Summary In this paper, we investigate the use of spectral decomposition and facies classification on time-lapse data related to a Brazilian pre-salt carbonate reservoir. Synthetic seismic data were generated through petro-elastic modeling (PEM), which is based on a representative geological model and flow simulator dynamic properties. Reservoir pressure and saturation distributions are used that corresponds to periods of time, in which, real time-lapse seismic data will be available. The spectral decomposition method used, is based on a modified matching pursuit algorithm. At this stage, we focus on the interpretation of the spectral decomposition, mainly on geometric effects.