Bayesian 4D inversion: integrating time-shift information

Abstract

A new inversion method for time-lapse seismic data is developed. The method is based on the 3-dimensional spatially coupled AVO inversion in the Fourier domain, describedin Buland et al. (2003b). Further, a technique for estimating the time-shift betweentwo seismic surveys, called non-rigid matching (Nickel & Sønneland, 1999), is an importantand integrated part of the method. In addition to using the technique to alinethe repeated survey with the baseline survey, the estimates produced by the non-rigidmatching are treated as observations together with the seismic observations.As the natural choice for many geophysical inversion problems, the inversion methodproposed is set in a Bayesian framework. Prior information in the form of well logscombined with seismic observations and an estimate of the time-shift, produce the posteriordistribution. The posterior distribution not only gives a best fit to the inversionproblem, but also a measure of the uncertainty.Synthetic data based on the Gullfaks field are used to test our time-lapse inversionmethod, including tests of robustness.