/

Fig. 5. Annual mean sea level (above ZHL) based on 1906–2008 hourly sea levelswith a 5yr moving average line.

6 I.B. Araújo et al. / Journal of Hydrology xxx (2013) xxx–xxx

4.3. Tides

The amplitude and phase of the main tidal constituents wereobtained by harmonic analysis performed on the hourly levels foreach calendar year. Their mean and respective standard deviationsare given in Table 2. Although MSL results are given for 1906–2008,those for tides are only shown from 1933, as a result of, for in-stance, an unexplained 22� (�44 min) phase shift in the semi-diur-nal lunar component. Note that timing problems do not have alarge effect on the annual mean sea level because they will tendto average out over the year.

Tides are dominated by M2 with mean amplitude and phase lagof 104.15 ± 0.68 cm and 74.4 ± 0.9� (relative to local time (UTC)),respectively. Yearly variability is illustrated in M2, S2 and Sa plotsin Fig. 6. Annual variation of mean sea level, as represented bySa, has an average amplitude of 5.21 cm. M2, S2, K2, K1 and O1 re-sults are consistent with those obtained by Reis (2005) from a har-monic analysis of a year of data (31 December 2002 to 30December 2003).

From the mid 50s, small trends are mostly found in S2 with a verysmall amplitude decrease (�0.01 ± 0.00 cm yr�1) alongside a smallincrease in phase (0.1 ± 0.0� yr�1). M2 phase lag also increases byan identical amount but no trends have been found in amplitude.These trends are influenced by instrumentation changes from1992 that increase M2 and S2 amplitudes by �1 cm and 0.1 cm,respectively, and phase lag by �1.5�. The increase in phase is clearlypronounced after sea level measurements are made inside the PVCtube installed in the stilling well. If the data after 1992 is disre-garded, then M2 and S2 amplitudes decrease by 0.04 ± 0.01 cm yr�1

and 0.02 ± 0.00 cm yr�1, respectively. Decreases in M2 amplitudeshave also been reported for Brest (Cartwright, 1972; Simon, 1982).

4.4. Extreme sea levels

Reduced annual percentiles of observed sea levels, obtained byremoving the annual median level from each annual percentile,

Table 2Mean (1933–2008) amplitude and phase, and their standard deviation (STD), for themain (>1 cm) tidal constituents at Leixões.

Mean amplitude (cm) STD Mean phase (�) STD

Sa 5.21 3.08 239.9 67.4Ssa 4.54 2.25 146.5 105.5MSf 1.29 0.69 169.3 108.2Mf 1.38 0.94 189.9 86.8Q1 1.93 0.24 266.5 6.9O1 6.28 0.18 317.2 1.5S1 1.26 0.48 138.6 159.7K1 6.92 0.15 58.7 1.0N2 22.23 0.27 56.0 1.1M2 104.15 0.68 74.4 0.9L2 2.42 0.37 89.8 7.8S2 36.64 0.30 102.2 1.1K2 10.38 0.22 99.4 1.5

Please cite this article in press as: Araújo, I.B., et al. Analysing the 100 year sea lej.jhydrol.2012.12.019

were analysed to determine changes in the frequency of measuredsea levels. The reduced level gives a measure of the percentile levelof a year relative to MSL, i.e. by reducing these levels it is possibleto understand if the forcing in the extremes is common to MSL.Moreover, it removes issues regarding vertical land movementand datum uncertainties (Woodworth and Blackman, 2004).

The 99.9, 99 and 95 percentile and the 0.1, 1 and 5 percentilecorrespond to the highest and lowest sea levels, respectively. Re-sults for 99 and 95 percentile will be particularly interesting asthey take into account >8 h of data (in a year), which will be lessprone to errors.

The interannual variability in all reduced percentile levels isinfluenced by a small nodal modulation (18.6 yr). These levels donot have similar time dependence to the median. The higher ob-served sea level levels, within the 99 and 95 percentile, decreaseat 0.09 mm yr�1, over the 1906–2008 period (Table 3). In the re-duced levels the decrease is only significant in the 95 percentile(0.02 ± 0.01 cm yr�1). This reduction is accompanied by an increasein the reduced 5-percentile (0.03 ± 0.01 cm yr�1) over the sameperiod (Table 3). These results show a decreasing tendency in themaximum and minimum extreme sea levels.

