www.cerf-jcr.org

The Effect of the 18.6-Year Lunar Nodal Cycle on Regional

Sea-Level Rise Estimates

Fedor Baart

{{

, Pieter H.A.J.M. van Gelder

{

, John de Ronde

{{

, Mark van Koningsveld

{1

,

and Bert Wouters

{{

{

Department of Hydraulic

Engineering

Faculty of Civil Engineering

and Geosciences

Delft University of Technology

Stevinweg 1

2628 CN Delft,

The Netherlands

{

Department Marine and

Coastal Systems

Deltares

P.O. Box 177, 2600 MH Delft,

The Netherlands

1

Department of Environmental

Engineering

Van Oord

P.O. Box 8574, 3009 AN

Rotterdam, The Netherlands

{{

Department of Climate

Research and Seismology

Royal Netherlands

Meteorological Institute

P.O. Box 201, 3730 AE De Bilt,

The Netherlands

ABSTRACT

BAART, F.; VAN GELDER, P.H.A.J.M.; DE RONDE, J.; VAN KONINGSVELD, M., and WOUTERS, B., 2012. The effect

of the 18.6-year lunar nodal cycle on regional sea-level rise estimates. Journal of Coastal Research, 28(2), 511–516. West

Palm Beach (Florida), ISSN 0749-0208.

Sea-level rise rates have become important drivers for policy makers dealing with the long-term protection of coastal

populations. Scenario studies suggest that an acceleration in sea-level rise is imminent. The anticipated acceleration is

hard to detect because of spatial and temporal variability, which consequently, have become important research topics. A

known decadal-scale variation is the 18.6-year nodal cycle. Here, we show how failing to account for the nodal cycle

resulted in an overestimation of Dutch sea-level rise. The nodal cycle is present across the globe with a varying phase and

a median amplitude of 2.2 cm. Accounting for the nodal cycle increases the probability of detecting acceleration in the

rate of sea-level rise. In an analysis of the Dutch coast, however, still no significant acceleration was found. The nodal

cycle causes sea level to drop or to rise at an increased rate; therefore, accounting for it is crucial to accurately estimate

regional sea-level rise.

ADDITIONAL INDEX WORDS: Sea level, subsidence, decadal, tide, trend estimate.

INTRODUCTION

The current and expected rates of sea-level rise are

important drivers for policy makers dealing with the long-

term protection of coastal areas and populations. An example of

an area where sea-level rise is important is the Dutch coast.

There are several measures planned to deal with the expected

acceleration in sea-level rise, which will cost up to J1.6 billion

y

21

until 2050 (Kabat et al., 2009). The long history of tidal

records and the economic value of the area below sea level make

the Dutch coast an interesting case for analyzing sea-level

measurements and scenarios and for comparing local estimates

with global estimates.

LOCAL TRENDS

Sea-level changes are usually reported in the form of trends,

often determined over a period of one or more decades. For The

Netherlands, an important trend was reported after the 1953

flood, when a relative sea-level rise of 0.15 to 0.20 cm y

21

was

estimated for the design of the Delta Works. The first Delta

Committee report (Deltacommissie, 1960) referred to this

change rate as ‘‘relative land subsidence’’. Relative sea level,

the current term, is the sea-level elevation relative to the

continental crust as measured by tide gauges. Absolute sea

level is relative to a reference ellipsoid and is measured by

satellites. A recent estimate (van den Hurk et al., 2007) showed

that relative sea level rose at a rate of 0.27 cm y

21

during the

period 1990–2005. The land subsidence at the Dutch coast

varies around 0.04 60.09 cm y

21

(mean 6standard error of the

mean [SEM]; Kooi et al., 1998).

Local Forecasts

Coastal policy is shifting from observation-based reactions to

scenario-based anticipation (Ministerie van Verkeer en Water-

staat, 2009); it is, therefore, interesting to compare observed

trends with predicted rates. Sea-level scenarios often predict

not only a sea-level rise but also an accelerated rise. The

earliest Dutch scenario, published after the 1953 storm,

forecasted a rise of several meters due to Greenland ice melting

over an unspecified period (Deltacommissie, 1960). Van

Dantzig (1956) used a more concrete number of 70 cm for the

next century in a related publication. The latest study by the

Royal Netherlands Meteorological Institute (KNMI) (van den

Hurk et al., 2007; Katsman et al., 2008) resulted in a low and a

high scenario. The low scenario estimates a rise of 0.25 cm y

21

in the period 1990 through 2050 and 0.32 cm y

21

for the period

DOI: 10.2112/JCOASTRES-D-11-00169.1 received 19 September

2011; accepted in revision 20 September 2011.

