Maritime traffic surveillance requires a very accurate and continuous analysis of the sea. This area consists of many different objects, actors and rules. Monitoring such a wide and complex area requires adapted visual tools, such as Maritime Surveillance Systems. These tools help identifying abnormal behaviours, which can lead to risky situations. We investigate the usability of visual analytics and geovisualization methods that will improve these systems: a better synthesis of the data and more effective tools lead to improved situation awareness. Visualization methods and needs of controllers being very specific, there is no single solution for modelling, visualizing and analysing maritime data. In this paper, we identify the limits of current research in geovisual analytics for maritime surveillance. A new approach for guiding the selection of visual analytics methods is proposed. The profile of the user, the purpose of use and the situation to analysis are considered together; a knowledge-based system will guide the user toward the most suitable visualization methods.