Non-tidal percentiles were also estimated (Table 3) to be able togain further insight into the forcing of these percentile levels disre-garding the tidal component. For this we followed identical proce-dure to Woodworth and Blackman (2004), in which a time serieswas obtained by subtracting tidal percentile values from the 99percentile observed values, both reduced. The difference obtained(not shown herein) was different from zero. The large sea level var-iability found at this site is translated in these results suggestingtides are less of a contributing factor than weather or other oceanicforcing.

The higher percentile levels in the non-tidal signal have nega-tive trends for the 95 and 99-percentiles and positive trends inall lower percentiles (Table 3), this can contribute to the decreasefound in the non-tidal residual standard deviations.

4.5. Atmospheric influences

In addition to the ‘inverted barometer effect’ on MSL, regionalclimatic patterns should also be considered. In the North Atlantic,sea level is strongly influence by the North Atlantic Oscillation(NAO) (Woolf et al., Wakelin et al., 2003), which is taken as a mea-sure of air pressure gradient and winds affecting Europe. The dif-ferent phases of the NAO are known to influence atmosphericand oceanic heat content, precipitation, salinity and river runoff(Hurrell and Deser, 2009; Otero et al., 2010; Trigo et al., 2004), con-tributing towards sea level change via changes in mass andvolume.

The Principal Component (PC) based NAO time series consists ofthe leading series of Empirial Ortogonal Functions (EOFs) of annualSea Level Pressure (SLP). Winter (December–March) NAO anoma-lies, were obtained from the Climate Analysis Section, NCAR, Boul-der, USA, (Hurrell et al., 2003), accessed 1-1-2012. The annual MSLand winter (DJFM) MSL values measure the sensitivity of sea levelto the annual and winter NAO index measured with the assump-tion that sea level is a linear function of the NAO index. The NAOaction during the winter period accounts for more than 1/3 ofthe total variance of Sea Level Pressure (SLP).

There is a �0.4 correlation between (detrended) winter sea lev-els and winter NAO index (�24.9 ± 10.6 mm change in level perunit winter NAO index).

The negative sensitivity to the NAO index found at Leixões is inagreement with results from the northern Iberian stations (Woolfet al., 2003; Fenoglio-Marc et al., 2005). According to Woolf et al.(2003) in southern Europe, sea level is lower during NAO-positiveyears with values for the northern Spanish coast ranging between

vel record of Leixões, Portugal. J. Hydrol. (2013), http://dx.doi.org/10.1016/

Fig. 6. Amplitude (in cm) and phase (in degrees) of M2 and S2 (main semi-diurnal) tidal constituents and annual solar constituent (Sa).

Table 3Trends in observed sea level and non-tidal percentiles and their reduced trends. Statistically significant trends are highlighted in bold italics.

Percentile Observed sea level trend(cm yr�1)

Reduced sea level trend(cm yr�1)

Non-tidal residual sea level trend(cm yr�1)

Reduced non-tidal residual level trends(cm yr�1)

99.9 �0.13 ± 0.05 �0.06 ± 0.04 �0.05 ± 0.04 �0.34 ± 0.3299 �0.09 ± 0.04 �0.02 ± 0.02 �0.04 ± 0.02 �0.79 ± 0.5595 �0.09 ± 0.03 �0.02 ± 0.01 �0.02 ± 0.01 �2.10 ± 1.165 �0.03 ± 0.03 0.03 ± 0.01 0.03 ± 0.01 3.10 ± 1.211 �0.06 ± 0.04 0.01 ± 0.02 0.05 ± 0.02 1.43 ± 0.590.1 �0.06 ± 0.05 0.01 ± 0.03 0.07 ± 0.03 1.01 ± 0.42

I.B. Araújo et al. / Journal of Hydrology xxx (2013) xxx–xxx 7

20 and 60 mm/unit NAO index. This negative response was alsofound between annual non-tidal residual levels and the NAO index(�4.73 ± 1.42 mm per NAO index).

