Climate Change 2001:
Working Group I: The Scientific Basis
Other reports in this collection

11.3.2 Mean Sea Level Changes over the Past 100 to 200 Years

11.3.2.1 Mean sea level trends


Figure 11.8:
Global mean sea level variations (light line) computed from the TOPEX/POSEIDON satellite altimeter data compared with the global averaged sea surface temperature variations (dark line) for 1993 to 1998 (Cazenave et al., 1998, updated). The seasonal components have been removed from both time-series.

The primary source of information on secular trends in global sea level during the past century is the tide gauge data set of the Permanent Service for Mean Sea Level (PSMSL) (Spencer and Woodworth, 1993). The tide gauge measurement is of the level of the sea surface relative to that of the land upon which the gauge is located and contains information on both the displacement of the land and on changes in ocean volume (eustatic changes). The land displacement may be of two types: that caused by active tectonics and that caused by glacial rebound. Corrections for these effects are required if the change in ocean volume is to be extracted from the tide gauge record. Both corrections are imperfectly known and are based on sea level observations themselves, usually from long geological records. Different strategies have been developed for dealing with these corrections but differences remain that are not inconsequential (see Table 11.9).

The sea level records contain significant interannual and decadal variability and long records are required in order to estimate reliable secular rates that will be representative of the last century. In addition, sea level change is spatially variable because of land movements and of changes in the ocean circulation. Therefore, a good geographic distribution of observations is required. Neither requirement is satisfied with the current tide gauge network and different strategies have been developed to take these differences into consideration. Warrick et al. (1996), Douglas (1995) and Smith et al. (2000) give recent reviews of the subject, including discussions of the Northern Hemisphere geographical bias in the historical data set.

In the absence of independent measurements of vertical land movements by advanced geodetic techniques (Section 11.6.1), corrections for movements are based on either geological data or geophysical modelling. The former method uses geological evidence from locations adjacent to the gauges to estimate the long-term relative sea level change which is assumed to be caused primarily by land movements, from whatever cause. This is subtracted from the gauge records to estimate the eustatic change for the past century. However, this procedure may underestimate the real current eustatic change because the observed geological data may themselves contain a long-term component of eustatic sea level rise (Section 11.3.1). The latter method, glacial rebound modelling, is also constrained by geological observations to estimate earth response functions or ice load parameters, which may therefore themselves contain a component of long-term eustatic sea level change unless this component is specifically solved for (Section 11.3.1).

A further underestimate of the rate of sea level rise from the geological approach, compared to that from glacial rebound models, will pertain in forebulge areas, and especially the North American east coast, where the linear extrapolation of geological data could result in an underestimate of the corrected rate of sea level change for the past century typically by 0.3 mm/yr because the glacial rebound signal is diminishing with time (Peltier, 2000). However, in areas remote from the former ice sheets this bias will be considerably smaller.

Also, in adding recent mass into the oceans, most studies have assumed that it is distributed uniformly and have neglected the Earth's elastic and gravitational response to the changed water loading (analogous to glacio-hydro-isostatic effect). This will have the effect of reducing the observed rise at continental margin sites from ongoing mass contributions by as much as 30% (cf. Nakiboglu and Lambeck, 1991).

Table 11.9 summarises estimates of the corrected sea level trends for the past century. Estimates cover a wide range as a result of different assumptions and methods for calculating the rate of vertical land movement, of different selections of gauge records for analysis, and of different requirements for minimum record length.


Figure 11.9:
Ranges of uncertainty for the average rate of sea level rise from 1910 to 1990 and the estimated contributions from different processes.

