Integrated Probabilistic Seismic Reservoir Characterization Workflow

The application of a multi-step process of reservoir characterization conditioned with seismic-derived attributes and associated uncertainty is presented. The workflow consists of: a log-facies classification integrating petrophysical properties derived from formation evaluation analysis and elastic properties computed through a rock physics model; a bayesian linearized seismic inversion; a probabilistic estimation of petrophysical properties; and a seismic facies classification. This methodology introduces some improvements with respect to traditional workflows: log-facies are discriminated and classified in a petro-elastic space, both in depth and time domain, handling scale changes for the elastic logs and the discrete log-facies; the rock properties distribution is described by a Gaussian Mixture Model, rather than a Gaussian Model; the conditional probabilities of elastic properties are estimated at coarse scale taking into account the uncertainty associated to the scale change. This conditional probability is combined with the probability of elastic properties from the Bayesian inversion to obtain the posterior probability of petrophysical properties. Then, litho-fluid classes are identified based on petrophysical properties probabilities and on log-facies classification. Since log-facies are coherent both in the geological and geophysical domain, probability volumes of petrophysical properties and reservoir log-facies are easily integrated in the hierarchical reservoir modelling workflow.