Joint facies and reservoir properties inversion

I have developed a joint inversion of seismic data for the simultaneous estimation of facies and reservoir properties, such as porosity, mineralogy, and saturation. The inversion method is a Bayesian approach for the joint estimation of facies and reservoir rock/fluid properties based on the statistical assumption of mixtures of nonparametric distributions of the model parameters. A mixture distribution is a convex combination of distributions. In the approach, I use a mixture of n nonparametric distributions, where n is the number of facies, and the weights of the combination represent the probability of the facies. The statistical assumptions in the proposed Bayesian inversion allow modeling multimodal and nonsymmetric distributions of the model parameters because they are not restricted to specific shapes of the probability distributions and allow modeling multimodal distributions caused by the presence of different facies and nonlinear relations between reservoir properties and elastic data. The inversion can be applied to elastic properties from well logs or derived from seismic data, and they can be combined with traditional Bayesian linearized AVO inversion. The method provides the point-by-point posterior distribution of facies and reservoir properties, as well as the most-likely models and its associated uncertainty, and it is successfully applied to real data.