Model entropy constraints in multi-parameter FWI: A promising tool for time-lapse FWI
Anton Ziegon, Kristopher A. Innanen
Seismic full-waveform inversion (FWI) holds significant potential for the cost-effective and reliable monitoring, measurement, and validation of the carbon dioxide plume within carbon sequestration projects. However, this form of seismic imaging has the drawback of non-unique solutions, meaning that infinite sets of model parameters explain the observed data sufficiently good. Therefore, model regularization has to be introduced to promote meaningful models, for example models close to a reference model or smooth models. This report explores the application of a relatively novel joint inversion approach, namely joint minimum entropy (JME) constraints, as a regularization scheme within multi-parameter elastic FWI. The synthetic study shows that entropy constrained FWI outperforms conventional minimum length regularization and cross-gradients structural coupling in terms of model reconstruction and data fit. The JME and minimum entropy (ME) regularization is identified as a promising tool to promote more focussed key features in the final images, which ultimately improves subsurface characterization. A simple synthetic carbon sequestration time-lapse imaging example showed considerable improvements in the time-lapse differencing when the inversion is regularized with model entropy constraints, and therefore underlining the potential for possible time-lapse applications. This study sets the basis to exploit the beneficial JME and ME property and the results suggest further testing on multi- and single-parameter time-lapse FWI, which will help to reveal the full potential of model entropy constraints.