Data Assimilation in a Regional Finite Element Sea-Ice Model for the Arctic - Application of the Singular Evolutive Interpolated Kalman Filter

The Arctic region is sensitive to climate change. Since the Arctic sea-ice cover influences the surface heat budget of the Earth the observed sea-ice decline is seen as an indication of global warming. Furthermore, the dynamics of sea ice plays an important role for the sea-ice mass distribution in the Arctic, for the production of dense, cold, and salty water in the Arctic Ocean, which contributes to the thermohaline circulation, and also for the freshwater budget of the Nordic Seas. Thus, a realistic description of sea-ice motion is important to draw conclusions for the mass transport and sea-ice mass distribution.The Finite-Element Sea-Ice Model simulates the large-scale physical sea-ice processes like the sea-ice growth and circulation realistically. The model domain covers the entire Arctic Ocean and its marginal seas. Together with the Singular Evolutive Interpolated Kalman (SEIK) Filter and remotely sensed sea-ice drift observations this sea-ice model is applied for data assimilation to investigate details of the sea-ice dynamics. So far, drift assimilation has been carried out to analyze and modify only the drift field with subsequent computation of the advection or redistribution of ice mass which corresponds more to the physical model behavior than a statistical analysis that the SEIK Filter provides. The sea-ice drift data assimilation with the SEIK Filter achieves drift modification and furthermore changes in the two other sea-ice variables concentration and thickness. The modifications of these "unobserved variables" (within the meaning of data assimilation) are validated and it is found that they are in good agreement for at least 2 months for the sea-ice thickness and even 4 months for the sea-ice concentration which is the longest period examined. The drift improvement is achieved due to the sea-ice concentration and thickness changes which leads to a sustainable effect for further sea-ice drift simulation. Furthermore, the assimilation results suggest a higher thickness variability that the model alone is not able to produce. A localized version of the SEIK Filter leads to a more pronounced drift correction which is not sustainable in course of further model integration because in this case the sea-ice concentration and thickness are not much affected by the assimilation method. This thesis describes the initial work of the sea ice drift assimilation with the SEIK Filter and further examines the ability of the SEIK Filter to modify the model state using the observation data. The applicability, capability, limitation of the assimilation method and suggestions are discussed.