In a well field of the Lyon metropolitan area, designed for drinking water supply, behaviour of each object (infiltration basins, wells, rivers, aquifer, and unsaturated zone) and their interactions are complex and not well-known. As a first step, infiltration tests at different spatial scales in one artificial basin were performed to better understand the basin operation and to estimate the hydrodynamic parameters of the unsaturated zone. Results of interpretation, using numerical methods based on Richards equation, reveal that infiltrated basin fluxes mainly depend on saturated hydraulic conductivity of a layer located just below the calibrated sand layer that cover the basin bottom. Indeed this layer has been estimated to be less permeable than the aquifer, which allows the existence of the unsaturated zone below. Pumping tests in the groundwater have been performed using production wells and a well specially implemented during this thesis work in order to estimate aquifer hydrodynamic parameters. Observations analysis and a conceptual modelling, in 2D and then in 3D, lead to a better understanding of the controlling mechanisms (stratification, input and output) and to simulate both basin infiltration rates and water table rise. Considering the whole basin scale, input fluxes are transient, related to the clogging statement of the infiltration area but also to the temperature of inflow water and the initial statement of the soil just below the basin. Sensibility analyses using the models highlight that the amount of the water table rise is mainly influenced by the aquifer saturated hydraulic conductivity and also by the location of imposed boundaries in the aquifer (rivers and pumping wells). The model properly accounts basin inflow fluxes and water table fluctuations. The model is able to verify if flows are reversed in relation to river exchanges, if wells are fed by infiltration basins and it highlights aquifer flows below the river. A 3D modelling has been realised in another area of the well field, comprising two infiltration basins, two river arms and pumping wells. In agreement with underflow in the aquifer, rivers are imposed in the model as third kind boundary conditions. Aquifer and river exchanges are calibrated with observed data of one aquifer flood-wave propagation. Significant differences of hydraulic heads have been observed at different depths of the aquifer using panpipes piezometers, specifically implemented, close to one infiltration basin. Theses differences are closely related to basin operation. These observations are properly calculated by the 3D model. Using the model, the effect of one sandy-clay heterogeneous layer (whose saturated hydraulic conductivity is lower than the ones of other aquifer lithologies) on aquifer flows (direction and flux) is notable. The model properly accounts basin inflow fluxes and water table fluctuations. The model is able to verify if flows are reversed in relation to river exchanges, if wells are fed by infiltration basins and it highlights aquifer flow below the river.
