"Decadal prediction has emerged nowadays as a main concern in the climate sciencecommunity. Improving the forecast on the time scales of 10-30 years is expected tocarry tremendous value for the society, supporting plans for infrastructure upgrades,financial decisions or energy policies.Decadal predictions are in between the seasonal forecast (6-12 months) and theclimate prediction (on the scale of a century) and share a number of physical anddynamical features of both. The decadal time range is at the confluence of forecasting, such as that performed in seasonal prediction, having a marked predictable signal in the initial state and the long term climate prediction driven mainly from external forcing and fully independent from the initial state. Refined observing networks for the ocean component are now available (see e.g. the ARGO project at www-argo.ucsd.edu, or SMOS mission of ESA) and contribute also to promising future improvements in forecast quality. Data assimilation is the field of geosciences that study the problem of state estimation of evolving, possibly chaotic, dynamical system on the base of incomplete, inhomogeneous and noisy observations. Data assimilation is regarded nowadays with strong interest from the climate science community, because of the potential to improve decadal prediction by introducing adequate initialization process for the whole climate system, making the best use of the current observations. Several fundamental unresolved questions need to be addressed to adapt existing data assimilation algorithms as well as to develop new strategies for the Earth system models used in decadal prediction. Testing different data assimilation approaches and contributing to the current open debate on how to improve initialized decadal prediction are the main motivations of the proposed research. The project ultimate goal is the implementation of a state-of-the-art data assimilation technique for the initialization of the IC3 EC-Earth model."