Published Pre-print online 15 December 2011.

’Coastal Education & Research Foundation 2012

Journal of Coastal Research 28 2 511–516 West Palm Beach, Florida March 2012

2050 through 2100. The high scenario predicts 0.58 cm y

21

and

0.77 cm y

21

for the same periods. A high-end estimate of

2.02 cm y

21

was reported by the second Delta Committee in

2008, based on the Intergovernmental Panel on Climate

Change (IPCC) A1FI scenario for the period 2050 through

2100 (Deltacommissie, 2008, see figure 4, page 24). This

extreme scenario was used to assess the sustainability of the

Dutch coastal policy.

Global Trends

The global measurement of relative sea level started in 1933

when the Permanent Service for Mean Sea Level (PSMSL)

began collecting sea-level data from the global network of tide

gauges (Woodworth and Player, 2003). Trends based on those

measurements vary around 0.17 cm y

21

. For example, Holgate

(2007) reported a 0.145 cm y

21

over the period 1954–2003 and

Church et al. (2008) reported 0.18 cm y

21

over the period 1961–

2003. With the launchof the TOPEX/Poseidon satellite in 1992,

measurements of absolute sea level became available, with

near global coverage and high resolution in time and space.

Those measurements were used in the latest estimates,

summarized in the IPCC report (Bindoff et al., 2007), giving a

0.31 cm y

21

absolute sea-level rise over the period 1993–2003.

Despite the apparent difference, tidal-station measurements

compare well with satellite data when accounting for correc-

tions, start of time window, and the geographical location

(Prandi, Cazenave, and Becker, 2009).

Global Forecasts

Of the global scenarios for future sea-level rise, the most

influential are the current model-based IPCC scenarios (Bind-

off et al., 2007). The estimated rise varies between 0.17 cm y

21

(lower B1) and 0.56 cm y

21

(higher A1FI) over the period 1980–

1999 through 2090–2099 (Meehl et al., 2007). All scenarios

result in a most likely sea-level rise that is higher than the

average rate of 0.18 cm y

21

over the period 1961 to 2003.

Detecting Acceleration

Even though sea-level rise acceleration was expected to

become apparent in the early years of this century (Woodworth,

1990), there is presently no overall, statistically significant

acceleration, other than that in the early 20th century (Church

and White, 2006; Jevrejeva et al., 2008). The probability of

detecting an acceleration in sea-level rise is low because of the

effect of decadal variations (Douglas, 1992; Holgate, 2007).

Accounting for decadal variations can, therefore, enhance our

ability to detect acceleration.

The Nodal Cycle

One such decadal variation is the lunar nodal cycle. The tide

on the Earth is driven by six different forcing components with

periods varying from 1 day to 20,940 years. The fifth component

is the 18.6-yearly lunar nodal cycle (Doodson, 1921). The term

nodal cycle is best explained while looking up from the Earth.

Consider the node as the intersection of the ecliptic plane,

which follows the path of the Sun, and the orbital plane, which

follows the path of the Moon. This node moves westward,

making a circle every 18.6 years.

The main effect of this cycle is that it influences the tidal

amplitude (Woodworth, 1999; Gratiot et al., 2008). There are

indications that the 18.6 yearly cycle also influences regional

mean sea level, for example, at the Dutch coast (Dillingh et al.,

1993) and at a collection of other tidal stations (Houston and

Dean, 2011; Lisitzin, 1957). Global variation studies on tide

gauges using spectral analysis by Trupin and Wahr (1990) and

on satellite data using harmonic analysis (Cherniawsky et al.,

2010) also indicate a cycle in regional mean sea levels.

Observed tide is often compared with the equilibrium tide. The

equilibrium tide is the tide that would exist if the earth were

completely covered by water and if there were no friction. The

equilibrium tide theory builds on the work of Doodson (1921),

Cartwright and Tayler (1971), and Cartwright and Edden (1973).