5. Discussion

Analysis of this long-term sea level record has revealed anunexpected decrease in MSL that is contrary to the consistent re-gional trend pattern of over 2 mm yr�1 found by Marcos et al.(2005) for North Iberian stations, during the second half of the20th Century. Wöppelmann and Marcos (2012) investigated theimportance of the non-climatic contribution of vertical land move-ment to observed rates of sea level change. After removing thatcontribution a sea level rise, in excess of 3.4 mm yr�1 (for the past70 yr), was re-estimated for the northern Iberian coast. Non-cli-matic contributions have an increasing effect in MSL rates in thisregion, however, it is important to note that the period analysedis short compared to the analysis presented herein.

Please cite this article in press as: Araújo, I.B., et al. Analysing the 100 year sea lej.jhydrol.2012.12.019

GIA influences on sea level have been neglected on the basisthat for this site the predicted average is within the standard errorof the estimated MSL trend. Other local vertical land movement, forinstance from land drainage, have also been neglected followingInSAR data that support the stability of the site.

MSL includes contributions from tides and surges and are af-fected by changes in atmospheric pressure, wind, density and/oroceanic/atmospheric circulation. Over the years the tidal ampli-tude and phase-lag have changed in the Leixões harbour. Correla-tion between tides and sea level are found when the entire dataseries is analysed. During 1906–2008 M2 decreases with increasesin MSL and non-tidal variations (Section 4.4) at a rate of 0.04 ± 0.01and�0.20 ± 0.07, respectively, though correlations are weak (<0.4).

Change in MSL affects sea level directly but also modifies thetide by changing the water depth. Depth increases give longer tidalwavelength hence affecting the tidal pattern with variations in thetidal levels. Increases in tidal amplitude and decreases in phase insingle inlet systems have also been related to a decrease in bottom

vel record of Leixões, Portugal. J. Hydrol. (2013), http://dx.doi.org/10.1016/

8 I.B. Araújo et al. / Journal of Hydrology xxx (2013) xxx–xxx

friction, changes in bathymetry and to other non-linear effects(Araújo et al., 2008). Dredging activity can increase depth andinfluence bed roughness, therefore altering bottom friction. ThePort of Leixões can be considered as an enclosed bay with a directopening to the ocean with dimension and bottom friction changesas a result of continued dredging, excavation and constructionwork.

In this sense, observed changes in tidal amplitude and phasewere investigated using a barotropic depth integrated tide model,following Egbert et al. (1994) on a 10� � 10� m grid for the area.The bathymetry was assumed constant inside each individual dock(refer to Fig. 1), while outside the harbour depth smoothly in-creases with increasing distance from the coast. The open oceanboundary was forced with the global ocean tide model FES2004(Lyard et al., 2006) and both a linear and a quasi-linearised qua-dratic function (Kabbaj and Le Provost, 1980) were implementedfor bottom friction.

Using this model, the effect of the evolution of the harbour (i.e.,the construction of the docks 1–4) on the M2 tide, was determined.In addition to using different harbour geometries, the depth in thevarious docks was varied between 2 and 10 m as was the value ofthe bottom friction coefficient. For the latter it was important touse the quasi-linearised bottom friction law because it uses theChevy coefficient for which the range is well known and thereforeallows a good estimation of its effect. Neither the geometry of theharbour, the change in depth or bottom friction generated changesin the tidal amplitude larger than 0.1 mm nor did they cause signif-icant changes in the phase lag.

The assumption so far has considered a constant density insidethe harbour. However, if water run-off from the river Leça is signif-icant and/or if fresh water discharges from the Douro River reachthe Port, than this assumption no longer holds. Fresh water inthe harbour forced by saline tides outside the harbour will alterthe tidal amplitude observed at the tide gauge. A more elaboratetide model covering the dynamics of the shelf plus harbour is re-quired to fully understand the changes documented herein.Though interesting, this is beyond the scope of this study and isalso a limited exercise in view of the scarce (local) data availableto validate such a model.

