Analysis and representation of regional sea-level variability from altimetry and atmospheric-oceanic data

A simple representation of the sea-level variability in time at low and medium frequencies and at large and medium spatial scales is investigated that accounts for the correlation between sea-level heights and other atmospheric and oceanic parameters. The selected fields are sea-level height, sea-surface temperature, wind speed and sea-surface atmospheric pressure; they are considered in the interval between 1992 and 1997. The sea-level height data correspond to a single space altimetry mission (Topex/Poseidon) and provide a homogeneous data set. The dominant characteristics of the variability of each field and the coupled variability between two fields are analysed in the European seas using the spectral analysis method and the statistical methods of principal component analysis and canonical correlation analysis. In the single-field study, both the spectral and the principal component analyses show different characteristics of the sea-level height variability in the three seas. The strongest annual and semi-annual signals are in the Mediterranean Sea, whereas in the other two seas, especially in the Baltic Sea, the dominant spectral components have comparable power. The trends are interpreted as interannual variability due to the shortness of the investigated time interval. The highest positive trends are observed in the eastern Mediterranean Sea. In the coupled-fields study, both the linear regression analysis and the canonical correlation analysis show a high correlation between the sea-level height and the sea-surface temperature in the Mediterranean Sea and between the sea-level height and the wind speed in the North Sea and the Baltic Sea. The Mediterranean Sea is particularly suitable for building sea-level height models from the statistical methods because the first four modes of both the single- and the coupled-fields models account for about 90 per cent of the variance of the sealevel height field. It is therefore chosen as a test area to check the accuracy of the models and the ability of the canonical correlation method to predict the sea-level variability. The accuracy of the statistical models is assessed by computing dual crossover height differences between the Topex/Poseidon and ERS-1 and ERS-2 sea-surface heights corrected using the variability models. Both the single- and the coupled-fields models are found to be a good representation of the sea-level variability in the Mediterranean Sea, whereas the extrapolated canonical correlation model, derived using the sea-surface temperature as a predictor, is less accurate, but still acceptable. Relative bias and drift between the Topex/Poseidon and ERS data result from the analysis and reflect a nonhomogeneous pre-processing of the altimetry data. The single-mission sea-level variability models are a first step towards the construction of a multimission sea-level model from unified multimission altimetry data.