Article of the Month -
The Difficulties in Using Tide Gauges to Monitor
Long-Term Sea Level Change
John HANNAH, New Zealand
This article in .pdf-format (pdf, 103
1) This paper has been presented at the XXIV FIG
Congress in Sydney 11-16 April 2010 in the session on Vertical Reference
Frame. The paper is a peer reviewed paper.
Handouts of this presentation as a .pdf file. John Hannah is a
Registered Professional Surveyor and is a former President of the NZ
Institute of Surveyors.
Climate change has a variety of important impacts, one of which is
reflected in sea levels. Indeed, long term rising trends in global sea
levels are often used to corroborate the assertion of long term climate
change. When tide gauge records are examined in order to determine the
long-term trends in sea level it is typical for a single number
representing the derived trend, to be quoted. However, the problems in
deriving such numbers are rarely, if ever, discussed. Indeed, there
appears to be a widespread ignorance as to the fragility of tide gauge
records and hence the accuracy of derived long-term sea level trends.
This paper uses specific examples from New Zealand to illustrate and
explain the problems that exist in deriving an accurate figure for the
eustatic changes in sea-level at a tide gauge site. It highlights the
importance of assessing accurately the influence of anthropological
factors, changes in tide gauge datums, and geophysical effects. These
factors, which can compromise or even completely invalidate a record,
must be able to be assessed over the entire history of the tide gauge
record (often 100+ years). This paper, after exploring these factors and
their potential influence, concludes by making recommendations on
procedures to be followed if we are to leave future generations better
quality sea level data than is often available at present.
Sea level change is an important climate-related signal, studies of
which have featured in all recent International Panel for Climate Change
(IPCC) scientific assessments (e.g., IPCC, 2001; IPCC, 2007). In
undertaking sea level change analyses, the data is typically drawn from
the Permanent Service for Mean sea Level (PSMSL) database at the
Proudman Oceanographic Laboratory. For each tide gauge this data is used
to derive a figure for sea level rise. In order to correct the derived
figure so that it reflects the eustatic component of sea level rise, a
great deal of attention has been given to the task of separating the
motion of the land and wharf structures (to which the tide gauge is
attached), from the observed sea level rise signal. This has resulted in
the increasingly widespread collocation of GPS receivers with tide
gauges (c.f., Woppelman, 2007). In addition to these land based studies,
satellite altimetry has advanced to the point whereby TOPEX and JASON 1
time series are now being used to assess long term sea level changes
over the open oceans. Such studies, while separate from the land based
tide gauge studies, are not independent in that the data from certain
coastal tide gauges have been used for altimeter calibration purposes
(e.g., Chambers et al, 1998; Nerem and Mitchum, 2001).
In nearly all of these studies, the tide gauge data is typically
assumed to be high quality and not subject to question. This important
assumption is rarely, if ever challenged. However, if such a high
quality record is to be obtained it is essential that issues such as the
datum history of the tide gauge and local wharf movements be well
documented and verified. Given that many gauges are located in port
facilities where wharf removal, development, and/or extensions occur,
this is easier said than done. Indeed, New Zealand experience indicates
that some primary gauges have been renewed, replaced or changed at least
five or six times in their 100 year history. In addition, it is not
uncommon for tide gauges to malfunction for significant periods of time
thus offering the possibility of a (potentially) biased tidal record.
This study, then, attempts to highlight the importance of the above
factors, giving specific examples of how the analyses of New Zealand’s
long term sea level trends have been influenced by them and illustrating
how a record can be invalidated by poor information. While the examples
have been drawn from New Zealand experience, they illustrate problems
that are generic in nature to much of the available tide gauge data.
Other issues that arise in long-term sea level change analyses
include the influence of geophysical effects and the length of the tide
gauge record, Douglas (1997) pointing out that a gauge record needs to
be at least 60 years in length if incorrect estimates of sea level
change are to be avoided.
The paper concludes by making some practical recommendations on
procedures to be followed if future generations of investigators are to
be left higher quality, long term data sets than are currently
2. POSSIBLE ERRORS IN THE SEA LEVEL ANALYSIS PROCESS
2.1 Tide Gauge Errors
Most pre-1980 sea level data have been collected from float activated
tide gauges, an example of which is shown in Figure 1 below. The sea
enters the float chamber via a small orifice at the bottom. The orifice
(and chamber), act as a mechanical low-pass filter that eliminates high
frequency wave action. As the float rises and falls with the tide, it is
connected by a pulley mechanism to a mechanical recording device that in
turn causes the tidal signal to be traced out on a paper graph. The
paper graph rotates (typically on a drum), that is in turn driven by a
mechanical clock. Much of the historical tide gauge data held by the
PSMSL will have been collected off paper charts produced by such
Figure 1. Float Activated Tide Gauge (Hydrographer of the Navy, 1969)
These particular gauges have a number of well known error sources,
- Sediment collecting in the bottom of the stilling well. This was
often evidenced by a flattened low water tidal curve – the float
would sit on the sediment at or near low water and fail to delineate
the change in the low water tide. Due to the biases likely to be
introduced, data demonstrating this behavior needs to be rejected
from any subsequent sea level trend analysis.