Meteorologically induced trends for the Northern Iberian sta-tion, estimated by Marcos et al. (2005), suggested that in this re-gion the meteorological forcing slightly slows sea-level rise,whilst the thermosteric effect is seen as responsible for the positivetrends. Approximately 1/3 of difference between stations is attrib-uted to spatial differences in the meteorological forcing.

The series of sea levels obtained after removing the MSL and pre-dicted tide levels from the observed sea levels is defined here as thenon-tidal residual, also known as meteorological surge. Non-tidalresidual standard deviations have decreased by 0.16 ± 0.06 mm yr�1

and 0.32 ± 0.12 mm yr�1, during 1906–2008 and 1956–2008,respectively. This reduction can be interpreted as a decrease in er-rors in the sea level observations or a decrease in meteorologicalsurge influences on sea level. Assuming that the editing describedin Section 3.2 eliminated most errors then storm surges would beinfluencing the decrease in observed sea levels.

The decreasing tendency for the higher non-tidal percentile lev-els to decrease while lower percentiles increased (Table 3) is inagreement with the decrease found in the non-tidal residual stan-dard deviations. Results from the extreme level analysis suggestthat tides contribute less to the changes found in sea level thanweather or other oceanic forcing. The NAO index and inversebarometer have shown that prevailing weather patterns and atmo-spheric pressure contribute towards some of the sea levelvariability.

Local/regional and seasonal changes in the oceanic circulationinduced by winds and/or upper ocean density gradients may also

Please cite this article in press as: Araújo, I.B., et al. Analysing the 100 year sea lej.jhydrol.2012.12.019

be contributing to sea level change. Although the contributionfrom thermohaline changes induced by poleward flow of warmsaltier water from the south; nearshore circulation related to localfreshwater outflows (Douro plume) and Ekman effect (upwelling/downwelling) may help understand the results obtained herein,this goes beyond the scope of the present study. Further research,is needed to understand how sea level is being affected as a resultof steric and mass changes influenced by changes in the localatmospheric and oceanic dynamics. Wind setup and tide surgeinteractions should also be assessed.

The Leixões series exemplifies how interannual and longer vari-ations of sea level can have large effect on trends computed evenfrom long series. Although this record was long enough to deter-mine long-term trends in sea level, the underlying interannualand some longer variations are not fully understood to reassurecorrect trend estimates. Although the use of the current (available)data from Leixões is not recommended for long-term (secular)trend studies, this does not undermine the validity of these datafor climate prediction or seismic, tsunami and coastal studies, tak-ing into account the points stressed throughout this paper.

This first analysis of the Leixões record is limited by the currentknowledge involving the tide gauge station. Future use of this dataseries will benefit from any ancillary information brought to light,specially documenting the gauge location, characteristics and da-tum. Additionally, relevant information might also be achievedby digitising the nearby equally long dataset from the nearby Cant-areira tide gauge, located at the mouth of the Douro (�5 km). Thedirect comparison between Cantareira and Leixões records maycontribute to the understanding of the variability found in the la-ter. For shorter and more recent time-scales a cross check is possi-ble between Leixões and Viana do Castelo dastasets (�56 km Northof Leixões).

6. Conclusions

Historical sea level (1890–2004) from paper records of the Leix-ões tide gauge have been recovered.

MSL between 1906 and 2008 has decreased by 0.70 ±0.27 mm yr�1. This result is neither consistent with the globalmean sea level trend nor with regional neighbouring stations. Onthe assumption that the data was not influenced by vertical landmovement and datum shifts that would affect MSL, we analysedthe contribution of atmospheric forcing and possible changes tothe tides as a result of the significant changes that occurred tothe Leixões Port.

The NAO index and inverse barometer have shown that prevail-ing weather patterns and atmospheric pressure contribute to thesea level variability at this site.

The evolution of the port construction works and changes to thegauge and stilling contribute to a reduction in the observed sea le-vel standard deviation between 1956 and 2008. Significant inter-annual/decadal variability in annual mean sea levels after 1992 isalso probably associated with the changes to the gauge (in the lastdecades) and a change to the stilling well (2004). GIA rates and ver-tical land displacement estimates for this site are not significant.