There have been several more studies since the SAR of trends observed in particular regions. Woodworth et al. (1999) provided a partial update to Shennan and Woodworth (1992), suggesting that sea level change in the North Sea region has been about 1 mm/yr during the past century. Lambeck et al. (1998) combined coastal tide gauge data from Fennoscandinavia together with lake level records and postglacial rebound models to estimate an average regional rise for the past century of 1.1 ± 0.2 mm/yr. Studies of the North American east coast have been particularly concerned with the spatial dependence of trends associated with the Laurentian forebulge. Peltier (1996) concluded a current rate of order 1.9 ± 0.6 mm/yr, larger than the 1.5 mm/yr obtained by Gornitz (1995), who used the geological data approach, and Mitrovica and Davis (1995), who employed Post Glacial Rebound (PGR) modelling. Note that the observations of thermal expansion (Section 11.2.1.1) indicate a higher rate of sea level rise over recent decades in the sub-tropical gyres of the North Atlantic (i.e., off the North American east coast) than the higher latitude sub-polar gyre. Thus the differences between three lower European values compared with the higher North American values may reflect a real regional difference (with spatial variations in regional sea level change being perhaps several tenths of a millimetre per year - see also Section 11.5.2). In China, relative sea level is rising at about 2 mm/yr in the south but less than 0.5 mm/yr in the north (National Bureau of Marine Management, 1992), with an estimated average of the whole coastline of 1.6 mm/yr (Zhen and Wu, 1993) and with attempts to remove the spatially dependent component of vertical land movement yielding an average of 2.0 mm/yr (Shi, 1996). The two longest records from Australia (both in excess of 80 years in length and not included in Douglas, 1997) are from Sydney and Fremantle, on opposite sides of the continent. They show observed rates of relative sea level rise of 0.86 ± 0.12 mm/yr and 1.38 ± 0.18 mm/yr over the periods 1915 to 1998 and 1897 to 1998 (Mitchell et al., 2000), corresponding to approximately 1.26 mm/yr and 1.73 mm/yr after glacial rebound correction using the Peltier ICE-4G/M2 model, or 1.07 mm/yr and 1.55 mm/yr using the corrections of Lambeck and Nakada (1990).

There have been only two analyses of global sea level change based on the PSMSL data set published since the SAR. Douglas (1997) provided an update to Douglas (1991) and applied the PGR model of Tushingham and Peltier (1991) to a selected set of twenty-four long tide gauge records, grouped into nine geographical areas, with minimum record length 60 years and average length 83 years. However, the only Southern Hemisphere sites included in this solution were from Argentina and New Zealand. The overall global average of 1.8 ± 0.1 mm/yr agreed with the 1991 analysis, with considerable consistency between area-average trends. The standard error of the global rate was derived from the standard deviation of regional trends, assuming that temporal and spatial variability is uncorrelated between regions. Peltier and Jiang (1997) used essentially the same set of stations as Douglas and a new model for postglacial rebound.

From Table 11.9 one can see that there are six global estimates determined with the use of PGR corrections derived from global models of isostatic adjustment, spanning a range from 1.4 mm/yr (Mitrovica and Davis, 1995; Davis and Mitrovica, 1996) to 2.4 mm/yr (Peltier and Tushingham, 1989, 1991). We consider that these five are consistent within the systematic uncertainty of the PGR models, which may have a range of uncertainty of 0.5 mm/yr depending on earth structure parametrization employed (Mitrovica and Davis, 1995). The average rate of the five estimates is 1.8 mm/yr. There are two other global analyses, of Gornitz and Lebedeff (1987) and Nakiboglu and Lambeck (1991), which yield estimates of 1.2 mm/yr, lower than the first group. Because of the issues raised above with regard to the geological data method for land movement correction, the value of Gornitz and Lebedeff may be underestimated by up to a few tenths of a millimetre per year, although such considerations do not affect the method of Nakiboglu and Lambeck. The differences between the former five and latter two analyses reflect the analysis methods, in particular the differences in corrections for land movements and in selections of tide gauges used, including the effect of any spatial variation in thermal expansion. However, all the discrepancies which could arise as a consequence of different analysis methods remain to be more thoroughly investigated. On the basis of the published literature, we therefore cannot rule out an average rate of sea level rise of as little as 1.0 mm/yr during the 20th century. For the upper bound, we adopt a limit of 2.0 mm/yr, which includes all recent global estimates with some allowance for systematic uncertainty. As with other ranges (see Box 11.1), we do not imply that the central value is the best estimate.



Other reports in this collection