Following Rossiter (1967), we used Equation (1) for the

equilibrium elevation fand the resulting nodal amplitude A(in

millimeters), with the Mmass of the moon in kilograms, the E

mass of the earth in kilograms, the emean radius of the earth

(in kilometers), the rmean distance between the earth center

and the moon center (in kilometers), the llatitude in radians,

and the Nlongitude of the Moon’s’ ascending node (from ’18u189

to ’28u369). The phase wis 0ufor |l|$35.3uand 180ufor |l|

,35.3u.

f~9

8

M

Eee

r

3

sin2l{1

3

cos N0|0:06552A~26:3 sin2l{1

3

M~5:9736e24

E~7:3477e22

e~6,371,000

r~384,403,000

ð1Þ

Proudman (1960) showed that the nodal tide should follow

the equilibrium tide for friction. The earth tide should also be

taken into account. Rossiter (1967) corrected by a factor of 0.7 to

allow for the effect of a yielding Earth. This is also the approach

used by Pugh (1987) and Cherniawsky et al. (2010). The

correction factor is based on the combined effect of the change

in the height of the equilibrium level above the solid earth,

given by the formula 1 2k2h(V

p

/g), where kand hare the

Love numbers (Love, 1909). The elastic response of the earth

has an amplitude of hV

p

/g, where his a known elastic constant,

V

p

/gis the gravitational potential, and gis the gravity constant.

When the tidal periods become longer, not only the elastic

response but also the viscose response is important, and,

therefore, the factor of 0.7 may not be appropriate (Pugh, 1987).

For regional sea-level rise estimates, the spatial variability of

the nodal cycle is relevant. This spatial variability is also

relevant for estimating the global mean sea level. The global

mean sea level itself is not affected by this cycle, but trend

estimates can be affected because both tide gauges and the

satellites have limited coverage of the world. Tide gauges have

higher coverage in the Northern Hemisphere, and the

altimetry satellites only cover the area between 264uand 64u.

Examining the agreement with the equilibrium tide and the

observed nodal cycle is relevant because it determines the best

512 Baart et al.

Journal of Coastal Research, Vol. 28, No. 2, 2012

method to estimate the local effects of the nodal cycle. Previous

comparisons with the equilibrium tide (e.g., Currie, 1976;

Trupin and Wahr, 1990) have shown agreement.

Accounting for the nodal cycle should increase the probabil-

ity of finding acceleration or deceleration in the rate of sea-level

rise (Baart et al., 2010; Houston and Dean, 2011). In this

article, we determine whether accounting for the nodal cycle

affects sea-level rise estimates locally and analyze how the

nodal cycle varies across the globe.

METHODS

The phase and amplitude of the nodal cycle are estimated by

multiple linear regression using Equation (2). Variable tis time

in Julian years (365.25 d) since 1970, b

0

is the initial mean sea

level (in centimeters), b

1

is the rise (centimeters per year), and a

and bcan be transformed into the amplitude A~ﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ

a2zb2

p(in

centimeters) and the phase w5(arctan a/b) in radians.

Acceleration is tested by comparing the regression model with

the quadratic term b

2

(centimeters per year

2

)withthe

regression model without the quadratic term.

htðÞ~b0

|{z}

mean level

zb1t

|{z}

trend

zb2t2

|ﬄﬄﬄ{zﬄﬄﬄ}

acceleration

zasin 2pt

18:6

zbcos 2pt

18:6

|ﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄ{zﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄ}

nodal cycle

:ð2Þ

We used a spectral analysis only to determine whether the

nodal cycle was the most dominant signal in the spectrum for

cycles with a period greater than a year. The stacking method

was not used because the ‘‘‘‘detrending before fitting the cycle’’’’

approach leads to an underestimate of the amplitude when the

time-series length is not several times as long as the 18.6-year

period. Therefore, we used the same harmonic analysis

approach as Battjes and Gerritsen (2002) and Houston and

Dean (2011).

RESULTS

Local Relative Sea Level

To determine the relevance of the nodal cycle at the Dutch

coast, a spectral analysis was carried out on the yearly means of

six main tidal gauges for the period 1890–2008. The data were

corrected for atmospheric pressure variation using an inverse

barometer correction. The spectral density shows a clear peak

at the 18.6-year period (Figure 1). The multiple linear

regression yields a sea-level rise (b

1

) of 0.19 60.015 cm y

21

(95%), an amplitude (A) of 1.2 60.92 cm, and a phase (w)of

21.16 (with 1970 as 0), resulting in a peak in February 2005

(Figure 2). No significant acceleration (inclusion of b

2

) was

found.

Probability of Acceleration Detection

The nodal cycle explains 9% of the variance in the detrended

mean sea level. Explaining more variance has the advantage

that other effects become clearer. We used this process to

determine the change in the probability of detecting an

acceleration in the rate of sea-level rise. This probability, the

statistical power, is calculated for the lower and higher KNMI

Figure 1. Spectral analysis of the mean of six tidal stations in the period

1890–2008. The dashed line marks 18.6 year.