- Friction in the float mechanism. The tidal constituents used in
harmonic analysis are comprised of the amplitude and phase of a
combination of cosine functions superimposed upon each other
(cf.,Pugh, 2004). The rate of change in the tide is zero at high and
low water and very close to zero in the near proximity. Excess
friction in the float mechanism (perhaps due to lack of maintenance)
results in the float ”sticking” at these points thus creating an
arbitrary flattening of the tidal curves at both high and low tide.
In the recent digitization of the sea level data at New Plymouth,
for example, this lack of definition was sometimes observed for days
on end. Unless there is a symmetry in these effects about the mean
tide, data demonstrating this type of behaviour should also be
- Clock errors. This problem arises either when the clock has been
incorrectly set or because it drifts with time. It is not uncommon
to find errors of one hour or more in the record. In mid-latitude
regions, such as New Zealand, this can easily produce a difference
between the observed tide (as delineated on the tide chart), and the
predicted tide of 700 mm – 800 mm. In reality, however, these timing
errors have little influence on the monthly and annual sea level
means that are used in long-term sea level trend studies.
- Gauge setting errors. The traditional mechanism for calibrating
a float activated tide gauge was to observe the water level on the
tide pole adjacent to the gauge and then to ensure that this reading
was reflected on the chart record. Some old gauges (e.g., Foxboro
gauges) could be set to little better than 0.2 ft. Setting errors of
0.2 ft – 0.3 ft (0.06 m – 0.09 m) appear to have been reasonably
common. Such settings would typically occur when the paper tide
graphs (rolls) were changed (i.e., anywhere between every two weeks
and two months). Assuming a standard deviation for the gauge setting
of 0.25 ft, (0.076 m) and a setting interval of one month, then the
contribution of this error to the standard deviation associated with
an annual MSL could be expected to be in the order of 0.022 m.
In recent years mechanical tides gauges have been replaced by
electronic tide gauges whereby the sea level is sensed by bouncing an
electronic signal off the water surface to an associated sensor unit
either above or below. Typically, a burst of readings are made over a
short time period (say 30 sec) and a mean figure for the sea surface
obtained. This occurs at some predetermined interval such as 10 minutes.
Such gauges are thus capable of producing much higher frequency, and
more precise data than was possible from the old tide charts. Other
gauge types exist such as pressure gauges and “bubbler” type gauges.
Douglas (2001) provides a brief summary of these.
However, even with electronic gauges, the calibration problem
(equivalent to the gauge setting error) remains. In addition some, such
as the quartz crystal pressure gauges, can drift severely with time.
Indeed, New Zealand experience with one such pressure gauge at Cape
Roberts in the Antarctic has shown that a calibration interval of two to
three years is inappropriate – the data being so contaminated with drift
errors as to be essentially unusable. While a calibration period of at
least six months is preferred, logistical constraints have limited the
calibration of the Cape Roberts gauge to 12 monthly intervals.
New Zealand experience further indicates that the most important
issue in obtaining high quality monthly or annual MSL data is the care
and maintenance of the gauge. Poor maintenance is often indicated by
long periods of gauge outage, frequent breaks in the tidal record,
timing errors and poor curve definition at high and/or low water. Where
a gauge has been well maintained (such as with the Auckland and
Wellington gauges), a posteriori error analysis undertaken on the full
sets of digital data collected over 100 years indicate that an annual
sea level means should be able to be given a standard deviation of
between 0.020 m and 0.025 m (Hannah, 2004).
2.2 Datum Errors
In attempting to derive a long-term sea level trend, datum errors,
generally arising from anthropological factors, are by far the most
important to resolve. Unlike gauge errors that are greatly reduced by
the quantity of data collected and the resulting meaning process, datum
errors can be subtle, tend to be systematic and, if not correctly
resolved, will completely invalidate a sea level record. Such errors can
arise from the following sources.
