Here we present results from a long-term model simulation of the 3-D coupled ecosystem model ECOSMO II for a North Sea and Baltic Sea set-up. The model allows both multi-decadal hindcast simulation of the marine system and specific process studies under controlled environmental conditions. Model results have been analysed with respect to long-term multi-decadal variability in both physical and biological parameters with the help of empirical orthogonal function (EOF) analysis. The analysis of a 61-year (1948–2008) hindcast reveals a quasi-decadal variation in salinity, temperature and current fields in the North Sea in addition to singular events of major changes during restricted time frames. These changes in hydrodynamic variables were found to be associated with changes in ecosystem productivity that are temporally aligned with the timing of reported regime shifts in the areas. Our results clearly indicate that for analysing ecosystem productivity, spatially explicit methods are indispensable. Especially in the North Sea, a correlation analysis between atmospheric forcing and primary production (PP) reveals significant correlations between PP and the North Atlantic Oscillation (NAO) and wind forcing for the central part of the region, while the Atlantic Multi-decadal Oscillation (AMO) and air temperature are correlated to long-term changes in PP in the southern North Sea frontal areas. Since correlations cannot serve to identify causal relationship, we performed scenario model runs perturbing the temporal variability in forcing condition to emphasize specifically the role of solar radiation, wind and eutrophication. The results revealed that, although all parameters are relevant for the magnitude of PP in the North Sea and Baltic Sea, the dominant impact on long-term variability and major shifts in ecosystem productivity was introduced by modulations of the wind fields.