Integration of 3D Seismic Data in Reservoir Stochastic Simulations: A Case Study

Abstract

Abstract Integration of seismic data in stochastic models is a key point for reducing uncertainties on reservoir properties between wells. Stochastic simulations of the average porosity of a reservoir in North Sea, constrained both by well and seismic derived information, are discussed in this paper. Firstly, a non parametric regression based on a multivariate gaussian segmentation is used to calibrate at wells a set of seismic attributes in terms of average porosity. This technique allows to account for multivariate non linear relationships between seismic and porosity information, and also to fully quantify the uncertainties attached to the porosity estimation. Two approaches to account for the seismic derived information in simulations were compared. The first one is based on error cokriging where the seismic porosities constrain the simulations depending on their attached uncertainties. Seismic derived data significantly reduce the dispersions of the simulations, by comparison to the case where only well data participate. In the second approach, a cokriging between well and seismic porosities is used in the simulations. This allows to account for the spatial structure of seismic porosities and to better characterize well porosity spatial structure from its correlation with the dense seismic information. The non parametric regression appears as a way to linearize the seismic information before cokriging with well information.