- When tide gauges are shifted from one wharf structure to another
and the new gauge zero differs by some unknown (or unrecorded)
quantity from the previous gauge zero. In recent attempts to
reconstruct the tidal record at New Plymouth it has become apparent
that the tide gauge had been moved from one wharf to another at
least four times since 1918. In the case of the Wellington gauge,
written records indicate that the gauge was moved between 1944 and
1945, but there is no record of a datum shift. The MSL data before
and after that date indicate with some clarity that such a datum
shift occurred (c.f., Hannah, 1990). An analysis of the data
indicated that the shift was in the order of 0.025 m.
- When a tide pole is replaced and the new pole is set at a
different level than the previous one. When it remembered that the
tide pole is the means by which tide gauges have historically been
calibrated, then it becomes clear that any unrecorded shift in the
tide pole will immediately translate into an unrecorded datum
change. Tide poles, which are attached to wharf structures, can
easily be damaged by vessels and are often obliterated by oil and
other port debris. It is likely that even a well built tide pole
will require replacement on a 20 year cycle. A recent detailed
analysis of the records relating to the well maintained Lyttelton
gauge, indicates unrecorded variations in the position of the tide
pole of 0.08 ft (0.024 m) over a 40-year period. The dates when
specific changes occurred are not known. In reality the tide pole is
the fragile link that holds a tidal record together. If the position
of the tide pole has not been monitored throughout the history of
the tidal record then the record must be subject to question as must
the accuracy of any subsequent long-term sea level analysis.
- When there is no consistent history of leveling from stable
benchmarks to the tide pole. Any local subsidence in a wharf
structure (and thus in the attached tide pole or tide gauge) will
only be detected if there has been a consistent history of leveling
to stable local benchmarks. For example, for many years, and in
earlier sea level analyses the Wellington gauge was assumed stable
(Hannah, 1990). However, by 2003, a sufficiently long time series of
local leveling data had been collected so as to indicate an apparent
long-term subsidence in the wharf structures of about 0.15 mm/yr
(Hannah, 2004). Conversely, at Dunedin, it has only become clear
recently that certain local bench marks are subsiding while the
wharf structures remain stable. The 2004 analysis of long-term sea
level change, which assumed both were subsiding, gave a result of a
sea level rise of 0.94 mm/yr (Hannah, 2004). The most recent
analysis (with this erroneous assumption corrected), now shows a sea
level rise of 1.3 mm/yr – a very significant difference.
- Changes in the setting of the gauge datum. It is altogether
possible that a gauge may exhibit none of the above three problems
but yet still exhibit obvious datum shifts. This typically happens
when some new (or different) figure is adopted for a gauge datum and
when the tidal recording device is reset accordingly. At New
Plymouth, for example, it is clear that changes in the gauge setting
of 1.0 ft (0.305 m), 1.5 ft (0.457 m), 2.0 ft (0.610 m) and 3.0 ft
(0.914 m) all occurred in the space of 10 years. In two such cases
there was no clear record of exactly why or when this had happened.
Indeed, it appears that there was some confusion between the Port
Authority (the owner and operator of the gauge) and the national
surveying and mapping organization (responsible for the tidal
predictions), as to what datum offset should have been set.
In summary, the New Zealand experience is clear. Given a reasonable
tidal record over a long period of time, a clear and unambiguous datum
history is the single most important issue to be resolved if there is to
be a reliable estimate for long-term sea level change. It is around this
issue that the greatest risk of unwarranted assumptions of reliability
and consistency exist – particularly from those who may not be
intimately familiar with the specifics of the history of a particular
2.3 Analysis Errors
While datum errors tend to be the most difficult to resolve (due to
the need for good record keeping over periods of many decades), analysis
errors, or other unmodelled systematic effects, can also intrude. The
removal of the tidal signal by a low-pass filter and by averaging
techniques is very effective. Other fluctuations in sea level such as
those arising from shallow water effects, storm surges and wind set-up,
while not easily modeled, are effectively eliminated by using monthly
and annual sea level means. However, there is real danger in seeking to
resolve accurately long-term sea level changes from data sets of less
than 60 years in length. Douglas (2001), for example, summarises
research showing that large variations in the estimates of sea level
rise can be explained in nearly all cases by the selection criteria used
by a particular investigator – short records being one of the most
important. It is vital that the periodic effects from such signals as
inter-decadal variability be eliminated (Holgate and Woodworth, 2004).
Studies in New Zealand by Goring and Bell (1999) and by Bell et al
(2000) reveal the importance of the Inter-decadal Pacific Oscillation
(IPO) and the El Nino Southern Oscillation (ENSO) on sea level
variability at Auckland and Tauranga. Indeed, it has been speculated
that such effects might extend as far south as Lyttelton.
