Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter.
Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 37
INTRODUCTION
Recent Changes in Sea Level
A Summary
TIM P. BARNETT
Scripps Institution of Oceanography
The purpose of this chapter is to review the behavior of
sea level over the past century. This review is prompted by
recent interest in global warming and the CO2 problem and
the possible increase in ocean level that might accompany
such a warming (e.g., Hansen et al., 1981~. The change in
sea level in such a scenario is due to both an increase in
ocean temperature and a decrease (melting) of landbound
ice. This change in relative sea level (RSL) will, in the
absence of compensating factors, be global in nature
(Barrett and Baker, 1981), although the variations need
not be uniform over the globe (e.g., Farrell and Clark,
1976; Clark and Lingle, 19771.
The review is organized as follows. The next section
summarizes research over the past 20 yr on changes in sea
level. Next, the data used to study these changes are
examined critically. Two different approaches to estimat-
ing RSL change are then reviewed using specific case
studies. These sections are intended to point out the
nonuniqueness in estimates of RSL change, plus the ap-
parent fact that sea level has not been rising at all points in
the world's oceans over the past century. The next section
summarizes some of the possible causes of RSL change
and thus provides an introduction to other chapters of this
volume. A concluding section summarizes briefly the
37
status of research on RSL. Given the review nature of this
chapter, it has been convenient to rely heavily on two
articles (Barrett, 1983a, 1984a) that will serve as basic
reference for the discussion of the two different approaches
to estimating RSL change and will offer abundant refer-
ences for readers interested in delving further into the
subject.
SOME PRIOR STUDIES OF "GLOBAL" SEA LEVEL
Sea-level stations are located mainly on coastal mar-
gins and a few islands. Over the vast interior of the
world's oceans there are no sea-level data. Thus any
analysis of mean sea-level (MSL) data that claims to rep-
resent a "global" condition is already attended by consid-
erable assumption regardless of the results of the analysis.
An effort is made throughout the text to remind the reader
of this fact.
At least six different estimates of the "eustatic increase"
in mean "global" sea level were made between 1940 and
1961. These estimates are summarized by Lisitzin (1974)
and are presented in Table 1.1. Perhaps the most notable
aspect of these early analyses is that all values agree
remarkably well, even though they were obtained by sev-
eral different methods and data sets (e.g., sea level or
cryological information). All of these studies were se
OCR for page 38
38
TABLE 1.1 Estimates of Mean Global Sea-Level Increase
Author
Rate (mm/yr) Method
Thorarinsson (1940)
Gutenberg ( 1941 )
Kuenen (1950)
Lisitzin (1958)
Wexler (1961)
Fairbridge and
Krebs (1962)
Emery (1980)
0.5
1.1 +0.8
1.2to 1.4
1.12 + 0.36
1.18
1.2
3.0
Gornitz et al. (1 982)a 1.2
Barnett ( 1 983a) 1.51 + 0. 15
Barnett (1984a) 1.43 to 2.27
Cryologic aspects
Sea level (many stations)
Different methods
combined
Sea level (six stations)
Cryologic estimates
Selected sea level stations
Sea level (many stations)
and selected stations
Sea level (many stations)
Selected sea level stations
Sea level (many stations)
aThe authors attempt a correction for crustal motion and find
. mm/yr. The 1.2 mm/yr is without this correction.
verely limited by geographic extent, time span, and accu-
racy. A second notable aspect of the early analysis pointed
out by Lisitzin was the existence of an apparent "break" in
temporal behavior of sea level. Prior to about 1890 to
1900, she found little or no trend in sea level. After that
time, however, sea level begins a slow, monotonic rise.
Fairbridge and Krebs (1962) attempted to create an
estimate of"global" sea-level change between 1860 and
1960. They paid careful attention to station location and
data quality. In addition, they showed the high correlation
that exists between nearby sea-level stations, thus allow-
ing one station to represent a large area. Their final esti-
mates show an increase for RSL of 1.2 mm/yr, although
they note the estimate is biased by an overabundance of
North Atlantic data. Their world sea-level curve shows, as
suggested by Lisitzin (1974), a period between 1890 and
1910 of little change. Unfortunately, they never stated
exactly how their"global" curve was obtained.
Emery (1980) used data from selected stations in the
worldwide tide gauge net to estimate a rise in "global" sea
level of 3 mm/yr over the past 40 yr. (Note: this is roughly
three times the rate estimated by earlier workers; see Table
1.1.) He pointed out that his study is hampered by lack of
data in the Southern Hemisphere. As noted above, this
problem attends all studies of "global" sea level. Emery
also found that the rate of sea-level rise between 1970 and
1974 is almost five times the longer-term value.
