FWI of Synthetic Data from a Physical Modelling Facility Channel Model

Xiaohui Cai, Kristopher A. Innanen, Daniel O. Trad, Joe Wong

Full-Waveform Inversion (FWI) is a powerful seismic imaging technique capable of delivering high-resolution subsurface characterizations. While its potential in complex geological scenarios is well-established, challenges remain in its application to simpler models, due to issues like dependence on low-frequency data and vulnerability to local minima. This study aims to develop an efficient and accurate FWI workflow adaptable to a variety of models. Initially, a frequency-domain multiscale FWI approach was employed to evaluate its ability to recover channel geometry, interface features, and internal velocity distributions. While effective, its computational cost was high. To address this, a time-domain FWI workflow was implemented, which achieved comparable accuracy with significantly improved efficiency. Results demonstrate that integrating low-frequency information and adopting a progressive frequency strategy are critical for enhancing inversion stability and accuracy. This work provides a foundation for robust FWI methodologies applicable to both physical modeling data and real-world seismic datasets.