Figure 2. Annual mean sea level averaged over six Dutch tidal stations

(black dots). Multiple linear regression with a nodal cycle (solid curve),

with a confidence interval (dotted curve), and a prediction interval (dashed

curve). Linear regression line through the period 1890–1990 (light gray).

Linear regression line through the period 1991–2009 (dark gray).

Sea-Level Rise and the Lunar Nodal Cycle 513

Journal of Coastal Research, Vol. 28, No. 2, 2012

scenarios (van den Hurk et al., 2007) for the Dutch coast. The

power was estimated using a simulation, with a generated data

set, based on the broken linear trends from the scenarios in van

den Hurk et al. (2007). In addition, the Dutch nodal cycle was

imposed but with a random, uniform distributed phase, as well

as a random, normal distributed error, based on the residuals

after fitting the nodal cycle for the mean of the Dutch tidal

station measurements. The simulation was performed with

200 samples per condition. The detection of the acceleration

was done by comparing the linear-regression model with a

model with an acceleration term included using an analysis of

variance (ANOVA) with 1 degree of freedom. The probability of

detecting sea-level acceleration for the lower scenario went up

from 46% without the nodal cycle to 48% with the nodal cycle.

The probability of detection in the high-end scenario went up

from 82% without the nodal cycle to 84% with the nodal cycle

included. Generally, 80% is considered an acceptable level.

Thus, it can be concluded that, even without accounting for the

nodal cycle, it is likely that the acceleration in the higher

scenario, if it were present, would have been found.

Local Absolute Sea Level

Repeating the previous analysis on the North Sea satellite

data yielded the same nodal cycle (Figure 3). By including the

nodal cycle, the absolute sea-level rise lowers from 0.23 cm y

21

to 0.07 cm y

21

because, coincidently, the time window starts at

the bottom and ends in the peak of the nodal cycle. This clearly

shows how including the nodal cycle may affect estimates of

sea-level rise.

Global Relative Sea Level

Now that it is known that the nodal cycle is important for

estimates of local sea-level rise, the next question is how the

nodal cycle varies across the globe. The variation in global

relative sea-level was analyzed using the PSMSL tidal gauge

data set. From the 1157 gauges, 511 were selected based on

their recorded history of at least 57 (3 319) years. The analysis

of the spectral densities at the tidal stations was skipped

because it has already been performed in detail (Trupin and

Wahr, 1990), showing a peak at 18.6 years.

Equation (2) was applied to the selected stations, in which 134

stations showed an amplitude (A) that was significantly different

from 0. This confirms the global presence of the effect of the

lunar nodal cycle, with a median amplitude of 2.2. The variation

in global phase and amplitudes are shown in Figure 4.

Global Absolute Sea Level

The phases found at the tidal stations were compared with the

phases found in nearby measurements from altimetry satellites

for verification. This data set was obtained from the Common-

wealth Scientific and Industrial Research Organisation web site

and consists of sea surface heights with inverse barometer (IB)

corrections, seasonal signals removed, and glacial isostatic

adjustments corrected. Because satellite data are only available

for one lunar nodal period, the results are susceptible to other

influences and are not yet stable. The variation in global phase is

plotted in Figure 4. Tidal gauge and satellite measurements

show a reasonable correspondence in the Atlantic Ocean but not

in the Pacific Ocean. The canonical correlation between the

amplitude and the phases of stations and of the nearby satellites

is 0.21, which is low yet statistically significant.

When to Include the Nodal Cycle

The two extra parameters, amplitude and phase, can result

in a less-accurate estimate of the sea-level rise parameter. One

way to approach this is by determining whether the variance

explained by the combination of the two extra parameters is

statistically significant (using an ANOVA with 2 degrees of

freedom). An alternative is to use the Akaike information

criterion.

Another simulation provides a general estimate of what

would be a good period for including a nodal cycle in estimates

of regional sea level. Here, we assumed a sea-level rise equal to

0.2 cm y

21

, a nodal amplitude of 2.2 cm, and a uniform

distributed random phase and a random error of 2.5 cm. By

varying the time period and comparing the root mean square

error of the estimate of nodal fit and linear fit, we find that,

with these conditions, it is useful to include the nodal cycle

terms starting with periods of 14 years and longer. This period

for which it is advisable to include the nodal cycle becomes

longer because the ratio between them is a function of the

random error and the amplitude of the nodal cycle increases. A

Figure 3. Absolute sea level in the North Sea. (Top figure) Linear

regression fitted through corrected satellite observations for the North Sea

from Topex (o), Jason1(n), and Jason2 (+). Dashed line represents the

confidence interval; dotted line represents the prediction interval. (Lower

figure) Seasonal regression (Equation 2) fitted through corrected satellite

observations for the North Sea from Topex (o), Jason1(n), and Jason2 (+).

