Recent notes(Older stuff is in the notebook.)
Image processing for faults
August 1, 2014
I am writing this note while travelling to Australia in the first leg of a worldwide lecture tour for the Society of Exploration Geophysicists. I am super excited to be doing this, for some obvious reasons, but also because this tour gave me a perfect excuse for new research of seismic image processing for geologic faults.
- Vertical slices of a 3D seismic image, before and after unfaulting.
The 3D seismic image of the onshore Schoonebeek oil field in the Netherlands was provided by Kees Rutten and Bob Howard, via TNO, a Dutch research organization. It's a geologically beautiful high-resolution image, with numerous intersecting faults. Displayed above are vertical slices of this 3D seismic image, before and after unfaulting. The vertical white line represents the intersection of orthogonal vertical slices through the 3D image.
The methods for locating faults and estimating fault slips are entirely automatic, and consist mostly of some well-known image processing techniques combined in ways that I describe in the SEG lecture.
Unfaulting here does not imply a reversal of geologic time, a restoration of the subsurface to some original unfaulted state. Rather, unfaulting is a simple way to assess the accuracy of the estimated fault locations and slips. After unfaulting, subsurface reflectors should appear more continuous, and this increased continuity is clearly apparent. But not everywhere, and shortcomings are opportunities for future research.
One shortcoming is that the footwall side of every fault has been constrained to not move during unfaulting. Where faults intersect, this constraint leads to some kinky unfaulted seismic reflectors. These do not affect our visual assessment of estimated fault slips, but how might we do better?
While thinking about that question, you can test this processing with your own 3D seismic images, because all of the image processing software used here is freely available with an open-source license. That license enables companies to use this software in commercial software products. Some have already done so, because an earlier version (now about two years old) has been available with the same open-source license.
If you download the latest version, note that it includes software for generating synthetic 3D images for testing. You can modify this software to answer some good questions. For example, what slips are estimated at locations where two conjugate faults intersect? (You can see such intersections in the image slices displayed above.)
If you do build and run the software, I hope you will let me know. And if you also attend the SEG lecture, please introduce yourself. I look forward to meeting you.