Résumé
Ecosystem models provide a platform allowing exploration into the possible responses of marine food webs to fishing pressure and various potential management decisions. In this study we investigate the particular effects of overfishing on the structure and function of the southern Benguela food web, using two models with different underlying assumptions: the spatialized, size-based individual-based model, OSMOSE, and the trophic mass-balance model, Ecopath with Ecosim (EwE). Starting from the same reference state of the southern Benguela upwelling ecosystem during the 1990s, we compare the response of the food web to scenarios of overfishing using these two modelling approaches. A scenario of increased fishing mortality is applied to two distinct functional groups: i) two species of Cape hake, representing important target predatory fish, and ii) the forage species anchovy, sardine and redeye. In these simulations, fishing mortality on the selected functional groups is doubled for 10 years, followed by 10 years at the initial fishing mortality. We compare the food web states before the increase of fishing mortality, after 10 years of overfishing and after a further 10 years during which fishing was returned to initial levels. In order to compare the simulated food web structures with the reference state, and between the two modelling approaches, we use a set of trophic indicators: the mean trophic level of the community and in catches, the trophic pyramid (biomass per discrete trophic level), and the predatory/forage fish biomass ratio. OSMOSE and EwE present globally similar results for the trophic functioning of the ecosystem under fishing pressure: the biomass of targeted species decreases whereas that of their potential competitors increases. The reaction of distant species is more diverse, depending on the feeding links between the compartments. The mean trophic level of the community does not vary enough to be used for assessing ecosystem impacts of fishing, and the mean trophic level in the catch displays a surprising increase due to the short period of overfishing. The trophic pyramids behave in an unexpected way compared to trophic control theory. because at least two food chains with different dynamics are intertwined within the food web. We emphasize the importance of biomass information at the species level for interpreting dynamics in aggregated indicators, and we highlight the importance of competitive groups when looking at ecosystem functioning under fishing disturbance. Finally, we discuss the results within the scope of differences between models, in terms of the way they are formulated, spatial dimensions, predation formulations and the representation of fish life cycles. (C) 2009 Elsevier B.V. All rights reserved.