Home ]

 

SWLGS Luncheon Topics

Updated August 05, 2003

September 2000

The Seismic Well Log: Wave of the Future?
Mike Graul, Texseis, Inc.


Abstract: Envision a large field of closely spaced seismic wiggly traces. Nothing unusual about that — it just describes a garden variety 3D seismic volume. But these wiggles are different; they are estimated well logs: SP, gamma ray, density, porosity, shale volume, …your choice. Not grossly smoothed traces with the usual coarse seismic resolution, but detailed enough to interest a geologist, maybe even a reservoir engineer.

Ridiculous, right? Maybe not. Recent developments in predicting rock properties and lithologic classification indicate real hope for high resolution seismic estimations. The Hampson Russell EMERGE process and the Seismic Attribute — Classification techniques of Tury Taner (Rock Solid Images, Inc.) are but two examples of the future having arrived ahead of schedule. The process, PreLog, described here, gives us a look at the inner workings and enormous potential of the trace-to-log process.

The skeptic may wonder how the seismic tool, which has been likened to a yardstick in measuring paper thickness, can suddenly become a microscope. Firstly, it hasn't been so sudden. Taner's work, for example, goes back to the early 70's. Secondly, the lengthy gestation has produced a synergistic blending of seismic attributes and modern statistical methods that remove the Bandwidth Barrier restricting resolution to a feeble 50 - 100 feet.

The usable seismic attributes are virtually unlimited: inversion, envelope, the stacked (migrated) trace, instantaneous phase and/or frequency, AVO attributes (e.g., intercept, gradient, Fluid Factor), inter alia. Each of these readily available attributes may be treated with such nonlinear operations as squaring or exponentiation to increase the list by an order of magnitude. Add to these trace attributes the volume characteristics, such as coherency, and the input possibilities are impressive. The statistical techniques include the mysterious and ever-popular neural networks, as well as the venerable multivariate linear regression, with an obligatory dose of convolution.

Designing and training the prediction operators with known well log data, validating with logs held in reserve, and finally, application to the seismic trace volume, has resulted in stand-alone estimates of such reservoir characteristics as shale volume, porosity, and pore fluid, which have proven surprisingly accurate. It would also appear that the technique will appeal to the accountant as well as the interpreter. Cheap and effective. In the right hands it could be a grease-finding juggernaut.

Biographical Sketch: Mike Graul has been at the seismic game for over 40 years, having begun his quest for knowledge as a Chevron geophysicist in New Orleans, in 1957 — at the age of 9. He led the typically nomadic life of a seismic crew worker, finally settling in Houston, where he completed his tour of duty with Chevron after some 23 years. Drawing upon his experience in research, processing, and interpretation, he formed a training/consulting company (Exploration Education Consultants, Inc.), and later, a seismic processing company, Texseis, Inc., where he serves as a partner, with responsibility for research and development. Mike has taught a wide variety of seismic courses for the SEG, AAPG, University of Houston, SPE, as well as numerous oil companies, large and small. He is a member of the SEG, EAEG, GSH, DGS, IEEE, and is a graduate of Rensselaer Polytechnic Institute.