The decrease in MSL is believed to be influenced by: (i) prevail-ing weather systems in the North Atlantic, especially in the winter,and local atmospheric pressure; (ii) by a decrease in observed levelstandard deviation; and (iii) changes to the gauge and stilling well,which have reduced the noise in the readings. Furthermore, theevolution of the port cannot be ignored when trying to understandsea level change.

Though beyond the scope of this work, further understandingand quantification of how sea level is affected by steric and masschanges influenced by changes in the local and North Atlanticatmospheric and oceanic dynamics, is needed.

vel record of Leixões, Portugal. J. Hydrol. (2013), http://dx.doi.org/10.1016/

I.B. Araújo et al. / Journal of Hydrology xxx (2013) xxx–xxx 9

The recovery of this dataset and its analysis has been useful indocumenting the potential sources of errors in the observed sealevels. Although we do not recommend that all the present (avail-able) data from Leixões be used in long-term (secular) trend stud-ies, these data are of great interest to climate, seismic and tsunamistudies. They should be used taking into account the pointsstressed throughout this paper.

Acknowledgements

We are grateful to the Port authority - Associação dos Portos doDouro e Leixões (APDL) and Eng. Brogueira for their interest andfinancing of the project which enabled the recovery of this histor-ical dataset. We would like to thank Eng. Miguel Lázaro (APDL), Dr.Joana Reis (Portuguese Hydrographic Office) and Eng. GonçaloCrisótomo (IGEO), for assisting with questions regarding the tidegauge; Eng. Carol Bos for developing the digitization software; Prof.Philip Woodworth and an anonymous reviewer for their valuablecomments which have helped improve this manuscript.

Support for the 20th Century Reanalysis Project dataset is pro-vided by the U.S. Department of Energy, Office of Science Innova-tive and Novel Computational Impact on Theory and Experiment(DOE INCITE) program, and Office of Biological and EnvironmentalResearch (BER), and by the National Oceanic and AtmosphericAdministration Climate Program Office.

This work has received support from FTC- Fundação para aCiência e Tencologia under PesT-C/MAR/LA0015/2011.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.jhydrol.2012.12.019.

References

Antunes, C., Taborda, R., 2009. Sea level at Cascais tide gauge: data, analysis andresults. J. Coast. Res. SI 56, 218–222.

Araújo, I.G.B., 2006. Sea level variability: examples from the Atlantic coast ofEurope. University of Southampton, Faculty of Engineering Science andMathematics, School of Ocean and Earth Science, Doctoral Thesis, p. 216.

Araújo, I.B., Pugh, D.T., 2008. Sea level at Newlyn 1915–2005: analysis of trends forfuture flooding risks. J. Coast. Res. 24, 203–212. http://dx.doi.org/10.2112/06-0785.1.

Araújo, I.B., Dias, J.M., Pugh, D.T., 2008. Model simulations of tidal changes in acoastal lagoon, the Ria de Aveiro (Portugal). Cont. Shelf Res. 28 (8), 1010–1025.http://dx.doi.org/10.1016/j.csr.2008.02.001.

Bell, C., Vassie, J.M., Woodworth, P.L., 2000. POL/PSMSL Tidal Analysis Software Kit(TASK Package). TASK-2000 version dated December 1998. <www.psmsl.org>.

Brockwell, P.J., Davis, R.A., 2002. Introduction to Time Series and Forecasting,second ed. Springer, New York, pp. 434.

Cartwright, D.E., 1972. Secular changes in the Oceanic Tides at Brest, 1711–1936.Geophys. J. Int. 30, 433–449.

Church, J.A., White, N.J., 2011. Sea-level rise from the late 19th to the early 21stCentury. Surv. Geophys. 32 (4–5), 586–602. http://dx.doi.org/10.1007/s10712-011-9119-1.

Colosi, J.A., Munk, W., 2006. Tales of the venerable Honolulu tide gauge. J. Phys.Oceanogr. 36, 967–996. http://dx.doi.org/10.1175/JPO2876.1.

Compo, G.P., Whitaker, J.S., Sardeshmukh, P.D., 2006. Feasibility of a 100 yrreanalysis using only surface pressure data. Bull. Am. Meteorol. Soc. 87, 175–190. http://dx.doi.org/10.1175/BAMS-87-2-175.

