The regional effects of climate change can be heavily modulated by internal variability on time scales of seasons to decades. This climate variability may either mitigate or exacerbate the impacts of global warming and therefore has profound implications for mitigation and adaptation planning. The CMIP experiments provide global climate scenarios in response to external forcing, but they are not initialized in agreement with the observed state of the climate. Over the past decade, initialized climate predictions have been developed and are included in the 5th IPCC assessment report. These show predictive skill in the North-Atlantic region. Furthermore, the Norwegian Climate Prediction Model (NorCPM) suggests that predictability may extend to the Nordic Seas using sea surface temperature. By further initializing sea ice properties, NorCPM may increase the chances of predicting regional atmospheric, oceanic and ecosystem changes at the sea ice edge. This is of importance for the resource management in the region. Marine ecosystems are vulnerable to climate variability and change through impacts on the horizontal distribution of heat and nutrients. In the North Atlantic and Nordic Seas, shifts in the boundaries between relatively cold and warm waters are associated with changes in the sub-polar gyre (SPG). Therefore, the prediction skill shown for the North Atlantic ocean heat content offers exciting possibilities for more rational management of the commercially important fisheries in the affected regions. They also potentially open the door to improved predictions of terrestrial climate over Norway and Western Europe, as shifts in the large-scale atmospheric circulation are also linked to ocean heat content variations. The resolution of global climate prediction models is too coarse to resolve e.g., extreme weather events and effects on marine ecosystems. Regional downscaling is necessary in order to obtain climate information on scales that are relevant to society, and dynamical downscaling has been demonstrated to increase the regional value of the global scenarios. Dynamical downscaling of seasonal-to-decadal predictions has the potential to bring gains on spatial and temporal scales not possible from standard climate projections. In addition, the application of calibration techniques can lead to improved reliability. This proposal addresses an area of research currently not addressed by the research community.