Main components of full-waveform inversion for reservoir characterization

Building a 3D reservoir model, which has become a key part of reservoir management, is a challenging task. Classic seismic inversion techniques, both deterministic and stochastic, have attempted to reduce the uncertainty in reservoir modelling. However, these inversion methods are typically based just on amplitude and make a number of critical simplifying assumptions, such as that migrated data are accurate and can be modelled by 1D convolution (using reflection coefficients computed from the Zoeppritz equation or its linear approximations). Migration algorithms themselves may suffer from inadequate amplitude and multiple-scattering treatments. Full-waveform inversion (FWI), specifically designed for the direct estimation of the reservoir parameters, is proposed here as an alternative method for seismic reservoir characterization. This is clearly an ambitious goal, and it is not our intention to claim we have already achieved it. Instead, we intend to introduce the main components of such reservoir-oriented inversion and discuss a strategy for elastic, anisotropic FWI constrained by rockphysics models and facies types. Among the many challenges, we focus mostly on understanding the physics and describing some elements of an efficient forward-modelling engine. In particular, we show how analysing the radiation patterns helps to optimise the parameterisation and could reduce the null space.