Maritime safety and security require a constant surveillance of the traffic at sea, and several human actors are in charge of this control. Potential risks alerts, suspicious vessels and description of the situation are managed by visual surveillance systems. But these systems do not provide real analysis tools for maritime traffic data. Geovisualization and visual analytics have shown to be very efficient to handle big sets of heterogeneous data, and to make knowledge discovery easier to users. But advanced visual analytics environment can appear to be too complex to be used, according to user's skills. Surveillance tasks need to be formalized for having proper analysis tools. In our research toward user's skill adapted geovisualization, we propose here a study of visual surveillance tasks in maritime surveil-lance in a high-level taxonomy. Spatialization of the data will be processed according to these tasks, for proposing adequate visualization methods.