Unfortunately, a number of problems arise in obtaining
these estimates: (1) the erratic distribution and strong
clustering of stations, plus the analysis method, make the
above number strongly biased to several specific regions
and individual stations exhibiting strong MSL increases;
(2) the recent estimates of RSL rise have not been cor
TIM P. BARNETT
reeled for atmospheric effects; simple interannual changes
In sea-eve pressure (aim) could easily cause part of the
observed change; and (3) the 5-yr averaging time (1970 to
1974) is a time scale typical of short-term climate change
and regimes. Changes in ocean temperatures and circula-
tion, for example, associated with these epochs, could also
account for much of the observed RSL trend (cf. Fair-
bridge and Krebs, 1962~.
Etkins and Epstein (1982) accepted the 3-mm/yr RSL
increase and assumed it applies globally. They then tried
to explain the increase in terms of ocean temperature change
and discharge from the polar ice sheets. Perhaps their
most interesting result relates the observed change in rota-
tion rate of the Earth to the estimated change in the polar
ice caps. The apparent agreement so obtained in this
comparison would be much less impressive, however, if
they had not used the 3-mm/yr RSL rise figure of Emery.
They also did not discuss the fact that the rotation rate of
the Earth increased between 1900 and 1930, and so the
length of day decreased. Sea-level arguments, such as
they presented, would dictate an increase in length of day
(see Etkins and Epstein, 1982, reference 8~. The entire
problem of the effect of sea level and ice-cap melting on
Earth's rotation characteristics has been described well in
the l950s. Further discussion of this subject is deferred to
a later section in this chapter. Additional discussion of the
Etkins and Epstein paper is given by Robuck (1983~.
Gornitz et al. (1982) also attempted to develop an esti-
mate of "global" MSL change. They developed large area
averages of sea level, thus avoiding the clustering problem
that plagued Emery. They generally ignored regions that
do not show an increasing trend (e.g., Alaska) and eventu-
ally ended up with a 1.2-mm/yr increase as a "global"
number. They also attempted to remove vertical crustal
motion based on geologic data and to find a RSL trend of
1 mm/yr on a "global" basis.
Review of the references supporting the development
of their land motion estimate suggests that their correc-
tions may be highly unstable. Indeed, the corrections are
generally major, relative to the uncorrected trend, and
unstable from millennium to millennium (Newman et al.,
19801. It may thus be fortuitous that the two global-trend
estimates come out as close as they do. There are other
problems in their analysis. (1) Eight of the 13 regions
used in the global average border the Atlantic Ocean, a
fact that biases their estimates of RSL changes to past
variability in that ocean. (2) The regional time series are
not temporally homogeneous; thus, for instance, the
"global" average between 1880 and 1900 is dictated al-
most entirely by European data. (3) The use of raw aver-
age trends for each region to develop a "global" trend
allows a few active regions to dominate the global number
(cf. Table 1 of Gornitz et al., 1982~.
_ _ _ _ 1 _ _ _ 1 _ _ _ _ _ ~ ~ ~N 1 1 · 1
OCR for page 39
RECENT CHANGES IN SEA LEVEL: A SUMMARY
THE DATA
Sources and Distribution
Sea-level data have been taken routinely since at least
1808 (Brest, France) and even before, e.g., Amsterdam,
1682 (R. Fairbridge, Columbia University, personal com-
munication). These data have been organized by The Per-
manent Service for Mean Sea Level, Institute of Oceano-
graphic Science, England (Lennon, 1975-1978), and are
available in printed volumes and on magnetic tape. The
distribution of stations is shown in Figure 1 of Emery
(19801. That illustration shows the highest station densi-
ties in the United States and western Europe followed by
Japan. Long-term records from the Southern Hemisphere,
e.g., South America and Africa, are few in number. Long-
term records from islands in the open ocean far from land
are almost nonexistent. Hence, any attempts to infer
"global" changes in sea level from this data set are at-
tended by substantial assumption-the distribution of data
is simply not adequate to represent global coverage.
Instruments
Interestingly, little was found in the literature that de-
scribed the instrumental methods. However, Lennon (1970)
provided a good overview of current instruments and their
possible problems and there were many.
Most modern tide data come from instrumental mea-
surement (e.g., a float in a "stilling well") with the mea-
surements continually recorded by a pen and ink device.
In recent years the data recording mode has become digi-
tal. Before 1900, however, the measurement methods
varied considerably. The long-term San Francisco record
was obtained in the above manner, even though the actual
measurement site was moved four times (S. Hicks, NOAA,
personal communication). The old MSL records from India
(e.g., Bombay) were obtained by human "tide parties"
whose members made frequent visual sightings on a gradu-
ated staff inserted in the water (B. Zetler, Scripps Institu-
tion of Oceanography, personal communication). Need-
less to say, the risk associated with the data increases
considerably as one reaches back beyond 1900.
