Next: VELOCITY ESTIMATION IS NOT Up: Claerbout: Everything depends on Previous: INTRODUCTION

AVOIDING CHIMERAS

Some would say that seismic structural imaging is finished. They would say the future lies in linking the seismic image with porosity or permeability. They would interpret the AVO (Amplitude Verses Offset) to estimate it.

I was present at the creation.'' I was a consultant to Chevron Overseas Inc. when Bill Ostrander did his founding work on AVO. People asked me then whether I didn't feel a conflict, whether I didn't want to pursue this exciting new area at the university. After all, I was in on the ground floor. My answer was no and remains no. Even at the beginning we quickly learned that AVO often fails. How often does it fail? At a university we cannot know. I suppose some companies keep good records of their AVO failures while others forget their failures as quickly as they can. The reason I didn't want to jump on the AVO bandwagon is that universities don't drill holes. We would never know if we were right or wrong.

Since then we have gotten a pretty clear idea why AVO often fails. We learned a lot from the PhD thesis of Einar Kjartansson. I took his key ideas and results for my textbook, Imaging the Earth's Interior. Both Einar's thesis and my book are freely available at the SEP web site. Maybe I'll attach a few of Einar's results here to remind you. The bottom line is this: After the seismic waves encounter some V(x,y,z) which isn't merely v(z), the AVO analysis becomes extremely tricky.

Simply stated, the goal of reservoir characterization is to depict the extent of the reservoir, mathematically this is the lateral heterogeneity of the reservoir. But to do this from the earth's surface, we need to be able to compensate for the lateral heterogeneity everywhere above the reservoir.

A bright young man recently showed us his innovative statistical analysis leading to a 3-D color display of clay probability .I was impressed. (I'm from the old school where the weather man predicts whether it will rain. He doesn't tell us the probability of rain.) We asked the young man, how did you find the impedance?" He replied that he had used a commercial package from a very respected firm. I do respect that firm so they must have a disclaimer somewhere that says, garbage in, garbage out". What happened to the young man is that he had a little carbonate above his reservoir. His image at small offset was different from that at larger offset. Not merely the Amplitude was different. The whole blessed image was different. The impedance program failed to tell the young man that his data was unsuitable for impedance estimation. Who'd want to purchase a program that failed to produce its product? Better to sell it with some fine print.

Some companies envision us as the ones who should solve the problem and pass the solution along to the contractors. An alternative would be for us to set our sights lower and try to establish standards for estimating the quality of other people's impedance estimates.

As I said, an academic like me cannot really give a fair and balanced picture of AVO. I want to give you a fair picture that is not wholly my own opinion. The biggest bloodbath in corporate research this year might tell us something. I was told that everyone in the imaging group got offers (though most spurned them), the rock physics group survived (though its lab is in moth balls), but everyone in the AVO group got fired (got the package''). I was surprised by this result. Doesn't the AVO group produce the much sought-after porosity?'' I asked. The answer given was this: AVO is easily over sold". So there you have it. In industry there is accountability, which is absent at universities. We play poker for pennies, but they gamble their careers. I'm agreeable if some of my students want to work on AVO, but only if they are among the strongest students.

Two-dimensional marine data often has regular data coverage. Land data and three-dimensional data never really does. To be honest about our day-to-day activities with seismic amplitude, we do not spend our time making small and subtle measurements of amplitude. We spend our time preventing massive amplitude fluctuations that are associated with irregular data coverage. We often spend our time trying to hide the data-acquisition footprint.

Never-the-less, AVO remains an active area of industrial activity and research. There must be successes and those successes may be more secret than the failures. Today AVO may give us rock properties in simple situations (e.g. v(z)). One of our challenges is to make those estimates reliable in more complex areas. We know that those estimates are a very sensitive function of V(x,y,z). That's my claim. Everything depends on V(x,y,z).

Next: VELOCITY ESTIMATION IS NOT Up: Claerbout: Everything depends on Previous: INTRODUCTION
Stanford Exploration Project
4/19/1999