Bayesian Networks are useful tools for modeling interactions and predictions in social-ecological systems since they offer a robust theoretical framework towards risk and uncertainty management problems through the use of probabilities. Furthermore, this theory gives the possibility to combine expert knowledge and data. That's why they have been successfully used for helping resource-management decision-making process in numerous case studies. We propose to apply this approach in order to deal with the marine protected areas governance issue. A first model of Bayesian Network has already been developed from the French Polynesia case study concerning fisheries response to regulations in Moorea Island. This step allowed us to think about a more comprehensive model, which would encompass the ecological, economical and institutional components that underlie the understanding of the marine protected areas governance issue. Therefore we derived a second model from six case studies: three Mediterranean and three in French overseas. The first objective was to draw through the structure of the Bayesian Network a synthetic and comparative framework that represents the expert knowledge relative to the marine protected areas implementations and their consequences on the different components of the social-ecological system. The second objective is to simulate governance scenario for a particular case study-as the impacts of different regulation measures on the resources and biodiversity conservation of the ecosystem or on the satisfaction of users like fishermen or tourists-once the parameters of the model have been set up by using both database and expert judgment.