Integration of Seismic Data and Uncertainties in the Facies Model – Application to the Snorre Field

Abstract

In this paper, we prove the benefit of a consistent and quantitative use of 3D seismic data to constrain the facies model built on a complex reservoir, Snorre, located in the Norwegian Sea. The workflow starts with the application of Bayesian seismic inversion to 3D angle stacks, allowing the estimation of elastic parameter probability distributions over the reservoir. Uncertainty decrease with respect to the prior model is quantified for each elastic parameter. The estimated uncertainties are then used input to design a filter which is applied to well logs to propagate seismic uncertainties in the next step: Seismic facies modeling. We use therefore a non-parametric probabilistic classification algorithm, supervised by the filtered well logs and blocked facies. The resulting sand and shale probabilities are then controlled at blind wells, with an overall match with facies proportions from wells around 70%. Finally, the probabilities are used as a quantitative 3D constraint to stochastic object-based channel modeling techniques. By drawing different realizations from this model, we prove that the use of seismic data allows better focusing the spatial distribution of the channels.