In the field of environmental management, inefficient piecemeal sectorial approaches are still widespread, particularly in water resources management, where concerns and decisions are usually split among a set of rather weakly linked stakeholders. Planning and management problems to be addressed are in most cases complex and need to be tackled in a global, integrated and holistic way, including information from the different areas of science and knowledge from the various stakeholders. Therefore, the sharing and integration of data, knowledge and models has become a major issue. A lot of effort has been made in recent years to develop such integrative tools, including information systems (IS), decision support systems (DSS) or integrated modelling frameworks (IMF). Although these tools achieved great progress in terms of linkages, there is currently no environment providing a full integration of the various kinds of data, models, and their visualisation. This is the general issue tackled by the present thesis, with developments based on a semantic, systemic formalism, and applications focusing on water management issues. The first development step taken was the creation of a systemic description of water management. Albeit the various aspects of water management are abundantly documented and well-known, an integrated description including interconnections was lacking. Therefore, a graphic representation of interconnected system elements annotated with definitions was created. This innovative result constitutes an ontology (a formal representation of a knowledge domain) providing a robust reference and a fairly exhaustive reminder of the complexity of the water management field, including transdisciplinary aspects. In the second step, a data model was created, based on the new SYSMOD language. The latter proposes a new graphical language uniting the ontological and systemic approaches, with a strong emphasis on interactions. Whereas existing data models are generally specific to a field or designed for a target application, this new data model is generic and transdisciplinary. It can handle any kind of systemically organised data, which potentially includes any sort of environmental data. To visualise and manage the systems-based, generic data stored within the database (created on the basis of the generic data model) a new kind of visualisation was developed, proposing diagrams of interconnected nodes. Whereas existing information systems are often data warehouses without a user-friendly graphical user interface, or geographic information systems focussing on spatial data, this new diagram visualisation proposes views including spatial and non-spatial system elements (such as lakes, power plants, stakeholders or laws) interconnected by interactions (such as fluxes or influences). Bound to the database, the tool was called an "information system on the system" (ISS). It provides a way of graphically managing and integrating the systemically organised, generic, transdisciplinary data. Furthermore, the systems approach fosters the apprehension and communication of complex situations in a holistic way (involving many different elements, influences and interactions from various fields), through their representation into thematic views. In complement to the diagram visualisation, a reporting tool was developed to display the data linked to system elements (numeric values, comments, files, etc.) in a web-browser fashion. To interoperate with external models, another module was developed on the top of the ISS. Finally, a fully-fledged general user interface was created to integrate all the previous developments into a single environment. To visualise the different facets of system elements, a geographic and a chart view were added, in complement to the reporting view. This final tool was called the generic decision support system (GenDSS), as its application to given situations can give rise to a series of views and linkages to models, thereby providing specific DSS functionalities for these situations. The successive developments were tested in Birmingham (United Kingdom). Test results indicated that the GenDSS is currently limited to managing a few hundred elements; beyond this quantity, the loading times and the user interface reactivity can become long and unresponsive. On the other hand, the application also demonstrated the strengths of the GenDSS which provides an innovative integrated solution to centralise transdisciplinary data in various formats (e.g. numeric values, texts, geometries, files), to holistically describe situations using complementary visualisations, and to test scenarios using externally-linked models. Thus, the GenDSS may prove an adequately versatile solution for large institutions or enterprises dealing with complex situations, heterogeneous and transdisciplinary data, and where requirements are continuously evolving, such as: a car manufacturing corporation (which would involve several different production and research sites in different countries with different regulations), a hospital, or a strategic environmental planning network (where the tool could help to break the transdisciplinary, sectorial barriers). In the case of water, the GenDSS system could prove very useful to achieve a truly integrated water management approach.