Contribution of thermal expansion to present-day sea-level change revisited

We investigate the thermosteric (i.e., due to temperature only) sea-level change over the last 50 years using two global ocean temperature data sets recently published (Levitus et al., 2000a [Levitus, S., Stephens, C.M., Antonov, J.I., Boyer, T.P., 2000a. Yearly and year-season upper ocean temperature anomaly fields, 1948 1998, pp. 23, U.S. Gov. Printing Office, Washington, DC] and Ishii et al., 2003 [Ishii, M., Kimoto, M., Kachi, M., 2003. Historical ocean subsurface temperature analysis with error estimates, Monthly Weather Rev., 131, 51 73]). These data sets which provide gridded temperatures, down to 3000 m and 500 m respectively, are based on interpolation schemes of raw historical profiles over 1950 1998. We find that the two data sets compare well over 1950 1990, both in terms of thermosteric sea-level trends and global mean. Some difference is noticed however beyond 1990, due to differences in the raw temperature data processing. Analyses based on \textquoteleftEmpirical Orthogonal Function\textquoteright show that the interannual variability of the thermosteric sea level is dominated by the signatures of El Ni\~no Southern Oscillation, Pacific Decadal Oscillation and influenced by North Atlantic Oscillation. As a result, regional thermosteric sea-level trends are not stationary on a century time scale and have a typical lifetime on the order of a decade. In terms of global mean, the rate of thermosteric sea-level change computed over 10-year windows displays high variability, with values reaching up to three times the 40-year (1950 1990) average at some periods. Even negative values are noticed at other periods. One important consequence is that the pattern of sea-level trends derived from Topex/Poseidon altimetry over 1993 2003, which is mainly caused by thermal expansion, is very likely a non-permanent feature. Thus past and future extrapolation based on this 10-year altimetry pattern should be considered with caution.