Potential Problems
The sea-level data have numerous potential difficulties.
Some of these are listed below. [See also Fairbridge and
Krebs (1962) for additional problems.]
1. Station locations were subject to position change,
for example, as harbors developed through the century.
39
These changes can add discontinuity to the time series
and, if undetected, can induce apparent long-term trends.
2. Instrument position changes can be accommodated
if the old and new locations are tied to a common geodetic
reference point. It is often difficult to discover if this was
done. But even the leveling (reference) operation is not
without difficulty (see below).
3. Tectonic uplift/subsidence of the land mass on which
the gauge sits will induce an apparent sea-level change. In
some parts of the world (e.g., Alaska and Scandinavia),
this signal is as large or larger than any possible oceano-
graphic change (e.g., Hicks and Shofnos, 1965~. [Note:
this effect is most pronounced in regions of glacial re-
bound (high latitudes) and near the mouths of large rivers
(e.g., in the Gulf of Mexico) where heavy sedimentation
causes subsidence.]
4. Mean sea level is affected by a variety of physical
processes (e.g., Lisitzin, 1974; Namias and Huang, 1972),
including changes in water temperature and salinity, river
runoff, local and nonlocal currents, wind, waves, and
atmospheric pressure. All of these can change sea level in
conflicting ways, and these changes can be as large as the
signal under study. Since all of the variables mentioned
above are subject to climatological variations, they char-
acteristically have "red spectra," which means they can
contribute to estimates of trends in sea level.
5. Geodetic leveling between sea-level stations appar-
ently cannot be used to remove relative vertical motion of
the land, if the stations are more than a few hundred
kilometers apart. Sturges (1966) showed that the 500-mm
north-south sea-level difference inferred from leveling data
to exist along the Atlantic seaboard was inconsistent with
oceanographic data. The same situation was found along
the west coast of the United States. More shockingly,
Balazs and Douglas (1979) reviewed five different at-
tempts between 1969 and 1978 to obtain the relative eleva-
tion difference between San Francisco and San Pedro,
California (about 700 km apart). According to those sur-
veys, the two cities had a relative vertical displacement
rate of 70 mm/yr. The RSL trend between the two was 2.1
mm/yr the difference is a factor of nearly 35. Thus, the
leveling data for both coasts of the United States may well
contain serious errors of unknown origin when the separa-
tion distance exceeds a few hundred kilometers.
6. The possible instrumental problems are legion. The
ability to speculate in this area far exceeds the available
information. Lennon (1970) gave a good overview of
possible problems with "modern" equipment. Errors of
180 mm or more are possible in sea-level measurement,
and they may be virtually impossible to detect in the data.
Contrasting this error with an often stated RSL rise of
1 mm/yr gives one an appreciation for some of the poten-
tial uncertainties in sea-level studies.
OCR for page 40
40
- ''-~¢
-(- 4~
fat ~ ~
:::
FIGURE 1.1 Location of primary (~) and secondary (x' stations
versus estimates of vertical crustal motion for the period 1000 to
ESTIMATES OF RELATIVE SEA-LEVEL CHANGE:
KEY STATION APPROACH
Methods
One approach to estimating change in global RSL is to
analyze data from key stations, arguing that changes at
these stations are representative of changes in RSL in
large (1000+ km) regions surrounding the stations. Ex-
amples of such analyses are found in Table 1.1 (e.g.,
Fairbridge and Krebs, 1962; Barnett, 1983a). The latter
work is summarized here.
The use of the key station approach imposes the follow-
ing series of constraints on the data to be analyzed.
1. High-quality, continuous measurement that reveals
no sudden shifts suggestive of station movements should
be used. The records should be as long as possible.
2. Station locations should be away from areas of
strong tectonic movement (e.g., separated from areas of
major deposition/uplift).
3. Stations should be unaffected by spurious physical
TIM P. BARNETT
-
-_ ~_-~_
12 ~_
TO -2; ~ ~
~~
_~1
~ ~-
1000-2000 yr BP
`~;]
J.~'\-\\
WAX ~ ) 1
I'
/ _
2000 yrBP. Contours are in millimeters per year. After Newman
et al. (1980).
processes (e.g., gauge exposed to strong freshwater inva-
sions).
4. The spatial density of stations by oceans should be
proportional to the relative areas of the respective oceans,
e.g., 2.4/1.5/1.0, for the Pacific/Atlantic/Indian (cf. Sverdrup
et al., 1942~. The stations should be synchronous in time.