A second analysis problem that can arise relates to the influence of
unmodelled hydrological effects. The Hunter River, for example, has had
an influence on the data produced by the Newcastle tide gauge on the
East Coast of Australia. Equally, one of New Zealand’s longest tidal
records (Westport) was compromised by similar effects. The Westport
gauge sits at the mouth of the Buller River. If climate change brings
with it changes in rainfall patterns (as is expected to happen), then
the prospect exists for apparent sea level change to be masked or
exacerbated by changes in river flow.
2.4 Geophysical Effects
Early sea-level change analyses showed wide variation in result
(e.g., Gornitz, 1995). However, much of this variation was subsequently
able to be explained by ensuring that the tide gauge records used met
five criteria. These were: (1) that the records be at least 60 years in
length, (2) that they not be from sites at collisional tectonic plate
boundaries, (3) that they be 80% complete or better, (4) that at low
frequencies they be in reasonable agreement with nearby gauges sampling
the same water mass, and (5) that they not be from areas deeply covered
by ice during the last glacial maximum (Douglas, 1997). Reasons (2) and
(5) are the issues addressed in this section.
2.4.1 Tectonic Motion at Plate Boundaries
New Zealand sits astride the boundary between the Pacific and Australian
tectonic plates. The lateral movement between these plates is known to
be in the order of 35-40 mm/yr (Bevan et al, 2002). Of the four New
Zealand tide gauges with 80 or more years of data, three (Auckland,
Lyttelton and Dunedin) sit well away from the collision zone. Only the
gauge at Wellington sits within the zone. Hannah (1990) has argued that
the high level of coherence between the long-term sea level trends at
these four gauges provides some comfort in the view that the
differential vertical uplift between these four gauges (should it exist)
is relatively small. However, until a sufficiently long time series of
data is available that provide accurate estimates of regional vertical
tectonic uplift, the influence of such effects remains unknown.
2.4.2 Glacial Isostatic Adjustment
The second geophysical effect to be considered in the interpretation of
any tide gauge record is that of glacial isostatic adjustment (GIA).
Vertical motions from this effect are estimated by using a geophysical
model (e.g., Peltier, 2001), the size of the motion varying according to
the model adopted. For example, GIA estimates for Auckland range from
0.1 mm/yr (from the ICE4G (VM2) model to 0. 55 mm/yr (from the JM120,1,3
model). Similar levels of variability in estimate are found at other New
Zealand tide gauges.
It is encouraging to note that the above two issues now show real
promise of resolution. While not able to be separated readily, it is
anticipated that their total effect will be able to be estimated by the
use of GPS receivers installed alongside (or in close proximity) to tide
gauges. It is the expectation that the daily position solutions from
these receivers, when examined over long periods of time (10 years or
longer), will allow an assessment to be made of any absolute land motion
whether it be from tectonic or GIA causes. In New Zealand GPS receivers
have been collocated with the tide gauges in the Ports of Auckland,
Wellington, Lyttelton and Dunedin since 1999. The tide gauge records at
each of these ports extend back to 1900. Preliminary results indicate
that the regional land motion is in the order of 1.00 mm/yr (Denys et
al, 2009). At this time experiments are underway to determine final
strategies for processing the GPS data.
3. LOOKING TO THE FUTURE
In looking to future long-term sea level change analyses, it is
crucial not only that reliable tide gauges be maintained at existing
sites where long records exist, but also that any datum changes be
carefully determined and recorded. In some cases this may require the
negotiation of agreements between port companies (typically with
commercial objectives) and national science/research agencies (with
research objectives). In New Zealand, in the 1990s, the failure to
resolve these issues adequately has had a damaging impact on some tide
gauge records. The recent resolution of these issues has improved the
situation considerably. However, constant vigilance is required. Nowhere
is this better illustrated than at Auckland where the tide gauge has
been moved three times since 2000 and where links to stable local bench
marks have not been adequately maintained due to local construction
The installation of CGPS receivers on or beside tide gauges a decade
ago now opens the possibility for an independent assessment of any
vertical land motion irrespective of cause. Having come thus far, it is
vital that such initiatives continue. Longevity of data set is a crucial
factor in accurate and sound future scientific analysis as is
consistency and accuracy in processing strategy.
It is an easy matter to adopt a tide gauge record from a national or
international database without any real thought as to its overall
quality. While temporary (or short term) operational difficulties in
collecting tide gauge data can usually be overcome by using monthly or
annual sea level means, datum shifts are not so able to be overcome. New
Zealand experience indicates that these are by far the most important
issues to be resolved if a high quality, consistent tide gauge record is
to be produced. By their very nature tide gauge records are fragile and
can only be considered to be usable for long-term sea level change
studies if there is a long, consistent and reliable datum history. In
addition, if there are to be reliable estimates of eustatic sea level
rise then any geophysical effects must be able to be removed. As New
Zealand reaches the anniversary of a decade of CGPS measurements, it is
encouraging to realize that this latter goal is now in sight.
