Influence de l'injection d'un traceur sur les courbes de restitution en écoulement naturel

Tracer tests are usually interpreted with analytical models based on simplifying hypothesis or on very simplified numerical models. The objectives of this master thesis were to evaluate the influence of 1) injection parameters on tracer test results and 2) numerical model complexity on numerical results. Injection parameters tested in laboratory and by numerical models were the volume, the concentration and the injection rate of the injection solution. The tracer tests were conducted in an artificial confined aquifer (small-scale model, homogenised sand) and the numerical models were developed to simulate this aquifer. The results were breakthrough curves C(t) in the injection and observation wells. These were interpreted by an analytical model based on the breakthrough curve form from which the mean arrival time, the mean velocity and the dispersivity were determined. The overall average of the mean arrival time was used to evaluate a mean effective porosity of 0.35. The mean dispersivity showed a rise with distance: 4.5 mm at 46 cm, 6.7 mm at 92.5 cm and 15.1 mm at 137.5 cm. However, as this aquifer is mainly homogenous, this rise is the result of the distortion of the flow field around observation wells. The tracer tests also showed that there were tendencies between the injection parameters and the tracer test results. These tendencies were almost all confirmed by the 2D and 3D numerical models. However, the 3D numerical results were nearer of the observed tendencies because they considered the partially penetrating wells and the gravity effects. Adding or not the observation well in 3D also had a major effect on the numerical results at small scale. Therefore, a detailed numerical modeling of the observation wells is important when interpretation with numerical model is used. Furthermore, if the wells are partially penetrating, a 2D model is not enough acurate, a 3D model is requested.