The present article gives an overview on the leading concepts and modelling approaches for marine ecosystems research including (1) The trophodynamic theory of pelagic ecosystems, (2) Compartment/network models, (3) Mesocosm experiments and (4) Individual based modelling approaches and virtual ecosystems (VE). The main research questions addressed, as well as the potential and limits of each approach, are summarized and discussed and it is shown how the concept of ecosystem has changed over time. Aquatic biomas spectra (derived from the theory of pelagic ecosystems) can give insight into the trophic structure of different systems, and can show how organism sizes are distributed within the system and how different size groups participate in the system s metabolism and production. Compartment/network models allow for a more detailed description of the trophic structure of ecosystems and of the energy/biomass fluxes through the explicit modelling of P/B-and food consumption rates and biomasses for each system compartment. Moreover, system indices for a characterization and comparison with other systems can be obtained such as average trophic efficiency, energy throughput, and degree of connectivity, degree of maturity, and others. Recent dynamic extensions of trophic network models allow for exploring past and future impacts of fishing and environmental disturbances as well as to explore policies such as marine protected areas. Mesocosm experiments address a multitude of questions related to aquatic processes (i.e. primary production, grazing, predation, energy transfer between trophic levels etc.) and the behaviour of organisms (i.e. growth, migration, response to contaminants etc.) under semi-natural conditions. As processes within mesocosms often differ in rate and magnitude from those occurring in nature, mesocosms should be viewed as large in vitro experiments designed to test selected components of the ecosystem and not as an attempt to enclose a multitude of interacting processes. Models that use individual organisms as units can provide insight into the causes of natural variability within populations (growth, phenotype, behaviour) and into the role of intraspecific variation for interspecific processes, succession, and feedback mechanisms. In biological oceanography, interdisciplinary research is increasingly using "Virtual Ecosystems" to simulate non-linear interactions between the dynamics of fluctuating ocean circulation, the physics of air-sea interaction, turbulence and optics, biogeochemistry, and the physiology and behaviour of plankton, which can be compared with real observations. The different approaches available for the analysis of aquatic ecosystems should be seen as complementary ways for the description and understanding of ecosystems. The modern view of marine ecosystems, as has emerged from ecosystem analysis over the last decades, is that of a composite of loosely coupled subsystems of desynchron dynamics which through their combined action maintain the fundamental structure and function of the whole