Henri Calandra

Technical Advisor, Depth Imaging and High Performance Computing, Total

Monday, February 4, 2008
3:30-4:30 PM
LWSN 3102A/B, Purdue University

Title: Seismic Processing Challenge and Computational Roadmap

This text will be replaced

Abstract

   The extraordinary challenge that the oil industry must face for hydrocarbon exploration requires the development of leading edge technologies to reconstitute the three-dimensional structure of the Earth. The use of seismic reflections is one of these essential and strategic methods. Reflection imaging is key to seismic exploration because it is used as the basis on which the geologist bases his or her search for new oil deposits. Today, this analysis is made possible thanks to the extraordinary revolution in depth imaging techniques over the last 20 years. The industry has seen a fantastic leap in the ability to process seismic data and to build an increasingly accurate image of the Earth's structure. Seismic Depth imaging and High performance computing are both key components of this revolution. Indeed the fast evolution of computers has enabled the development of specialized algorithms allowing the processing of increasingly large volumes of data generated by seismic acquisitions.
   Depth imaging processing is an inverse problem. The objective is to reconstruct one velocity model representation of the sub-surface explaining the seismic input data. Inverse problems means iterative process loop. Today this loop is divided in 2 parts : Pre Stack Depth Migration (PSDM) and velocity model update based on Ray Tomography. The PSDM is currently the most common method employed for seismic depth imaging and the most CPU intensive. Ray Tomography is less CPU intensive than the PSDM but requires more human interactivity. Reducing the seismic processing cycle time is very important in exploration and requires integrated tools which allow the geophysicist and geologist to build the most accurate interpretation of the subsurface. This integration is made possible through a global methodology. This integrated vision of the depth imaging loop addresses the global data work flow and can have consequences in data organization: storage, communication, accessibility, but also with the development of new depth imaging technology.
   The development of these different technologies has followed very closely the progress in high performance computing. The massively parallel intrinsic nature of seismic data allows the geophysicist to develop very efficient algorithms. Parallelism can be applied on individual seismic traces, trace collections (e.g. shot gathers), or frequency planes when frequency domain methods are used. The last decade has seen successively the introduction of the full 3D Kirchhoff Pre-SDM, 3D SHOT profile PSDM based on more expensive approximations of the wave equation and more recently 3D Reverse Time Migration based on a full finite-difference solution of the acoustic wave equation. The implicit structure of these algorithms makes implementation very effective on a wide variety of architectures, including SMP, clusters and clusters of SMP. By introducing several levels of parallelism, we can push the limit of computations and of optimization. Recent progress both on interconnect topology and new accelerating technology (FPGA, GPGPU, CELL) open new R&D directions to push the limits of computation.

   Henri Calandra obtained his M.Sc. in mathematics in 1984 and a Ph.D. in mathematics in 1987 from the Universite des Pays de l'Adour in Pau, France. He joined Cray Research France in 1987 and worked on seismic applications for 2 years. In 1989 he joined the applied mathematics department of the French Atomic Agency. In 1990 he started working for Total SA. After 12 years of work in high performance computing and as project leader for Pre-stack Depth Migration Research, he became head of Total USA's Geophysics Research Group for 3 years in 2002 and coordinated Depth Imaging Research for the worldwide group until mid 2007. He is now technical advisor in depth imaging and High performance computing.

Co-sponsored by The Cyber Center, The Rosen Center for Advanced Computing, The Advanced Computer Systems Laboratory and The Computational Science and Engineering Programs.
Go back to CRI Seminar Home CRI Seminar #1 Spring 2008