Application of an Ensemble Optimal Interpolation in a North/Baltic Sea model: Assimilating temperature and salinity profiles

Abstract

This paper describes the implementation of an Ensemble Optimal Interpolation (EnOI) in a two-way nested North/Baltic Sea model for assimilating temperature and salinity profiles. In the EnOI, the state vector is extended to include variables from both the fine and coarse resolution models. In the stationary ensemble, the annual and semi-annual cycles in temperature and sea level are removed. Its necessity is demonstrated by comparing the spatial correlations computed with/without annual and semiannual cycles. Anisotropic and inhomogeneous features of background error covariances produced by the ensemble are shown for temperature and salinity in a few areas, e.g., the central part of the North Sea and in the upstream of the Skagerrak. Some features are coherent with the known circulation of the area. The assimilation experiments are carried out from January 8, 2005 to April 25, 2005 with an operational forecasting model. The impact of profile assimilation is examined by comparison with observations. The results indicate that the assimilation of the temperature and salinity profiles can significantly improve the ocean forecasts. The sea surface temperature is greatly improved along the Norwegian coast and in the Skagerrak. The pronounced salinity front in the North Kattegat, which is difficult to model, is also improved. The root mean square differences (RMSD) between the forecasts and observations for temperature and salinity have been reduced by 25% and 32% during the experiment period, respectively. For sea level, comparisons with independent tide gauge data reveal positive improvement with RMSD reduced about 1–3 cm. The impacts of assimilated initial state on the forecast are also investigated. Starting from an initial state with assimilation, the model produces better forecasts than the unassimilated one. Moreover, the impact could persist for nearly 3 weeks.