Résumé
Drinkable water storage is almost completly composed of subterranean water. A great part of this water comes from the karstic aquifer which is developed in carbonated rocks, leading to huge cavities. The originality of the karstic aquifer comes from its heterogeneous structure. Because of this heterogeneousness, classical analytic methods can't be used with success. Therefore we have applied a neural network based approach for modelling this well studied problem and achieved interesting results