- Bell, R.G., D.G. Goring, De Lange, W.P, (2000). Sea-level change
and storm surges in the context of climate change, Trans.,
27(1/Gen), 10 pp., Inst. of Professional Engineers N.Z.
- Bevan, J., Tregoning, P., Bevis, M., Kato, T., Meertens, C.,
(2002). Motion and rigidity of the Pacific Plate and implications
for plate boundary deformation. J. Geophys. Res., 107(B10), 2261.
- Chambers D.P., Ries, J.C., Shum, C.K., Tapley, B.D., (1998). On
the use of tide gauges to determine altimeter drift. J. Geophys.
Res., 103(C6), 12,885-12,890.
- Denys, P., Hannah, J., Beavan, J., (2009). Tide gauge – CGPS
measurements in New Zealand. Proceedings, IAG 2009: Geodesy for
planet earth. In press.
- Douglas, B. C., (1997). Global sea level rise: A
redetermination. Surveys Geophys. 18, 279-292.
- Douglas, B. C., (2001). Sea level change in the era of the
recording, in Sea level rise - history and consequences, edited by
B.C. Douglas et al, Academic Press, New York.
- Gornitz, V., (1995). Sea-level rise: A review of recent past and
near-future trends, Earth surface processes and landforms, 20, 7-20.
- Goring, D.G., Bell, R.G., (1999). El Niño and decadal effects on
sea-level variability in northern New Zealand: A wavelet analysis,
N.Z. Jour. of Marine and Freshwater Research, 33, 587-598,
- Hannah, J., (1990). Analysis of MSL data from New Zealand for
the period 1899-1988. J. Geophys. Res., 95(B6), 12,399-12,405.
- Hannah, J., (2004). An updated analysis of long-term sea level
change in New Zealand. Geophys. Res. Letters, 31, L03307.
- Holgate S.J., Woodworth, P.L., (2004). Evidence for enhanced
coastal sea level rise during the 1990s. Geophys. Res. Letters 31,
- Hydrographer of the Navy (1969). Admiralty Manual of
Hydrographic Surveying, Volume 2.
- IPCC (2001), Climate Change 2001, The Scientific Basis, 3rd
Assessment Report, Cambridge University Press.
- IPCC (2007), Climate Change 2007, The Physical Science Basis,
4th Assessment Report, Cambridge University Press.
- Nerem, S.R., Mitchum G.T., (2001). Obervations of sea level
change from satellite altimetry, in Sea level rise history and
consequences, edited by B.C. Douglas et al., Academic Press, San
- Peltier, W.R., (2001). Global glacial isostatic adjustment and
modern instrumental records of relative sea level history, in Sea
level rise history and consequences, edited by B.C. Douglas et al.,
Academic Press, San Diego.
- Pugh, D., (2004). Changing Sea Levels, Effects of Tide, Weather
and Climate. Cambridge University Press.
- Woppelmann, G., (2007). An Inventory of Collocated and
Nearly-Collocated CGPS stations and Tide Gauges,
http://www.sonel.org/stations/cgps/surv_update.html. Accessed 14
John Hannah BSc, DipSci, MSc, PhD, MNZIS, RPSurv, completed his first
two degrees at the University of Otago, New Zealand. Two years later, in
1974, he became a Registered Surveyor. In 1976 he began study at The
Ohio State University, completing an MSc and a PhD, both in Geodetic
Science. From 1982 until 1988 he was Geodetic Scientist, and then
subsequently, Chief Geodesist/Chief Research Officer with the Department
of Lands and Survey, New Zealand. After a 17 month appointment to the
Chair in Mapping, Charting and Geodesy at the US Naval Postgraduate
School, California, he returned to New Zealand as Director of Geodesy
and subsequently, Director of Photogrammetry for the Dept. of Survey and
Land Information. In 1993 he joined the School of Surveying, as
Professor and Head of Department, becoming its Dean in 2001. He
relinquished this administrative role at the end of 2004 in favour of
more teaching and research. His publications reflect his research
interests in sea level change and surveying education. He is a
Registered Professional Surveyor and is a former President of the NZ
Institute of Surveyors.
Professor John Hannah
School of Surveying
University of Otago
PO BOX 56 Dunedin New zealand
Ph. 0064 3 479 9010
Fax.0064 3 479 7586