Computational frameworks for modelling, migration and inversion

Daniel O. Trad, Tianze Zhang, Ivan Sanchez

Over the past decade, scientific computing has witnessed significant advancements, including the introduction of new programming languages, pre-built libraries optimized for high-performance computing, and the rise of generative artificial intelligence, which provides innovative approaches to solving computational problems. Traditional methods for developing migration and inversion algorithms—built step-by-step in low-level languages—now face competition from modern frameworks that abstract many computational details, allowing researchers to focus on core challenges. While these frameworks often simplify initial development, their benefits may diminish as projects demand more detailed solutions and optimizations. However, they also introduce fresh ideas and possibilities that extend beyond conventional approaches.

This talk will explore current trends in modeling, migration (RTM), and Full Waveform Inversion (FWI) across diverse programming environments. We will discuss traditional frameworks, such as open-source compiled languages, for acoustic and elastic modeling, RTM and FWI from topography. Additionally, we will examine modern approaches, including Devito, Python-based methods with and without machine learning, physics-informed neural networks (PINNs), neural operators, generative artificial intelligence for programming and ongoing efforts to advance 3D modeling and FWI.