Assimilation of satellite altimeter data into an eddy-resolving primitive equation model of the North Atlantic Ocean

• Main Author: Oschlies, Andreas
• Format: Thesis
• Language: unknown
• Published: 1994
• Subjects: North Atlantic
• Online Access: http://oceanrep.geomar.de/43651/; http://oceanrep.geomar.de/43651/1/Diss.%201994%20Oschlies,%20A.pdf

The aim of the present study is to develop an efficient assimilation scheme suitable for assimilating satellite altimeter data into a basin-scale eddy-resolving primitive equation model with active thermodynamics. Two alternative methods, referred to as extended nudging and re-initialization, respectively, will be presented and intercompared. The first approach essentially extrapolates the surface observations into the ocean interior using a priori correlations provided by the model climatology. In contrast, the second assimilation procedure emphasizes dynamical relationships and in particular the conservation of temperature and salinity on isopycnals, rather than relying heavily on uncertain statistical correlations. Both methods are tested and found to perform about equally well by running identical twin experiments, in which the model is assumed perfect and generates its own data for assimilation. Twin experiments are further used to compare the assimilation of the altimeter data directly along the satellite ground tracks with the assimilation of preprocessed maps combining all the data within some chosen time window. The results of these experiments indicate that - as far as the assimilation techniques proposed in this study are concerned - assimilation of quasi-synoptic maps is preferential. The actual assimilation experiments are performed with objectively analyzed sea surface height anomalies measured by the U.S. Navy's satellite GEOSAT. Adding the model's mean sea surface to obtain an estimate of absolute dynamic sea level, these maps are then assimilated into the WOCE (World Ocean Circulation Experiment) CME (Community Modelling Effort) model of the North Atlantic Ocean at 5-day intervals covering the year 1987. While the results of the two alternative assimilation procedures were almost indistinguishable in the identical twin experiments using simulated data, marked differences emerge when real altimeter data are used. These differences are shown to arise from different physical and statistical implications of the two assimilation methods. It turns out that a correct treatment of water mass properties is of crucial importance for the success of the assimilation scheme. Only the re-initialization method, which explicitly conserves temperature and salinity on isopycnals, leads to reasonable hydrographic situations, whereas the extended nudging scheme, that uses statistical correlations to infer temperature and salinity from sea surface height measurements, is found to introduce completely unrealistic water masses.