Compo, G.P., Whitaker, J.S., Sardeshmukh, P.D., Matsui, N., Allan, R.J., Yin, X.,Gleason, B.E., Vose, R.S., Rutledge, G., Bessemoulin, P., Bronnimann, S., Brunet,M., Crouthamel, R.I., Grant, A.N., Groisman, P.Y., Jones, P.D., Kruk, M., Kruger,A.C., Marshall, G.J., Maugeri, M., Mok, H.Y., Nordli, O., Ross, T.F., Trigo, R.M.,Wang, X.L., Woodruff, S.D., Worley, S.J., 2011. The 20th Century ReanalysisProject. Q. J. R. Meteorolog. Soc. 137, 1–28. http://dx.doi.org/10.1002/qj.776.

Egbert, G.D., Bennett, A.F., Foreman, M.G.G., 1994. TOPEX/POSEIDON tidesestimated using a global inverse model. J. Geophys. Res. 99, 24821–24852.http://dx.doi.org/10.1029/94JC01894.

Please cite this article in press as: Araújo, I.B., et al. Analysing the 100 year sea lej.jhydrol.2012.12.019

Ekman, M., 1988. The world’s longest continued series of sea level observations.Pure Appl. Geophys. 127 (1), 73–77. http://dx.doi.org/10.1007/BF00878691.

Fenoglio-Marc, L., Tel, E., Garcia, M.J., Kjaer, N., 2005. Interannual to decadal sealevel change in south-western Europe from satellite altimetry and in-situmeasurements. In: Jekeli, C., Bastos, L., Fernandes, J. (Eds.), Gravity, Geoid andSpace Missions. Springer, Berlin, Germany, pp. 242–247.

Holgate, S.J., 2007. On decadal rates of sea level change during the 20th Century.Geophys. Res. Lett. 34, L01602. http://dx.doi.org/10.1029/2006GL028492.

Hurrell, J.W., Deser, C., 2009. North Atlantic climate variability: the role of the NorthAtlantic oscillation. J. Mar. Syst. 78 (1), 28–41.

Hurrell, J. W., Kushnir, Y., Ottersen, G., Visbeck, M., 2003. The North AtlanticOscillation: Climate Significance and Environmental Impact, GeophysicalMonograph, 134. American Geophysical Union, Washington, USA, p. 279.

IOC, 1985. Manual on Sea Level Measurement and Interpretation, IOC Manuals andGuides, No. 14. Intergovernmental Oceanographic Commission, Paris, p. 83.

Kabbaj, A., Le Provost, C., 1980. Nonlinear tidal waves in channels: a perturbationmethod adapted to the importance of quadratic bottom friction. Tellus 32 (2),143–163. http://dx.doi.org/10.1111/j.2153-3490.1980.tb00942.x.

Lyard, F., Lefevre, F., Letellier, T., Francis, O., 2006. Modelling the global ocean tides:modern insights from FES2004. Ocean Dynam. 56, 394–415. http://dx.doi.org/10.1007/s10236-006-0086-x.

Marcos, M., Gomis, D., Monserrat, S., Álvarez-Fanjul, E., Pérez, B., García-Lafuente, J.,2005. Consistency of long sea-level time series in the northern coast of Spain. J.Geophys. Res. 110, C03008. http://dx.doi.org/10.1029/2004JC002522.

Marcos, M., Puyol, B., Wöppelmann, G., Herrero, C., García-Fernández, M.J., 2011.The long sea level record at Cadiz (southern Spain) from 1880 to 2009. J.Geophys. Res. –Atmos., 116, C12003, doi:10.1029/2011JC007558.

Mathers, E.L., Woodworth, P.L., 2001. Departures from the local inverse barometermodel observed in altimeter and tide gauge data and in a global barotropicnumerical model, J. Geophys. Res., 106(C4), pp. 6957–6972, doi:10.1029/2000JC000241.

Otero, P., Ruiz-Villarreal, M., Peliz, A., Cabanas, J.M., 2010. Climatology andreconstruction of runoff time series in northwest Iberia: influence in the shelfbuoyancy budget off Ría de Vigo. Sci. Mar. 74 (2), 1156–1180. http://dx.doi.org/10.3989/scimar.2010.74n2247.

