Karst aquifers represent an important groundwater resource world-wide. They are highly vulnerable to contamination due to fast transport through the system and limited attenuation of contaminants. The two main hydrogeological approaches developed for studying flow and transport are: inference of the system structure from karst spring hydrographs and chemographs; numerical modelling of flow and transport using a theoretical distribution of flow and transport field parameters. These two approaches lack of validation by detailed field measurements and observations. The main objective of this thesis is to "fill the gap" existing between field and model data. Observations of flow and transport parameters at several locations within the system were used to develop a conceptual model. This model was then compared to the existing models. The main field test site is the Milandre karst aquifer, located in the Swiss tabular Jura. Natural tracers (major ions, oxygen-18, specific conductance) and discharge were measured on the underground river, its main tributaries, percolation waters, and the main spring. These data were collected on a long-term basis in order to assess the spatial variability of the parameters, and on a short time scale (i.e. flood events) in order to investigate the dynamic processes. Complementary sites (Brandt and Grand Bochat) were used for more observations at the base of the epikarst. The proposed conceptual model considers four sub-systems: the soil zone, the epikarst, the unsaturated zone, and the phreatic zone. Each has its own specificity with respect to flow and transport. The soil zone controls the actual infiltration into the system. It contributes efficiently to groundwater storage. It mixes quickly stored water with fresh infiltrated water. Its thickness determines land-use: thick soils are generally cultivated whereas thin soils are under forested areas. The solutes concentration of soil waters depends on land-use for pollution-related paramet.