SC-11: Full-Waveform Inversion in Practice: Black Box or Pandora's Box?

The course targets a range of experience levels, from students and young professionals to more experienced practitioners, looking to deepen their understanding of current full-waveform inversion (FWI) implementations and FWI products. Both background theory and practical considerations will be covered to support this objective. The course is not aimed at experienced researchers, and it will not cover algorithmic optimisation and coding details.

Introduction — What is FWI, and how does it relate to seismic imaging and inversion methods?

FWI Implementations — This section reviews the different FWI implementations, their applications, and their requirements. We then step through the key stages of the algorithm:
• Seismic wavefield modelling and the definition of the objective function.
• From the FWI gradient to a model update, including examples of optimisers.
• Constraints, pre-conditioners, and crosstalk considerations.
• Output properties and resolution expectations.

Practical Considerations — This section focuses on how to use FWI in practice:
• Input data requirements, including how to prepare the seismic data and starting models, as well as crucial wavelet considerations.
• Key parameters to set FWI up for success.
• How seismic acquisition design impacts FWI, and how FWI requirements shape acquisition, including bandwidth and sampling considerations.
• Computational requirements.
• Multi-level quality control.

Conclusions — Summary and future directions for FWI.

The concept of full-waveform inversion (FWI) to recover subsurface information from seismic data was published more than 40 years ago. Thanks to the advances in algorithm design and compute power, practical applications have gradually emerged in the last 15 years. From diving-wave FWI, now an established seismic velocity model building tool, to recent implementations of elastic multi-parameter FWI that directly solve for rock properties from field data, FWI has realised its original potential and evolved into what is now considered the future of seismic imaging. The objective of this course is to review the practical implementation of FWI, from input data requirements and parameterisation to expected outputs and quality control, while also covering relevant theoretical aspects. The course explores when, why, and how to use FWI for seismic imaging, structural and quantitative interpretation, thereby providing the geophysicist with a clear understanding of FWI-derived products and their applications.