Computational Techniques in Earth Sciences: from Modelling to Inversion

The EAGE Local Chapter Paris is pleased to welcome you to this webinar focusing on recent advances in computational techniques for Earth science applications. As the complexity and scale of geophysical datasets continue to grow, so does the need for innovative modelling and inversion strategies that can meet modern scientific and computational challenges.

This session will feature two talks delivered by experts and will highlight cutting-edge approaches at the intersection of numerical modelling, data analysis and experimental seismology. Together, these contributions showcase the potential of advanced computations to transform how we model, simulate, and interpret processes in the Earth’s interior.

Speakers: Dorian Soergel (Viridien), Dirk-Jan van Manen (ETH Zurich)

Agenda

Immersive Experimentation: Enabling Hybrid Physical-Numerical Wave Propagation Experiments in Real-Time (30 min)

by Dirk-Jan van Manen (ETH Zurich)

This talk presents an overview of the research carried out at the Centre for Immersive Wave Experimentation, an ETH Zurich lab established in 2016 to investigate the immersion of acoustic and elastic wave propagation experiments in numerically simulated environments using so-called immersive boundary conditions. In this approach, waves propagating in the physical experiments drive the numerical simulations and, vice-versa, waves propagating in the simulations drive the experiments.

Enabled by a combination of active wave control on the boundaries and by exchanging boundary conditions in real-time through a low latency high-performance computing system, not only are boundary reflections suppressed, but reflections from scatterers in the virtual numerical environment included. Our exact, full-wavefield approach even facilitates multiple scattering between any physical and virtual scatterers. Several examples of acoustic and elastic immersive wave experiments will be presented.

Simultaneous Bayesian Inversion: adapting Bayesian Inversion to Big Data (30 min)

by Dorian Soegel (Viridien)

Bayesian Inversion is a very powerful technique in the presence of non-linearity and complex model spaces, but its computational cost is high. This talk will focus on one of the methods recently developed to circumvent the limitations of bayesian inversion for large datasets, which is called simultaneous bayesian inversion. The strengths and limitations of the method are showcased on a new global shear-wave velocity tomography of the upper mantle.