Tide Gauge Records Suitable For Global Sea Level Analyses

The preceding section has shown how interannual and longer sea level variations create difficulties in determining underlying long-term trends of sea level. But determining an accurate trend is not enough, because even for very long records, the trend reflects processes other than a global increase of sea level. We turn now to the problem of identifying suitable long records from the very large data base of tide gauge data at the PSMSL. Figure 3.12 shows the most optimistic view of the situation. Plotted here are all of the PSMSL RLR tide gauge sites with records >20 years in length as of the 1997 data release. There are more than 500 sites on this map, but there is a very great location bias for the Northern Hemisphere and an absolute dominance of coastal positions rather than islands. (As an aside, this map illustrates why altimetric satellites such as TOPEX/Poseidon and JASON are required. There are simply not enough island sites to provide the necessary spatial coverage of the broad oceans for studies of basin-scale sea level variability and circulation.)

Figure 3.12 considerably overstates the number of potentially useful measuring sites (at least for the next 3 or 4 decades) because it includes tide gauge sites with records as short as 20 years. As the analysis of the previous section showed, much longer records are required to determine the underlying trend of sea level change at a site. Very long records go a long way toward relieving

Longitude, degrees

Figure 3.12 RLR tide gauge locations with records longer than 20 years.

Longitude, degrees

Figure 3.12 RLR tide gauge locations with records longer than 20 years.

the often-cited problem of the poor sampling of the oceans by tide gauges, at least at time scales approaching a century. We saw in Fig. 3.11 how long records gave GIA-corrected trends that converged toward a common value for the trend of sea level regardless of location. Groger and Plag (1993) and others have rightly pointed out the poor geographic distribution of tide gauge locations, but have failed to consider a critical factor. It is that the longer the period of a sea level variation, the greater the spatial extent of that signal. Thus very long records do not require the coverage needed by shorter ones to establish a value for global change. For the 20th century at least, Douglas (1991, 1992) has shown that widely separated tide gauge records show very consistent trends and (lack of) acceleration. Woodworth (1990) and Gornitz and Solow (1991) also observed the latter. Thus on a 100-year temporal scale, the lack of uniform global coverage appears to be much less important than usually supposed.

Sometimes tide gauge records may be unsuitable for mechanical or other causes affecting the sea level signal. This is where the use of multiple records from relatively nearby sites, as in the cases of San Francisco/San Diego and the U.S. middle Atlantic series previously discussed, can be decisive. The problem of the disagreement of sea level records at Sydney and Newcastle (Douglas, 1991) has already been mentioned; if this could be straightened out, a valuable long sea level record in the data-sparse Southern Hemisphere would be gained.

As an example of an obvious problem visible in a single record, consider Fig. 3.13 for the long series at Manila, the Philippines. The Manila record has

Figure 3.13 Manila mean annual relative sea level.

been used by many investigators of sea level rise, but obviously something happened in the early 1960s to cause an abrupt order-of-magnitude increase in the rate of rise there. The sudden increase is not plausibly oceanographic in origin. In the documentation accompanying this record, the PSMSL reports that harbor reclamation projects may be responsible. Whatever the reason, the Manila sea level record, although enticing for its length of nearly a century, is unsuitable for scientific use.

Sometimes problems with sea level time series have only a short duration, such as at Brest, France, during the World War II years. In this case, the anomalous values are recognizable by disagreement with the Newlyn, England, values, which otherwise are highly correlated (Douglas, 1992) over their common observation time. Subsequent to the problem with the Brest record, there is a gap of several years and then sea levels vary again in concert with those of Newlyn. In the documentation on their Web site, the PSMSL notes that gaps often are indicative of problems in series of sea level data and should be carefully investigated.

