Abstract EN:

Ground Penetrating Radar is used in near surface acquisitions to determine several characteristics from the soil. The analysis of the data is generally complicated by the heterogeneity of the near surface layer. The influence of water is also very strong and a change is the water content of the soil can change the results. In order to improve the data analysis, we use a modelling program with a finite differences in time domain (FDTD) algorithm. We determine the shape of the source impulse to be as near as possible from the real results. At first, we use this program to model the structure in a peat basin in the French Alps. We discover a geological structure which may result from a progressive sinking of the basin nearby the cliff. Another study helps us to discriminate between karsts filled with clay and hollow karsts in limestone. Finally, we can follow the changes affecting a snow layer at several times during the winter. We mounted a laboratory experiment in order to reproduce a groundwater level in a sand layer at varying depths. We had no signal reflection on the top of the saturated sand layer. We show that the 3 commonly used relations linking the water content to the relative dielectric permittivity give similar results, both on the water contents and on the variations of the water volume in the sand. With the modelling of the profiles from the sans box, we can give some estimations for the dielectric parameters of the sand. At last, we simulated an accidental fuel pollution from the surface of the sand box. We show that the fuel is not homogeneously distributed in the sand. We have no reflection on the fuel saturated layer, and only the wave velocity is clearly affected by the presence of fuel. The plume is more concentrated in the vicinity of the injection point and is the origin of water movements in the sand. New signals appear after a longer time due to these changes. A modelling gives a good idea from the shape of the plume.