Dashed line represents the confidence interval; dotted line represents the

prediction interval.

514 Baart et al.

Journal of Coastal Research, Vol. 28, No. 2, 2012

similar discussion can be found in Blewitt and Lavalle

´e (2002)

for the comparable problem of fitting geodetic velocities.

If the goal is to develop an unbiased local estimate of the sea-

level rise parameter, the simplest approach is to use time series

of multiples of 18.6 +9.3 years (integer plus a half). When the

goal is to develop a good estimate of the level or acceleration,

this approach cannot be used.

The goal of including the nodal tide is to fit the nodal cycle, not

other decadal cycles. Therefore, the nodal cycle should only be

incuded in the 18.6 year is within the modal frequency bin of the

multi-year spectrum. In addition, it is advisable to check for the

reliability of the fit, by, for example, splitting up the tidal signal

into two separate parts, which should yield the same nodal cycle

phase. If the estimate of the nodal cycle is based on satellite

measurements, it should be verified using local tide gauges.

CONCLUSIONS

Coastal management requires estimates of the rate of sea-

level rise. The trends found locally for the Dutch coast are the

same as have been found in the past 50 years (Deltacommissie,

1960; Dillingh et al., 1993). Even though including the nodal

cycle made it more likely that the high-level scenarios would

become apparent in the observations, no acceleration in the

rate of sea-level rise was found. The higher, recent rise (van den

Hurk et al., 2007) coincides with theup phase of the nodal cycle.

For the period 2005 through 2011, the Dutch mean sea-level is

expected to drop because the lunar cycle is in the down phase.

This shows the importance of including the 18.6-year cycle in

regional sea-level estimates. Not doing so on a regional or local

scale for decadal length projections leads to inaccuracies.

There is a difference between the nodal cycle phase expected

from the equilibrium and the nodal cycle phase found in tidal

records. This is inconsistent with the results from Trupin and

Wahr (1990), possibly because of the difference between the

stacking approach and the harmonic approach. The difference

here is similar to the differences found by Cherniawsky et al.

(2010). The cause for the difference between the observed nodal

cycle and the equilibrium nodal cycle is not known. It could be a

physical effect but could also be the result of the way our mean

sea levels are measured and computed.

Whatever the cause, if there is a known decadal signal in the

sea-level records, it should be taken into account. Doing so will

provide better estimates of local sea-level rise, but only if it is

determined that the nodal fit is clearly present.

Although the nodal tide does not affect the true global mean

sea level, it can affect global mean sea level estimates. In sea-

level trends from satellites, if one assumes the equilibrium

nodal phase, one would expect a small nodal cycle in the mean

because of the phase distribution in combination with the

limited spatial coverage of the altimetry satellite. The sea-level

trends based on tide gauges can also be affected by the nodal

cycle because of irregular spatial sampling of the tidal gauges.

The observed nodal cycle shows a pattern that is more E–W,

rather than the equator–poles pattern of the equilibrium. The

nodal cycle can thus be safely ignored for global mean sea-level

estimates based on satellites. For global mean sea-level

estimates based on tidal gauges, the distribution of nodal cycle

phases could be checked for approximate, circular uniformness.

Globally, the 18.6-year cycle is observable in one-fourth of the

selected tidal stations, with a varying phase. The phases found,

based on tidal records and satellite data, show a weak

association, probably because of the short period of the satellite

measurements. It is not yet possible to give an accurate

estimate of the effect of the cycle across the globe. Just like a

sea-level rise trend can be very sensitive to the window of

observation, an estimate of a cycle is highly sensitive to

peaks. Without removing such effects, for example, the El

Nin

˜o–Southern Oscillation, short series like the satellite

measurements are not very representative of the effect of the

nodal cycle. A logical follow-up to this research would be to

simulate the effect of the nodal cycle using a global tide model.

Figure 4. Nodal cycle, estimated using Equation (2). Amplitudes in cm

(size of circles) and phases in years (color) of the lunar nodal cycle. (Top

figure) Tide gauges with at least 57 y of measurements. (Middle figure)

Altimetry satellites. (Bottom figure) Equilibrium.

Sea-Level Rise and the Lunar Nodal Cycle 515

Journal of Coastal Research, Vol. 28, No. 2, 2012