As important as data quality considerations may be, many high-quality and long tide gauge records for which an adequate GIA correction exists still show a large scatter in their trends. In most cases, plate tectonics is the reason. Douglas (1991) examined sea level trends from regional groupings of tide gauges that would be expected to show the same trend internal to each group based on oceanographic considerations and small differences of GIA. He found that Indian, Japanese, U.S. Pacific Northwest, and other locations associated with converging tectonic plates showed an inconsistency of sea level trends even over a small area. As an example, consider Fig. 3.14, which shows 5-year average relative sea levels for Seattle and Neah Bay, in Washington State in the United States. This is an area of converging tectonic plates, with associated seismicity and volcanism. The records are obviously morphologically similar at an interannual scale—note for example the usual U.S. west coast evidence of the 1982-1983 ENSO event. But the trends of sea level at these nearby sites are very different. Long-term relative sea level is increasing at Seattle and falling at Neah Bay. Geodesists have used the differing trends from these and other tide gauge records in the area to study vertical crustal movements, while at the same time the trends have been used by some investigators to study sea level rise. It should be clear by now that a tide gauge record cannot simultaneously serve geodesy and oceanography!

We turn now to the matter of providing a list of tide gauge records useful for determining a 20th-century rate of global sea level rise. From the preceding remarks and examples, it is clear that the tide gauge records must be long (>60-70 years at a minimum), be free of vertical crustal movements due to plate tectonics, be adequately correctable for GIA, and have trends insensitive to small changes in their record length or be capable of being edited in a justifiable manner based on oceanographic considerations. This is a long list

7200 i-

7200 i-

6950 -

1935 1940 1945 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995

Figure 3.14 Relative annual sea level at Seattle and Neah Bay, Washington.

of restrictions, and only a small number of sites qualify. Table 3.2 gives the locations and trends of relative sea level for suitable 20th-century tide gauge records exceeding 70 years in length. Even though many of these records began before 1900, that date was selected as a starting point because of the evidence previously mentioned for a sharp increase in the rate of sea level in the mid-19th century, but without further significant acceleration over the entire 20th century (Douglas, 1992).

The groupings of the tide stations in Table 3.2 are presented according to oceanographic region. For example, U.S. east coast records are divided into sites north and south of Cape Hatteras, where the Gulf Stream leaves the coast and turns toward Europe. As remarked earlier, it is essential to group sea level trend measurements this way to prevent regions with a large number of gauges from having an excessive weight in the global solution.

The sea level trends in Table 3.2 have a significant scatter, but at least all are positive. In addition, they are not yet corrected for GIA. In Chapter 4 we shall see that applying a GIA correction, based on a new solution by W. R. Peltier, gives in general highly consistent results for sea level rise regardless of geographic location. However, in the Mediterranean, where the GIA correction is small (a few tenths of a millimeter per year), the sea level trends remain systematically small. Douglas (1997) noted that there has been no apparent increase of sea level at Marseille, Genova, and Trieste in the last 50 years, but could provide no explanation.

Table 3.2

20th-century Relative Sea Level Trends from Records > 70 Years

Table 3.2

20th-century Relative Sea Level Trends from Records > 70 Years

Groups

Trend (mm/yr)

Start

End

Span (yr)

Aberdeen I+ 11

0.7

1900

1997

97

Newlyn

1.7

1915

1997

82

Brest

1.3

1900

1991

91

Cascais

1.6

1903

1991

88

Lagos

1.4

1909

1992

83

Marseille

1.2

1900

1996

96

Genova

1.2

1900

1992

92

Trieste

1.1

1905

1997

92

Auckland1

1.3

1904

1989

85

Dunedin1

1.4

1900

1989

89

Lyttelton1

2.3

1904

1989

85

Wellington1

1.7

1901

1988

87

Honolulu

1.5

1905

1997

92

San Francisco2

1.8

1900

1980

80

San Diego2

1.9

1906

1980

74

Balboa2

1.5

1908

1980

72

Buenos Aires2

1.1

1905

1980

75

Pensacola

2.1

1923

1996

73

Key West

2.2

1913

1997

84

Fernandina

2.0

1897

1993

96

Charleston

3.3

1922

1997

75

Baltimore

3.1

1903

1997

94

Atlantic City

4.0

1912

1997

85

New York

3.0

1900

1997

97

Boston

2.7

1921

1997

76

Portland

1.9

1912

1997

85

Halifax

3.4

1920

1996

76

1 New Zealand trends from Hannah (1990).

2 Data after 1980 eliminated to avoid effect of 1982/83 and 1997/98 ENSO events on trends.

Was this article helpful?

0 0

Post a comment