9.5.2 Sea Level

A precondition for attributing changes in sea level rise to anthropogenic forcing is that model-based estimates of historical global mean sea level rise should be consistent with observational estimates. Although AOGCM simulations of global mean surface air temperature trends are generally consistent with observations (Section 9.4.1, Figure 9.5), consistency with surface air temperature alone does not guarantee a realistic simulation of thermal expansion, as there may be compensating errors among climate sensitivity, ocean heat uptake and radiative forcing (see, e.g., Raper et al., 2002, see also Section 9.6). Model simulations also offer the possibility of attributing past sea level changes to particular forcing factors. The observational budget for sea level (Section 5.5.6) assesses the periods 1961 to 2003 and 1993 to 2003. Table 9.2 evaluates the same terms from 20C3M simulations in the MMD at PCMDI, although most 20C3M simulations end earlier (between 1999 and 2002), so the comparison is not quite exact.

Simulations including natural as well as anthropogenic forcings (the ‘ALL’ models in Table 9.2) generally have smaller ocean heat uptake during the period 1961 to 2003 than those without volcanic forcing, since several large volcanic eruptions cooled the climate during this period (Gleckler et al., 2006). This leads to a better agreement of those simulations with thermal expansion estimates based on observed ocean warming (Section 5.5.3) than for the complete set of model simulations (‘ALL/ANT’ in Table 9.2). For 1993 to 2003, the models that include natural forcings agree well with observations. Although this result is somewhat uncertain because the simulations end at various dates from 1999 onwards, it accords with results obtained by Church et al. (2005) using the PCM and Gregory et al. (2006) using HadCM3, which suggest that 0.5 mm yr^–1 of the trend in the last decade may result from warming as a recovery from the Mt. Pinatubo eruption of 1991. Comparison of the results for 1961 to 2003 and 1993 to 2003 shows that volcanoes influence the ocean differently over shorter and longer periods. The rapid expansion of 1993 to 2003 was caused, in part, by rapid warming of the upper ocean following the cooling due to the Mt. Pinatubo eruption, whereas the multi-decadal response is affected by the much longer persistence in the deep ocean of cool anomalies caused by volcanic eruptions (Delworth et al., 2005; Gleckler et al., 2006; Gregory et al., 2006).

Both observations and model results indicate that the global average mass balance of glaciers and ice caps depends linearly on global average temperature change, but observations of accelerated mass loss in recent years suggest a greater sensitivity than simulated by models. The global average temperature change simulated by AOGCMs gives a good match to the observational estimates of the contribution of glaciers and ice caps to sea level change in 1961 to 2003 and 1993 to 2003 (Table 9.2) with the assumptions that the global average mass balance sensitivity is 0.80 mm yr^–1 °C^–1 (sea level equivalent) and that the climate of 1900 to 1929 was 0.16°C warmer than the temperature required to maintain the steady state for glaciers (see discussion in Section 10.6.3.1 and Appendix 10.A).

Calculations of ice sheet surface mass balance changes due to climate change (following the methods of Gregory and Huybrechts, 2006 and Section 10.6.3.1) indicate small but uncertain contributions during 1993 to 2003 of 0.1 ± 0.1 mm yr^–1 (5 to 95% range) from Greenland and –0.2 ± 0.4 mm yr^–1 from Antarctica, the latter being negative because rising temperature in AOGCM simulations leads to greater snow accumulation (but negligible melting) at present. The observational estimates (Sections 4.6.2 and 5.5.6) are 0.21 ± 0.07 mm yr^–1 for Greenland and 0.21 ± 0.35 mm yr^–1 for Antarctica. For both ice sheets, there is a significant contribution from recent accelerations in ice flow leading to greater discharge of ice into the sea, an effect that is not included in the models because its causes and mechanisms are not yet properly understood (see Sections 4.6.2 and 10.6.4 for discussion). Hence, the surface mass balance model underestimates the sea level contribution from ice sheet melting. Model-based and observational estimates may also differ because the model-based estimates are obtained using estimates of the correlation between global mean climate change and local climate change over the ice sheets in the 21st century under SRES scenarios. This relationship may not represent recent changes over the ice sheets.

Summing the modelled thermal expansion, global glacier and ice cap contributions and the observational estimates of the ice sheet contributions results in totals that lie below the observed rates of global mean sea level rise during 1961 to 2003 and 1993 to 2003. As shown by Table 9.2, the terms are reasonably well reproduced by the models. Nevertheless, the discrepancy in the total, especially for 1961 to 2003, indicates the lack of a satisfactory explanation of sea level rise. This is also a difficulty for the observational budget (discussed in Section 5.5.6).

A discrepancy between model and observations could also be partly explained by the internally generated variability of the climate system, which control simulations suggest could give a standard deviation in the thermal expansion component of ~0.2 mm yr^–1 in 10-year trends. This variability may be underestimated by models, since observations give a standard deviation in 10-year trends of 0.7 mm yr^–1 in thermal expansion (see Sections 5.5.3 and Section 9.5.1.1; Gregory et al., 2006).

Since recent warming and thermal expansion are likely largely anthropogenic (Section 9.5.1.1), the model results suggest that the greater rate of rise in 1993 to 2003 than in 1961 to 2003 could have been caused by rising anthropogenic forcing. However, tide gauge estimates suggests larger variability than models in 10-year trends, and that rates as large as that observed during 1993 to 2003 occurred in previous decades (Section 5.5.2.4).

Overall, it is very likely that the response to anthropogenic forcing contributed to sea level rise during the latter half of the 20th century. Models including anthropogenic and natural forcing simulate the observed thermal expansion since 1961 reasonably well. Anthropogenic forcing dominates the surface temperature change simulated by models, and has likely contributed to the observed warming of the upper ocean and widespread glacier retreat. It is very unlikely that the warming during the past half century is due only to known natural causes. Lack of studies quantifying the contribution of anthropogenic forcing to ocean heat content increase and glacier melting, and the fact that the observational budget is not closed, make it difficult to estimate the anthropogenic contribution. Nevertheless, an expert assessment based on modelling and ocean heat content studies suggests that anthropogenic forcing has likely contributed at least one-quarter to one-half of the sea level rise during the second half of the 20th century (see also Woodworth et al., 2004).

Anthropogenic forcing is also expected to produce an accelerating rate of sea level rise (Woodworth et al., 2004). On the other hand, natural forcings could have increased the rate of sea level rise in the early 20th century and decreased it later in the 20th century, thus producing a steadier rate of rise during the 20th century when combined with anthropogenic forcing (Crowley et al., 2003; Gregory et al., 2006). Observational evidence for acceleration during the 20th century is equivocal, but the rate of sea level rise was greater in the 20th than in the 19th century (Section 5.5.2.4). An onset of higher rates of rise in the early 19th century could have been caused by natural factors, in particular the recovery from the Tambora eruption of 1815 (Crowley et al., 2003; Gregory et al., 2006), with anthropogenic forcing becoming important later in the 19th century.