Modelling complex systems that cover multiple domains, for their better understanding, increasingly demands collaboration between different disciplines. However, these disciplines do not necessarily share the same points of view on the real objects of the system, and these can be complementary. In addition, the representation of such systems requires multi-scale description implying at least the local (individual), global and underlying environment. This PhD work proposes (1) a conceptual framework for complex systems analysis and representation from different points of view at the global and the local levels while taking into account the environment and (2) its application to the representation and analysis of the carbon dynamics from plot to village levels in the West African savannas (WAS). Using multi-agent system (MAS) organizations-centered approach, the Organization-Role- Entity-Aspect (OREA) meta-model has been proposed to represent a complex system from different points of view. At the global level a point of view is reified as an organization made of the roles the entities can play in the organization; at the local level the points of view decompose an entity internal structure in a set of aspects. Through the concept of role, an entity can play different roles in different organizations. Through the concept of aspects, an entity can play a role in different ways. OREA is an extension of the Agent-Group-Role meta-model where: (1) roles are not limited to agents but can be assigned to any kind of object (including the environment), (2) the decomposition does not only apply to the organizations only, but also to the entities themselves, (3) the use of the framework for knowledge representation rather than a pure software engineering paradigm is emphasized. OREA provides a framework to specify explicitly and separately the macro and the micro levels. The macro-level in OREA is specified without any assumption on the micro-level. The macro-level is relevant to