Résumé
At present, essentially two multi species models are used in fisheries science: Multi-species Virtual Population Analysis (MSVPA) and the Ecopath family of models (with EcoSim and EcoSpace). Developed in a complementary way to those models, the OSMOSE model (Object-oriented Simulator of Marine ecOSystems Exploitation) aims at representing and studying the spatial, age- and size-structured dynamics of marine fish assemblages. Its structure and formalism allows to: associate and articulate species and ecosystem dynamics; investigate the role of biodiversity by means of different diversity indices, either cardinal (e.g. species richness, Shannon index, evenness) or ordinal (e.g. size spectrum and derived indices); simulate different fishing scenarios and management measures (minimal length at catch, target species, fishing mortality, quotas, MPAs); provide output variables that can be compared to those produced by existing multi-species models such as MSVPA (fish numbers by age and size, recruitment, mortality rates); use biological and ecological data and information that are widely available in the literature or in databases such as FishBase. The OSMOSE model is based on the hypothesis that predation is an opportunistic process, depending on fish size rather than on its taxonomy. Predation opportunism is implemented by allowing predators to feed on any fish species provided that the prey size does not exceed a threshold value, and that it is located in its vicinity. From FishBase data, the upper predator/prey size ratio is estimated at about 3.5 (Froese Pauly, 2000). Such a predation process can account for cannibalism, omnivory and variability, which are frequently observed in fish diets (Smith Re ay, 1991; Rice, 1995). Furthermore, the strength of both predation and competition relationships vary according to changes in relative species abundance. Thus, rather than being represented from pre- established species interactions, trophic webs are considered as being fundamentally dynamic.