Non-Linear Full-Waveform Inversion Using Geological Prior Knowledge

Abstract

One of the important challenges in seismic inversion is to resolve finer structures from band-limited seismic data. Reservoir-oriented full-waveform inversion has the potential to deliver high-resolution quantitative images and is a promising technique to obtain macro-scale physical properties of the subsurface(Ashnashari et al., 2012). Because full-waveform inversion accounts for the entire wavefield, the seismic modelling embedded in the inversion algorithm honors the full physics of wave propagation (Virieux and Operto, 2009). This makes the technique potentially an effective instrument for improving the characterization of complex geological settings (Plessix et al., 2010). Like for most geophysical application, prior information such as data collected in wells is available and should be used to improve the result. For this purpose we propose a new strategy for including geological prior knowledge in full-waveform inversion, which will ensure an even higher resolution in the final images. This new scheme does not constrain the inversion but uses blocky models drawn from the prior distribution as a starting point for the inversion. After an unconstrained inversion, the non-blocky result is re-interpreted in terms of the prior model. This can be seen as a Bayesian update in iterative non-linear inversion, this process is repeated after every iteration. This updated blocky model will be used as a starting model for the next iteration. This leads to a guided, nonlinear inversion process, where a geological scenario is proposed between two linear iteration steps. Given the prior probabilities and covariances, we are able to interpret the presence or absence of thin layers that otherwise cannot be detected using only band-limited seismic data. This scheme is demonstrated on a high-resolution synthetic model based on the Book cliff outcrop in Utah(USA).

AAPG Datapages/Search and Discovery Article #90216 ©2015 AAPG Annual Convention and Exhibition, Denver, CO., May 31 - June 3, 2015