Estimating structural uncertainties in seismic images using equi-probable tomographic model

Assessing the uncertainty on the structural information contained in seismic images is critical for risk analysis in reservoir delineation, reserve estimation, and well planning. We propose here an original approach aiming at assessing structural uncertainties associated to ray based tomography. While it has similarities with formerly published approaches it is based on the random generation of equi-probable tomographic models rather than on randomly sampling the a posteriori “probability density function”. Moreover it is associated with non-linear slope tomography which allows considering some non-linear aspects of the problem. We think these two aspects offers significant advantages in terms of efficiency and accuracy. In this paper we carefully review the concepts and definitions (in particular the notions of confidence region and error bars), and then present our approach and discuss its advantages. We finally present an application to a North Sea dataset where we estimate errors bars for a target horizon.