Peltier, W.R., 2004. Global glacial isostasy and the surface of the ice-age earth: theICE-5G (VM2) model and GRACE. Ann. Rev. Earth Planet. Sci. 32, 111–149.http://dx.doi.org/10.1146/annurev.earth.32.082503.144359.

Pugh, D.T., 1987. Tides, surges and mean sea level: a handbook for engineers andscientists. John Wiley & Sons, Chichester, p. 472.

Reis, J.M., 2005: Caracterização da Maré do Porto de Leixões. Relatório de Progressode trabalhos, REL.PT OC 05/05, Instituto Hidrográfico, Lisbon, p. 45.

Sousa, J., Hooper, A., Hanssen, R., Bastos, L., Ruiz, A., 2011. Detection of grounddeformation in the Porto metropolitan area with persistent scatterinterferometry (PSI). In: FRINGE 2011 Workshop, Advances in the Science andApplications of SAR Interferometry and Sentinel-1 Preparatory Workshop,September 2011, ESA-ESRIN, Frascati (Rome), Italy.

Trigo, R.M., Pozo-Vázquez, D., Osborn, T.J., Castro-Diez, Y., Gámiz-Fortis, S., Esteban-Parra, M.J., 2004. North Atlantic oscillation influence on precipitation, river flowand water resources in the Iberian Peninsula. Int. J. Climatol. 24, 925–944.http://dx.doi.org/10.1002/joc.1048.

van Veen, J., 1954. Tide-gauges, subsidence-gauges and flood-stones in TheNetherlands. Geol. Mijnbouw (Nw. Ser.) 16, pp. 214–219.

Wakelin, S.L., Woodworth, P.L., Flather, R.A., Williams, J.A., 2003. Sea-leveldependence on the NAO over the NW European continental shelf. Geophys.Res. Lett. 30 (7), 1403. http://dx.doi.org/10.1029/2003GL017041.

Whitaker, J.S., Compo, G.P., Wei, X., Hamill, T.M., 2004. Reanalysis withoutradiosondes using ensemble data assimilation. Mon. Weather. Rev. 132, 1190–1200, http://dx.doi.org/10.1175/1520-0493(2004) 132<1190:RWRUED>2.0.CO.

Woodworth, P.L., 1999. High waters at liverpool since 1768: the UK’s longest sealevel record. Geophys. Res. Lett. 26 (11), 1589–1592. http://dx.doi.org/10.1029/1999GL900323.

Woodworth, P.L., Blackman, B.L., 2004. Evidence for systematic changes in extremehigh waters since the mid-1970s. J. Clim. 17, 1190–1197, http://dx.doi.org/10.1175/1520-0442(2004) 017<1190:EFSCIE>2.0.CO;2.

Woodworth, P.L., Player, R., 2003. The permanent service for mean sea level: anupdate to the 21st century. J. Coast. Res. 19 (2), 287–295.

Woodworth, P.L., Gehrels, W.R., Nerem, R.S., 2011. Nineteenth and 20th centurychanges in sea level. Oceanography 24 (2), 80–93. http://dx.doi.org/10.5670/oceanog.2011.29.

Woolf, D.K., Shaw, A.G.P., Tsimplis, M.N., 2003. The influence of the North Atlanticoscillation on sea level variability in the North Atlantic region. J. Atmos. OceanSci. 9 (4), 145–167. http://dx.doi.org/10.1080/10236730310001633803.

Wöppelmann, G., Marcos, M., 2012. Coastal sea level rise in southern Europe andthe nonclimate contribution of vertical land motion. J. Geophys. Res. 117,C01007. http://dx.doi.org/10.1029/2011JC007469.

Wöppelmann, G., Pouvreau, N., Simon, B., 2006. Brest sea level record: a time seriesconstruction back to the early 18th century. Ocean Dynam. 56 (5–6), 487–497.http://dx.doi.org/10.1007/s10236-005-0044-z.

vel record of Leixões, Portugal. J. Hydrol. (2013), http://dx.doi.org/10.1016/