Geographic information systems (GIS) are powerful tools to manage, analyse and represent spatial data. They are used in various disciplines, including marine biology. One of the most important phenomena intensively studied by marine biologists is the dynamics of fish. This is partly because there is an increasing need for sustainable management of fisheries which are very important in the economy of coastal zones. Fish aggregations are a fundamental component of these dynamics and should be better understood to establish efficient recovery strategies in the context of declining aquatic resources. However, the traditional representations of fish aggregations do not model those aggregations explicitly as spatial objects. Moreover, despite many interesting capabilities of current GIS, these tools are unable to handle the tridimensional, dynamic and fuzzy nature of fish aggregations. The main objective of this thesis is to propose new approaches to improve the representation of fish aggregations at the regional and local scales. In 2D, the proposed approach is based on the fuzzy spatial objects models, which are based on the fuzzy sets theory. It uses a vector data structure to delineate the maximal and minimal extents of fish aggregations and a raster data structure to model the gradual transition which exists between these boundaries. In 3D, the proposed approach for the representation of fish aggregations is based on the dynamic Delaunay tetrahedralisation and the 3D α-shapes clustering algorithm. The integrated algorithm allows automatic detection of the fish aggregations contained in a dataset. 3D models also allow amongst other things the study of the morphological properties of the different aggregations. Testing these approaches with fisheries data (e.g. datasets from scientific surveys, acoustic datasets) revealed several benefits and limitations which are discussed throughout this thesis.