FWT2D: A massively parallel program for frequency-domain full-waveform tomography of wide-aperture seismic data-Part 2

  • Authors:
  • Florent Sourbier;Stéphane Operto;Jean Virieux;Patrick Amestoy;Jean-Yves L'Excellent

  • Affiliations:
  • Géosciences Azur - CNRS - IRD - UNSA - UPMC, Sophia-Antipolis, France;Géosciences Azur - CNRS - IRD - UNSA - UPMC, Villefranche/mer, France;Laboratoire Géophysique Interne et Tectonophysique, BP 53, 38041 Grenoble Cedex 9, France;ENSEEIHT-IRIT, BP 7122 - F31071, Toulouse Cedex 7, France;INRIA, Laboratoire de l'Informatique du Parallélisme, Université de Lyon, CNRS-ENS Lyon-INRIA-UCBL, France

  • Venue:
  • Computers & Geosciences
  • Year:
  • 2009

Quantified Score

Hi-index 0.00

Visualization

Abstract

This is the second paper in a two-part series that describes a massively parallel code that performs 2D full-waveform inversion of wide-aperture seismic data for imaging complex structures. We present several numerical validation of the full-waveform inversion code with both canonical and realistic synthetic examples. We illustrate how different multiscale strategies can be applied by either successive mono-frequency inversions or simultaneous multifrequency inversions and their impact on the convergence and the robustness of the inversion. We present a scalability analysis using a real marine data set recorded by a dense array of ocean bottom seismometers to image the crustal structure of a subduction zone. We obtained a speedup of 20 when using 50 processes on a PC cluster which allowed us to iteratively invert 13 frequencies of the full data set in less than 2 days. This computational performance will allow in the future more extensive analysis of full-waveform tomography methods when applied to representative case studies or when considering 3D geometries.