The three sources of information about the subsurface are well logs, seismic data and geological experience. This paper presents an integrated approach in which these data are calibrated and combined to improve the characterization of the reservoir. The whole processing consists of three major steps: - calibration of poststack seismic wavelets using borehole data; - lateral extrapolation and/or interpolation of impedance well logs using seismic data and geological experience (modelling and inversion); - automatic seismic facies recognition using statistical pattern recognition techniques to provide a classification of the traces that will be related to the geological properties of the wells. The studied reservoirs covered by a 2D seismic data set correspond to several levels of channels in a fluvial environment. The main lithologies are various types of sandstones, shales, lignite and carbonate facies related to pedogenetic alteration.