We present the results of a hydraulic tomography led on a 60 × 40 m² fractured and karstic field in Southern France in order to image, in a model, its transmissivity field. The dataset employed for the tomography consists in drawdown responses to cross-boreholes pumping tests reaching pseudo steady-state, with 8 different pumping wells and 22 measurement boreholes. The inversion of the dataset was led on a 2D model coupling a discrete network and a continuum, by following the Discrete Network Deterministic Inversion (DNDI) method. This method permits an optimization of both the transmissivity distribution and the structural geometry of the discrete network, which represents in this case the interconnected fractures and conduits in the aquifer. The optimized model obtained after inversion allows reproducing the observed drawdown in the field, and proposes a contrasted imaging of the hydraulic properties, as awaited in such fractured site. The fracture network in the optimized model also shows coherent orientations of fracturing, compared to the orientations effectively observed on the field, even though this information was not included in the inversion. A comparison of the results obtained with this coupled model to results obtained on the same data with equivalent porous media model (without integration of a discrete network) shows that the integration of a discrete network in the model greatly improves the ability of the model to reproduce the flows existing in such fractured fields, and thus the observed drawdowns.