Full-Waveform Inversion for time-lapse seismic data

This application is available only in the software release for members of SINBAD consortium.

This package is an application for full-waveform inversion of time-lapse seismic data sets using a modified Gauss-Newton inversion framework. Ideas from distributed compressive sensing and stochastic optimization are exploited to create improved time-lapse models and speed-up the inversion, respectively.

Author: Felix Oghenekohwo (foghenekohwo@eos.ubc.ca)

Date: February, 2015

Contents

Downloading & Dependencies

The code can be found in the SLIM sofware release under /applications/WaveformInversion/TimeLapseFWI.

The code has been tested with Matlab R2013a.

This code uses the following packages, also found in the tools part of the SLIM software release.

Running & Parallelism

All the examples and results are produced by the scripts found in this software release under applications/WaveformInversion/TimeLapseFWI/. Start matlab from that directory or run startup in that directory to add the appropriate paths.

To run the scripts follow the instructions in the README file enclosed with the code. The scripts can be run in serial mode.

Read more about SLIM's approach to parallel computing in Matlab.

Examples and results

An example of our joint FWI on the BG time-lapse model is shown here.

References

[1] Felix Oghenekohwo, Ernie Esser, and Felix J. Herrmann [2014]. Time-lapse seismic without repetition: reaping the benefits from randomized sampling and joint recovery. Presented at the 76th EAGE Conference and Exhibition.

[2] Felix J. Herrmann, Xiang Li, Aleksandr Y. Aravkin, and Tristan van Leeuwen, A modified, sparsity promoting, Gauss-Newton algorithm for seismic waveform inversion, in Proc. SPIE, 2011, vol. 2011.

[3] Xiang Li, Aleksandr Y. Aravkin, Tristan van Leeuwen, and Felix J. Herrmann, Fast randomized full-waveform inversion with compressive sensing. 2011. Geophysics.

Acknowledgements

Thanks to our sponsors and NSERC for their financial support. The synthetic time-lapse model was provided by BG Group.