Hard-rock aquifers (HRA) have long been considered to be two-layer systems, with a mostly capacitive layer just below the surface, the saprolite layer, and a mainly transmissive layer underneath, the fissured layer. Although this hydrogeological conceptual model now attracts a wide consensus within the scientific community, it is difficult to prove that it can be efficient in terms of deterministic modelling, especially with an equivalent porous medium model, w hich would not be the first choice for “fractured aquifers”. In order to investigate the storage capacities of both layers and the recharge, a deterministic hydrogeological model was used on a HRA to calibrate the specific yield of each layer. The application site is the Plancoët migmatitic aquifer located in north-western Brittany, France, with piezometric data from 40 observation wells surveyed every two weeks for eight years. When the variance, the most common quality-of-fit criterion used in inverse calibration methods, was found to be inefficient for calibrating the specific yield and the recharge together, Advar , a new quality-of-fit criterion based on the seasonal piezometric amplitude variation, was developed and calculated for a set of 512 simulation results. Contrary to the variance, it is able to define the best values for the recharge and the specific yield in each layer. Adding to the conclusion that this two-layer model is well suited to such a HRA, it is demonstrated that the saprolite is about 2.5 times more capacitive than the fissured layer.