5.5.3 Ocean Density Changes

Sea level will rise if the ocean warms and fall if it cools, since the density of the water column will change. If the thermal expansivity were constant, global sea level change would parallel the global ocean heat content discussed in Section 5.2. However, since warm water expands more than cold water (with the same input of heat), and water at higher pressure expands more than at lower pressure, the global sea level change depends on the three-dimensional distribution of ocean temperature change.

Analysis of the last half century of temperature observations indicates that the ocean has warmed in all basins (see Section 5.2). The average rate of thermosteric sea level rise caused by heating of the global ocean is estimated to be 0.40 ± 0.09 mm yr^–1 over 1955 to 1995 (Antonov et al., 2005), based on five-year mean temperature data down to 3,000 m. For the 0 to 700 m layer and the 1955 to 2003 period, the averaged thermosteric trend, based on annual mean temperature data from Levitus et al. (2005a), is 0.33 ± 0.07 mm yr^–1 (Antonov et al., 2005). For the same period and depth range, the mean thermosteric rate based on monthly ocean temperature data from Ishii et al. (2006) is 0.36 ± 0.12 mm yr^–1. Figure 5.19 shows the thermosteric sea level curve over 1955 to 2003 for both the Levitus and Ishii data sets. The rate of thermosteric sea level rise is clearly not constant in time and shows considerable fluctuations (Figure 5.17). A rise of more than 20 mm occurred from the late 1960s to the late 1970s (giving peak 10-year rates in the early 1970s) with a smaller drop afterwards. Another large rise began in the 1990s, but after 2003, the steric sea level is decreasing in both estimates (peak rates in the late 1990s). Overlapping 10-year rates from these two estimates have a very high temporal correlation (r = 0.97) and the standard deviation of the rates is 0.7 mm yr^–1.

Figure 5.19. Global sea level change due to thermal expansion for 1955 to 2003, based on Levitus et al. (2005a; black line) and Ishii et al. (2006; red line) for the 0 to 700 m layer, and based on Willis et al. (2004; green line) for the upper 750 m. The shaded area and the vertical red and green error bars represent the 90% confidence interval. The black and red curves denote the deviation from their 1961 to 1990 average, the shorter green curve the deviation from the average of the black curve for the period 1993 to 2003.

The Levitus and Ishii data sets both give 0.32 ± 0.09 mm yr^–1 for the upper 700 m during 1961 to 2003, but the Levitus data set of temperature down to 3,000 m ends in 1998. From the results of Antonov et al. (2005) for thermal expansion, the difference between the trends in the upper 3,000 m and the upper 700 m for 1961 to 1998 is about 0.1 mm yr^–1. Assuming that the ocean below 700 m continues to contribute beyond 1998 at a similar rate, with an uncertainty similar to that of the upper-ocean contribution, we assess the thermal expansion of the ocean down to 3,000 m during 1961 to 2003 as 0.42 ± 0.12 mm yr^–1.

For the recent period 1993 to 2003, a value of 1.2 ± 0.5 mm yr^–1 for thermal expansion in the upper 700 m is estimated both by Antonov et al. (2005) and Ishii et al. (2006). Willis et al. (2004) estimate thermal expansion to be 1.6 ± 0.5 mm yr^–1, based on combined in situ temperature profiles down to 750 m and satellite measurements of altimetric height. Including the satellite data reduces the error caused by the inadequate sampling of the profile data. Error bars were estimated to be about 2 mm for individual years in the time series, with most of the remaining error due to inadequate profile availability. A close result (1.8 ± 0.4 mm yr^–1 steric sea level rise for 1993 to 2003) was recently obtained by Lombard et al. (2006), based on a combined analysis of in situ hydrographic data and satellite sea surface height and SST data (Guinehut et al., 2004). It is presently unclear why the latter two estimates are significantly larger than the thermosteric rates based on temperature data alone. It is possible that the in situ data underestimate thermal expansion because of poor coverage in Southern Oceans, and it is interesting to note that a model based on assimilation of hydrographic data yields a somewhat higher estimate of 2.3 mm yr^–1 (Carton et al., 2005). Published estimates of the steric sea level rates for 1955 to 2003 and 1993 to 2003 are shown in Table 5.2.

We assess the thermal expansion of the upper 700 m during 1993 to 2003 as 1.5 ± 0.5 mm yr^–1, and that of the upper 3,000 m as 1.6 ± 0.5 mm yr^–1, allowing for the ocean below 700 m as for the earlier period (see also Section 5.5.6, Table 5.3).

Antonov et al. (2002) attributed about 10% of the global average steric sea level rise during recent decades to halosteric expansion (i.e., the volume increase caused by freshening of the water column). A similar result was obtained by Ishii et al. (2006) who estimated a halosteric contribution to 1955 to 2003 sea level rise of 0.04 ± 0.02 mm yr^–1. While it is of interest to quantify this effect, only about 1% of the halosteric expansion contributes to the global sea level rise budget. This is because the halosteric expansion is nearly compensated by a decrease in volume of the added freshwater when its salinity is raised (by mixing) to the mean ocean value; the compensation would be exact for a linear state equation (Gille, 2004; Lowe and Gregory, 2006). Hence, for global sums of sea level change, halosteric expansion cannot be counted separately from the volume of added land freshwater (which Antonov et al., 2002, also calculate; see Section 5.5.5.1). However, for regional changes in sea level, thermosteric and halosteric contributions can be comparably important (see, e.g., Section 5.5.4.1).