Use of global ocean reanalyses for reconstructing sea level variability patterns over the last 40 years: methods, results and limitations
Philippe Rogel(1) , Clément Ubelmann(1) , Anthony Weaver(1) , and Nicolas Daget(1)
42 Av Coriolis,
A 40-year reanalysis (1962-2001) of the ocean circulation, carried out in the framework of the European ENACT project, has been used to investigate how sea level variability can be reconstructed. This reanalysis has been obtained by the assimilation of historical in situ temperature measurements into a global low-resolution version of the OPA ocean GCM. The observational data set is the quality controlled in situ data gathered for the ENACT project by the Metoffice (Ingleby and Huddelston, 2005). The assimilation scheme is the 3D-Variational scheme described e.g. in Ricci et al. (2005), with a full multivariate background error covariance term, which enables the correction of all model variables at each 10-day analysis window.
The model’s hypothesis of a local free surface, but assuming a constant total volume of the ocean, does not allow to investigate global mean sea level variations over the period, but only regional patterns of variability. It furthermore imposes rigorous constraints for any comparison to observations such as altimetry or tide gauges, and, theoretically, for the computation of innovations in the context of data assimilation. Therefore, the protocol for a rigorous comparison is exposed in the poster, and results using historical tide gauges and precise Topex/Poseidon-Jason altimetry are shown. This validation shows that, though the best agreement is in the tropical regions, low frequency signals at the midlatitudes in sufficient agreement to be analysed over long periods.
Over such a long period, the assimilation system causes long term drifts which have regional signatures, mostly due to the act that salinity variations are not sufficiently constrained by the assimilation scheme. Fortunately, the more-than-a-decade record of altimetry helps identifying them using a statistical method. An application is shown in the North Atlantic region, where the drift is a 15-year regular change of the east-west sea level gradient. Finally, corrected sea level anomalies are analysed in this region through Empirical Orthogonal Functions. The dominant mode is close to the tripole observed in winter SST anomalies. Link with winter North Atlantic Oscillation forcing is evidenced.