Such a distribution will allow equal weight to be given to
all oceans and years in the subsequent analysis, thereby
avoiding biasing problems.
5. Finally, the selected stations must represent large
. . .
geographic regions.
The locations of a set of stations that generally satisfy
these criteria are shown in Figure 1.1 and are listed in
Table 1.2.
A method of inferring, quantitatively, the existence of a
signal that is coherent over all portions of a data field
using empirical orthogonal function (EOF) analysis (cf.
Backus and Preisendorfer, 1978; Barnett, 1978) was ap-
plied to the primary key station set and to the secondary
station set to check sensitivity of the results to data pertur-
bations. The sea-level data at position i are represented by
OCR for page 41
RECENT CHANGES IN SEA LEVEL: A SUMMARY
hi(t), i = 1, 2, . .
in the study. Define
and
. NP, where NP is the number of stations
hi = (hift)>t'
o2=,, (1.1)
where ~ >, denotes a time averaging operation. The data
were normalized to have unit variance and zero mean, i.e.,
h (t)- h
h' (t) = i I . ( 1.2)
csi
This is a crucial step for later interpretation of the results,
for now each station will have equal weight in the subse-
quent analysis. This, coupled with equal time and area
weighting in the key station selection, ensures a more or
less equal geographic weighting for the analysis.
Next form the correlation matrix:
Cij = (`h if t~h~(t)~. (1.3)
Denote the eigenvectors and eigenvalues of Eq. (1.3) by
Bni and An, respectively, and the associated principal com-
ponents:
Antt) = ~Bni h'(t) (1.4)
Finally, it follows that
1
h (t) = MA n (t)Bni ~ 1 .5
n
Special properties of the (Bn, An) are given in the above
references.
The EOF representation of the RSL field is interpreted
as follows:
Results
41
Figures 1.2 through 1.5. For the longer time period, the B
all have the same sign (Figure 1.2), indicating the exis-
tence of a coherent pattern of variation in the data set.
Note all stations analyzed are not contributing equally to
this pattern since all By are not the same size. Figure 1.3
shows that the coherent change is associated with increas-
ing RSL. Further, the time variation appears well fit by a
simple linear trend ( 1.51 + 0.15 mm/yr).
Analysis of the more recent period suggests RSL is not
rising uniformly within the data set analyzed (Figure 1.41.
Indeed the implied decrease in RSL in the eastern Asian
region TON is supported by independent hydrographic
observations (e.g., White et al., 1979~. The implied RSL
trend in Figure 1.5 is 1.79 + 0.22 mm/yr. (Note that
neither of the temporal reconstructions suggests an accel-
erating rate of RSL in recent years.)
Consideration of the results and the data shows that the
EOF analysis is being dominated by temporal trends.
Stations with positive trends have B. components of one
sign; those with negative trend have By components of the
opposite sign. The analysis method is extracting a single
trend common to all stations that accounts for as much of
the total field variance as possible. Extraneous noise,
which would normally contaminate study of data from
individual stations, is therefore effectively filtered out of
data prior to trend estimation.
TABLE 1.2 Sea-Level Stations
Station
PRIMARY
San Francisco
Honolulu
Tonoura, Japan
Balboa, C.Z.
Sydney
Bombay
Baltimore
Cristobal, C.Z.
Cascais, Port.
Takoradi
1. The Bni represent average patterns of spatial covari
ability between the NP-members of the sea-level field. If
all members are fluctuating in unison, as would be the case
for a global change in RSL, then all components of Bit, the
first, most energetic eigenvector, will have the same sign.
2. The Angst) modulate, in time, the intensity of the
spatial patterns of By . For instance, a secular trend in the
hi coherent at all stations will be manifested by a secular
trend in Alit).
3. Using Eq. (1.5) and an average value of the Hi from SECONDARY
Eq. (1.1), it is possible to reconstruct the behavior of Ketchikan
global sea-level change if such a creature exists. Hosojima
Aden
Aberdeen
Helsinki
Montevideo
Ystad
The results of the above analysis for two different time
periods, 1903 to 1969 and 1930 to 1975 are shown in
Missing
Abbreviation Start End Years
SFO
HNL
TON
BCZ
SYD
BOM
BAL
CRS
CAS
TAK
MON
1862 1874
1905 1975
1894 1975
1908 1969
1897 1977
1878 1964
1903 1975
1909 1969
1882 1975
1930 1975
1
1
4
1
1919 1975
1894 1977
1879 1960 45
1862 1965 1
1879 1970
1938 1970
1887 1974
