Marine ecosystem-based management requires end to end models, which are models capable of representing the entire ecosystem including physical, chemical and biological processes, anthropogenic activities, and multiple species with different sizes, life histories and from different trophic levels. To adequately represent ecosystem dynamics in shallow coastal regions, end-to-end models may need to include macrobiota species (wild and aquacultured) and may have to allow feedbacks (i.e. two-way coupling) between macrobiota and planktonic ecosystem dynamics. This is because the biomass of macrobiota can locally exceed the biomass of plankton, thus influencing the distribution of planktonic ecosystem tracers and altering the overall food web structure. Here, I describe a hybrid (Eulerian/Individual-Based) ecosystem framework, implemented in the Regional Ocean Modeling System (ROMS), a state-of-the-art 3-D ocean circulation model. The framework was applied to a model of a synthetic embayment containing seagrass, rockweed and kelp beds, a wild oyster reef, a mussel ranch and a fish farm. I found that two-way coupling is essential to reproduce expected spatial patterns of all variables and to conserve mass in the system. I also developed a shellfish ecophysiology model (SHELL E) and compared its results against water samples collected over 5 years in Ship Harbour, a fjord with mussel aquaculture in Nova Scotia, Eastern Canada. Also, from a high-resolution bio-optical survey of the fjord, I found that mussels decrease phytoplankton biomass inside the farm, but also cause a bloom of phytoplankton outside the farm. Using ROMS/SHELL-E, I determined that the increase of phytoplankton around the farm is caused by the waste products of the farmed bivalves, which have a fertilization effect, enhancing phytoplankton production outside the farm during nutrient-limited and light-replete conditions (i.e. late spring to late fall in Ship Harbour). The main conclusion of this thesis is that—in shallow coastal regions—ecosystem models must represent bilateral interactions between macrobiota and physical-planktonic dynamics, in a spatially-explicit setting, to adequately represent mass flows and ecosystem dynamics. The hybrid end-to-end modelling system provides a computationally efficient framework for describing these interactions and, through careful comparisons against observations, can be a powerful tool to test hypotheses and generate insights into coastal ecosystems.

L’objectif de cette thèse est d’améliorer la compréhension du fonctionnement et de la structure trophique du golfe de Gabès en Tunisie. Afin de concilier exigences écologiques et exploitation des ressources marines, différents modèles écosystémiques ont été développés pour étudier sa dynamique trophique et contribuer à la réflexion sur la mise en place de plans de gestion. Un modèle trophique d’équilibre de masse « Ecospace » a été construit afin d’évaluer les conséquences écosystémiques de différentes mesures de gestion. Les résultats des simulations ont permis d’explorer les interactions entre la pêche côtière et la pêche au chalut benthique et d’identifier des zones où les mesures de gestion sont effectives. Un modèle end-to-end a également été appliqué pour expliciter la dynamique des espèces considérées, depuis le forçage climatique jusqu'à la pêche. Cette approche de modélisation consiste à forcer le modèle individu-centré « OSMOSE » par un modèle biogéochimique « Eco3MMed ». Ce modèle a permis d’établir une représentation cohérente du réseau trophique et de simuler des scénarios de gestion théoriques de mise en réserve. Le modèle end-to-end a également été utilisé pour étudier la sensibilité d’un ensemble d’indicateurs écologiques à la pression de pêche. Les résultats ont révélé que les indicateurs de taille sont les plus adaptés pour faire le suivi de l’impact de la pêche dans le golfe de Gabès. Au final, une approche comparative entre plusieurs écosystèmes méditerranéens a été mise en place avec le modèle EcoTroph pour comparer leurs structures trophiques et explorer les effets de plusieurs niveaux d’exploitation par l’analyse de leurs spectres trophiques. / The objective of this thesis is to improve the understanding of trophic structure and functioning of the gulf of Gabes in Tunisia. In order to reconcile environmental concerns and exploitation of marine resources, different ecosystem models have been developed to study the ecosystem dynamics and contribute to the discussion on the implementation of management plans. A spatial and temporal dynamic model “Ecospace” was built to evaluate the ecosystem consequences of different management measures based on scenarios derived from the current regulation. The results of simulations allowed to investigate the interactions between coastal and benthic trawl fishing and to identify areas where management measures are effective. An end-to-end model has been applied to the gulf of Gabes ecosystem to represent the dynamics of 11 high trophic level species, from climate forcing to fishing.This modelling approach consists in forcing the individual-based model "OSMOSE" by a biogeochemical model "ECO3M-Med". This model allowed to establish a coherent representation of the food web and simulate theoretical management scenarios of spatial fishing closure. The end-to-end model has also been used to study the sensitivity of a set of ecological indicators to fishing pressure. The simulation of different levels of fishing mortality showed that size indicators were the most relevant to monitor the impact of fishing in the gulf of Gabes. Finally, a comparative approach between several Mediterranean ecosystems was applied using the EcoTroph model to compare their trophic structures and explore the effects of different levels of fishing pressure through the analysis of their trophic spectra.